mentalhealth andabortionreportofthe apataskforceon mentalhealth andabortion...

Reportof the

APATaskForceon

MentalHealthandAbortion

APATask Force onMental Health and Abortion

Brenda Major, PhD, Chair

Mark Appelbaum, PhD

Linda Beckman, PhD

Mary Ann Dutton, PhD

Nancy Felipe Russo, PhD

Carolyn West, PhD

A copy of the report is available online at http://www.apa.org/pi/wpo/mental-health-abortion-report.pdf

Suggested bibliographic reference: American Psychological Association, Task Force on Mental Health and Abortion. (2008). Report of the Task Force onMental Health and Abortion. Washington, DC: Author. Retrieved from http://www.apa.org/pi/wpo/mental-health-abortion-report.pdf

Also APA reports synthesize current psychological knowledge in a given area and may offer recommendations for future action. They do not constituteAPA policy nor commit APA to the activities described therein.

Copyright © 2008 by the American Psychological Association. This material may be reproduced in whole or in part without fees or per-mission provided that acknowledgment is given to the American Psychological Association. This material may not be reprinted, translated,or distributed electronically without prior permission in writing from the publisher. For permission, contact APA, Rights and Permissions,750 First Street, NE, Washington, DC 20002–4242.

Printed in the USA.

REPORTOFTHEAPATASK FORCEONMENTALHEALTHANDABORTION

Table of Contents

Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Definitions and Scope of Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Questions Addressed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Importance of Recognizing Variability in the Abortion Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Conceptual Frameworks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10AbortionWithin a Stress-and-Coping Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Abortion as a Traumatic Experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

AbortionWithin a Sociocultural Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Abortion and Co-Occurring Risk Factors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Systemic risk factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Personal risk factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Summary of Conceptual Frameworks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Methodological Issues in Abortion Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Comparison/Contrast Groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Co-Occurring Risk Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Sampling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Measurement of Reproductive History and Problems of Underreporting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Attrition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

OutcomeMeasures:Timing, Source, and Clinical Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Other Statistical Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Interpretational Problems and Logical Fallacies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Summary of Methodological Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Review of Scientific Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Search Strategy and Criteria for Inclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Descriptive Overview of Literature Identified for This Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Review of Comparison Group Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Record-Based Studies and Secondary AnalysesWith Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Medical records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Secondary analyses of survey data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Evaluation of record-based and secondary analyses studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

Summary of medical-record and secondary analyses studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

Report of the APA Task Force onMental Health and Abortion 1

Comparison Group Studies: Primary Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Description of findings:U.S. samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Description of findings:Non-U.S. samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Evaluation of primary data comparison group studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Studies of Abortion for Reasons of Fetal Abnormality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Description of findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

Evaluation of fetal abnormality studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Review of Abortion-Only Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72Prospective Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Retrospective Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Summary and Evaluation of Abortion-Only Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87The Relative Risks of Abortion Compared to its Alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Prevalence of Mental Health Problems AmongWomen in the United StatesWho Have Had an Abortion. . . . . . . . . . 90

Predictors of Individual Variation in Responses Following Abortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

Conclusions and Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Endnotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

List of TablesTable 1A:Medical-Record Studies:U.S. Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

Table 1B:Medical-Record Studies: International Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Table 2: Secondary Analyses of Survey Data:U.S. Samples and International Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Table 3A:Primary Data Comparison Group Studies:U.S. Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

Table 3B: Primary Data Comparison Group Studies: International Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

Table 4:Abortion for Reasons of Fetal Anomaly. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Table 5:U.S. Samples of Abortion Group(s) Only:No Comparison Groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Table 6:Population Estimates of Proportion of allWomen andWomen Identified as Having Been PregnantExceedingCES-D Clinical Cutoff Score,National Longitudinal Survey of Youth: 1992 . . . . . . . . . . . . . . . . . . . . . . 91

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

2 Report of the APA Task Force onMental Health and Abortion


EXECUTIVE SUMMARY

The Council of Representatives of the American Psy-chological Association charged the Task Force onMental Health and Abortion (TFMHA) with “collect-ing, examining, and summarizing the scientific re-search addressing the mental health factors associatedwith abortion, including the psychological responsesfollowing abortion, and producing a report basedupon a review of the most current research.” In con-sidering the psychological implications of abortion,the TFMHA recognized that abortion encompasses adiversity of experiences. Women obtain abortions fordifferent reasons; at different times of gestation; viadiffering medical procedures; and within different per-sonal, social, economic, and cultural contexts. All ofthese may lead to variability in women’s psychologicalreactions following abortion. Consequently, globalstatements about the psychological impact of abortionon women can be misleading.

The TFMHA evaluated all empirical studies publishedin English in peer-reviewed journals post-1989 thatcompared the mental health of women who had an in-duced abortion to the mental health of comparisongroups of women (N=50) or that examined factorsthat predict mental health among women who havehad an elective abortion in the United States (N=23).This literature was reviewed and evaluated with re-spect to its ability to address four primary ques-tions: (1) Does abortion cause harm to women’smental health? (2) How prevalent are mental healthproblems among women in the United States whohave had an abortion? (3) What is the relative riskof mental health problems associated with abortioncompared to its alternatives (other courses of actionthat might be taken by a pregnant woman in similarcircumstances)? And, (4) What predicts individualvariation in women’s psychological experiences fol-lowing abortion?

A critical evaluation of the published literature re-vealed that the majority of studies suffered frommethodological problems, often severe in nature.Given the state of the literature, a simple calculationof effect sizes or count of the number of studies thatshowed an effect in one direction versus another wasconsidered inappropriate. The quality of the evidencethat produced those effects must be considered toavoid misleading conclusions. Accordingly, the


TFMHA emphasized the studies it judged to be mostmethodologically rigorous to arrive at its conclusions.

The best scientific evidence published indicates thatamong adult women who have an unplanned pregnancythe relative risk of mental health problems is no greaterif they have a single elective first-trimester abortion thanif they deliver that pregnancy. The evidence regardingthe relative mental health risks associated with multipleabortions is more equivocal. Positive associations ob-served between multiple abortions and poorer mentalhealth may be linked to co-occurring risks that predis-pose a woman to both multiple unwanted pregnanciesand mental health problems.

The few published studies that examined women’s re-sponses following an induced abortion due to fetal ab-normality suggest that terminating a wanted pregnancylate in pregnancy due to fetal abnormality appears to beassociated with negative psychological reactions equiva-lent to those experienced by women who miscarry awanted pregnancy or who experience a stillbirth ordeath of a newborn, but less than those who deliver achild with life-threatening abnormalities.

The differing patterns of psychological experiences ob-served among women who terminate an unplannedpregnancy versus those who terminate a planned andwanted pregnancy highlight the importance of takingpregnancy intendedness and wantedness into accountwhen seeking to understand psychological reactions toabortion.

None of the literature reviewed adequately addressedthe prevalence of mental health problems amongwomen in the United States who have had an abortion.In general, however, the prevalence of mental healthproblems observed among women in the United Stateswho had a single, legal, first-trimester abortion for non-therapeutic reasons was consistent with normative ratesof comparable mental health problems in the generalpopulation of women in the United States.

Nonetheless, it is clear that some women do experiencesadness, grief, and feelings of loss following terminationof a pregnancy, and some experience clinically signifi-cant disorders, including depression and anxiety. How-ever, the TFMHA reviewed no evidence sufficient tosupport the claim that an observed association betweenabortion history and mental health was caused by theabortion per se, as opposed to other factors.

This review identified several factors that are predictiveof more negative psychological responses following

first-trimester abortion among women in the UnitedStates. Those factors included perceptions of stigma,need for secrecy, and low or anticipated social supportfor the abortion decision; a prior history of mentalhealth problems; personality factors such as low self-es-teem and use of avoidance and denial coping strategies;and characteristics of the particular pregnancy, includ-ing the extent to which the woman wanted and feltcommitted to it. Across studies, prior mental healthemerged as the strongest predictor of postabortion men-tal health. Many of these same factors also predict nega-tive psychological reactions to other types of stressfullife events, including childbirth, and, hence, are notuniquely predictive of psychological responses followingabortion.

Well-designed, rigorously conducted scientific researchwould help disentangle confounding factors and estab-lish relative risks of abortion compared to its alterna-tives, as well as factors associated with variation amongwomen in their responses following abortion. Even so,there is unlikely to be a single definitive research studythat will determine the mental health implications ofabortion “once and for all” given the diversity and com-plexity of women and their circumstances.



INTRODUCTION

Although the U.S. Supreme Court legalized abortionin the United States more than 35 years ago (Roe v.Wade, 1973), it continues to generate enormous emo-tional, moral, and legal controversy. Over the last twodecades, one aspect of this controversy has focused onthe effects of abortion on women’s mental health(Bazelon, 2007; Cohen, 2006; Lee, 2003). Public de-bate on this issue can be traced to 1987, when then-President Ronald Reagan directed then-SurgeonGeneral C. Everett Koop to prepare a Surgeon Gen-eral’s report on the public health effects (both psycho-logical and physical) of abortion. After conducting acomprehensive review of the scientific literature, Dr.Koop declined to issue a report; instead, he sent a let-ter to President Reagan on January 9, 1989, in whichhe concluded that the available research was inade-quate to support any scientific findings about the psy-chological consequences caused by abortion (Koop,1989a). In subsequent testimony before Congress, Dr.Koop stated that his letter did not focus on the physi-cal health risks of abortion because “obstetricians andgynecologists had long since concluded that the physi-cal sequelae of abortion were no different than thosefound in women who carried to term or who hadnever been pregnant” (Koop, 1989, p. 195). Dr. Koopalso testified that although psychological responsesfollowing abortion can be “overwhelming to a givenindividual,” the psychological risks following abortionwere “miniscule” from a public health perspective(Koop, 1989b, p. 241).

Dr. Koop’s letter and an unofficial draft of his reportread into the Congressional Record were cited by bothabortion opponents and proponents to claim both thepresence and absence of scientific evidence showing adetrimental effect of abortion on women’s mentalhealth (see Wilmoth, deAlteriis, & Bussell, 1992). Rec-ognizing the importance of this issue, the AmericanPsychological Association (APA) convened a panel ofscientific experts in February 1989 to conduct a re-view of the scientific literature on psychological re-sponses to abortion. The panel focused on studieswith the most rigorous research designs, reportingfindings on the psychological status of women whohad legal, elective, first-trimester abortions in theUnited States. Based on their review of this literature,the task force concluded that the most methodologi-cally sound studies indicated that “severe negative re-

actions after legal, nonrestrictive, first-trimester abor-tion are rare and can best be understood in the frame-work of coping with a normal life stress” (Adler,David, Major, Roth, Russo, & Wyatt, 1990, pp. 43;see also Adler, David, Major, Roth, Russo, & Wyatt,1992). The task force recognized that some individualwomen experience severe distress or psychopathologyfollowing abortion. However, the task force also notedthat it was not clear that these symptoms are causallylinked to the abortion. The conclusions of Dr. Koopand the 1989 APA Task Force have been widely re-garded as the definitive scientific statements on thelink between abortion and mental health.

Since publication of Koop’s letter and unofficial draftreport (1989a, 1989b) and the 1989 Task Force Re-port (see Adler et al., 1990), a number of new studieshave been published in peer-reviewed journals that ad-dress the association between abortion and women’smental health. Some of these studies support the con-clusions of the 1989 Task Force Report (e.g., Cohan,Dunkel-Schetter, & Lydon, 1993; Gilchrist, Han-naford, Frank, & Kay, 1995; Russo & Dabul, 1997;Russo & Zierk, 1992), whereas others challenge them(e.g., Cougle, Reardon, & Coleman, 2003; Fergusson,Horwood, & Ridder; 2006; Gissler, Kauppila, Meri-lainen, Toukomaa, & Hemminki, 1997; Reardon &Cougle, 2002a). Reviewers of this emerging literaturehave reached differing conclusions. Based on their re-view of the post-1990 literature, for example, Brad-shaw and Slade (2003) concluded that “Theconclusions drawn from the recent longitudinal stud-ies looking at long-term outcomes following abortion,as compared to childbirth, mirror those of earlier re-views (e.g., Adler et al., 1992; Wilmoth et al., 1992),with women who have abortions doing no worse psy-chologically than women who give birth to wanted orunwanted children” (p. 948). In contrast, in testimonyintroduced in support of a law that would havebanned all abortions in South Dakota except for thosein which the mother’s life was in danger, Coleman(2006b) concluded that the scientific evidence showsthat “abortion poses significant risk to women’s men-tal health and carries a greater risk of emotional harmthan childbirth.”

Recognizing the need for a critical review of the recentliterature, in 2006 the Council of Representatives ofAPA established a new Task Force on Mental Healthand Abortion (TFMHA) composed of scientific expertsin the areas of stigma, stress and coping, interpersonalviolence, methodology, women’s health, and reproduc-tive health. The APA Council charged the TFMHAwith “collecting, examining, and summarizing the


scientific research addressing the mental health factorsassociated with abortion, including the psychologicalresponses following abortion, and producing a reportbased upon a review of the most current research.”This report summarizes the findings of the 2006TFMHA. This report updates rather than duplicates ef-forts of the 1989 Task Force. We refer the reader toAdler et al. (1992) for a discussion of APA’s involve-ment in abortion-related issues, the history and statusof abortion in the United States, and a methodologicalcritique of the literature on abortion prior to 1990 (seealso the fall 1992 issue of the Journal of Social Issues).

In preparing this report, the TFMHA recognized thatdiffering moral, ethical, and religious perspectives af-fect how abortion is perceived. Furthermore, it recog-nized that members of APA are likely to have a widerange of personal views on abortion. Irrespective oftheir views on the morality of abortion, however, APAmembers are united in valuing carefully and rigorouslycollected and interpreted scientific evidence. TheTFMHA considered its mission not only to review, butalso to critically and objectively evaluate the quality ofthe scientific evidence without regard to the directionof its findings in order to draw conclusions about themental health implications of abortion based on thebest scientific evidence available. This TFMHA reportrepresents the most thorough, current, and criticalevaluation of the literature published since 1989 (seeBradshaw & Slade, 2003; Coleman, Reardon, Stra-han, & Cougle, 2005; Dagg; 1991; Posavac & Miller,1990; Stotland, 1997; Thorp, Hartmann, & Shadi-gian, 2003, for prior published reviews of this litera-ture).

OverviewWe begin this report by defining terms, outlining thescope of the TFMHA report, and specifying the ques-tions that the research literature has been used to ad-dress (Section I). Next, we discuss conceptualframeworks important for understanding the empiri-cal literature on abortion and mental health (SectionII) and important methodological issues to consider inevaluating this literature (Section III). We then turn tothe core of our report (Sections IV and V): a reviewand evaluation of empirical studies published in Eng-lish in peer-reviewed journals post-1989 that com-pares the mental health of women who have had anelective abortion to the mental health of various com-parison groups (see detailed inclusion criteria below).We reviewed only peer-reviewed studies in order to in-clude only research findings that stood the test of in-

dependent scrutiny of qualified scientific experts. In afollowing section (Section VI), we review researchpublished post-1989 in the United States that has ad-dressed factors that predict mental health amongwomen who have had an elective abortion. We endwith a summary and conclusions based on our re-view (Section VII).

Definitions and Scope of ReportThere are multiple ways to conceptualize the mentalhealth implications of abortion and many empiricalliteratures that are relevant to this topic. Studies exam-ining the mental health implications of childbearing,particularly of unwanted childbearing, or of singleparenting, for example, are relevant for comparisonpurposes (see Barber, Axinn, & Thornton (1999) forinformation on mothers with unwanted births). So,too, are studies of the effects on children of being bornunwanted (see David, Dytrych, & Matejcek, 2003) oron women of being denied abortion (see Dagg,1991). To review all of those literatures in this report,however, would be a massive undertaking beyond thescope and charge of this task force. To keep its taskmanageable, the TFMHA limited its review and evalu-ation to the empirical literature on the implications ofinduced or intentional termination of pregnancy forwomen’s mental health. We do not consider the impli-cations of abortion for the mental health of fathers,other children or family members, or clinic workers.Although these are important questions worthy ofstudy, they are beyond the scope of this report.

Our review is limited to studies examining the mentalhealth implications of induced abortion. In some stud-ies, induced termination of pregnancy is not differenti-ated from spontaneous termination of pregnancy(spontaneous abortion, or miscarriage). Althoughspontaneous abortion may have mental health conse-quences, we consider those consequences only whenthey are compared with those of induced abortion.Other terms used to indicate induced abortion includeelective abortion, voluntary abortion, and therapeuticabortion. These distinctions can be important. Giventhat abortion involves a medical procedure, the termtherapeutic would seem to apply to all abortions.However, typically the term is applied to abortions in-duced for medically related reasons, such as to protectthe mother’s health or because of severe fetal abnor-malities. This term also was used to describe abortions


performed for psychiatric reasons prior to legalizationof abortion in the United States. Almost all abortions(92% according to the 2002 National Survey of Fam-ily Growth) in the United States are of unintendedpregnancies, pregnancies that are not induced for ther-apeutic reasons (Finer & Henshaw, 2006a). A late-term induced abortion of an intended pregnancy mayhave very different implications for mental health thana first-trimester induced abortion of an unintendedpregnancy.

We also limited our review to studies examining theimplications of induced abortion for mental healthoutcomes. Other outcomes potentially related to abor-tion (either as antecedents or consequences), such aseducation, income, occupational status, marital status,and physical health, are beyond the scope of this re-port. We conceptualized mental health broadly, relyingon the World Health Organization (WHO) definitionof mental health as a “state of well-being in which theindividual realizes his or her own abilities, can copewith the normal stresses of life, can work productivelyand fruitfully, and is able to make a contribution tohis or her community” (World Health Organization[WHO], 2007). This report thus considers a widearray of outcomes related to mental health, includingmeasures of psychological well-being (e.g., self-esteem,life satisfaction), emotions (e.g., relief, sadness), prob-lem behaviors (e.g., substance abuse, child abuse), andmeasures of severe psychopathology. In consideringthe mental health implications of abortion, it is crucialto distinguish between clinically significant mental dis-orders, such as major depression, generalized anxietydisorder, or posttraumatic stress disorder, and a nor-mal range of negative emotions or feelings one mightexperience following a difficult decision, such as feel-ings of regret, sadness, or dysphoria. While the latterfeelings may be significant, by themselves they do notconstitute psychopathology. In this report, we use theterm mental health problems to refer to clinically sig-nificant disorders assessed with valid and reliablemeasures or physician diagnosis. We use the term neg-ative psychological experiences or reactions to refer tonegative behaviors (e.g., substance use) and emotions(e.g., guilt, regret, sadness), and the term psychologi-cal well-being to refer to positive outcomes, such asself-esteem and life satisfaction. Because most studiespublished during the review period framed their re-search in terms of mental health problems and the

negative experiences or reactions of women, this re-port, of necessity, emphasized these outcomes ratherthan psychological well-being following abortion.

Our core review and evaluation was also limited tostudies that met the following inclusion criteria: (a)empirical research, (b) published in English, (c) inpeer-reviewed journals, (d) subsequent to 1989, (e)measuring a mental health relevant outcome subse-quent to abortion, and (f) including a comparisongroup of women (see details on selection criteria,below).

In addition to these core studies, the TFMHA re-viewed studies based on U.S. samples that met theabove inclusion criteria but did not include a compari-son group of women. Because such studies do not in-clude a comparison group, they cannot be used todraw conclusions about relative risks of abortion com-pared to its alternatives. Nonetheless, these studiesprovide important insight into sources of variability inwomen’s experiences of abortion in the U.S. context.

Questions AddressedWhen considering the empirical literature on the asso-ciation between abortion and mental health, it is use-ful to keep in mind four primary questions that thisliterature addressed: (1) Does abortion cause harm towomen’s mental health? (2) How prevalent are mentalhealth problems among women in the United Stateswho have had an abortion? (3) What is the relativerisk of mental health problems associated with abor-tion compared to its alternatives (other courses of ac-tion that might be taken by a pregnant woman insimilar circumstances)? And, (4) What predicts indi-vidual variation in women’s psychological experiencesfollowing abortion? As we discuss below, each of thesedifferent questions requires a different research ap-proach. Some of these questions are scientificallytestable; others are not.

Does abortion cause harm towomen’smentalhealth?Although this is the question that is posedmost often in public debates, this question is not scien-tifically testable as stated. An adequate answer to thisquestion requires a randomized experimental designthat would rigorously define the experimental, con-trol, and outcome variables and specify any limita-tions in generalizing the results. Unlike many other


areas of research, however, the study of abortion isnot open to the methodologies of randomized clinicaltrials. For obvious reasons, it is neither desirable norethical to randomly assign women who have un-wanted pregnancies to an abortion versus delivery ver-sus adoption group. Thus, although people havefrequently used the existing literature to make causalstatements, inferences of cause from this literature areinappropriate.

Howprevalent aremental health problems amongwomen in theUnited Stateswho have had an abor-tion?This question focuses attention on the extent towhich abortion poses a threat to women’s mentalhealth, i.e., is associated with a clinically significantmental disorder (see Wilmoth et al., 1992 for a discus-sion of this issue). In order to answer this question, re-search must have several key characteristics. First, theresearch must be based on samples of women represen-tative of the women to whom one wants to generalize.Thus, to address whether abortion poses a threat to themental health of women in the United States requires astudy based on a nationally representative sample ofwomen in the United States. Highly selected samples,biased samples, samples with considerable attrition orunderreporting, or samples of women in other culturesand social contexts are not appropriate for answeringthis question. As will be discussed below, samplingproblems are a serious concern in abortion research.Second, an adequate answer to the prevalence questionalso requires a clearly defined and agreed-upon defini-tion of a “mental health problem” and a valid, reliable,and agreed-upon measurement of that problem. Feel-ings of sadness or regret within the normal range ofemotion are not clearly defined and agreed-upon men-tal health problems. Mental health outcomes that meetestablished criteria for clinically significantdisorders are. Third, researchers must know the preva-lence of the same mental health problem in the generalpopulation of U.S. women who share characteristicssimilar to the abortion group, e.g., women who are ofa similar age and demographic profile. Such informa-tion is essential for interpreting the significance of find-ings. For example, if 15% of women in a nationallyrepresentative sample who had had an abortion werefound to meet diagnostic criteria for depression, themeaning of this would be more a cause for concern ifthe base rate for clinical depression among women inthe general population of a similar age and demo-graphic profile was 5% than if it was 25%.

What is the relative risk ofmental health problemsassociatedwith abortion compared to its alternatives(other courses of action thatmight be taken by apregnant woman in similar circumstances)? Thisquestion addresses relative risk. It focuses attention onthe crucially important but frequently overlookedpoint that the outcomes associated with elective abor-tion must be compared with the outcomes associatedwith other courses of action that might be taken by apregnant woman in similar circumstances (i.e., facingan unwanted pregnancy). Once a woman is pregnant,there is no mythical state of “unpregnancy.” Ques-tions of relative risk include: How does the mentalhealth of a woman who has an abortion compare tothe mental health that a woman in comparable cir-cumstances would experience were she not to have anabortion or were she to be denied an abortion? Arenegative feelings that may accompany abortion of anunwanted pregnancy more severe than alternative so-lutions, such as giving up a child for adoption or rais-ing a child a woman does not want or feelsemotionally, physically, or financially unable to carefor? Only research designs that include a comparisongroup that is clearly defined and otherwise equivalentto women who have an elective abortion are appropri-ate for answering this primary question. Otherwise,any previously existing group differences associatedwith the outcome variable may bias conclusions. Aswill be discussed below, few studies examining themental health implications of abortion include appro-priate comparison groups for answering this question.

What predicts individual variation in women’s psy-chological experiences following abortion? This lastquestion addresses the substantial individual variationobserved in women’s psychological experiences fol-lowing abortion. Rather than focusing on how the“typical” woman responds following a “typical” abor-tion, this question asks why some women experienceabortion more or less favorably than do others. Thisquestion is important to address because the propor-tion of women who have negative mental health issuesafter having an abortion will vary depending on thecharacteristics of each woman as well as the character-istics of her circumstances—there is no one answerthat applies to all women. Because this question fo-cuses on within-group variability rather than on differ-ences between the abortion group and another group,research designed to answer this question does not re-quire a comparison group of women who do not have


abortions, or a nationally representative sample. Re-search designed to answer this question, however,should at minimum be prospective and longitudinaland use reliable and valid measures of mental health.

Variability in the Abortion ExperienceIn considering the psychological implications of abor-tion, it is important to recognize that the term abor-tion encompasses a diversity of experiences and meansdifferent things to different women. Women obtainabortions for a variety of reasons, at different times ofgestation, via differing medical procedures, all ofwhich may affect the experience of abortion. Women’sresponses after abortion do not only reflect the mean-ing of abortion to her; they also reflect the meaning ofpregnancy and motherhood, which varies amongwomen. Furthermore, women obtain abortions withinwidely different personal, social, economic, religious,and cultural contexts that shape the cultural meaningsand associated stigma of abortion and motherhood aswell as others’ responses to women who have abor-tion. All of these may lead to variability in women’spsychological experiences to their particular abortionexperience. For these reasons, global statements aboutthe psychological impact of abortion on women canbe misleading.

Women obtain abortions for different reasons. Thevast majority of abortions are of unintended pregnan-cies—either mistimed pregnancies that would havebeen wanted at an earlier or later date or unwantedpregnancies that were not wanted at that time or atany time in the future (Henshaw, 1998; Torres & For-rest, 1988). Approximately half of women in theUnited States will face an unintended pregnancy dur-ing their lifetime, and about half of those who unin-tentionally become pregnant resolve the pregnancythrough abortion (Finer & Henshaw, 2006a). The rea-sons that women most frequently cite for terminatinga pregnancy include not being ready to care for a child(or another child) at that time, financial inability tocare for a child, concern for or responsibility to others(especially concerns related to caring for a future childand/or for existing children), desire to avoid singleparenthood, relationship problems, and feeling tooyoung or immature to raise a child (Finer, Frowirth,Dauphinee, Singh, & Moore, 2005). Some pregnan-cies are terminated because they are a consequence ofrape or incest. Very few (<1%) women cite coercionfrom others as a major reason for their abortion (Finer

et al., 2005). A very small percentage of abortions areof planned and wanted pregnancies. Women who ter-minate wanted pregnancies typically do so because offetal anomalies or risks to their own health.

Gestational age at time of abortion varies. The vastmajority (over 90%) of abortions in the United Statesoccur in the first trimester of pregnancy (Boonstra,Gold, Richards, & Finer, 2006). Later-trimester abor-tions occur for a variety of reasons. In some cases,particularly involving teenagers, a woman may be un-aware that she is pregnant until the second trimesteror must go through legal proceedings (e.g., judicial by-pass) in order to obtain an abortion (Boonstra et al.,2006). Later-trimester abortions also are performedafter discovery of fetal abnormalities or risks to themother’s health.

Abortion procedures vary as well. Although mostfirst-trimester abortions are performed using electricvacuum aspiration (EVA), nonsurgical methods in-volving use of a drug or combination of drugs to ter-minate pregnancy (e.g., mifepristone) are increasinglybeing used. Nonsurgical abortions comprised 14% ofnonhospital abortions in 2005 as compared to 6% in2001 (Jones, Zolna, Henshaw, & Finer, 2008). Proce-dures for abortions later than the first trimester in-clude dilation and evacuation and induction of labor.

The experience of abortion may also vary as a func-tion of a woman’s ethnicity and culture. The UnitedStates is home to a growing number of ethnic and im-migrant populations, including Hispanic (13%),African American (12.9%), and Asian and Pacific Is-landers (4.2%). According to the 2000 Census data,African American women are more than three times aslikely as White women to have an abortion (Dugger,1998). Latinas are approximately two times as likelyas White women to have an abortion, although thereare important subgroup differences. Based on esti-mates from the Hispanic Health and Nutrition Exami-nation Survey, among Latinas, Mexican women usedabortion least; Puerto Rican women used abortionmore than Mexican women, and Cubans used abor-tion the most (Erickson & Kaplan, 1998). The over-representation of ethnic minority women among thosewho obtain abortions in the United States may repre-sent the general problem of greater poverty and re-duced access to health care, including reproductivehealth services, among women of color. Although


there appears to be a strong influence of traditionalAfrican American and Latino cultural andreligious values on women’s use of abortion, this influ-ence varies by age, country or area of ancestry or ori-gin, level of acculturation, socioeconomic status, andeducational and occupational attainment (Dugger,1998; Erickson & Kaplan, 1998). Thus, it appearsthat for women of color, moral and religious values in-tersect with identities conferred by race, class, or eth-nicity to influence women’s likelihood of obtaining anabortion and, potentially, their psychological experi-ences following it. Historical linkages between coer-cive abortion and sterilization practices and theeugenics movements may lead some poor women andwomen of color to feel ambivalent on the issue ofabortion despite understanding the importance of re-productive choice (Dugger, 1998; Erickson & Kaplan,1998).

Women’s experience of abortion may also vary as afunction of the developmental phase of the life cycle inwhich it occurs. A teenager who terminates her firstpregnancy, for example, may experience different psy-chological effects compared to an adult woman whoterminates a pregnancy after giving birth to severalchildren.

Women’s experience of abortion may also vary as afunction of their religious, spiritual, and moral beliefsand those of others in their immediate social context.There are religious denominational differences in so-cial attitudes toward abortion (e.g., Bolzendahl &Brooks, 2005). Women who belong to religiousgroups that oppose abortion on moral grounds, suchas Evangelical Protestants or Catholics, may be moreconflicted about terminating a pregnancy throughabortion. Religiosity and religious beliefs are likely toshape women’s likelihood of having an abortion, aswell as their responses to abortion.

In summary, women’s psychological experience ofabortion is not uniform, but rather varies as a func-tion of characteristics and events that led up to thepregnancy; the circumstances of women’s lives and re-lationships at the time that a decision to terminate thepregnancy was made; the reasons for, type, and timingof the abortion; events and conditions that occur inwomen’s lives subsequent to an abortion; and thelarger social-political context in which abortion takesplace. This variability is an important factor in under-

standing women’s psychological experiences followingabortion.

CONCEPTUAL FRAMEWORKS

Much of the research examining the psychological im-plications of abortion has been atheoretical (Posavac &Miller, 1990). Nonetheless, several different perspec-tives have shaped understanding of potential associa-tions between abortion and mental health outcomes.These perspectives are not necessarily mutually exclu-sive and are often complementary. Yet, they lead to dif-ferent questions and different methodologicalapproaches and can lead to different conclusions.

AbortionWithin a Stress and Coping PerspectiveOne frequently used framework for understandingwomen’s psychological experience of abortion is de-rived from psychological theories of stress and coping(e.g., Lazarus & Folkman, 1984). This perspectiveviews abortion as a potentially stressful life eventwithin the range of other normal life stressors (Adleret al., 1990, 1992). Because abortion occurs in thecontext of a second stressful life event—a pregnancythat is unwanted, unintended, or associated withproblems in some way—a stress and coping perspec-tive emphasizes that it can be difficult to separate outpsychological experiences associated with abortionfrom psychological experiences associated withother aspects of the unintended pregnancy (Adler etal., 1990, 1992). Abortion can be a way of resolvingstress associated with an unwanted pregnancy, and,hence, can lead to relief. However, abortion can alsoengender additional stress of its own.

A hallmark principle of psychological theories of stressand coping is variability (e.g., Billings & Moos, 1981;Lazarus & Folkman, 1984). From this perspective, al-though unwanted pregnancy and abortion can posechallenges and difficulties for an individual woman,these events will not inevitably or necessarily lead tonegative psychological experiences for women. Stressemerges from an interaction between the person andthe environment in situations that the person appraisesas taxing or exceeding his or her resources to cope. Awoman’s psychological experience of abortion will bemediated by her appraisals of the pregnancy and abor-


tion and their significance for her life, her perceivedability to cope with those events, and the ways inwhich she copes with emotions subsequent to the abor-tion. These are shaped by conditions of the woman’senvironment (e.g., age, material resources, presence orabsence of a supportive partner) as well as by charac-teristics of the woman herself (e.g., her personality, at-titudes, and values). Thus, for example, a woman whoregards abortion as conflicting with her own and herfamily’s deeply held religious, spiritual, or cultural be-liefs but who nonetheless decides to terminate an un-planned or unwanted pregnancy may appraise thatexperience as more stressful than would a woman whodoes not regard an abortion as in conflict with her ownvalues or those of others in her social network.

Research derived from a stress-and-coping perspectivehas identified several factors that are associated withmore negative psychological reactions among womenwho have had an abortion. These include terminatinga pregnancy that is wanted or meaningful; perceivedpressure from others to terminate a pregnancy; per-ceived opposition to the abortion from partners, fam-ily, and/or friends; and a lack of perceived socialsupport from others. Other factors found to be associ-ated with more negative postabortion experiences in-clude personality traits (e.g., low self-esteem, apessimistic outlook, low- perceived control) and a his-tory of mental health problems prior to the pregnancy(see Adler et al., 1992; Major & Cozzarelli, 1992;Major et al., 2000 for reviews).

Importantly, many of the same individual and inter-personal factors that predict how women will ap-praise, cope with, and react psychologically toabortion are also predictors of how women will ap-praise, cope with, and react psychologically to othertypes of stressful life events, including unwantedmotherhood or relinquishment of a child for adoption.For instance, low-perceived social support, low self-es-teem, and pessimism also are risk factors for postpar-tum depression (Beck, 2001; Grote & Bledsoe, 2007;Logsdon & Usui, 2001). Consequently, the same riskfactors for adverse reactions to abortion can also berisk factors for adverse reactions to its alternatives.

Abortion as a Traumatic ExperienceWhereas the above framework views abortion withinthe range of normal life stressors, an alternative per-spective views abortion as a uniquely traumatic experi-

ence. This perspective argues that abortion is traumaticbecause it involves a human death experience, specifi-cally, the intentional destruction of one’s unborn childand the witnessing of a violent death, as well as a vio-lation of parental instinct and responsibility, the sever-ing of maternal attachments to the unborn child, andunacknowledged grief (e.g., Coleman, Reardon, Stra-han, & Cougle, 2005; MacNair, 2005; Speckhard &Rue, 1992). The view of abortion as inherently trau-matic is illustrated by the statement that “once a youngwoman is pregnant…it is a choice between having ababy or having a traumatic experience” (original ital-ics; Reardon, 2007, p. 3). The belief that women whoterminate a pregnancy typically will feel grief, guilt, re-morse, loss, and depression also is evidentin early studies of the psychological implications ofabortion, many of which were influenced by psychoan-alytic theory and based on clinical case studies of pa-tients presenting to psychiatrists for psychologicalproblems after an abortion (see Adler et al., 1990).

Speckhard and Rue (1992; Rue, 1991, 1995) positedthat the traumatic experience of abortion can lead toserious mental health problems for which they coinedthe term postabortion syndrome (PAS). They concep-tualized PAS as a specific form of posttraumatic stressdisorder (PTSD) comparable to the symptoms experi-enced by Vietnam veterans, including symptoms oftrauma, such as flashbacks and denial, and symptomssuch as depression, grief, anger, shame, survivor guilt,and substance abuse. Speckhard (1985,1987) devel-oped the rationale for PAS in her doctoral dissertationin which she interviewed 30 women specifically re-cruited because they deemed a prior abortion experi-ence (occurring from 1 to 25 years previously) to havebeen “highly stressful.” Forty-six percent of thewomen in her sample had second-trimester abortions,and 4% had third-trimester abortions; some had abor-tions when it was illegal. As noted above, this self-se-lected sample is not typical of U.S. women who obtainabortions. PAS is not recognized as a diagnosis in theDiagnostic and Statistical Manual of the AmericanPsychiatric Association (American Psychiatric Associa-tion, 2002).

AbortionWithin a Sociocultural ContextA third perspective emphasizes the impact of thelarger social context within which pregnancy andabortion occur on women’s psychological experienceof these events. Unwanted pregnancy and abortion do


not occur in a social vacuum. The current sociopoliti-cal climate of the United States stigmatizes somewomen who have pregnancies (e.g., teen mothers) aswell as women who have abortions (Major & Gram-zow, 1999). It also stigmatizes the nurses and physi-cians who provide abortions. From a socioculturalperspective, social practices and messages that stigma-tize women who have abortions may directly con-tribute to negative psychological experiences postabortion.

The psychological implications of stigma are pro-found (see Major & O’Brien, 2005, for a review). Ex-perimental studies have established thatstigmatization can create negative cognitions, emo-tions, and behavioral reactions that can adversely af-fect social, psychological, and biological functioning.Effects of perceived stigma include cognitive and per-formance deficits (Steele & Aronson, 1995), increasedalcohol consumption (Taylor & Jackson, 1990), so-cial withdrawal and avoidance (Link, Struening,Rahav, Phelan, & Nuttbrock, 1997), increased de-pression and anxiety (Taylor, Henderson, & Jackson,1991), and increased physiological stress responses(Blascovich, Spencer, Quinn, & Steele, 2001). Societalstigma is particularly pernicious when it leads to “in-ternalized stigma”—the acceptance by some membersof a marginalized group of the negative societal be-liefs and stereotypes about themselves. Women whocome to internalize stigma associated with abortion(e.g., who see themselves as tainted, flawed, ormorally deficient) are likely to be particularly vulnera-ble to later psychological distress.

A sociocultural context that encourages women to be-lieve that they “should” or “will” feel a particularway after an abortion can create a self-fulfillingprophecy whereby societally induced expectancies canbecome confirmed. Mueller and Major (1989) demon-strated experimentally the effect of expectancies onwomen’s psychological experiences after abortion.They randomly assigned women prior to their abor-tion to one of three short counseling interventions.One intervention focused on improving women’s self-efficacy for coping with abortion (creating positivecoping expectancies), another focused on reducing theextent to which women attributed their pregnancy totheir character (as opposed to their behavior), and thethird focused on birth control. Women exposed to theself-efficacy intervention were significantly less likely

to display depressed affect following the abortion thanthose in the other two conditions. Societal messagesthat convey the expectation that women will copepoorly with an abortion would be expected to havethe reverse effect; i.e., by creating negative coping ex-pectancies, they may cause women to feel bad follow-ing an abortion.

Whether or not a particular behavior or attribute isstigmatized often varies across cultures and time(Crocker, Major, & Steele, 1998). Actions that oncewere viewed benignly can become stigmatized (e.g.,smoking), and others that once were highly stigma-tized (e.g., sex out of wedlock, divorce, cohabitation)can become less so. As society’s views of a behaviorchange, so too will the appraisals and responses ofthose who engage in that behavior. Hence, the socio-cultural context can shape a woman’s appraisal ofabortion not only at the time that she undergoes theprocedure, but also long after the abortion. Socialmessages that encourage women to think about (reap-praise) a prior abortion in more negative ways (as asin, as killing a child) may increase women’s feelingsof guilt, internalized stigma, and emotional distressabout an abortion they had long ago. In contrast, so-cial messages and support groups that encouragewomen to cognitively reappraise an abortion in amore positive or benign way may lead to improvedemotional responses (Trybulski, 2006).

Abortion and Co-Occurring Risk FactorsA fourth conceptual framework for understandingwomen’s postabortion mental health emphasizes sys-temic, social, and personal factors that are precursorsto unintended pregnancy and, hence, place women atrisk for having abortions and/or predispose them toexperience mental health problems, regardless of preg-nancy and its resolution. From this perspective, mentalhealth problems that develop after an abortion maynot be caused by the procedure itself, but instead re-flect other factors associated with having an unwantedpregnancy or antecedent factors unrelated either topregnancy or abortion, such as poverty, a history ofemotional problems, or intimate-partner violence.This co-occurring risk perspective emphasizes that as-pects of a woman’s life circumstances and psychologi-cal characteristics prior to or co-occurring with herpregnancy must be considered in order to make senseof any mental health problems observed subsequent toabortion.


Unwanted pregnancies are not random events. Thelives of women who have unwanted pregnancies orabortions differ in a variety of ways from the lives ofwomen who do not have unwanted pregnancies orabortions, and do so before, during, and after preg-nancy occurs. These differences may have implicationsfor later functioning apart from any influence fromthe experience of unwanted pregnancy and/or abor-tion. The necessity of considering preexisting or co-oc-curring group differences is widely recognized byresearchers who study the consequences of nonmaritaland adolescent births (e.g., Moore, 1995). As de-scribed below, substantial research literature hasshown that systemic and personal characteristics thatpredispose women to have unintended pregnanciesalso predispose them to have psychological and behav-ioral problems. Consequently, correlations betweenabortion status and mental health problems observedafter an abortion may be spurious due to their jointassociation with similar risk factors present prior tothe pregnancy. We briefly review evidence consistentwith this perspective below.

Systemic risk factors.Poverty is a systemic risk fac-tor for unplanned pregnancy and for abortion.Women at particularly high risk for unintentionalpregnancy and women who obtain abortions tend tobe young, unmarried, poor, and women of color(Finer & Henshaw, 2006a; Jones, Darroch, & Hen-shaw, 2002a, 2002b; Jones & Kost, 2007). In 2000,women with resources below the federal povertylevel represented 57% of all abortions (Jones, Dar-roch, & Henshaw, 2002b). Exposure to sexual orphysical abuse during childhood and exposure to in-timate partner violence including rape also are asso-ciated with greater likelihood for both unintendedpregnancy and abortion (e.g., Boyer & Fine, 1992;Dietz et al., 1999; Gazmararian, Lazorick, Spitz,Ballard, Saltzman, & Marks, 1996; see Coker,2007; Pallitto & O’Campo, 2005; Russo & Denious,1998b for reviews).

From a co-occurring risks perspective, the greater ex-posure to adverse life circumstances (poverty, abuse,and intimate violence) among the group of womenwho have abortions compared with other women mayunderlie a positive correlation observed between abor-tion and mental health problems. Given the former’sgreater exposure to adversity, the absence of such anassociation would be noteworthy.

Indeed, these same systemic factors shown to be asso-ciated with increased risk for unintended pregnancyand abortion have also been shown to be associatedwith increased risk for mental health problems. Forexample, studies based on nationally representativesamples show that poverty is strongly related to an in-creased likelihood of psychiatric disorder (e.g., Kessler,et al., 1994; Robins & Regier, 1991). Children whogrow up in poor neighborhoods are at higher risk forteen pregnancy, substance abuse, obesity, smoking,and dropping out of school, all of which are risk fac-tors for psychological problems (Mather & Rivers,2006; Messer, Kaufman, Dole, Savitz, & Laraia,2006). Exposure to domestic (intimate) violence alsois a strong and well-documented predictor of physicaland mental health problems, including suicide, post-traumatic stress disorder, depression, and substanceabuse (see Golding, 1999, for a meta-analysis and re-view). The more violence-related events a woman hasexperienced and the more stressful life events she hasexperienced in general, the greater her risk for devel-oping a mental disorder (Breslau, Kessler, Chilcoat,Schultz, Davis, & Andreski, 1998; Brown & Harris,1978; Golding, 1999).

Personal risk factors. In addition to systemic factors,personality or behavioral factors may also predisposea woman to unplanned pregnancy and abortion, aswell as to mental health problems. There is substantialevidence that problem behaviors tend to co-occuramong the same individuals. For example, high schoolstudents who report engaging in early sexual activityalso are more likely to report smoking; using alcohol,marijuana, and hard drugs; minor delinquency; and,to a lesser extent, major aggression and gambling(Willoughby, Chalmers, & Busseri, 2004). Womenwho have unintended pregnancies and abortions aremore likely than other women to have previously en-gaged in a behaviors such as smoking, using alcoholand illicit drugs, engaging early in sexual intercourse,and having unprotected sexual intercourse (Costa,Jessor, & Donovan,1987).

One explanation for this pattern is that involvementin problem behaviors follows definite pathways inwhich specific factors place the individual who hasparticipated in one behavior (e.g., drug use) at risk ofinitiating another (e.g., early sexual activity), whichputs that person at risk for another event (unintendedpregnancy), which in turn puts that person at risk for


another event (abortion) (e.g., Kandel, 1989). A longi-tudinal study based on data from the National Longi-tudinal Study of Youth (NLSY) showed that drug useamong young women greatly increased their risk ofearly sexual activity (before age 16) when other im-portant risk factors were controlled (Rosenbaum &Kandel, 1990). In a subsequent study also based ondata from the NLSY, Mensch and Kandel (1992)showed that drug use was uniquely predictive of bothsubsequent premarital teen pregnancy and the decisionto terminate a premarital teen pregnancy. Toavoid confounding antecedents of pregnancy with itsconsequences, they restricted their analyses to theyoungest birth cohorts in the sample. This ensuredthat the measurement of the independent variables(e.g., drug use) preceded the events of interest (pre-marital teen pregnancy and abortion). They foundthat the risk of premarital teen pregnancy was nearlyfour times as high for women who had used illicitdrugs other than marijuana as it was for women withno history of prior substance involvement. Further-more, early illicit drug use was the strongest predictorof a later abortion. Another prospective longitudinalstudy found that women who at age 18 (none ofwhom had had a pregnancy or abortion) had reportedsmoking or using drugs had an increased likelihood ofa subsequent unplanned pregnancy and, as a result,higher rates of abortion by age 29 compared towomen who at age 18 had not reported using thesedrugs (Martino, Collins, Ellickson, & Klein, 2006).

An alternative explanation for the co-occurrence ofproblem behaviors is that individuals who engage inproblem behaviors such as alcohol or drug use share aset of personality characteristics that predisposes themto engage in risky behaviors that increase the likeli-hood of other problems (e.g., unplanned pregnancy;Jessor & Jessor, 1977; see Dryfoos, 1990, for a re-view). For example, scoring high on a measure of “un-conventionality” has been found to positively predictboth abortion and unplanned pregnancy (Martino,Collins, Ellickson, & Klein, 2006). Personality factorsthat diminish a person’s ability to regulate negativeemotion may also put him or her at risk for engagingin problem behaviors. In a longitudinal study of a rep-resentative sample of 1,978 Black and White adoles-cents, Cooper, Wood, Orcutt, and Albino (2003)found that high impulsivity and an avoidance style ofcoping with negative emotions were risk factors for in-volvement in a wide range of problem behaviors, in-

cluding risky sexual behavior, substance use, delin-quent behavior, and educational underachievement.Furthermore, an avoidance coping style prospectivelypredicted initial or increasing involvement in all ofthese problem behaviors among individuals with noprior experience with that behavior. Thus, for exam-ple, girls high in avoidance coping who had little or noprior sexual experience were subsequently more likelyto engage in risky sexual behavior than girls lower inavoidance coping. Because early sexual activity andrisky sexual behavior are risk factors for unintendedpregnancy, which in turn is a risk factor for abortion,being high in avoidance styles of coping with negativeemotion may be a predisposing risk factor for the ex-perience of abortion.

Importantly, many of these personal characteristicsthat put women at risk for problem behaviors and un-planned pregnancy also put them at risk for mental orphysical health problems, whether or not a pregnancyis aborted or carried to term. For example, a numberof studies demonstrate that using avoidant forms ofcoping with negative emotions is associated withpoorer mental health and exacerbates adjustment dif-ficulties over time, even after controlling for prior lev-els of adjustment (Aldwin & Revenson, 1987; Major,Richards, Cooper, Cozzarelli, & Zubek, 1998). Thebest predictor of mental health problems later in life isa prior occurrence of mental health problems. For ex-ample, Kessler, Avenevoli, and Merikangas (2001) re-ported that 50% of adolescents who had anoccurrence of major depression and 90% of adoles-cents who experienced mania during their adolescencecontinued to have recurrences of these disorders inadulthood.

Summary of Conceptual FrameworksThe four perspectives summarized above can be com-plementary ways of understanding underlying causesof women’s psychological experience of abortion. Thefirst perspective regards abortion as a stressful lifeevent similar to other types of stressful life events awoman may experience. According to this perspective,women will vary markedly in how they appraise, copewith, and adjust to unwanted pregnancy and abortion,just as people vary widely in how they respond toother types of stressful life events. A stress-and-copingperspective thus does not rule out the possibility thatsome women may experience severe negative psycho-logical experiences following abortion, but locates


such reactions in women’s appraisals and copingprocesses and the personal and social factors thatshape those, rather than in the nature of the eventitself. In contrast, the second perspective suggests thatdue to its unique features, abortion is likely to be ex-perienced as traumatic by most women. Thus, in con-trast to other perspectives discussed, this particularframework suggests that most women will have nega-tive psychological experiences subsequent to abortion.

The sociocultural perspective emphasizes thatwomen’s psychological experiences of abortion areshaped by the immediate and larger sociocultural con-text within which the abortion occurs. From this per-spective, social and cultural messages that stigmatizewomen who have abortions and convey the expecta-tion that women who have abortions will feel badmay themselves engender negative psychological expe-riences. In contrast, social and cultural messages thatnormalize the abortion experience and convey expec-tations of resilience may have the opposite effect.

The co-occurring risk perspective emphasizes that pre-existing and/or ongoing conditions may account fordifferences in mental health or problem behaviors ob-served between women who have had an abortion andwomen who have not. Unwanted pregnancy and abor-tion are correlated with preexisting and/or ongoingconditions (e.g., poverty), life circumstances (e.g., ex-posure to violence), problem behaviors (e.g., druguse), and personality characteristics (e.g., avoidancestyle of coping with negative emotion) that can haveprofound and long-lasting negative effects on mentalhealth. These conditions may predispose women tounintended pregnancies and abortion and have nega-tive effects on mental health regardless of reproductivehistory and outcomes. From this perspective then,mental health and problem behaviors observed afterabortion are often a byproduct of conditions and char-acteristics that preceded or coexist with the unin-tended pregnancy and abortion.

METHODOLOGICAL ISSUES IN ABORTION RESEARCH

Many scholars have noted that research on the mentalhealth implications of abortion is plagued by numer-ous methodological problems (see, e.g., Adler et al.,

1992; Koop, 1989; Wilmoth et al., 1992).These prob-lems continued to be reflected in most of the studiesreviewed by the current task force and limited conclu-sions that could be drawn from this literature. In thefollowing discussion, we highlight the problems thatwe encountered most often in our review of the post-1989 literature. We do not recapitulate all of the de-tails presented in previous methodological discussions(see McCall & Appelbaum, 1991, for further discus-sion of some of these issues). The primary issues weaddress are those of comparison and contrast groups,co-occurrence of risk factors, sampling, measurementof reproductive history and underreporting, attrition,statistical treatment of data, outcome measurement,and clinical relevance. These issues are not independ-ent of each other. Indeed, the complex interactionsamong these factors can make it difficult to sort outtheir separate and combined effects.

Comparison/Contrast GroupsIn order for empirical research to address the relativerisk of elective abortion compared to alternativecourses of action that a pregnant woman facing an un-wanted pregnancy might take, clearly defined and oth-erwise equivalent comparison groups are essential.Otherwise, any previously existing group differencesassociated with the outcome variable may badly biasconclusions. One appropriate comparison groupwould be women who are denied or unable to obtainan abortion and who, hence, must carry to term anunwanted pregnancy. Other appropriate comparisongroups would be women who deliver an unwantedpregnancy and either give the child up for adoption orraise it. By at least partly controlling for the “wanted-ness” of the pregnancy, such comparisons provide as-surance that the women being compared face a similarsituation. Unfortunately, very few studies used appro-priate comparison groups.

One way researchers attempted to solve this problemwas by using covariate adjustments to try to make“nonequivalent” groups “equivalent.” The analysis ofcovariance, however, can be extremely sensitive to vio-lations of its assumptions, and these assumptions areparticularly liable to violation when used to try to ad-just for initial group differences (see, e.g., Elashoff,1969). One violation occurs when the covariate(s) aremeasured after the treatment—a problem characteristicof retrospective studies of abortion, in which the co-variates are assessed after the abortion. A second viola-


tion occurs when the relationship between the covari-ate and the outcome differs across groups. A third vio-lation occurs when the relationship between thecovariate and the outcome is nonlinear. Unfortunately,tests of the validity of these assumptions were rarelyencountered in the published literature onabortion. Consequently, caution should be exercisedin accepting the findings of studies in which initially in-comparable groups were compared (adjusted for co-variates) without a test of the validity of theassumptions.

Co-Occurring Risk FactorsUnfortunately, very few studies encountered in our re-view of the literature adequately assessed and con-trolled for co-occurring risks. As discussed above,there are naturally occurring interrelations amongmany of the phenomena associated with elective abor-tion that make it difficult to tease apart the causalchains that might be operating. Elective abortion com-monly co-occurs with unwanted or unintended preg-nancy, and unwanted/unintended pregnancy is oftenassociated with adverse circumstances and characteris-tics that may be associated with mental health prob-lems. Because few studies adequately controlled forthese co-occurring risks, it is almost impossible fromthe available literature to distinguish outcomes thatflow from abortion per se from outcomes that mightappear to be associated with abortion, but in actualityhave their origins in the unwanted/unintended preg-nancy (or some other co-occurring risk), which ismore highly represented in the abortion group than inthe comparison group. It was particularly difficult todetect these co-occurring conditions and their conse-quences from secondary data analyses of data sets col-lected for other purposes because potential confoundsthat were not of interest in the initial data collectionwere unlikely to have been adequately assessed.

SamplingProblems of sampling characterized most of the stud-ies reviewed. Two basic designs in the abortion litera-ture presented sampling problems. The first occurredwhen convenience samples of women were recruitedspecifically for the study without concern for the de-gree to which they represented a definable population,for example, women seeking pregnancy testing at ahealth clinic (Cohan, Dunkel-Schetter, & Lydon,1993), women waiting to see their doctor (Williams,2001), or pregnant teens residing at a maternity home

(Medora, Goldstein, & von der Hellen, 1993). Oftenthe samples were extremely small (< 30; e.g., Cohan etal., 1993). In many cases, little, if anything, was re-ported about the inclusion rates of the women in ei-ther the abortion group or the comparison groups orthe context of their situations, information necessaryto establish the representativeness and generalizabilityof the data. Sometimes data were based on volunteersamples of women who responded to mailed question-naires about their reproductive history (Reardon &Ney, 2000). Such volunteers do not represent an unbi-ased sample representative of the population as awhole and cannot be used as evidence to establishprevalence rates or normative responses.

The second and equally problematic situation oc-curred when subsamples were selected for analysisfrom extant studies that were initially conducted forother purposes. This characterized most of the studiesbased on secondary analyses of medical records orpublic survey data sets. Many of the studies with thelargest sample sizes that have been used to makeclaims about the effects of abortion are of this type—e.g., studies based on the National Longitudinal Studyof Youth (NLSY) (e.g., Reardon & Cougle, 2002a;Russo & Zierk, 1992), National Survey of FamilyGrowth (NSFG) (e.g., Cougle, Reardon, & Coleman,2005), or the National Longitudinal Study of Adoles-cent Health (Coleman, 2006). In these studies, subsetsof the complete sample were taken to allow certaincomparisons of interest to be made. For example, onlywomen who reported terminating or delivering a firstpregnancy might be selected (e.g., Cougle et al., 2003).

There are a number of serious problems with selectingsubsamples from the larger data set in this way: (a)The secondary sampling destroys the sampling proper-ties that might have originally characterized the sam-ple, particularly if population-based sampling weightswere not properly taken into account. Distorted sam-pling weights (or non-use of sampling weights) canlead to inaccurate estimations when the results areused to estimate prevalence of mental health problemsin the general population following abortion. (b) Sam-pling on certain characteristics (e.g., first pregnancy;Cougle et al., 2005; Schmiege & Russo, 2005) may af-fect other characteristics of the sample, thereby com-promising generalizability. For example, women whohave an abortion on their first pregnancy are morelikely to be younger and to be unmarried than women


who have their first abortion on a later preg-nancy. (c) In some studies, additional sources of non-equivalence between abortion and comparisongroups were created by selecting a first “target” preg-nancy occurring in a specified time period of data col-lection (e.g., the latter 6 months of 1989). This wasto create abortion and delivery comparison groupswithout attention to reproductive history differencesbetween these groups, when reproductive history is afactor affecting retention in the population sampled(e.g., Cougle, Reardon, & Coleman, 2003; Reardon& Coleman, 2006; Reardon & Cougle,2002a). (d) Serious violation of sampling principlesalso occurs when differential exclusion is used in con-structing comparison groups such that one group isadvantaged relative to the other (e.g., Coleman et al.,2002; Cougle, et al., 2005).

Measurement of Reproductive Historyand Problems of UnderreportingMany of the studies reviewed were characterized byinaccuracy in the information available regarding awoman’s reproductive history, particularly her abor-tion history. In some studies, a woman’s abortion sta-tus was verifiable (e.g., data were collected at the timethat she sought an abortion at a clinic or from hermedical records). More typically, however, abortionstatus was established based on self-report. For exam-ple, in all of the studies based on a secondary analysisof survey data, abortion status was established by ask-ing women to indicate, either on a questionnaire orverbally, to an interviewer whether or not they hadhad an abortion in the past. Women’s reports of anearlier abortion were then correlated with currentmental health/emotional status, with the latter attrib-uted to the former (e.g., Coleman, Reardon, Rue, &Cougle, 2002a; Cougle et al., 2005).

This approach has many problems. Abortion, likeother stigmatized conditions, is typically underre-ported (Jones & Kost, 2007). It has long been recog-nized that individuals are unlikely to frankly answerquestions that have the potential to be embarrassing,overly self-disclosing, or in other ways reflect nega-tively on them. One of the earliest applications of astatistical model designed for reducing bias in obtain-ing answers to sensitive questions—the so-called ran-domized response methodology—was for estimatingthe mean number of abortions in an urban populationof women (Greenberg, Kuebler, Abernathy, &

Horvitz, 1971). The percentage of women reportingan abortion on surveys is consistently lower than thenumber expected based on estimates made from na-tional provider data, sometimes markedly so (Jones &Forrest, 1992; Jones & Kost, 2007). Absent the use oftechniques such as randomized response methodologyor the selection of highly disclosing samples, one islikely to obtain biased estimates of prevalence rates.Generally, there are two types of underreporting: fail-ure to acknowledge having had any abortions andhaving had multiple abortions but reporting onlysome of them (Jones & Kost, 2007).

Underreporting of abortion in surveys is of particularconcern when there is differential underreporting bysubgroups of women (Fu, Darroch, Henshaw, &Kolb, 1998; Jones & Forrest, 1992). Women morelikely to underreport include those who are unmar-ried, Black or Hispanic, Catholic, low-income, andaged 20–24 (Jones & Kost, 2007). Underreporting canintroduce systematic bias into a study. Only a fewstudies reviewed attempted to test for possible under-reporting biases. For example, Schmiege and Russo(2005) examined and compared the relation of abor-tion versus delivery to depression (CESD cutoff score)in the NLSY data set among groups known to vary inunderreporting (e.g., White married women, unmar-ried Black women, Catholics). Their analyses sug-gested that at least in the NLSY data set,underreporting by specific subgroups did not appearto introduce systematic bias into observed associationsbetween abortion and a mental health outcome.

In general, the nature of the potential bias introducedby underreporting (i.e., whether it biases toward over-estimating or underestimating adverse impact of abor-tion) is unclear. It is possible that women who feelmost distressed by an abortion are less likely to reportit to others; as a consequence, they may be underrep-resented in the abortion group, biasing results towardunderestimating negative effects. It is also possible thatresponse biases in the other direction may be ob-served. For example, women who are experiencingdistress may view the survey as an opportunity forcatharsis and hence be more likely to disclose theirabortion than women less distressed. In addition,women most willing to report one “problem” (e.g.,depression, anxiety, abuse) may also be those mostable to recall or willing to report another “problembehavior” (abortion), biasing results toward overesti-


mating negative effects. Many scholars have noted theproblem of selective recall bias in surveys on the partof individuals experiencing a disorder who may (1)more thoroughly scrutinize their history in an effort toexplain their disorder and (2) more accurately recallstigmatizing events, such as abortion, than individualsnot experiencing a disorder (e.g., Neugebauer & Ng,1990; Chouinard & Walter, 1994). Recall biases canexplain, for example, why a positive relationship be-tween abortion history and breast cancer has been ob-served in retrospective surveys but is absent inprospective studies (American Cancer Society:http://www.cancer.org/). Specifically, breast cancer pa-tients seeking to understand their disease are thoughtto be more motivated to search their memories as wellas more willing to report socially stigmatizing condi-tions (such as abortions or sexually transmitted infec-tions) to a health care provider than are healthywomen, leading to a spurious relationship.

Measurement of abortion also typically suffers fromunderspecification. Many studies lack important infor-mation about the abortion, such as length of gestation,type of procedure, or whether the abortion was per-formed for therapeutic reasons, all of which may affecthow women respond emotionally and physically afteran abortion. For example, abortions performed beyondthe first trimester involve a more risky medical proce-dure and more pain, which may have negative ef-fects. They also occur at a more advanced stage ofdevelopment, which may change the meaning ofthe pregnancy, making abortion more stressful (Major,Mueller, & Hildebrandt, 1985). Delay may also reflectambivalence toward the pregnancy or indicate that awanted pregnancy was terminated because of discoveryof a health problem or fetal defect. It is also unclear towhat extent research on earlier surgical methods ofabortion applies to newer nonsurgical methods ofabortion, which are used at the earliest stages of gesta-tion and differ from traditional methods in other waysas well, although studies suggest comparable postabor-tion emotional adjustment for women experiencingeach method (Ashok, Hamoda, Flett, Kidd, Fitzmau-rice, & Templeton, 2005; Howie, Henshaw, Najo,Russell, & Templeton, 1997; Lowenstein et al., 2006;Sit, Rothschild, Creinin, Hanusa, & Wisher, 2007).

AttritionAnother potentially serious methodological confoundencountered was attrition—loss of cases during the

course of an investigation. Attrition has been a long-standing concern in the study of abortion (see for ex-ample, Adler, 1976). The consequences of attritionrange from potentially serious loss of power to biasingof results when attrition is not random (i.e., biased ina specific direction) and differs by group. In the caseof abortion, for example, underestimation of theprevalence of distress in the final sample would occurif women who were most upset by the abortion weremore likely to be lost to a follow-up than those whowere retained in the sample. Similarly, overestimationof the prevalence of distress would occur if womenwho were least distressed by the abortion were morelikely to be lost to a follow-up. Consequently, it is es-sential that researchers test for biases in attrition. Onlya few studies reviewed did so. One study that did testfor attrition (Major et al., 2000) found that amongwomen who had a first-trimester abortion, thosewho were retained in the sample at the 2-yearpostabortion measurement period did not differ signif-icantly from those who were lost to attrition on anydemographic or psychological variable assessed eitherprior to the abortion, immediately post abortion, or 3months post abortion. Thus, at least in this sample, noevidence of systematic bias in attrition was observed.

OutcomeMeasures: Timing,Source,and Clinical SignificanceProblems of outcome measurement also were fre-quently encountered in this literature. It is vital that themeasures of mental health are valid and reliable. Insome studies reviewed, claims of mental health impact(or no impact) were made on the basis of psychometri-cally poor measures, including one-item measures (e.g.,Coleman, 2006a; Reardon & Ney, 2000). For exam-ple, Reardon and Ney (2000) measured substanceabuse with yes/no responses to the single question“Have you ever abused drugs or alcohol?” This is nota reliable measure of substance abuse. A clinically rele-vant measure (as opposed to a scale score withoutknown clinical relevance) should be the minimal stan-dard for measuring impact. In addition, claims of im-pact should be accompanied by epidemiologicallymeaningful effect size indicators such as odds ratios,which provide clinically relevant measures of impact.Odds ratios should be presented in conjunction withdata of the rates or proportions of women affected(i.e., a finding of 3 to 1 in 100 women presents a differ-ent level of threat than 3 to 1 in 1 million women). Ab-solute and relative levels of the effect should be clear.


An associated problem encountered in both primaryand secondary studies was related to the timing ofmeasurement. Some studies first contacted theirparticipants months or years (or an unspecified timeinterval) after the target abortion and engaged them inretrospective reporting of their preabortion status(e.g., Bradshaw & Slade, 2005; Cougle et al., 2005) ortheir mental health/emotional status at selected pointsafter the event (e.g., Kersting et al., 2005). Retrospec-tive reporting is subject to a large number of distor-tions and biases. There is agreement amongmethodologists that measures taken nearer an eventare more likely to be accurate than measures taken ata time distant from the event.

Finally, assessing the clinical significance of abortion,as with any other medical procedure, requires asking“what is the benefit?” as well as “what is the harm?”of the procedure. Many of the abortion studies re-viewed focused only on negative outcomes. Focusingsolely on adverse effects can create a distorted pictureof the information needed to provide complete and ac-curate informed consent. It is akin to focusing on therisks of chemotherapy without addressing its potentialbenefits for curing cancer. For example, in separate re-ports based on the same sample, one research team re-ported a negative association between abortion andmental health (Fergusson, Horwood, & Ridder,2006) and a positive association between abortion andother life outcomes (e.g., education, employment; Fer-gusson, Boden, & Horwood, 2007). The authors con-cluded that there is a “need for further research intothe risks and benefits associated with abortion as ameans of addressing the issues raised by unwanted ormistimed pregnancies” (Fergusson et al., 2007, p. 11).

Other Statistical IssuesMany of the studies included in our review were char-acterized by statistical problems. One frequently en-countered problem, especially in the studies based onsecondary data analyses, was inflation of the probabil-ity of making a Type I error in inference by perform-ing many significance tests at the same level onewould if there were to be only a single test. This ap-peared in two forms. The first form occurred when theinitial sample (often a reasonably large sample) wasdivided into smaller and smaller subsets, and thesesubsets were then used to test for differences betweenabortion and nonabortion cases within each subsetwithout any overall control for the number of signifi-

cance tests conducted (e.g., Coleman, Reardon, &Cougle, 2002; Reardon & Ney, 2000). This practiceincreases the probability of a statistically significantdifference occurring due to chance. The second formencountered was the ad hoc search for covariates. Inmany studies, especially those based on analyses ofsecondary data sets, the data analyst began with a setof all possible covariates (usually defined by the meas-ures available in the data set) and tested each covari-ate for significance (testing the partial regressioncoefficients for significance). The analyst then pro-ceeded to conduct analyses using only the significantcovariates (e.g., Coleman, Maxey, Rue, & Coyle,2005). Without any correction for chance via alpha-level control, this completely ad hoc, atheoretical ap-proach also capitalizes on chance. Furthermore, thechoice of covariates to include in analyses can play akey role in how much variance in the outcome vari-able is explained by pregnancy outcome.

Interpretational Problems and Logical FallaciesIn addition to the methodological problems describedabove, the TFMHA also encountered a numberof cases in which data were incorrectly interpreted orgeneralized, if not in the actual research reports them-selves, then in reviews, summaries, and press releasesbased on that research. Accordingly, the TFMHA feltit important to point out several logical fallacies thatmust be guarded against in drawing conclusions fromthis literature.

The first logical fallacy is the tendency to infer causa-tion from correlation. Frequently, significant correla-tions observed between abortion history and othervariables (e.g., substance abuse, depression, higher ed-ucational outcomes) were misinterpreted as evidencethat abortion caused these variables to occur. Suchcausal claims are unwarranted, as the relationshipsmay be spurious, the causal direction may be reversed,or the relationship may be due to a third variable thatis associated with both abortion and the outcome vari-able (e.g., poverty). It is sometimes argued that a casefor causality is stronger in abortion studies that estab-lish (a) time precedence of the abortion before an out-come variable, (b) covariation of abortion and theoutcome variable, and (c) lack of plausible alternativeexplanations or control of third variables associatedwith both abortion and the outcome variable. These,however, are only necessary but not sufficient condi-tions to establish causality. Furthermore, although


some of the studies reviewed did meet criteria (a) and(b), the TFMHA could identify no study reporting asignificant association of abortion with a mentalhealth outcome that met criterion (c).

A second logical fallacy is the tendency to confuse arisk and a cause. For example, some writers appearedto assume that if a prior history of abortion was foundto be a “risk factor” for a certain outcome (e.g., vio-lent death), then a prior history of abortion is a“cause” of violent death. Many things can serve asmarkers for causes or may be associated with causeswithout themselves being a part of the causal mecha-nisms in play. For example, age is the most importantknown risk factor for Alzheimer’s disease (AD), but itis not the mechanism that causes people to developAD. Rather, age is a statistical predictor in a popula-tion of who in that population is at risk, that is, morelikely (older versus younger) to develop AD(http://www.nia.nih.gov/Alzheimers/). The steps thatlink risks and causes must be explicitly developed anddemonstrated before one can validly make the asser-tion that removing a particular risk factor will lead toa desired outcome.

A third and very serious logical fallacy is the “inter-ventionist fallacy”—the belief that if a relationship isobserved between two variables, the form or magni-tude of the relationship will remain unchanged if anintervention changes some part of the current state ofaffairs. For example, because there is a substantialpositive relationship between family income and chil-dren’s school performance, it is tempting to think thatincreasing family income would lead to improved chil-dren’s school performance. Such a conclusion, how-ever, does not logically follow. It might be that whatdrives the relationship between family income andschool performance is the family expenditure onbooks. Were one to intervene and supplement familyincome, it does not necessarily follow that the familywould increase its expenditure on books, which are (inthis example) the actual component that drives thechild’s school performance, and, hence, the interven-tion might fail.

As applied to the case of abortion, one example ofthe interventionist fallacy would be the belief that ifabortion and depression are related, then reducingaccess to abortion would reduce the prevalence ofdepression. A change in the availability of elective

abortion, however, would have many consequences.It would mean that women who want to terminatean unwanted pregnancy would now be forced to de-liver. As a consequence, the characteristics of thepopulation of women who delivered children wouldchange. Characteristics previously prevalent amongwomen who had an abortion (e.g., greater poverty,exposure to violence) would now be prevalentamong the delivery group. The portrait of the men-tal health of mothers might reasonably be expectedto be worse. This potential change in the profileof women giving birth does not include any newmental health problems that might develop fromstresses associated with raising a child a womanfeels unable to care for, or may not want, or fromrelinquishing a child for adoption. Thus, reducingaccess to abortion would be likely to result inpoorer mental health among women who deliver.Hence, rather than reducing the prevalence of de-pression among women, this intervention could po-tentially increase it.

Summary ofMethodological IssuesMost of the studies published on postabortion mentalhealth contain one or more of the methodological orinterpretational problems discussed above. Conse-quently, reviews of the literature that simply count thenumber of studies that show one effect versus anotheror that calculate effect sizes without carefully consider-ing and weighing the quality of the evidence that pro-duced the effect are inappropriate and oftenmisleading. It is essential to keep the methodologicaland interpretational points discussed above in mindwhen considering the literature on postabortion men-tal health reviewed below.

It is also important to recognize, however, that not alldesign problems are equally serious. The extent towhich a design flaw affects the merits of a particularstudy depends in part on the goal of the study. For ex-ample, the lack of a comparison group is notoverly limiting when the researcher’s goal is to under-stand predictors of response among women who haveabortions. Some flaws can be compensated for by lim-iting generalization or interpretation. However, otherflaws are so serious that they limit any conclusionsthat can be drawn from the study (e.g., differential ex-clusion of women from one group but not the compar-ison group on a variable known to be related to theoutcome variable).


REVIEWOF SCIENTIFIC LITERATURE

Search Strategy and Criteria for Inclusion

In order to evaluate the scientific literature on mentalhealth effects of abortion, the TFMHA searchedPsycINFO and Medline for English-language peer-re-viewed articles published between 1990 and 2007based on human subjects. Research conducted withU.S. as well as non-U.S. samples was searched. Key-word combinations paired abortion with each of thefollowing words: anxiety, depression, mental disor-ders, mental health, trauma, PTSD, domestic violence,drug abuse, emotions, employment, life satisfaction,self-esteem, somatoform, stigma, substance abuse, sui-cide, acute psychosis, schizophrenia, psychiatric symp-toms, and psychosocial factors. In addition,postabortion syndrome, postabortion adjustment, andtherapeutic abortion were also used as search terms.The search results were supplemented by a manualsearch of reference sections of reviewed articles. Thissearch strategy resulted in an initial set of 216 uniquereferences. Seven additional references were broughtto the attention of the task force by reviewers.

Our review process consisted of four steps. In the firststep of review, the abstract of each article in the initialset was reviewed independently by two task forcemembers according to the following inclusion criteria:(1) The study reported empirical data of a quantitativenature (qualitative studies were omitted). (2) Thestudy was published in a peer-reviewed journal (disser-tations, letters to editors, reviews, book chapters, andconference proceedings were omitted). (3) The studyincluded at least one postabortion measure related tomental health (those that considered only mentalhealth prior to the abortion were omitted). (4) Thestudy focused on induced abortion [those that focusedsolely on “spontaneous” abortions (miscarriages) orthat did not differentiate miscarriage from inducedabortion were omitted].

Those articles that appeared to meet all of the abovecriteria were included for further review. In the secondstep, a minimum of two task force members independ-ently read all articles identified in our first step. Onlyarticles judged to have met all of the above inclusioncriteria were retained. In the third step, all studies thatmet criteria for inclusion were coded, summarized,

and evaluated independently by at least two membersof the task force, with the restriction that task forcemembers did not evaluate their own work.

In a final step, articles were categorized according towhether or not they included a comparison group ofwomen who did not have an abortion. Only studiesthat include a comparison group are capable of ad-dressing the question of relative risk. Accordingly, ourcore review focused only on studies that includedcomparison groups. Studies without a comparisongroup have the potential to address predictors of indi-vidual variation in women’s responses following abor-tion. They also are capable of addressing the questionof prevalence of mental health problems amongwomen who have abortions, but only to the extentthat they are based on a sample representative of thepopulation to which one intends to generalize. Ac-cordingly, in a separate section we review such studies,but only when based on a U.S. sample.

DescriptiveOverview of Literature Identifiedfor This ReviewThrough the process described above, 50 papers wereidentified that compared psychological experiences ofwomen after abortion to psychological experiences of acomparison group of women. These 50 include studiesbased on U.S. and international samples. The restric-tion of empirical studies to those published in Englishresulted in a relatively narrow slice of internationalcontexts represented in this report. One should not as-sume that this small set is representative of the globalexperience of abortion and mental health, as laws, cus-toms, and contexts vary widely. Twenty-five paperscompared women who had an abortion to womenwho had a different reproductive history (e.g., a deliv-ery, miscarriage, no pregnancy) by performing second-ary analyses of public data sets or records originallycollected for other purposes; 18 of these papers werebased on U.S. samples; the remaining papers werebased on samples from New Zealand (1) and Finland(6). These are summarized in Tables 1 and 2. A secondset of papers (N=19) described original studies con-ducted primarily for the purpose of comparing re-sponses of women who had a first-trimester abortion(or an abortion of unspecified gestation) to responsesof women who had a different reproductive history.Most of these studies were based on samples collectedat clinics or physicians’ offices; some were retrospec-tive. Seven were conducted in the United States, the


remainder in other countries. These studies are summa-rized in Tables 3a and 3b. A third set of papers (N=6)consisted of studies comparing psychological experi-ences of women who had a late-trimester abortion of apregnancy for reasons of fetal anomaly to anothergroup of women. All but one was conducted on non-U.S. samples. These studies are summarized in Table 4.These 50 papers constitute the core of our review. Ourliterature search also identified 23 papers based on U.S.samples that did not include a comparison group butmet all other inclusion criteria. These papers are sum-marized in Table 5.

REVIEWOF COMPARISONGROUP STUDIES

Record-Based Studies and Secondary AnalysesWithComparisonGroups

The major change in the scientific literature during thetime period encompassed by our review compared tothe literature reviewed by the first APA task force wasthe publication of 25 papers in peer-reviewed journalsbased on secondary analyses of publicly available datasets. The studies are of two types: (a) analyses of databased on medical records and (b) analyses of data setscollected for purposes other than analyzing the rela-tionship between pregnancy outcome and mentalhealth. Because publicly available data sets often in-clude questions about reproductive histories, includingpregnancy outcomes (abortion, delivery, miscarriage),they provide an opportunity for comparing womenwho report having had an abortion to other groups ofwomen. Utilizing existing data sets, particularly longi-tudinal data sets, also has the advantage of being ableto ask and answer questions without having towait the years it takes to conduct a prospective studyfocused specifically on abortion. Findings based onnational probability samples potentially may be gener-alized more widely than those based on conveniencesamples and may be more useful for estimating nor-mative effects. Nonetheless, as pointed out above inthe methodological issues section of this report, thereare many serious limitations of this approach that se-verely constrain conclusions that can be drawn fromthese studies (see also McCall & Appelbaum, 1991).In the following discussion, we provide a brief descrip-tion of these studies, followed by an evaluation of

their methodology. Table 1 and Table 2 provide a de-scription of the key methods, measures, and findingsof these studies, as well as their limitations.

Medical records. Ten papers were published based onmedical records. Four papers were based on analyses ofmedical records from California’s state-funded insur-ance program (Medi-Cal). This program provideshealth care for low-income children and families, aswell as elderly, blind, and disabled persons in the stateof California. These “at-risk” women may be facing awide range of challenges that compromise their physicaland mental health. Six reports were based on officialhealth register data drawn from medical records and onthe entire population of Finland (See Table 1).

All four Medi-Cal studies focused on an initial targetpregnancy event (abortion vs. delivery) in the last halfof 1989 and after excluding women with subsequentabortions only from the delivery group, examined therecords of the remaining sample of women for subse-quent death (Reardon et al., 2002), outpatient admis-sions (Coleman, Reardon, Rue, & Cougle, 2002b),inpatient admissions (Reardon, Cougle, Rue, Shuping,Coleman, & Ney, 2003), and sleep disturbances(Reardon & Coleman, 2006). All four papers reportedhigher rates of negative outcomes in the abortiongroup compared with the delivery group.

In considering the weight of the evidence with regardto the mental health implications of abortion, itshould be kept in mind that the Medi-Cal studies arenot independent of each other because the samplesoverlap, and most of the outcomes examined are cor-related. Strengths of the Medi-Cal studies include anobjectively verifiable abortion history and the use ofdiagnostic codes for assessing mental illness. Nonethe-less, these papers are characterized by a number ofmethodological limitations that make it difficult to in-terpret the results. These include differential exclusionof women with subsequent abortions from the deliv-ery group but not from the abortion group, a sam-pling strategy that both advantaged the delivery groupand rendered generalizability of the findings problem-atic; lack of basic demographic information known tobe associated with mental health, including maritalstatus and race; lack of information about previous re-productive history, lack of adequate assessment ofprior mental health history, lack of adequate informa-tion about co-occurring risks (e.g., health status, vio-



Table 1:Medical Records–Abortion vs.Comparison GroupsU.S.STUDIES

Medi-Cal Data SetGeneral Description:Medi-Cal is California’s state-funded medical insurance program for low-income individuals; 249,625 women identified ashaving a“short paid claim”for Medi-Cal funding for either an abortion or delivery (pregnancy event) in 1989; for most studies 194,694 women wereidentified as citizens with valid SSN. Samples for the studies below were based on this subgroup. In 1989, pregnant women were Medi-Cal-eligible iffamily income was less than 185% of federal poverty level.

Limitations Common to All Studies Based on this Data Set: Pg intendedness or wantedness not controlled; basic covariate info (e.g., race,marital status, # births & abortions) unknown; inadequate controls for prior mental illness; sample representativeness suspect, even for generalizing tolow-income population—more than 20% of the sample excluded before samples of specific studies selected. Pg outcomemay affect eligibility in differ-ent ways: having a baby may qualify a woman for Medi-Cal, independent of her own characteristics,while women who remain on Medi-Cal postabortionwould have to qualify for other reasons.

Causality direction ambiguous—women with poor health status may be more likely to choose abortion.Misleading“First Pregnancy” label used to iden-tify target population cannot be specified; N is so large that minute differences can be statistically sig. Impact of controlling months of eligibility not clearas womenmay have lapses of coverage during period examined.Given poorer health of low income populations, inability to separate therapeutic fromelective abortions a particular limitation.

Citation

Reardon,D.C.,Ney,P.G.,Scheuren,F.,Cougle, J.,Coleman,P.K.,& Stra-han, T.W. (2002).Deathsassociated with preg-nancy outcome:Arecord linkage study oflow incomewomen.(2002) SouthernMedicalJournal,95, 834-841.

Coleman,P. K., ReardonD.C.,& RueV.M.(2002).State-funded abortionsversus deliveries:Acomparison of outpa-tient mental healthclaims over 4 years.American Journal ofOrthopsychiatry,72,141-152.

Sample &Procedure

Medi-Cal records linkedto death certificates be-tween 1989-89; afterscreening for“aberrant,indeterminate, and out-of-scope data”173,279(1,294 deaths) casesused in primary analy-ses; some analysesexcludedwomenwithinpatient and outpa-tient psychiatric claimsin preceding 6-18months & womenwithsubsequent abortionfrom delivery group.

Differs from generaldescription above inreporting 193,794women as having avalid SSN; after screen-ing exclusions,womenwith target Pg event inthe last half of 1989selected;womenwithsubsequent abortionexcluded from deliverygroup;womenwithboth inpatient or out-patient psychiatric ad-missions claim in yrpreceding target Pgevent excluded; finalsample = 54,419.

Sample Sizes

1.AB N= 30,260DEL N= 83,6902.AB N= 41,956DEL N= 17,472

AB N= 14,297DEL N= 40,122

Primary Outcome

Rates of causes of deathreported on Californiadeath certificate be-tween 1989 and 1998.

1.Cumulative rates ofoutpatient psychiatricadmission claims at90d,180d,yr1,& yrs 1-4after target Pg event;2.Rates of disorder in13 groups of selectedI CD-9 diagnostic cate-gories.

Key Findings

1.Age-adjusted risk ofdeath significantlyhigher in AB group fromviolent causes but notfor nonviolent causes.2.Womenwith subse-quent abortions wereexcluded only fromDELgroup and number ofpsychiatric claims inprevious yr controlled,age-adjusted risk ofdeath significantlyhigher in AB group forboth violent andnonviolent causes.

1. Significantly highercumulative rates of out-patient claims for ABgroup controlling forage,number of Pgevents,&months ofMedi-Cal eligibility.2. Of 13 comparisons,AB group rates signifi-cantly higher in 4 cate-gories (adjustmentreactions;bipolar disor-der; neurotic depres-sion; schizophrenicdisorders);marginallysignificant in 2 (anxietystates; alcohol & drugabuse).

Additional Limita-tions Specificto Study Listed

39,329 (65%) of ex-cluded sample is with-out explanation; 8-yearperiod in which deathsidentified not congru-ent with ACOG defini-tions of pregnancy-related or even preg-nancy-associated death.Differential exclusionadvantages deliverygroup.

Differential exclusion ofwomenwith subse-quent abortion fromDEL grp; inadequatecontrol through exclu-sion for prior mentaldisorder; flaw of“valida-tion by cross-quotation”found in claim of evi-dence for causal model“accumulating”basedon citation toresearch that doesnot warrant that claim(p.149).

lence exposure), lack of information about criticalcharacteristics of the abortion decision context(e.g., whether the pregnancy was initially intended andterminated because of fetal anomalies ), and inclusionof covariates across analyses and studies that variedfor unspecified reasons (see Table 1). Yet anotherproblem with this data set is that women who delivera child are more likely to be eligible for Medi-Cal be-cause they have a baby, independent of their owncharacteristics. Women who have an abortion mayqualify for the abortion, but those who remain onMedi-Cal post abortion (and who hence would bepicked up in the follow-up measurement) would haveto have other characteristics besides motherhood to

qualify (e.g., mental illness, other illness, poverty notassociated with parenthood).

The Medi-Cal findings with regard to cause of death(Reardon et al., 2002) can be compared with record-based studies conducted in Finland that are basedon the entire population of the nation (Gissler, Hem-minki, & Lonnqvist, 1996; Gissler et al., 1997),albeit from a differing cultural context. These studiesalso found significantly higher rates of pregnancy-associated deaths for natural and violent causes(including accidents, homicide and suicide) in theabortion group compared with a delivery group.Like the Medi-Cal studies, these studies also had


Table 1:Medical Records–Abortion vs.Comparison GroupsU.S.STUDIES (continued)

Citation

Reardon,D.C.,Cougle, J.R., RueV.M., ShupingM.W.,Coleman P.K.,& NeyP.G. (2003).Psychiatricadmissions of low-in-comewomen followingabortion and childbirth.CanadianMedical Asso-ciation Journal,168,1253-1256.

Reardon,D.C.,& Cole-man, P.K. (2006). Rela-tive treatment for sleepdisorders followingabortion and child de-livery: A prospectiverecord-based study.Sleep,29, 105-106.

Sample &Procedure

After screening exclu-sions, womenwith tar-get Pg event in the lasthalf of 1989 were se-lected; womenwith in-patient psychiatricadmissions claim inyear preceding targetPg event excluded;womenwith subse-quent abortion ex-cluded from deliverygroup; final sample56,741.

After screening exclu-sions, womenwith ahistory of treatment forsleep disorder ex-cluded; womenwithsubsequent abortionexcluded only fromdelivery group; finalsample = 56,824 cases.

Sample Sizes

AB N= 15,299DEL N= 41,442

AB N= 15,345DEL N= 41,479

Primary Outcome

1.Cumulative rates ofinpatient psychiatricadmission claims at90d,180d,& yr1 aftertarget Pg event; 1sttime rates in yr 1,2,3,&4 after target Pg event;2.Rates of disorder in 9groups of selected ICD-9 diagnostic categories.

Cumulative rates oftreatment for categoryrepresenting nonor-ganic sleep disorderand sleep disturbancesat 180d,yr1, and 1-4years after target Pgevent; 1st time rates yr1 through 4 after targetPg event.

Key Findings

1.Controlling for ageandmonths ofMedi-Caleligibility to the endofthe timeperiod ana-lyzed, theABgrouphadsignificantly higher ratesfor both cumulative and1st time rates of inpa-tient claims for ABgroupat timeperiods listed;2.Of 9 comparisons,ratesof ABgroupwere signifi-cantly higher in 4 cate-gories (adjustmentreaction;depressive psy-chosis, single episode;depressive psychosis,re-current episode;bipolardisorder).

Controlling for age andnumber ofmonths ofMedi-Cal eligibility, sig-nificantly higher treat-ment rates in AB groupat 180 d ,y1 & yr 4,& sig-nificantly higher 1sttime rates in yr 3,butnot yrs 2 & 4.


Reluctance to hospital-ize newmothers couldaccount for lower post-delivery admissionrates.Misleading use ofterm“first admission”because only mentalhealth claims for oneyear prior to Pg wereexamined. Inadequatecontrols for prior men-tal illness.

Impact of controllingfor months of eligibilityis not clear as authorsnote that somewomenhad lapses of coverageduring the period ex-amined.


Table 1:Medical Records–Abortion vs.Comparison GroupsINTERNATIONAL STUDIES—Finland

General Description:National data registers based onmedical recordsmake it possible to examine health status of the entire population of the countryso underreporting bias not amajor issue.These studies provided inspiration for Medi-Cal studies.Note outcomes based on ACOG definitions of Pg-associateddeaths (occurring within one year of end of pregnancy, regardless of cause of death) vs.Pg-related deaths (occurring within one year of end of Pg from anycauses related to or aggravated by their Pg or its management,but not from accidental or incidental causes) differ from definitions in Medi-Cal studies.

Limitations Common toAll Studies Based on this Data Set: Neither intendedness nor wantedness of Pg controlled; information on age,maritalstatus, and reproductive history lacking; low rates of unintended pregnancy and ready access to abortion in Finlandmake it likely most births are wanted.

Citation

Gissler,M.,Hemminki, E.,& Lonnqvist, J. (1996).Suicides,1987-94: regis-ter linkage study.BritishMedical Journal 313,1431-1434.

Gissler,M.,Kauppila,R.,Merilainen,J.,Toukomaa,H.,&Hemminki,E.(1997).Pregnancy-associateddeaths in Finland1987-1994--Definitionproblemsandbenefits of record link-age.ActaObstetGynecolScand,76,651-637.

Gissler,M.,& Hemminki,E. (1999).Pregnancy-re-lated violent deaths.Scand J Public Health,1,54-55. [Letter to editor].

Sample &Procedure

Death register recordsfor 1347 suicides werelinked to birth, abortion,and hospital dischargerecords, identifying 73deaths occurring within1 year of a birth orabortion.

Record linkage study ofwomen of reproductiveage between 1987-1994; 281 deaths identi-fied as Pg-associated.

Additional analyses ofviolent death identifiedrecord linkage study ofviolent deaths amongthe 281 Pg-associateddeaths identified inGissler et al (1997).

Sample Sizes

AB N= 29DEL N= 30

AB N= 84Miscarriage N= 40DEL N= 137

AB N= 84Miscarriage N= 40DEL N= 138

Primary Outcome

Suicide rates

Rates of causes of death

Rates of causes ofviolent death

Key Findings

Suicide rate significantlyhigher in ABgroup:Di-vorcedwomen andwomen in the lower so-cial classeswere over-represented in theABsuicide group vs.womenin the abortion registeroverall.

Higher age-adjustedrates for overalldeaths,natural deaths,accidents, suicides,&homicides in AB group.

Higher age-adjustedrates of accidents,suicides,& homicides inAB group.


Given findings on classandmarital status in ABgroup,lack of control forwantedness,exposure toviolence,class,parity,andcircumstances of the Pgmakes comparisons be-tweenAB andDELgroups problematic.

Pg-related deaths notidentified.Only agecontrolled.

Only age controlled.Thesedata arebasedon thesame records asGissler etal.(1997) & apparentlywere an attempt tocounter claims thatGissleret al (1996) implied causa-tion. Authors emphasizethepoint that given the“finding that the risks foraccidental death andhomicide also increaseafter an inducedabortionandourprevious findingsthatwomen from lowersocial classes and singlewomenareover-repre-sented amongwomencommitting suicides afteran inducedabortion,donot support thehypothe-sis that abortion itselfcauses suicides”(p.55).


Table 1:Medical Records–Abortion vs.Comparison GroupsINTERNATIONAL STUDIES—Finland (continued)

Citation

Gissler,M.,Berg C.,Bou-vier-Colle, M.H.,&Buekens,P. (2004).Meth-ods to or identifyingpregnancy-associateddeath:population-based data from Finland1987-2000.Pediatric andPerinatal Epidemiology,18, 448-455.

Gissler,M.,Berg C.,Bou-vier-Colle M.H.,Buekens P. (2004). Preg-nancy-associatedmor-tality after birth,spontaneous abortion,or induced abortion inFinland,1980-2000.American Journal of Ob-stetrics andGynecology,190, 422-427.

Gissler,M.,Berg C.,Bou-vier-Colle M.H.,&Buekens,P. (2005). Injurydeaths, suicides, andhomicides associatedwith pregnancy,Fin-land, 1987-2000. Euro-pean Journal of PublicHealth,15, 459-463.

Sample &Procedure

Record linkage study ofwomen of reproductiveage between 1987-2000; 419 deaths identi-fied as Pg-associated.

Record linkage study ofPg-associated deaths1987-2000;of the15,823 womenwhodied,419 of the deathswere Pg-associated;.

Record linkage study ofPg-associated deaths1987-2000 from exter-nal causes; of the 5,299womenwho died,212of the deaths were Pg-associated;.

Sample Sizes

AB N= 129DEL N= 224

AB N= 129DEL N= 224

AB N= 92DEL N= 81

Primary Outcome

Rates of causes of Pg -associated deaths

Rates of natural andviolent causes ofPg-associated andPg-related deaths

Pg-associated deathsfrom external causes

Key Findings

Higher Pg-associatedmortality rates for abor-tion compared to birth

1.Pg-associated deathrates fromnaturalcauses (particularly nat-ural causes unrelated toPg) & from violentcauses higher in ABgroup.Direct Pg-relatedcauses higher in DELgroup,but significancenot reported (3.9 &1.3/100,000 Pg).2.When therapeuticabortions excluded,Pg-associatedmortality rateshigher in theDELgroup.

2.Death rates higher inAB group then DELgroup for all externalcauses, includingrates for unintentionalinjuries, suicide,&homicide.


Pg-related deaths notidentified;nothing wascontrolled.

Only age controlled.These findings include1987-2000 cases usedin previous studies, soare not independent.Therapeutic abortionsin early Pg likely underidentified.

These findings include1987-2000 cases usedin previous studies, soare not independent.Therapeutic abortionsin early Pg likely underidentified.Only agecontrolled.Authorsstate that their findingsdo not warrant causalconclusions and em-phasize the need formore information onrelevant covariates, in-cluding“mental health,social well-being,substance abuse, andsocio-economic cir-cumstances” in furtheranalyses (p. 462.)

Notes: AB = Abortion DEL =Delivery; Pg = pregnancy;ACOG=American College of Obstetricians and Gynecologists; ICD - International Classification of Diseases;Grp = Group;Sig = Significance


Table 1B: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsU.S.STUDIES

National Longitudinal Survey ofYouth (NLSY)General Description: The National Longitudinal Survey of Youth (NLSY) is based on annual interviews with a stratified,multi-stage national proba-bility sample of noninstitutionalized civilian men and women aged 14-21 as of 1979,with oversampling of Blacks,Hispanics, and poorWhites. Relevantmeasures include: an abbreviated version of the Rotter internal-external locus of control scale (IRotter, 1966; IE assessed in 1979); global self esteem(Rosenberg, 1979; RSE assessed in 1980 & 1987); Center for Epidemiological Studies-Depression Scale (Radloff, 1977; CESD assessed in 1992); reproductivehistories were first taken in 1982 and updated every 2 years subsequently.

Limitations Common to All Studies Based on this Data Set: No study used sampling weights so that normative statements are inappropri-ate and alpha levels are likely to be elevated, increasing probability of identifying difference due to chance as a reliable difference.Underreporting ofabortion raises question of possible reporting bias but direction of reporting bias unclear as womenmay be less likely to report stigmatized experiences(having an abortion,mental problems, experiencing violence), but those who are willing to report one stigmatized condition may be more willing to re-port others, increasing the likelihood of finding a correlation between 2 stigmatized events.Ns of analyses vary depending on covariates so are not alwaysclear. Large sample sizes mean that small effects are statistically significant.

Citation

Russo,N.F.,& Zierk,K.L.(1992).Abortion, child-bearing, and women’swell-being.ProfessionalPsychology: Researchand Practice,23,269-280.

Russo,N.F.& Dabul,A. J.(1997).The relationshipof abortion to well-being. Do race and reli-gionmake a difference?Professional Psychology:Research and Practice,28,23-31.

DataSource/PopulationStudied

1. 5,295 women forwhom there were NLSYinterviews involving theassessment of well-being in 1987;773 hadat least one abortion;233 had repeat abor-tions.2. Additional analysesbased on 4502 womenwho had no abortionsbefore their 1980 inter-view.

1. 4913 women drawnfrom the sample of5,295 women describedabove (3572White &1341 Black); 721 had aleast one abortion,175had repeat abortions.2. Additional analysesbased on 4336 women(3,147White & 1,189Black) who had noabortions prior to 1980interview.

Sample Sizes

1.AB N = 733Other N = 45622.AB N = 317Other N =4185

1.AB N = 721Other N =41922.AB N = 317Other N =4502

Primary Outcomes

1987 Globalself-esteem (RSE)


Key Findings

1.Womenwho had 1abortion had higher SEthan other two groups;when childbearing andresource variables werecontrolled,neitherhaving 1 abortion norhaving repeat abortionswere significantlyrelated to RSE.Totalabortions correlatedwith total unwantedbirths (r=.11).2.1980 RSEwas thestrongest predictor of1987 SE (partial r=.38).

Primary findings did notvary across groupsknown to vary in under-reporting.1.When childbearingand resource variableswere controlled,neitherhaving 1 abortion norhaving repeat abortionssignificantly related toRSE, regardless of raceor religion.2.1980 SEwas thestrongest predictor of1987 SE (partial r=.39-42) regardless of raceor religion.

Notes andAdditional Limita-tions Specificto Study Listed:

No clinical cut off score& clinical significance ofscores is unknown;large samplemeanssmall effects statisticallysignificant. Limited towomen under 33 yearsof age in 1987.

Religionmeasured in1979 only; highly com-mitted fundamentalistwomen not identified;sample does not in-clude Asians or NativeAmericans. Limited towomen under 33 yearsof age in 1987.


Table 1B: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsU.S.STUDIES (continued)

Citation

Reardon ,D.C.,&Cougle, J.R. (2002).Depression and unin-tended PG in theNational LongitudinalSurvey of Youth: Acohort study. BritishMedical Journal,324,151-152.

Sample &Procedure

Two samples weredrawn due to coding is-sues in the initial study;both the initial and cor-rected sample ns are re-ported here.1. Initial sample:421women identified as re-porting a first unin-tended Pg between1980 and 1992 that re-sulted in abortion ((N=293) or deliverywith no subsequenthistory of abortion inthe delivery grp(N=128 ).2. Corrected sample:1076 women identifiedas reporting a first unin-tended Pg between1980 and 1982 that re-sulted in abortion ((N=293) or deliverywith no subsequenthistory of abortion inthe delivery grp ((N=783). Results weresimilar in both samples& only results of cor-rected sample pre-sented here.

Sample Sizes

1. AB N=293DEL N=1282. AB N=293DEL N=783

Primary Outcome

%women exceedingthe 1992 CESD cut-offscore (>15).

Key Findings

AB grp had higher %scoring >=16 on CES-Din 1992 (27% vs.25%),controlling for family in-come, education, race,age at 1st Pg,and 1979I-E score .

Significantly higher riskfor AB grp amongmar-ried women (26% vs.19%),but not amongunmarried women(36% vs,29%,ns), con-trolling for family in-come, education, race,age at 1st Pg., and 1979I-E score.


Note:Differs from RSEstudies in focusing onoutcome of 1st Pg.Subsequent reanalysisby Schmiege & Russo(2005) showed thatfindings in correctedsample still based onmiscoded data.Exclusion of womenwith subsequent his-tory of abortion fromthe delivery group.Uses I-E score as a con-trol for pre-existingmental health but scaleis not ameasure ofmental health.CESDcontroversial due tocutoff at >15 yieldinghigh rate of false posi-tives and lack of speci-ficity of measurement.Generalizing to all 1stPg is inappropriate - re-stricting sample to onlythose womenwho hadcompleted the Rotter I-E scale in 1979,effec-tively eliminatedmost(339 of 425) of theteenagers who had de-livered; women in thepre-1980 DEL grp thatwas eliminated had thehighest % exceedingCESD cut-off (34%)compared to pre-1980AB (27%) and post-1980 AB ( 24%) & DEL(24%) grps. Limited towomen under 38 yearsof age in 1992.Variableused to define race in-cluded nonBlack andnonHispanic minoritiesin theWhite category.



Citation

Cougle, J.R., Reardon,D.C.,& Coleman, P.K.(2002).Depression as-sociated with abortionand childbirth:A long-term analysis of theNLSY cohort.MedicalScienceMonitor,9,CR105-112.

Sample &Procedure

Drawn from a largersubsample of 1,884womenwith first abor-tion or first delivery be-tween 1980 and 1992andwho had com-pleted both the 1979Rotter I-E scale and the1992 CES-D scale; totalAB & DEL grp ns not re-ported; average agefigure based on 884women (AB = 293;DEL = 591); subsamplens varied from 1031 -1361 depending on theanalyses.

Sample Sizes

AB N= 131 - 164DEL N= 877 - 1197

Primary Outcome

%women exceedingthe 1992 CESD cut-offscore (>15).

Key Findings

AB grp had higher %scoring >=16 on CES-Din 1992 (27% vs.21%),controlling for age, race,education, income,mar-ital status,history of di-vorce, and abbreviatedI-E score.

AB group had higherdepression risk amongwomenwhowereWhite,married,andwho did not have a 1stmarriage ending in di-vorce, controlling forrelevant covariates.

Significant differencesnot found amongBlack/Hispanic women,unmarried women,orwomenwith a 1st mar-riage ending in divorce,controlling for relevantcovariates.


This study is similarlydesigned and based onthe women erroneouslyidentified in first set ofanalyses in Reardon &Cougle (2002) , exceptthat womenwho hadintended pregnanciesare now added to DELgroup, reducing% ex-ceeding cut-off score.Reasons for discrepan-cies in AB & DEL groupsfrom previous study notclear,possibly due todifferent covariates(age vs.age at 1st Pg)used in the two studiesfor unknown reasons.Average age based on884 women so difficultto understand where nsexceeding that n in theregression analysescame from given age isa covariate in thoseanalyses.Variable usedto define race includednon-Black and non-His-panic minorities in theWhite category.

methodological limitations, including lack of informa-tion about pregnancy wantedness and lack of assess-ment of other critical variables known to co-varywith both pregnancy outcome and mental health(e.g., prior reproductive history, prior mental healthproblems, violence exposure, etc).

The largest and most methodologically rigorous Fin-land study used definitions provided by the AmericanCollege of Gynecology (ACOG) to analyze directpregnancy-related deaths (deaths occurring withinone year of end of pregnancy from causes related toor aggravated by the pregnancy or its management,

but not from accidental or incidental causes) sepa-rately from pregnancy-associated (deaths occurringwithin one year from end of pregnancy, regardless ofcause of death) (Gissler, Berg, Bouvier-Colle, &Buekens, 2004b).These analyses revealed that womenin the abortion group had lower rates of pregnancy-related deaths than women in the delivery group (1.3vs. 3.9 per 100,000 pregnancies), but higher rates ofpregnancy-associated deaths. However, when thera-peutic abortions were excluded from the categoryof pregnancy-associated deaths, women in the abor-tion group no longer had higher pregnancy-associateddeath rates than women in the delivery group.

This study affirms the importance of making a dis-tinction between pregnancy-related and pregnancy-associated deaths in drawing valid conclusions aboutthe association between abortion (vs. delivery)

and subsequent risk for various causes of death andalso establishes the importance of separating thera-peutic from elective abortions when attempting todraw such conclusions.



Citation

Schmiege,S.,& Russo,N.F. (2005).Depressionand unwanted firstpregnancy: Longitudi-nal cohort study. BritishMedical Journal,331,1303-1305.

Sample &Procedure

Two samples weredrawn due to coding is-sues in the initial study;both the initial and cor-rected sample ns are re-ported here.1. Initial sample:1247women identified asreporting a first un-wanted PG between1970 and 1992 thatresulted in abortion(N=479) or delivery(N=768).2. Corrected sample:1744 women identifiedas reporting a first un-wanted Pg 1970 & 1992that resulted in abor-tion (N=461) or delivery(N=1283).Results weresimilar in both samples& only results fromcorrected samplepresented here.

Sample Sizes

1.AB N=479DEL N=7682.AB N=461DEL N=1283

Primary Outcome

Both%women exceed-ing the 1992 CESD cut-off score (>15) andcontinuous 1992 CESDscores reported. Educa-tion, income,and familysize also examined.

Key Findings

%exceeding cutoffscore on 1992 CESD didnot significantly differfor AB vs DEL groups,controlling for age at1st Pg., race,marital sta-tus, education,and fam-ily income, in either thefull sample (25% vs.28%) or the post-1979subsample (23% vs.23%) for all women.

AB sig.associatedwithlower education and in-come and larger familysize,all risk factors fordepression. Additionalanalyses published in re-sponse to debates overpoints of design did notchange the pattern ofresults. The only sig.dif-ference betweenAB&DEL grps foundwas inunadjusted analyseswhen subsequent abor-tions excluded fromboth groups (AB = 21%>15 vs.DEL = 28% >15);the differencewas notsig.when covariatescontrolled.

Findings did not varyacross groups known tovary in underreporting,includingmarriedwhite women,umarriedWhite women,unmar-ried Black women,non-Catholics, and Catholics.


Note:NLSY staff pro-vided coding to ensureproper identification ofsample,but last line ofcode inadvertentlyomitted in initial analy-ses. Differs from otherstudies in focusing onunwanted 1st Pg.Studycriticized for not con-trolling same variabls asprevious studies, result-ing in a series of analy-ses, including thoselimited to post-1980 AB& DEL grps. Althoughunderreporting bias aconcern, findings didnot differ among grpsknown to vary in suchbias. Limited to womenunder 38 years of age in1992.



Citation

Reardon,D.,Colemen,P.K.,& Cougle, J.R. (2004).Substance use associ-ated with unintendedpregnancy outcomes inthe National Longitudi-nal Study of Youth.American Journal ofDrug andAlcohol Abuse,30, 369-383.

Sample &Procedure

After excluding allwomen Pg before 1980,identified 1748 womenreporting a first unin-tended PG between1980 and 1988 that re-sulted in abortion(N=213) or delivery(N=535) ,or had neverbeen Pg. ( N= 1144); asubsample of womenresponded to alcoholquestions, alcoholanalyses appear to bebased on 1243 women.

Sample Sizes

AB N= 213DEL N= 535Never Pg N= 1144

Primary Outcome

11 yes/no items relatedto alcohol abusesymptoms;4 related tosubstance use (# daysdrank in last mo;# drinks consumed ondays when drank; ifever usedmarijuana orcocaine in last mo).

Key Findings

Controlling for age, race,marital status, income,education, pre-Pg RSEand pre-Pg I-E.,no sig.differences amonggroups on # of drinks; in% scoring 2 or more or% scoring 4 or more onitems related to alcoholabuse; in the numberof drinks consumed,orin the use of cocaine.AB grp drank sig.moredays in last mo (6.36)than DEL grp (4.79) butnot than Nev Pg grp(5.93); and weremorelikely to usemarijuanain last month (18.6%)than the DEL or Nev Pggrps (7.9%).


Exclusion of women Pgbefore 1980makessample unrepresenta-tive and generalizationto unintended first Pginappropriate as notedabove.The large num-ber of tests performed,single itemmeasures ofkey dependent vari-ables, and small magni-tude of effects limitconclusions that can bedrawn from this study.Drinking on an averageof 6.36 (AB) vs.4.79(DEL) days per mo.notindicator of cliniciallysignificant alcoholabuse.Variable used todefine race includednonBlack and nonHis-panic minorities in theWhite category.

The most consistent findings across the Medi-Cal andFinland record-based studies were the higher rates of vio-lent death for women in the abortion group. In the Fin-land study described above, women in the abortion grouphad higher rates of violent pregnancy-associated deaths,and a higher proportion of their overall pregnancy-associ-ated deaths were due to violent causes (Gissler et al.,2004b). In interpreting this finding, it is useful to recallthe distinction between risk and cause discussed above.Abortion is a marker of risk for violence, not a cause ofviolence. Thus it is important to control for violence ex-posure in studies of pregnancy outcome.

Secondary analyses of survey data. Fifteen papersbased on secondary analyses met inclusion criteria forour review. These were based on nine data sets. Eightdata sets were from the United States: Five were basedon U.S. national probability surveys, and three werebased on local metropolitan area surveys. One paper

was based on analyses of the longitudinal NewZealand Christchurch Health and Development sur-vey. Key findings and methodological limitations ofthese studies are summarized in Table 2.

National Longitudinal Survey of Youth (NLSY).The NLSY has been the data set used most fre-quently to examine the relationship of abortion tomental health outcomes. The NLSY is a longitudinalnational survey of a cohort of males and femalesaged 14-21 years in 1979. Papers meeting our inclu-sion criteria assessed the following outcome vari-ables: self-esteem measured in 1987 (2 studies), riskfor depression measured in 1992 (3 studies), andsubstance use measured in 1988 (1 study). This setof papers demonstrates the problems of trying tobase conclusions about the mental health effectsof abortion on secondary analyses of data sets col-lected for other purposes. Conclusions of researchers

analyzing this same data set and even the same de-pendent variable varied markedly depending on sam-pling and analytic strategy.

Self-esteem. The first of the abortion studies to bebased on this data set focused on self-esteem asmeasured by the Rosenberg self-esteem scale (RSE;Rosenberg, 1965). This first study (Russo & Zierk,1992) analyzed a total sample of 5,295 women (773of whom reported having at least one abortion).Women who had an abortion had mean RSE scorescomparable to those of all women (33.3 vs. 33.2,respectively); women who had one abortion alsohad significantly higher RSE in 1987 than the other

two groups (women with no abortions, women withrepeat abortions), although the relationship wasextremely small. When contextual variables werecontrolled (education, income, employment, mar-riage, number of children, whether the pregnancywas wanted or unwanted), however, neither havingone abortion nor repeat abortions was related tosubsequent self-esteem. After eliminating from thestudy women who had an abortion before RSEwas measured in 1980, further analyses found thatpreexisting self-esteem was the most importantpredictor of 1987 RSE, followed by having moreeducation, higher income, employment, and fewerchildren.



National Longitudinal Study of Adolescent Health (ADD-HEALTH)General Description: ADD-HEALTH uses a multi-stage, school-based, longitudinal design in which data were three times: initial (1994-1995), and ap-proximately 1 year (1996), and 6 years (2001-2002) later. AtWave I all participants were in grades 7-12.All Wave I (N= 90,118) completed an in-school ques-tionnaire;a subsample (N=12,105)) completed an additional computer-assisted in-home interview that included questions about sexual history andreligion.This subsample was chosen by identifying a group of students who were representative of the adolescent population in grades 7-12 during the1994-1995 school year; in addition, adolescents who were disabled,African American students from well-educated families, Chinese, Cuban, Puerto Rican,living with twin, living with a full sibling, living with a half sibling, living with a non-related adolescent, and siblings of twins were oversampled.

Limitations Common to All Studies Based on this Data Set: School-based population does not include students who drop out due to Pg;ethnic minorities in sample may be particularly unrepresentative of the adolescent population as a whole. 1-itemmeasures psychometrically weak.

Citation

Coleman,P.K. (2006).Resolution of unwantedpregnancy during ado-lescence through abor-tion versus childbirth:Individual and familypredictors and conse-quences. Journal ofYouth andAdolescence,35, 903-911.

Data Source/PopulationStudied

130 adolescents ingrades 7-11 who com-pleted bothWaves I & IIand experienced a Pg.described as“notwanted”or“probablynot wanted”.

Sample Sizes

AB N = 65DEL N = 65

Primary Outcome

Single-itemmeasuresof counseling,12-month trouble sleeping,30-day cigarette use,30-daymarijuana use,12-month alcohol use,problems with parentsandwith school due toalcohol use.

Key Findings

Controlling for risk tak-ing and desire to leavehome,AB groupmorelikely to have counsel-ing, trouble sleeping,and usemarijuana inpast 30 days (problemswith parents due to al-cohol use approachedsignificance).


Number of total preg-nancies unknown,butsmall n’s raise questionsabout underreportingand drop-out rates.Sin-gle item outcomemeasures psychometri-cally weak.Percentagesand ns for outcomevariables not reportedso frequency of prob-lem unknown;previousmental health prob-lems not controlled.Given the large numberof variables in the dataset,why these particu-lar variables were in-cluded is unclear.



National Survey of Family Growth (NSFG)General Description: The NSFG Cycle V sample is a subsample of 10,847 women aged 15-44 drawn from the larger national probability sample of theNational Health Interview Survey.The NSFG is thus a stratified,multistage design involving individual sampling rates that requires using sampling weightsin computing statistics.

Limitations Common to All Studies Based on this Data Set: Retrospective data that may involve recall of events occuring decades previously.

Citation

Cougle, J., Reardon,D.C.,Coleman,P.K.,& Rue,V.M. (2005). General-ized anxiety associatedwith unintended preg-nancy: A cohort studyof the 1995 NationalSurvey of FamilyGrowth. Journal of Anxi-ety Disorders,19,137-142


Study sample: (1) allwomen having an unin-tended Pg ending inabortion for their firstPg event and (2) allwomen having an unin-tended Pg ending inlive birth delivery fortheir first Pg event whohad no abortions afterthat Pg.Womenwhoexperienced a pro-longed period of anxi-ety previous to or at thesame age as the Pgevent were excludedfrom the sample.

Sample Sizes

AB N = 1033DEL N = 1813

Primary Outcome

Dichotomousmeasure(yes/no) of generalizedanxiety (GE)

Key Findings

Significantly higher rateof GE in abortion vs.de-livery group (13.7% vs.10.1%), controlling forrace and age at inter-view. In stratified sub-analyses, difference sig.for unmarried or under20 at 1st Pg,but not formarried women.

Additional Limita-tions Specificto Study Listed:

Women reporting pre-Pg anxiety excluded socannot generalize to allfirst unintended preg-nancies; misleading lan-guage impliesgeneralized anxiety dis-order (GAD) is assessed,but items used to con-struct generalized anxi-ety variable are notcongruent with DSMdefinitions of general-ized anxiety disorder,making clinical implica-tions problematic; dif-ferential exclusion fromwomenwith subse-quent abortions fromdelivery but not abor-tion group; samplingweights not used in sta-tistical analyses; stratifi-cation used rather thancontrolling for relevantvariables; analyses notconducted to deter-mine the contributionof abortion to varianceover and above otherrelevant predictor vari-ables.

This study reported a number of relationships that haveimplications for what should be controlled when analyz-ing NLSY data, especially the importance of controllingfor wantedness of pregnancy and separating womenwith one abortion from those having repeat abor-tions. The number of abortions was slightly but signifi-cantly and positively correlated with unwanted births

(r = .11). Furthermore, repeated unwanted pregnancy,regardless of pregnancy outcome (birth or abortion),was significantly correlated with greater likelihood ofliving in poverty (r = .15) and lower education (-.13).1

Depression risk. Using a very different approach,three studies focused on the effects of first pregnancy


outcome (abortion vs. delivery) on risk for subsequentdepression (measured in 1992 by the Center for Epi-demiological Studies-Depression scale (CES-D; Radloff,1977). Reardon and Cougle (2002a) focused on unin-tended first pregnancy outcome (abortion vs. delivery).After correcting an initial coding error, they reportedanalyses controlling for age at first pregnancy, race,marital status, and whether the woman was in her firstmarriage. They also attempted to control for prior men-tal health by including only women who had completedan abbreviated Internal-External Locus of Control scale(I-E Scale; Rotter, 1966), assessed in 1979, prior to hav-ing a first pregnancy. Among all women, 25% of thedelivery group exceeded the CES-D cutoff score for de-pression (>15) compared to 27% of the abortion group,a nonsignificant difference. Among married women inthis subsample, a significantly higher percentage ofwomen in the abortion group (26%) than in the deliv-

ery group (19%) exceeded the CES-D cutoff score.Among unmarried women in this subsample, the find-ings were reversed, although not statistically significant(36% vs. 29%).

Cougle et al. (2003) published another paper alsofocusing on firstpregnancy outcome (abortion vs.delivery) relative to the same outcome variable, 1992CES-D. This study is based on essentially the samesample as the previous one with the primary differencebeing that women with wanted pregnancies werealso included in the delivery group. Again, a largerpercentage of women in the abortion group exceededthe CES-D cutoff score for depression compared withwomen in the delivery group.

Both of these studies are characterized by a numberof problems, the most important of which are the

Citation

Russo N.,& Denious, J.(2001).Violence in thelives of women havingabortions: Implicationsfor practice and publicpolicy.Professional Psy-chology: Research andPractice,32,142-150.


Secondary analyses arandom householdtelephone survey ofover 2,500 women and1,000men aged 18 orover and residing in thecontinental U.S., con-ducted in 1993.Analy-ses based on responsesof 2,525 women,324 ofthem identified ashaving had at least1 abortion;ns varieddepending onmissingdata.

Sample Sizes

AB (N= 324)Others (N= 2,201)

Primary Outcome

Global self esteem;abbreviated CES-D (6items); 1-itemmeasuresof suicidal ideation inpast year; if told by doc-tor she had anxiety/de-pression in past 5 years,1-item life satisfactionmeasure

Key Findings

AB correlated positivelywith CESD (.08),havingsuicidal thoughts (.08),being told by a doctorhad anxiety/depression(.08) & negatively withlife satisfaction (-.06).Also correlated with ex-periencing rape (.06),childhood physical (.15)& sexual (.18) abuse,having a violent partner(.11),& a partner whorefused to use condom(.06).Controlling forrace,education, chil-dren living at home,marital status, and part-ner and violence vari-ables, abortion notsignificantly related toany outcome variable.


Outcome and violencemeasures psychometri-cally weak.Timing ofevents vis-à-vis abortionunknown.Only womenmarried or living as acouple were askedabout partner violence.Limited generalizabilityof study group:havetelephone, youngerteenagers not included,older age (median 40-44),57%married.Lowreported abortion rate(13%) could reflect un-derreporting and/or re-call bias.Only onequestion asked aboutabortion history; repeatabortions not identified.Comparison is withother women,notwomenwith unin-tended Pg.


Commonwealth FundHealth of AmericanWomen Survey


Citation

Coleman,P.,Reardon,D.C.,& Cougle,J.R. (2005).Substanceuse among pregnantwomen in the contectof previous reproduc-tive loss and desire forcurrent pregnancy.British Journal of HealthPsychology,10,255-268

Coleman,P.,Maxey,C.D.,Rue,V.M.,& Coyle,C.T.(2005).Associationsbetween voluntary andinvoluntary forms ofperinatal loss and childmaltreatment amonglow incomemothers.Acta Pediatricia,94,1476-1483.


Data drawn from thepublic release data setthat resulted from theWashington,D.C.Metro-politan Area Drug Study(CD*MADS). The initialsample,constructed tooversample for lowbirth weight,pre-term,and admittedmaternaldrug use,consisted of1,020woman givingbirth inWashington,DCarea hospitals in 1992.The initial sample waspredominantly nevermarried,Black,between19 and 34 years of age,high school or less edu-cation, and of relativelylow family income(under $20,000). Ofthese cases, thosewithknownmedical out-comes of previous preg-nancies were selectedfor further analysis.

Data drawn from Fertil-ity and ContraceptionAmong Low IncomeChild Abusing andNeglectingMothers inBaltimore,MD,1984-1985,a study of familypatterns and contra-ceptive use amongmaltreatingmothers.Sample of 518mothers(Age range 18-50; 79%Black; 6.8% employed)whowere receivingAFDC.All women inter-viewed in home

Sample Sizes

Sample sizes for the sev-eral reportedanalysesdif-fer fromoneanalysis toanother. The key compar-isons reported inTable 3,inwhich theodds ratiosfor druguseduring thecurrentpregnancyas afunctionofprior abortionhistory seems tobebaseduponcomparisonsof 144womenwho re-portednoprior abortionsand282womenwho re-portedoneormoreabor-tionsprior to the indexdelivery. [Thesenumberswerenotdirectly re-ported in thepaperbutweredeterminedthroughanexaminationof thepublic releasedataset used in these analy-ses. Thenumbers are es-sentially consistentwithpercentages andmeth-ods reported in thepaper.]

118 physically abusivemothers and 119 ne-glectingmothers se-lected fromcohortreceivingChild Protec-tive Services (CPS) and281motherswithoutmaltreatment offences.In interview,100women reported 1abortion,59 reported2+ (abortion ave 6.5years earlier),99 re-ported 1miscarriage orstillbirth,34 reported2+ (ave 7.1 yrs earlier).

Primary Outcome

Differential odds ratiosfor the use of mari-juana, cigarettes, alco-hol, crack cocaine,othercocaine,and any illicitdrugs are reported for 1previous abortion vs noabortion history and 2or more abortions vs noabortion history afterstatistical adjustmentfor number of priorbirths,miscarriages, andstill births; age; educa-tion; number of peoplethe respondent liveswith; and a binary indi-cator reflecting if pre-natal care was soughtin the first trimester.

Association betweenself-reported abortionor miscarriage/stillbirthhistory and being in thephysically abusing orneglecting groups.Logistic analyses con-trolled for covariates(single-itemmeasures)associated withmal-treatment (e.g.,morechildren,history of de-pression, worries aboutincome,etc).

Key Findings

Adjusted for covariates,a statistically higherodds ratio was reportedfor the use of legal andillegal substances dur-ing the index pregancyif the woman had aprior history of abor-tion.

Adjusted for covariates,women reporting 1 abor-tionwere notmore likleythan those reporting noabortions to be in childneglect group,butweresigmore likely to be inphysical abuse group.His-tory ofmultiple inducedabortions not related toincreased risk for eitherabuse or neglect.Maternalhistory ofmultiplemiscar-riages and/or stillbirthscompared to nohistorywas associatedwith in-creased risk of physicalabuse andneglect.


The sample very spe-cialized. No indicationthat sampling fractionsused in analysis toreweight sample. Manyof the illegal substancecategories are fairly rare(e.g., there are only 58cases of any reportedcrack cocaine use dur-ing Pg among the sub-set of cases who hadusable data on abortionhistory.) Results lookvery different for covari-ate adjusted analysesand unadjusted analy-ses. No regression di-agnostic results arereported.

Retrospective self-re-ports of abortion in in-terview unreliable.Abortion likely underre-ported. Sample notrepresentative of U.S.women.No info aboutnature of abortion.Single-itemmeasuresof covariates.Causaldirection ambiguous.Same factors (e.g.,poverty;drug use)may contribute toincreased risk of childmaltreatment andabortion.


Washington,DC,Metropolitan Area Drug Study

miscoding of the first pregnancy variable and the dif-ferential exclusion of women having subsequent abor-tions only from the delivery group (see Table 2 fordetails).

In an effort to redress these problems, Schmiege andRusso (2005) reexamined depression risk in the NLSY.Using codes provided by the NLSY staff, they identi-fied a sample of 1744 women as having an unwantedfirst pregnancy. (They, too, had a coding error in theirinitial article, but it did not affect the pattern and sig-

nificance of their findings when corrected. After a se-ries of interchanges in which they addressed criticismsof their approach, we report here the findings based onthe corrected codes verified by the NLSY staff andpublished with the analyses.) First, Schmiege andRusso found that the sampling strategy that Reardonand Cougle (2002a) and Cougle et al. (2003) had usedto control for prepregnancy psychological state (whichwas to include only those women who had completedthe Rotter I-E scale in 1979 prior to their first preg-nancy) resulted in excluding from their sample the


Citation

Coleman P.K., ReardonD.C.,Rue,V.M.,& Cougle,J. (2002).A history of in-duced abortion in rela-tion to substance useduring subsequentpregnancies carried toterm.American Journalof Obstetrics and Gyne-cology, 187,1673-1678.


Data drawn from theNational Pregnancy andHealth Survey con-ducted in 1992 whosepurpose was to assessdrug and alcohol con-sumption in a nationalsample of pregnantwomen (N= 2,613).Hospitals with < 200annual births were se-lected in the first stageof sample selection; in-dividual mothers withinhospitals were ran-domly selected in thesecond stage.Soonafter delivery womenwere interviewed aboutreproductive historyand completed a druguse questionnaire an-swer sheet in responseto interviewer ques-tions. Samples used inanalyses were limited towomenwho recentlyhad given birth, andhad one previous in-duced abortion,oneprevious birth or noprevious births or abor-tions.

Sample Sizes

The primary sampleof women with a re-cent delivery (N= 607)has two subgroups:74 women with oneprevious inducedabortion and 531women with oneprevious birth.Thesecondary sampleincluded 738 first-time mothers with noprevious abortions.Both groups were pri-marilyWhite,married,and employed full-time.The average ageof the two groupsrespectively was 26.5and 23.4 years.

Controls/Covariates

Association betweenprevious reproductiveoutcome and usage ofalcohol or illicit drugsduringmost recentpregnancy.Differentialodds rates for use ofany illicit drugs,mari-juana, cigarettes and al-cohol reported for 1previous abortion vs.1previous birth group,and 1 previous abortionvs. first birth group. Ad-justed for covariates bystratifying covariates re-lated to substance usetype and running sepa-rate analyses.

Primary Outcome

Womenwith a previousabortion had higherrates of any illicit druguse,marijuana use andalcohol use thanwomenwith a previouslive birth.Differencesbetween reproductivehistory groups ap-peared greater whentime since previouspregnancy was longer(3-5 vs.< 2 years).Theabortion group also re-ported higher rates ofillicit drug use,mari-juana, and alcohol usethan first-timemothers.

Results

Samples analyzed notrepresentative of totalNPHS sample or of U.S.women giving birth.Retrospective self-re-ports of abortionmaybe unreliable.Abortionlikely underreported.Single-item outcomemeasures. No statisticaladjustment for numberof significance tests.Confounds not con-trolled. Small size ofabortion group led tomany cell counts <5 insubgroup analyseswhichwere intended tocontrol for confounds.Differences found couldbe due to other unmea-sured factors such aswhether pregnancy in-tended, domestic vio-lence or sexual abuse.Comparisons betweenprevious abortion andprevious birth groupscould be explained bychild-care demands onmothers or differentialstress of first versus latercompleted pregnancy.


Washington,D.C.Metropolitan Area Drug Study

women who had the highest risk for depression—thosewho had delivered at a younger age. Significantly morewomen who had delivered pre-1980 exceeded theCESD cutoff score (33.5%) than who had an abortionpre-1980 (26.5%). Like Cougle et al. (2003), they con-trolled for age of first pregnancy, race, education, andfamily income. However, instead of excluding womenbased on previous marriage, they considered it more

appropriate to maximize generalizability by controllingfor marital status. When Schmiege and Russo analyzedthe full sample (not restricted on the basis of I-Escores), they found no significant differences in depres-sion between the abortion and delivery groups whenrace, age at first pregnancy, 1992 marital status,education, and family income were controlled: 28.3%of women in the delivery group exceeded the CESD


Table 1B: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsINTERNATIONAL STUDIES–NEWZEALAND

Christchurch Health andDevelopment StudyGeneral Description: The Christchurch Health and Development Study a longitudinal study of a cohort of 1,265 children born in 1977 in theChristchurch,New Zealand, urban region who were studied from birth to age 25, including 630 females. Information was obtained on: (a) the history ofPG/abortion for female participants over the interval from 15-25 years; (b) measures of DSM-IV mental disorders (including major depression, overanxiousdisorder, generalised anxiety disorder, social phobia, & simple phobia), and suicidal behaviour for intervals 15-18, 18-21 and 21-25 years; and (c) childhood,family and related confounding factors.

Limitations Common to All Studies Based on this Data Set: Common to All Studies Based on this Data Set:Neither intendedness nor wanted-ness of Pg controlled; in New Zealand to obtain a legal abortion, a woman is referred to two specialist consultants by her doctor; the consultants must agreethat either (1) the Pg would seriously harm the life or the physical or mental health of the woman or baby; (2) the Pg is the result of incest; or (3) the womanis severely mentally handicapped.An abortion will also be considered on the basis of age or when the Pg is the result of rape.Comparisons with populationdata suggest abortion is underreported.Measures of child abuse psychometrically weak and it is likely underreported.

New Zealand

Fergusson D.M.,Hor-wood, L.J.,& Ridder, E.M.(2006).Abortion inyoungwomen andsubsequentmentalhealth. Journal of ChildPsychology & Psychiatry,47, 16-24.


Forty-one percent ofwomen Pg on at leastone occasion prior toage 25;14.6% have atleast one abortion Sam-ple sizes in analysesranged from 506 and520 depending on thetiming of assessment.Ns for prospectiveanalyses were providedin personal communi-cation from the author.

Sample Sizes

Concurrent analyses:AB N= 74DEL N= 131Never Pg N= 301

Prospective analysis:AB N= 48DEL N= 77Never Pg N= 367

Primary Outcome

In concurrent analyses,yes/no diagnosis ofmajor depression,anxi-ety disorder, alcoholand illicit drug depend-ence, suicidal ideationin previous 12mo., andtotal # of disorders. Inprospective analysis,total number of disor-ders from 21-25 yrs.

Key Findings

In concurrent analyses,controlling for covari-ates, AB grp had sig(p<0.05) higher rates ofdepression, suicidalideation, illicit drug de-pendence, & total men-tal health problemsthan the DEL grp & ex-cept for alcohol andanxiety disorder, signifi-cantly higher rates ofdisorder than the NeverPg grp. A prospectiveanalysis used Pg/abor-tion history prior to age21 to predict mentalhealth outcomes from21-25 years. Similarly,after covariate adjust-ment, the AB grp had asig.higher total # of dis-orders than the othergrps,which did not sigdiffer from each other.


Although a longitudinalstudy,most results re-ported involved theconcurrent assessmentof Pg status andmentalhealth.The oneprospective analysiswas limited to numberof disorders owing tothe relatively sparsedata for specific disor-ders over the interval21-25 years and thesmaller number ofwomenwho becamepregnant by age 21.

cutoff score compared to 25% of the abortion group, anonsignificant difference.

They also examined the implications of the practiceof differentially excluding all women who had subse-

quent abortions from only the delivery group (but notfrom the abortion group) by comparing abortion anddelivery groups with women having subsequent abor-tions excluded from both groups. Using this approach,significantly more women in the delivery group


New Zealand

Fergusson,D.M.,Boden,J.M.,& Harwood,L.J.(2007).Abortion amongyoungwomen and sub-sequent life outcomes.Perspectives on Sexualand ReproductiveHealth,39, 6-12.


492 women for whomfull information on Pghistory,education, in-come, welfare depend-ence, employment andpartnership variables toage 25 was availableclassified in 3 groups:abortion before age 21(AB); Pg but no abortionage 21 (DEL)(77); andnever Pg before age 21(Never Pg).; 125 hadhad at least one Pgby age 21;of 172 Pgreported,55% endedwith live birth,31%by abortion,& 14% inmiscarriage.

Two sets of analyses:(1) one based on 1st Pgoutcomes,AB vs DEL;(2) Pg-no abortion vs.Pg with abortion as cor-related dichotomouspredictor variables totake into account possi-ble overlap betweenabortion and Pgwith-out abortion.

Sample Sizes

AB N= 48DEL N= 77Never Pg N= 367

Primary Outcome

Social and economicoutcomes at ages21–25:4 educationalvariables; family in-come, welfare depend-ence, employment,partner violence (itemsfrom the Conflict TacticsScale), relationshipquality (items fromIntimate RelationsScale) & relationshipsatisfaction.

Key Findings

AB grp sigmore likelythan DEL grp to have at-tended university,gained a university de-gree, & gained a tertiaryqualification other thana university degree,&less likely to have beenwelfare-dependent.Also had sig highermean personal income& experienced sig. lowermean level of partnerviolence.AB grp not sigdifferent fromNever Pggroup on all educationoutcomes,mean familyincome,and both part-nershipmeasures.Women in the DEL grphad sig lower intelli-gence scores and levelsof educational achieve-ment in childhood &weremore likely todrop out of school.

Most differences ex-plained by pre-Pg family,social and educationalcharacteristics,exceptAB grp continued tohave sig higher levels ofsubsequent educationalachievement thanDELgrp.For all outcomes,DEL grp fared sig lesswell thanNever Pg grp.The pattern of resultswas similar across thetwo forms of analysis.


Comparisons based onrelatively small num-bers of women.

Table 1B: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsINTERNATIONAL STUDIES–NEWZEALAND (continued)

Notes:AB = Abortion group;DEL = delivery group;Pg = pregnancy


Table 2: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsU.S.STUDIES

National Longitudinal Survey ofYouth (NLSY)General Description: The National Longitudinal Survey of Youth (NLSY) is based on annual interviews with a stratified,multi-stage national probabil-ity sample of noninstitutionalized civilian men and women aged 14-21 as of 1979,with oversampling of Blacks,Hispanics, and poorWhites. Relevantmeasures include: an abbreviated version of the Rotter internal-external locus of control scale (IE, Rotter, 1966; assessed in 1979); global self esteem (RSE,Rosenberg, 1979; assessed in 1980 & 1987); Center for Epidemiological Studies-Depression Scale (CES-D, Radloff, 1977; assessed in 1992); reproductive his-tories were first taken in 1982 and updated every 2 years subsequently.

Limitations Common to All Studies Based on this Data Set: No study used sampling weights so that normative statements are inappropri-ate and alpha levels are likely to be elevated, increasing probability of identifying difference due to chance as a reliable difference.Underreporting ofabortion raises question of possible reporting bias, but direction of reporting bias unclear as womenmay be less likely to report stigmatized experiences(having an abortion,mental problems, experiencing violence), but those who are willing to report one stigmatized condition may be more willing to re-port others, increasing the likelihood of finding a correlation between 2 stigmatized events.Ns of analyses vary depending on covariates and are not al-ways clear. Large sample sizes mean that small effects are statistically significant. CES-D controversial due to cutoff at >15 yielding high rate of falsepositives and lack of specificity of measurement.Generalization limited to restricted age range (women 14-24 in 1979).

Citation

Russo,N.F.,& Zierk,K.L.(1992).Abortion, child-bearing, and women’swell-being.ProfessionalPsychology: Researchand Practice,23,269-280.

Russo,N.F.,& Dabul,A.J.(1997) The relationshipof abortion to well-being: Do race and reli-gionmake a difference?Professional Psychology:Research and Practice,28,23-31.


1.5,295 women forwhom there were NLSYinterviews involving theassessment of well-being in 1987; 773 hadat least one abortion;233 had repeat abor-tions.2.Additional analysesbased on 4502 womenwho had no abortionsbefore their 1980 inter-view.

1.4913 women drawnfrom the sample of5,295 women describedabove (3572White &1341 Black); 721 had aleast one abortion,175had repeat abortions.2.Additional analysesbased on 4336 women(3,147White & 1,189Black) who had noabortions prior to 1980interview.

Sample Sizes

1.AB N= 733Other N= 45622.AB N= 317Other N= 4185

1.AB N= 721Other N= 41922.AB N= 317Other N= 4502

Primary Outcome



Results

MRSE= 33.2 & 33.3 for allwomen vs.womenhavingat least 1 abortion;1.Womenwhohad 1 abor-tion hadhigher RSE thanno abortion ormultipleabortion groups;whenchildbearing and resourcevariableswere controlled,neither having 1 abortionnor having repeat abor-tionswere significantly re-lated to RSE.Totalabortions correlatedwithtotal unwantedbirths(r=.11).2. in subsample 1980 RSEwas the strongest predic-tor of 1987 SE (partialr=.38).-

Primary findingsdidnotvary across groups knownto vary inunderreporting.1.When childbearing andresource variableswerecontrolled,neither having1abortionnor having repeatabortions significantly re-lated toRSE,regardless ofraceor religion.2.1980 SEwas thestrongest predictor of 1987SE (partial r=.39-42) regard-less of raceor religion.


No clinical cut off scorefor RSE & clinical signifi-cance of scores is un-known; large samplemeans small effects sta-tistically significant.Agerange of sample limitedto women 22- 33 in1987.

Religionmeasured in1979 only; highly com-mitted fundamentalistwomen not identified;sample does not in-clude Asians or NativeAmericans.Age rangeof sample limited towomen 22-33 in 1987.


Table 2: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsU.S.STUDIES (continued)

Citation

Reardon,D.C.,& Cougle,J.R. (2002).Depressionand unintended preg-nancy in the NationalLongitudinal Survey ofYouth: A cohort study.BritishMedical Journal,324,151-152.

Sample &Procedure

Two samples weredrawn due to coding is-sues in the initial study;both the initial and cor-rected sample ns are re-ported here.1. Initial sample:421women identified as re-porting a first unin-tended Pg between1980 and 1992 that re-sulted in abortion(N=293) or deliverywith no subsequenthistory of abortion inthe delivery grp(N=128).2.Corrected sample:1076 women identifiedas reporting a first unin-tended Pg between1980 and 1992 that re-sulted in abortion(N=293) or deliverywith no subsequenthistory of abortion inthe delivery grp(N=783). Results weresimilar in both samples& only results fromcorrected sample arepresented here.

Sample Sizes

1.AB N=293DEL N=1282.AB N=293DEL N=783

Primary Outcome

Percent of womenexceeding the 1992CES-D cut-off score(>15).

Key Findings

ABgrp hadhigher%scoring>15 onCES-D in1992 (27%vs.25%),con-trolling for family in-come, education,race,age at 1st Pg,and 1979 I-E score .Sig higher riskfor ABgrp amongmar-riedwomen (26%vs.19%),but not amongunmarriedwomen (29%vs.36%), controlling forfamily income,educa-tion, race,age at 1st Pg,and 1979 I-E score.


Note: Differs from RSEstudies in focusing onoutcome of 1st Pg.Sub-sequent reanalysis bySchmiege & Russo(2005) showed thatfindings in correctedsample still based onmiscoded data.Ex-cludedwomenwithsubsequent history ofabortion only from thedelivery grp.Used I-Escore as a control forpre-existingmentalhealth but scale is not ameasure of mentalhealth.Generalizing toall 1st Pg is inappropri-ate - sample restrictedto only womenwhohad completed the Rot-ter I-E scale in 1979,ef-fectively eliminatingmost (339 of 425) of theteenagers who had de-livered; women in thepre-1980 DEL grp thatwas eliminated had thehighest % exceedingCES-D cut-off (34%)compared to pre-1980AB (27%) and post-1980 AB (24%) & DEL(24%) grps.Variableused to define race in-cluded nonBlack andnonHispanic minoritiesin theWhite category.Age range of samplelimited to women 27-38in 1992.

(28.1%) than the abortion group (20.7%) exceededthe CESD cutoff score (p. <01). These analysesillustrate that the sampling and exclusion strategiesresearchers use to analyze secondary data sets can dra-matically alter the conclusions reached regarding therelative risks for depression accompanying childbirthversus abortion. When attempting to examine the ef-fects of first pregnancy outcome, it is important tocontrol for both number of subsequent abortions andnumber of subsequent births in both groups.

Substance use. Reardon et al. (2004) used NLSY datato examine substance abuse among 535 women whohad terminated a first unintended pregnancy com-pared with 213 women who had delivered a first unin-tended pregnancy and 1144 women who had neverbeen pregnant. These researchers again excludedwomen pregnant before 1980 (i.e., those known to beat a significantly higher risk for depression than otherwomen in the sample and more likely to be found inthe delivery group; Schmiege & Russo, 2005). They



Citation

Cougle, J.R., Reardon,D.C.,& Coleman, P.K.(2003).Depression as-sociated with abortionand childbirth:A long-term analysis of theNLSY cohort.MedicalScienceMonitor,9,CR105-112.


Based on a larger sub-sample of 1,884 womenwith first abortion orfirst delivery with nosubsequent abortionsbetween 1980 and1992 andwho hadcompleted both the1979 Rotter I-E scaleand the 1992 CES-Dscale; total AB & DELgrp ns not reported,butreports an average agefigure based on 884women (AB = 293;DEL = 591); subsamplens reported as varyingfrom 1031-1361depending on theanalyses.

Sample Sizes

AB N= 131 - 164DEL N= 877 - 1197

Primary Outcome

Percent of womenexceeding the 1992CES-D cutoff score(>15).

Results

Final corrected table:ABgrp hadhigher% scoring>15 onCES-D in 1992,controlling for age,race,education, income,andabbreviated I-E score.Higher depression riskfound for ABgroupamongwomenwhowereWhite,married,andwhodid not have a firstmar-riage ending in divorce,controlling for relevant co-variates. Sig differencesnot found amongBlack/Hispanicwomen,unmarriedwomen,orwomenwith a firstmar-riage ending in divorce,controlling for relevantcovariates.


This study is similarlydesigned and based onthe women erroneouslyidentified in first set ofanalyses in Reardon &Cougle (2002) , exceptthat womenwho hadintended pregnanciesare now added to DELgroup, reducing% ex-ceeding cut-off score.Reasons for discrepan-cies in AB & DEL groupsfrom previous study notclear,possibly due todifferent covariates(age vs.age at 1st Pg)used in the two studiesfor unknown reasons.Average age figurebased on 884 womenso not clear how ns inthe regression analysesdetermined,given theyexceed that numberand age is a covariate inthose analyses.Variableused to define race in-cluded non-Black andnon-Hispanic minoritiesin theWhite category.Age range of samplelimited to women 27-38in 1992.

also excluded women who had subsequent abortionsfrom only the delivery group. In this subsample, con-trolling for prepregnancy I-E and RSE, age, race, mari-tal status, income, and education, few significant

differences were found between groups in reportedsubstance use. The exceptions were that women in theabortion group reported drinking on more days in thelast month than the delivery group (6.4 vs. 4.8), but



Citation

Schmiege,S.,& Russo,N.F. (2005).Depressionand unwanted firstpregnancy:Longitudi-nal cohort study.BritishMedical Journal,331,1303-1305.

Sample &Procedure

Two samples weredrawn due to coding is-sues in the initial study;both the initial and cor-rected sample ns are re-ported here.1. Initial sample:1247women identified as re-porting a first un-wanted Pg between1970 and 1992 that re-sulted in abortion(N=479) or delivery(N=768).2.Corrected sample:1744 women identifiedas reporting a first un-wanted Pg between1970 & 1992 that re-sulted in abortion(N=461) or delivery(N=1283).Results weresimilar in both samplesand only results fromcorrected sample arepresented here.

Sample Sizes

1.AB N=479DEL N=7682.AB N=461DEL N=1283

Primary Outcome

Both%women exceed-ing the 1992 CES-Dcutoff score (>15) andcontinuous 1992 CES-Dscores reported.Education, income,andfamily size examinedas outcomes.

Key Findings

Percentage exceedingcutoff score on 1992CES-D did not sig differfor AB vs.DEL groups,controlling for age at1st Pg, race,education,marital status, and fam-ily income, in full sam-ple (25% vs.28%) orpost-1979 subsample(23% vs.23%).AB sig as-sociated with lower ed-ucation and incomeand larger family size.Additional analysespublished in responseto debates over pointsof design did notchange the pattern ofresults,with only sig dif-ference found betweenAB & DEL grps in unad-justed analyses whensubsequent abortionsexcluded from bothgroups; risk was lowerin the AB grp (AB = 21%>15 vs.DEL = 28%>15); the difference wasnot sig when covariatescontrolled.Patterns offindings similar acrossgroups known to varyin underreporting.Womenwho refused tofill out the confidentialabortion card had siglower CES-D scoresthan womenwho com-pleted the card (13% vs.25% ).


Note:NLSY staff pro-vided coding to ensureproper identification ofsample,but last line ofcode inadvertentlyomitted in initial analy-ses, subsequently cor-rected. Differs fromother studies in focus-ing on unwanted firstPg.Study criticized fornot controlling samevariables as previousstudies, resulting inpublication of a seriesof analyses, includingthose limited to post-1980 AB & DEL grps.Al-though underreportingbias a concern, the pat-tern of findings did notdiffer among grpsknown to vary in under-reporting. However,lower CES-D scoresamongwomenwho re-fused to fill out the con-fidential abortion cardsuggests that depres-sionmight be overesti-mated in the abortiongroup.Age range ofsample limited towomen aged 27-38years in 1992.

not on more days than the never pregnant group(5.9%). They were also more likely to report usingmarijuana in the last month (18.6%) than did womenin the delivery (7.9%) or never pregnant (7.9%)groups. These researchers did not control for historyof drug use prior to the first pregnancy in their analy-ses despite the availability of this information in thedata set and despite published findings in the literaturethat linked such drug abuse to later reproductive out-comes including likelihood of having an abortion(Mensch & Kandel, 1992; Rosenbaum & Kandel,1990).

Evaluation of NLSY studies. Conclusions drawn from theNLSY about the mental health effects associated withabortion vary markedly by analytical strategy. Al-though the design of NLSY is longitudinal, like all sur-vey data, it is correlational, making causal claimsinappropriate. Collectively, these studies have a num-

ber of methodological limitations beyond those de-scribed above that make it difficult, if not impossible,to interpret the meaning of the correlations that arereported (see Table 2). Perhaps most importantly,none of these studies adequately controls for preexist-ing mental health or other important co-occurring riskfactors prior to abortion or delivery (the Rotter I-E isnot a measure of prior mental health), making it diffi-cult to interpret the meaning of correlations observedbetween abortion and a mental health outcome. Co-variates included in analyses varied across studies forunspecified reasons. Likewise, some contextual vari-ables, such as marital status, that were shown in somestudies to moderate results were not examined asmoderators in other studies, compounding difficultiesof comparing across studies. Further, some variablesthat were present in the NLSY and known to be re-lated to the outcome variable under consideration(e.g., prior substance abuse) were omitted as covari-



Citation

Reardon,D.,Coleman,P.K.,& Cougle, J.R. (2004).Substance use associ-ated with unintendedpregnancy outcomes inthe National Longitudi-nal Study of Youth.American Journal ofDrug andAlcohol Abuse,30,369-383.


After excluding allwomen Pg before 1980,identified 1748 womenreporting a first unin-tended Pg between1980 and 1988 that re-sulted in abortion(N=213) or delivery(N=535),or had neverbeen Pg (N=1144); asubsample of womenresponded to alcoholquestions; alcoholanalyses appear to bebased on 1243 women.

Sample Sizes

AB N=213DEL N=535Never Pg N=1144

Primary Outcome

Eleven yes/no itemsrelated to alcohol abusesymptoms;4 related tosubstance use (# daysdrank in last mo;# drinks consumedon days when drank;if ever usedmarijuanaor cocaine in last mo).

Results

Controlling for age,race,marital status, income,ed-ucation, pre-Pg RSE andpre-Pg Rotter I-E score,nosig differences amonggroups on # of drinks; in%scoring 2 ormore,or%scoring 4 ormore onitems related to alcoholabuse; in the number ofdrinks consumed,or in theuse of cocaine.AB grpdrank sigmore days in lastmo (6.36) thanDEL grp(4.79) but not thanNeverPggrp (5.93);andweremore likely to usemari-juana in lastmonth(18.6%) than theDEL orNever Pggrps (7.9%).


Exclusion of women Pgbefore 1980makessample unrepresenta-tive and generalizationto unintended first Pginappropriate as notedabove.The large num-ber of tests performed,single-itemmeasures ofkey dependent vari-ables, and small magni-tude of effects limitconclusions that can bedrawn from this study.Drinking on an averageof 6.36 (AB) vs.4.79(DEL) days per mo.notindicator of clinicallysignificant alcoholabuse.Variable used todefine race includednon-Black and non-His-panic minorities in theWhite category.

ates in analyses of that outcome variable. Analyseswere often based on small subgroups or subgroups forwhich no sample size was provided. On the otherhand, the overall large sample sizes used for someanalyses mean that small effects that are statisticallysignificant may have little clinical significance.

Although initially based on a national probability sam-ple, the ability to assess prevalence of mental healthproblems among women who have abortions from thisdata set is limited because (1) abortion has been under-reported in the NLSY compared with national norms;

(2) sample weights, required to construct populationestimates from the data, were not used in the analysesof any of the studies; and (3) the measurement of men-tal health outcomes was limited to self-esteem, depres-sion risk, and substance abuse. No actual measures ofpsychopathology were included.

The potentially strongest designs focused on mentalhealth outcomes associated with unintended first preg-nancy. However, the practices of excluding womenwho became pregnant at a young age (before 1979 or1980) and differentially excluding women having



National Longitudinal Study of Adolescent Health (ADD-HEALTH)General Description: ADD-HEALTH uses a multi-stage, school-based, longitudinal design which collected data in three waves: initial (1994-1995), andapproximately 1 year (1996), and 6 years (2001-2002) later. AtWave I all participants were in grades 7-12.All Wave I (N= 90,118) completed an in-schoolquestionnaire; a subsample (N=12,105) completed an additional computer-assisted in-home interview that included questions about sexual history andreligion.This subsample was chosen by identifying a group of students who were representative of the adolescent population in grades 7-12 during the1994-1995 school year; in addition, adolescents who were disabled,African American students from well-educated families, Chinese, Cuban, Puerto Rican,living with twin, living with a full sibling, living with a half sibling, living with a nonrelated adolescent, and siblings of twins were oversampled.

Limitations Common to All Studies Based on this Data Set: School-based population does not include students who drop out due to Pg;ethnic minorities in sample may be particularly unrepresentative of ethnic minorities in the adolescent population as a whole. 1-itemmeasures psycho-metrically weak.

Citation

Coleman,P.K. (2006).Resolution of unwantedpregnancy during ado-lescence through abor-tion versus childbirth:Individual and familypredictors and conse-quences. Journal ofYouth andAdolescence,35,903-911.

Sample &Procedure

One hundred and thirtyadolescents in grades7-11 who completedbothWaves I & II andexperienced a Pg de-scribed as“not wanted”or“probably notwanted”.

Sample Sizes

AB N= 65DEL N=65

Primary Outcome

Single-itemmeasuresof counseling,12month trouble sleeping,30 day cigarette use,30 daymarijuana use,12month alcohol use,problems with parentsandwith school due toalcohol use.

Key Findings

Controlling for risk tak-ing and desire to leavehome,AB groupmorelikely to have counsel-ing, trouble sleeping,and usemarijuana inpast 30 days (problemswith parents due to al-cohol use approachedsignificance).


Number of total Pgs un-known, but small nsraise questions aboutunderreporting anddrop-out rates.1-itemoutcomemeasures psy-chometrically weak.Percentages and ns foroutcome variables notreported so frequencyof problem unknown;previousmental healthproblems not con-trolled. Given the largenumber of variables inthe data set,why theseparticular variableswere included is un-clear. Not clear whencounseling occurred.

abortions subsequent to first pregnancy from the de-livery group but not the abortion group were shownto bias results toward overestimating adverse effects ofabortion in this data set. In the one study focusing onfirst pregnancy that did not use differential exclusionand was based on codes provided by NLSY staff, theproportion of women who met or exceeded the CESDcutoff scores did not significantly differ between abor-tion (25%) and delivery (28.3%) groups (Schmiege &Russo, 2005).

Washington, DC, Metropolitan Area Drug Study.Coleman, Reardon, and Cougle (2005) used this publicrelease data set to examine substance use during preg-nancy as a function of reported reproductive history.

The initial sample, which consisted of 1,020 womeninterviewed after giving birth in Washington, DC, areahospitals in 1992, was predominantly never married,Black, of low socioeconomic status, and oversampledfor low birth weight and preterm infants, and self-re-ported drug use. Of these cases, Coleman et al. (2005)selected those who in their interview reported no abor-tions, one abortion, or multiple abortions prior to theirrecent pregnancy and examined their reported drug useduring their recent pregnancy (see Table 2). Adjustedfor age, income, and number of people living in thehouse, a statistically higher odds ratio was reported forthe use of legal and illegal substances during the indexpregnancy if the woman had reported one prior abor-tion compared with no abortions, but not if she had



Citation

Hope,T.L.,Wilder, E.I.,&Watt,T.T. (2003).The re-lationships among ado-lescent pregnancy,pregnancy resolution,and juvenile delin-quency. SociologicalQuarterly,44,555-576.


ADD-HEALTH data fromWaves I & II used to ex-amine the relationshipsamong adolescent Pg,Pg resolution,anddelinquent behavior.Womenwho experi-enced Pg prior toWaveI,miscarried,or werestill Pg atWave II ex-cluded; 360 ever Pgadolescents who hadan abortion or keptbaby and did notchoose adoption wereidentified.Longitudinalanalysis based on 156womenwho becamePg betweenWaves I & IIreported here.Al-though adoption grphad sig higher delin-quency rate than Keptbaby group, the small n( 4),precluded inclusionin longitudinal analyses.

Sample Sizes

Longitudinal analysis:AB N=87Kept baby N=69

Primary Outcome

Comparing AB vs.Keptbaby groups,3.6% vs.15.0% onwelfare;39.9% vs.23.9% in in-tact families.These vari-ables not controlled.Most relevant here: lon-gitudinal analyses of re-lationship between Pgoutcome & cigarettesmoking or marijuanause on at least 1 day inthe past 30 days.

Results

ABgrp reported higherrates of cigarette smokingandmarijuana use thanthosewho kept babyboth prior to their Pg(Wave 1) and subsequentto their Pg (Wave II).Keep-ing baby associatedwith adecrease in cigarette andmarijuana use after Pg;nosig change in such usewas foundbefore vs.afterABgrp.


Number of total Pgs un-known, but small nsraise questions aboutunderreporting anddrop-out rates that mayadvantage Kept babygroup; measures psy-chometrically weak andof unknown clinical sig-nificance. Percentagesand ns for outcomevariables not reported,so bases for % of prob-lems in various grps un-clear. The extent towhich delinquentmothers may havehigher drop out ratesthan other mothers isunknown.Althoughadoption grp not ana-lyzed due to low n, thesig higher overall rateof delinquency for thatgrp emphasizes impor-tance of recognizingheterogeneity inwomenwho deliver.

reported multiple abortions compared with no abor-tions (with the exception of use of cigarettes duringpregnancy). Notably, these analyses did not control forhistory of drug use prior to the pregnancy. They alsodid not control for the wantedness of the pregnancy,although those data were available in the data set. Be-cause this study is based on a specialized sample, esti-mates of mental health problems among women in the

United States who have an abortion cannot be deter-mined from this study.

National Pregnancy and Health Survey. Coleman,Reardon, Rue, and Cougle (2002a) used data fromthis survey conducted in 1992 to examine the associa-tion between retrospective reports of a previous abor-tion and use of alcohol, cigarettes, or illicit drugs



National Survey of Family Growth (NSFG)General Description: The NSFG Cycle V sample is a subsample of 10,847 women aged 15-44 drawn from the larger national probability sample of theNational Health Interview Survey.The NSFG is thus based on a complex stratified,multistage design that requires using sampling weights in computingstatistics.

Limitations Common to All Studies Based on this Data Set: Retrospective self-report data that may involve recall of precise timing of keyvariables (e.g., abortion, onset of anxiety symptoms), occurring decades previously.

Citation

Cougle, J., Reardon,D.C.,Coleman,P.K.,& Rue,V.M. (2005).Generalizedanxiety associated withunintended pregnancy:A cohort study of the1995 National Survey ofFamily Growth. Journalof Anxiety Disorders,19,137-142.

Sample &Procedure

Study sample:1. All women having anunintended Pg endingin abortion for their firstPg event.2.All women having anunintended Pg endingin live birth delivery fortheir first Pg event whohad no abortions afterthat Pg.Womenwhoexperienced a pro-longed period of anxi-ety previous to or at thesame age as the Pgevent were excludedfrom the sample.

Sample Sizes

AB N=1033DEL N=813

Primary Outcome

Dichotomousmeasure(yes/no) of anxietysymptoms.

Key Findings

Sig higher rate of anxi-ety symptoms in AB vs.DEL group (13.7% vs.10.1%), controlling forrace and age at inter-view. In stratified sub-analyses, differencesig for unmarried orwomen under 20 at 1stPg,but not for marriedwomen.


Women reporting pre-Pg anxiety excluded socannot generalize to allfirst unintended preg-nancies; misleading lan-guage impliesgeneralized anxiety dis-order (GAD) is assessed,but items used to con-struct anxiety variableare not congruent withDSM definitions of gen-eralized anxiety disor-der, making clinicalimplications problem-atic; differential exclu-sion of womenwithsubsequent abortionsfromDEL but not ABgrp; sampling weightsnot used in statisticalanalyses; stratificationused rather than con-trolling for relevant vari-ables. No attempt tocontrol for any violencehistory although ques-tions re rape experi-ence available in dataset.

during the most recent pregnancy. The initial sampleconsisted of 2,613 women who participated shortlyafter giving birth in hospitals within the UnitedStates. The women wrote down answers in responseto interviewer questions; responses were concealedfrom the interviewer. Samples selected for analysiswere limited to three groups who had recently givenbirth: women with one previous pregnancy resultingin an induced abortion (n = 74), women with one pre-vious pregnancy resulting in live birth (n = 531), andwomen with no previous pregnancies (n = 738). Themajority of the women were White, married, and em-ployed full-time. Dichotomous measures of drug andalcohol use during most recent pregnancy were usedas outcome variables. Analyses revealed that women

who reported a previous abortion also reported higherrates of any illicit drug use, marijuana use, and alco-hol use than did women who had one previous livebirth or were first-time mothers. The researchers ad-justed for sociodemographic covariates by stratifyingthose related to substance use outcomes and conduct-ing separate analyses for each level of these vari-ables. Although these analyses identified somedifferences in the relationship of reproductive historyto alcohol and drug use for different levels of maritalstatus, income, and other demographic variables, find-ings are suspect because of the small number of partic-ipants in the abortion group and the failure to correctfor the relatively large number of significancetests. Other limitations include the absence of controls



Commonwealth FundHealth of AmericanWomen Survey

Citation

Russo N.F.,& Denious J.(2001).Violence in thelives of women havingabortions: Implicationfor practice and publicpolicy.Professional Psy-chology: Research andPractice,32,142-150.


Secondary analyses of arandom householdtelephone survey ofover 2,500 women and1,000men aged 18 orover and residing in thecontinental U.S., con-ducted in 1993.Analy-ses based on responsesof 2,525 women,324of them identified ashaving had at least1 abortion;ns varieddepending onmissingdata.

Sample Sizes

AB N=324Others N=2,201

Primary Outcome

Global self-esteem(RSE); abbreviated CES-D (6 items); 1-itemmeasures of suicidalideation in past year; iftold by doctor she hadanxiety/ depression inpast 5 years, 1-item lifesatisfactionmeasure.

Results

AB correlated positivelywith CES-D (.08),havingsuicidal thoughts (.08),being told by a doctorhad anxiety/depression(.08) & negativelywith lifesatisfaction (-.06).Also cor-relatedwith experiencingrape (.06), childhoodphysical (.15) & sexual (.18)abuse,having a violentpartner (.11) & a partnerwho refused to use con-dom (.06).Controlling forrace,education,childrenliving at home,marital sta-tus, andpartner and vio-lence variables,abortionnot sig related to any out-come variable.


Outcome and violencemeasures psychometri-cally weak.Timing ofevents vis-à-vis abor-tion unknown.Abbrevi-ated CES-D used;Onlywomenmarried or liv-ing as a couple wereasked about partner vi-olence. Limited gener-alizability of studygroup:have telephone,younger teenagers notincluded;older age(median 40-44), 57%married.Low reportedabortion rate (13%)could reflect underre-porting and/or recallbias.Only one questionasked about abortionhistory; repeat abor-tions not identified.Comparison is withother women,notwomenwith unin-tended Pg.

for wantedness of the recent pregnancy, history ofdrug use prior to the pregnancy, or previous mentalhealth.

Fertility and Contraception Among Low-IncomeChild Abusing and Neglecting Mothers in Baltimore,MD, 1984-1985 (Baltimore Study). Coleman,Maxey, Rue, and Coyle (2005) analyzed this dataset to examine the association between self-reportedabortion or miscarriage/stillbirth history and child

abuse and/or neglect, as identified by Child Protec-tive Services. The purpose of the original study hadbeen to study family patterns and contraceptive useamong maltreating mothers. Samples of 118 physi-cally abusive mothers, 119 neglecting mothers, and281 mothers without maltreatment offences were se-lected from a sample of 518 mothers who were re-ceiving Aid to Families With Dependent Children(79.9% Black and 93.2% unemployed). In an in-home interview, 159 of these women reported having



Harvard Study ofMoods and Cycles

Citation

Harlow,B.L.,Cohen,L.,Otto,M.W., Spiegelman,D.,& Cramer,D.W.(2004).Early life men-strual characteristicsand pregnancy experi-ences amongwomenwith andwithout majordepression:The Harvardstudy of moods and cy-cles. Journal of AffectiveDisorders,79,167-176.

Sample &Procedure

Subsample drawn froma cross-sectional sam-ple of 4,161 women be-tween 36-45 years ofage residing in sevenBostonmetropolitanarea communities con-sisting of 332 womenwho had a past or cur-rent history of majordepression asmeas-ured by DSM criteriaand 644 womenwithno such history.

Sample Sizes

Comparisons madebetween 332depressed and644 nondepressedwomen.

Primary Outcome

Percentage of womenwho reported experi-encing at least oneabortion for depressed(DEP) and nonde-pressed (NDEP) groups.

Key Findings

Percentage of womenhaving had at least oneabortion 34.1% and24.1%, for DEP & NDEPgrps, respectively;higher % of abortionsin the DEP group re-flected a higher % ofwomen havingmultipleabortions (14.8% vs.6.2%).Controlling forage,age at menarche,educational attainment,andmarital experience,no sig differences be-tween% of womenwith a lifetime historyof dep (19.3%) and nohistory of dep (17.9%)reporting at least oneabortion.Womenwithlifetime history of majordep upon study enroll-ment were 3 timesmore likely to reporthaving hadmultipleabortions before theirfirst onset of depressionthan were nonde-pressed women. Alsofound a strong associa-tion between dep andmarital disruption.


Direct comparisons be-tweenwomen report-ing abortion vs.deliverywere not conducted.Wantedness of Pg notassessed.Associationbetween dep andmari-tal disruption under-scores importance ofcontrolling for maritalstatus when seeking toassess the independentcontribution of abor-tion to depression risk.Retrospective repro-ductive history anddepression onset data.Researchers suggestvariety of unassessedantecedent conditionsmay underlie results,including involvementin abusive relationships.

had at least one abortion, and 133 reported at leastone miscarriage or stillbirth (both occurring on aver-age 6-7 years earlier). Controlling for a large numberof single-item covariates found in preliminary analy-ses to be associated with maltreatment (and that var-ied depending on their association with the outcomevariable, e.g., education was controlled only in theanalyses on physical abuse; employment controlled

only in the analyses on neglect), women reportingone abortion were not more likely than those report-ing no abortions to be in the child neglect group butwere significantly more likely to be in the physicalabuse group. History of multiple induced abortions,however, was not related to increased risk for eitherabuse or neglect. In contrast, maternal history ofmultiple miscarriages and/or stillbirths compared



National Pregnancy andHealth Survey

Citation

Coleman,P.K., Reardon,D.C.,& Cougle, J. (2002).A history of inducedabortion in relation tosubstance use duringsubsequent pregnan-cies carried to term.American Journal of Ob-stetrics and Gynecology,187, 1673-1678.


Data drawn from theNational Pregnancyand Health Survey con-ducted in 1992 whosepurpose was to assessdrug and alcohol con-sumption in a nationalsample of pregnantwomen (N= 2,613).Hospitals with < 200annual births were se-lected in the first stageof sample selection; in-dividual motherswithin hospitals wererandomly selected inthe second stage. Soonafter delivery,womenwere interviewedabout reproductivehistory and completeda drug-use question-naire answer sheet inresponse to inter-viewer questions. Sam-ples used in analyseswere limited to womenwho recently had givenbirth and had one pre-vious induced abor-tion, one previousbirth, or no previousbirths or abortions.

Sample Sizes

The primary sampleof women with a re-cent delivery (n = 607)had two subgroups:74 women with oneprevious inducedabortion and 531women with one pre-vious birth.The sec-ondary sampleincluded 738 first-time mothers with noprevious abortions.Both grps were prima-rilyWhite,married andemployed full-time.Average age of thetwo grps was 26.5 and23.4 yrs, respectively .

Primary Outcome

Association betweenprevious reproductiveoutcome and usage ofalcohol or illicit drugsduringmost recentpregnancy.Differentialodds ratios for use ofany illicit drugs,mari-juana, cigarettes and al-cohol reported for 1previous abortion vs.1previous birth group,and 1 previous abortionvs. first birth group.Ad-justed for covariates bystratifying covariates re-lated to substance usetype and running sepa-rate analyses.

Results

Womenwith a previousabortion hadhigher ratesof any illicit drug use,mari-juana use and alcohol use,thanwomenwith a previ-ous live birth.Differencesbetween reproductivehistory groups appearedgreaterwhen time sinceprevious pregnancywaslonger (3-5 vs.< 2 years).The abortion group alsoreported higher rates of il-licit drug use,marijuana,and alcohol use than first-timemothers.


Samples analyzed notrepresentative of totalNPHS sample or of U.S.women giving birth.Retrospective self-re-ports of abortionmaybe unreliable.Abortionlikely underreported.Single-item outcomemeasures.No statisticaladjustment for numberof significance tests.Confounds not con-trolled. Small size ofabortion group led tomany cell counts <5 insubgroup analyseswhich were intended tocontrol for confounds.Rates of use not re-ported. Differencesfound could be due toother unmeasured fac-tors such as whetherpregnancy intended,partner violence,or sex-ual abuse.Comparisonsbetween previous abor-tion and previous birthgroups could be ex-plained by child caredemands onmothersor differential stress offirst vs. later completedpregnancy.

with no history was associated with increased risk ofboth child physical abuse and neglect. Because thisstudy is based on a highly specialized sample, find-ings cannot be generalized to the population ofwomen in the United States.

Health of American Women Survey. Russo and De-nious (2001) used data from this survey, sponsored bythe Commonwealth Fund, to examine correlationsamong abortion history, violence history, and mentalhealth outcomes. This telephone survey was based ona national sample of men and women 18 years of ageor older, with oversampling of ethnic minorities.Among the 2,525 women surveyed, 324 reported

having had an abortion to the interviewer. Comparedwith other women, a larger percentage of women inthe abortion group reported experiencing suicidalthoughts in the past year and having a doctor givethem a diagnosis of anxiety or depression in the past 5years. Having an abortion was also slightly but signifi-cantly correlated with higher depressive symptomsand lower life satisfaction. When violence history andrelevant demographic and partner variables were con-trolled, however, abortion was no longer significantlyrelated to diagnoses of depression or anxiety, CES-Dscore, or the life satisfaction measure. This study, likethe others of this type, has several limitations. Abor-tion history was assessed through self-report (in this



Washington,DC,Metropolitan Area Drug Study

Citation

Coleman,P., Reardon,D.C.,& Cougle, J.R. (2005).Substance use amongpregnant women inthe context of previousreproductive loss anddesire for current preg-nancy. British Journal ofHealth Psychology,10,255-268.

Sample &Procedure

Data drawn from thepublic release data setof theWashington DCMetropolitan Area DrugStudy (CD*MADS).Theinitial sample, con-structed to oversamplefor low birth weight,pre-term,and admittedmaternal drug use, con-sisted of 1020 womangiving birth inWashing-ton, DC area hospitalsin 1992.The initial sam-ple was predominantlynever married,Black,between 19 and 34years of age,highschool or less educa-tion, and of relativelylow family income(under $20,000). Ofthese cases, those withknownmedical out-comes of previouspregnancies wereselected for furtheranalysis.

Sample Sizes

Sample sizes variedacross analyses. Keycomparisons in Table3, in which odds ratiosfor drug use duringthe current Pg as afunction of abortionhistory, appear basedon 144 women re-porting no prior abor-tions vs. 282 womenreporting one or moreabortions prior to theindex delivery. [These# not directly re-ported in paper butwere determinedthrough examinationof the public releasedata set used in theseanalyses.Numbersessentially consistentwith %d &methodsreported in thepaper.]

Primary Outcome

Differential odds ratiosfor the use of mari-juana, cigarettes, alco-hol, crack cocaine,othercocaine,and any illicitdrugs are reported for 1previous abortion vs.noabortion history and 2or more abortions vs.no abortion historyafter statistical adjust-ment for number ofprior births,miscar-riages, and still births;age; education;numberof people the respon-dent lives with; and abinary indicator reflect-ing if prenatal care wassought in the firsttrimester.

Key Findings

Adjusted for covariates,a statistically higherodds ratio was reportedfor the use of legal andillegal substances dur-ing the index preg-nancy if the womanhad a prior history ofabortion.


The sample very spe-cialized. No indicationthat sampling fractionsused in analysis toreweight sample.Ratesof use not reported forcomparison grps.Manyof the illegal substancecategories are fairly rare(e.g. there are only 58cases of any reportedcrack cocaine use dur-ing Pg among the sub-set of cases who hadusable data on abortionhistory).Results lookvery different for covari-ate-adjusted analysesand unadjusted analy-ses. Intendedness of Pgnot used as co-variatein abortion analyses.

case over the phone), and the rate of reported abor-tions was low compared with national norms, raisingconcerns about biases associated with underreporting.It cannot be determined from this data set whether theabortion took place before or after the violence oc-curred, or whether diagnoses of anxiety or depressionoccurred pre- or post abortion. In addition, samplingweights were not used.

National Survey of Family Growth (NSFG). Cougle etal. (2005) used data from the 1995 NSFG to examinethe association between outcome of first- unintendedpregnancy (abortion vs. delivery) and an occurrenceof “generalized anxiety” lasting more than 6 monthsdefined by a cutoff score). All variables—reproductivehistory, episodes of anxiety, as well as the timing ofthose episodes with respect to pregnancy— were deter-

mined retrospectively via self-reports, raising ques-tions about reliability and underreporting of abortion.As in their earlier studies, women with subsequentabortions were differentially excluded from the deliv-ery group but not the abortion group. Controlling forrace and age at interview, women in the abortiongroup were more likely to be classified as having hadan episode of generalized anxiety postpregnancy thanwomen in the delivery group (13.7% vs. 10.1%).Sample weights were not used, so these percentagescannot be used for normative estimates. Although in-formation on rape history, known to be related toboth unintended pregnancy and anxiety, was in thedata set, it was not controlled. The anxiety items werenot congruent with the DSM definition of generalizedanxiety disorder, raising questions about the clinicalsignificance of the outcome variable.



Fertility andContraceptionAmongLow IncomeChildAbusingandNeglectingMothers inBaltimore,MD,1984-1985

Citation

Coleman,P.,Maxey,C.D.,Rue,V.M.,& Coyle,C.T.(2005).Associationsbetween voluntary andinvoluntary forms ofperinatal loss and childmaltreatment amonglow incomemothers.Acta Pediatricia,94,1476-1483.


Data drawn from Fertil-ity and Contraceptionamong Low IncomeChild Abusing and Ne-glectingMothers in Bal-timore, MD,1984-1985,a study of family pat-terns and contraceptiveuse amongmaltreatingmothers.Sample of 518mothers (age range18-50; 79% Black; 6.8%employed) whowerereceiving AFDC.Allwomen interviewed inhome.

Sample Sizes

One hundred andeighteen physicallyabusive mothers and119 neglecting moth-ers selected fromcohort receiving childprotective services(CPS) and 281 moth-ers without maltreat-ment offences. Ininterview, 100 womenreported 1 abortion,59 reported 2+ (abor-tion ave 6.5 years ear-lier), 99 reported 1miscarriage or still-birth, 34 reported 2+(ave 7.1 yrs earlier).

Primary Outcome

Association betweenself-reported abortionor miscarriage/stillbirthhistory and being in thephysically abusing orneglecting groups.Logistic analyses con-trolled for covariates(single-itemmeasures)associated withmal-treatment (e.g.,morechildren,history of de-pression, worries aboutincome,etc).

Results

Adjusted for covariates,women reporting 1 abor-tionwere notmore likelythan those reporting noabortions to be in childneglect group,butweresigmore likely to be inphysical abuse group.His-tory ofmultiple inducedabortions not related toincreased risk for eitherabuse or neglect.Maternalhistory ofmultiplemiscar-riages and/or stillbirthscompared to nohistorywas associatedwith in-creased risk of physicalabuse andneglect.


Retrospective self-reports of abortion ininterview unreliable.Abortion likely underre-ported. Sample notrepresentative of U.S.women.No info aboutnature of abortion.Sin-gle-itemmeasures ofcovariates.Causal direc-tion ambiguous.Samefactors (e.g.,poverty;drug use) may con-tribute to increased riskof child maltreatmentand abortion. Intend-edness of Pg not as-sessed, and given thepoor health amongthis study population,lack of informationabout whether the pre-vious abortion was fortherapeutic reasons is aparticular limitation.

National Longitudinal Study of Adolescent Health(ADD-Health). Two studies were based on the ADD-Health data set, a longitudinal, nationally representa-tive, school-based survey of adolescents. Coleman(2006a) analyzed data from the ADD-Health to exam-ine the relationship between reproductive history andvarious problems in adolescents. From a much largersample of students who had completed an in-schoolquestionnaire at Wave I (N = 90,118) and a computer-

assisted home interview at Wave II (N =12,105), Cole-man selected adolescents in grades 7 through 11 whohad completed both Wave I and Wave II and who re-ported experiencing a pregnancy they described as“not wanted” or “probably not wanted” that was re-solved through abortion (n = 65) or delivery (n = 65).She then examined the likelihood that adolescentswho reported abortion versus delivery also reportedreceiving counseling for psychological or emotional


Table 2: Secondary Analyses of Survey Data–Abortion vs.Comparison GroupsINTERNATIONAL STUDIES–NEWZEALAND

Christchurch Health andDevelopment StudyGeneral Description: The NSFG Cycle V sample is a subsample of 10,847 women aged 15-44 drawn from the larger national probability sample of theNational Health Interview Survey.The NSFG is thus based on a complex stratified,multistage design that requires using sampling weights in computingstatistics.

Limitations Common to All Studies Based on this Data Set: Retrospective self-report data that may involve recall of precise timing of keyvariables (e.g., abortion, onset of anxiety symptoms), occurring decades previously.

New Zealand

Fergusson D.M.,Hor-wood, L. J.,& Ridder, E.M. (2006).Abortion inyoungwomen andsubsequentmentalhealth. Journal of ChildPsychology & Psychiatry,47,16-24.


The ChristchurchHealth and Develop-ment Study is a longitu-dinal study of a cohortof 1,265 children bornin 1977 in theChristchurch,NewZealand,urban regionwhowere studied frombirth to age 25, includ-ing 630 females; 41% ofwomen Pg on at leastone occasion prior toage 25; 14.6% had atleast one abortion.Sample sizes in analysesranged from 506 and520 depending on thetiming of assessment.Details on Ns forprospective analyseswere provided in per-sonal communicationfrom the author.

Sample Sizes

Concurrent analyses:AB N= 74DEL N= 131Never Pg N= 301

Prospective analysis:AB N=48Del N= 77Never Pg N= 367

Primary Outcome

DSM-IVmental disor-ders (includingmajordep,overanxious disor-der, GAD, social phobia,& simple phobia, andsuicidal behavior for in-tervals 15-18,18-21 and21-25 years, controllingfor childhood, family,and related confound-ing factors.Outcomesfor concurrent analyses:yes/no diagnosis ofmajor dep,anxiety dis-order, alcohol and illicitdrug dependence, suici-dal ideation in previous12mo., and total # ofdisorders; in prospec-tive analysis, total num-ber of disorders from21-25 yrs.

Results

In concurrent analyses,controlling for covari-ates, AB grp had sig(p<0.05) higher rates ofdepression, suicidalideation, illicit drug de-pendence, & total men-tal health problemsthan the DEL grp & ex-cept for alcohol andanxiety disorder, signifi-cantly higher rates ofdisorder than the NeverPG grp. A prospectiveanalysis used Pg/abor-tion history prior to age21 to predict mentalhealth outcomes from21-25 years.Similarly,after covariate adjust-ment, the AB grp had asig higher total # of dis-orders than the othergrps,which did not sigdiffer from each other.


Neither intendednessnor wantedness of Pgcontrolled; screeningcriteria related tomen-tal health for legal abor-tion in New Zealandmay bias portrait ofoutcomes.Abortion isunderreported.N toosmall for multiple abor-tions to be analyzedseparately.Although alongitudinal study,most results reportedinvolved the concurrentassessment of Pg statusandmental health.Theprospective analysiswas limited to numberof disorders owing tothe relatively sparsedata for specific disor-ders over the interval21-25 years and thesmall number ofwomenwho becamepregnant by age 21.


problems, having trouble sleeping during the past year,using cigarettes or marijuana during the past 30 days,using alcohol during the past year, or reported havingproblems with parents because of alcohol use. Alloutcomes were assessed with single-item measures.Adjusted for covariates previously shown to differbetween the two groups (risk-taking and desire toleave home), girls who reported an abortion weremore likely than girls who delivered to say they hadever had counseling, trouble sleeping during the pastyear, and used marijuana in past 30 days. No differ-ences were observed on frequency of alcohol use orcigarette smoking.

Strengths of this study included the use of a compari-son group of girls who delivered unwanted pregnan-cies, the weighting of design factors in the analyses,and efforts to enhance the accuracy of self-reports ofsensitive topics (respondents listened to prerecordedquestions through earphones and entered their own an-swers). Nonetheless, problems of sampling and meas-urement limit the utility of this study. The extremelysmall number of girls in the eventual sample analyzed(N=130), especially given the very large original sample(of approximately 6,000 girls), raises questions aboutunderreporting, drop-out rates, and exclusion criteria.Given that the sample is school-based, adolescents whodrop out of school to care for a child would not be in-cluded in the study. The single-item measures of psy-chological problems are psychometrically weak andclinically suspect. Because the percentages and Ns foroutcome variables were not reported, the frequencywith which problems occurred cannot be determined.Furthermore, the measure of counseling asked whetherthe respondent had ever received counseling for psy-chological or emotional problems—it cannot be deter-mined from this item whether counseling occurredprior or subsequent to the pregnancy.

Hope, Wilder, and Watt (2003) used data from theADD-Health study (Waves I and II) to examine therelationships among adolescent pregnancy, pregnancyresolution, and delinquent behavior. Although delin-quency includes behaviors that are not part of themental health focus of this review (e.g., lying to par-ents/guardian, taking part in a fight), one domain ofdelinquent behavior examined (alcohol use, use of ille-gal substances) is within the purview of this review.Thus, we focus here on longitudinal analyses examin-ing the relationship between pregnancy resolution and

substance use. In a set of prospective analyses focusingon adolescent girls who became pregnant betweenWave I and II of the survey, Hope et al. examined therelationship of pregnancy resolution (abortion vs. keptbaby) to reports of having smoked cigarettes or mari-juana at least 1 day in the past 30 days. These com-parisons of the abortion and “kept baby” groupsexcluded girls who experienced pregnancies prior toWave I as well as those who miscarried or were stillpregnant at Wave II.

Young women who had abortions reported higherrates of cigarette smoking and marijuana use thanyoung women who kept their baby, both prior to theirpregnancy (Wave I) and subsequent to their preg-nancy (Wave II). Keeping the baby was associatedwith a decrease in reported cigarette or marijuana usebetween the two waves of data collection, leading theauthors to conclude that adolescent motherhood func-tions as a social control on delinquent behavior. Incontrast, having an abortion was not associated with achange in rates of smoking or marijuana use fromWave I to Wave II, leading the authors to concludethat terminating a pregnancy through abortion doesnot increase the likelihood of delinquent behavior orsubstance use.

In addition to strengths and weaknesses of the ADD-Health school-based database described above, thisstudy is limited by single-item measures of cigaretteand marijuana use that are psychometrically weak.Furthermore, despite the large initial sample size ofover 6,000 girls, the number of pregnant girls (69 whohad abortions, 87 who kept their baby) in the finalsample was small.

The Harvard Study of Moods and Cycles. Harlow,Cohen, Otto, Spiegelman, and Cramer (2004) useddata from a cross-sectional sample of 4,161 womenbetween 36-45 years of age residing in the Bostonmetropolitan area to examine the relationship of earlylife menstrual-cycle characteristics and reproductivehistory to onset of major depression later in life. Theyanalyzed data from a subsample of 332 women whomet DSM criteria for having had major depressionand 644 women with no current or past history ofmajor depression. In-person interviews were used toestablish mental health status and to gather informa-tion on demographic and lifestyle characteristics, men-strual and reproductive history, past and current


medical conditions, and use of hormonal and nonhor-monal medications. Relevant analyses examined thelink between lifetime history of depression and abor-tion history. There were no significant differences be-tween the proportion of women with a lifetime historyof major depression (19.3%) who reported having oneabortion and the proportion of women with no his-tory of depression (17.9%) who reported having hadone abortion. However, women with a lifetime historyof depression were significantly more likely to reporthaving had multiple abortions before their first onsetof depression than were nondepressed women, con-trolling for age, age at menarche, educational attain-ment, and marital disruption. Direct comparisonsbetween women reporting abortion versus deliverywere not conducted. The researchers also reported astrong association between depression and marital dis-ruption, underscoring the importance of controllingfor marital status when seeking to assess the independ-ent contribution of abortion to depression risk. Theresearchers pointed out that the higher proportion ofwomen with multiple abortions found in the depressedversus nondepressed group may reflect a variety of an-tecedent conditions that were not assessed in thestudy, including involvement in abusive relationships.A particular strength of this study is its measurementof a clinically significant mental health disorder (de-pression) with established diagnostic criteria. In addi-tion to the usual issues involved with a cross-sectionalstudy that relies on retrospective self-report, study lim-itations include the possibility of a selective recall biason the part of depressed women, and lack of informa-tion on pregnancy intention or wantedness, whetheror not abortions were for therapeutic reasons, andwomen’s exposure to violence.

New Zealand Christchurch Health and DevelopmentStudy. The most comprehensive of the secondaryanalysis studies in terms of assessment of mentalhealth outcomes was conducted in New Zealand(NZ). Fergusson et al. (2006) analyzed data from a25-year longitudinal study of a cohort of children (in-cluding 630 females) born in 1977 in theChristchurch, NZ, urban region who were studiedfrom birth to age 25 years. Information was obtainedon (a) the self-reported reproductive history of partici-pants from 15-25 years (abortion, delivery, or neverpregnant); (b) measures of DSM-IV mental disorders(including major depression, overanxious disorder,generalized anxiety disorder, social phobia, and simple

phobia) and suicidal behavior for intervals 15-18, 18-21, and 21-25 years; and (c) childhood, family, and re-lated confounding factors, including measures of childabuse.

In a series of concurrent analyses adjusting for covari-ates such as greater childhood social and economicdisadvantage, family dysfunction, and individual ad-justment problems in the abortion group, Fergusson etal. (2006) found that women in the abortion grouphad significantly higher rates of concurrent depres-sion, suicidal ideation, illicit drug dependence, andtotal number of mental health problems than the de-livery group. Concurrent analyses also indicated thatexcept for alcohol and anxiety disorder, the abortiongroup had significantly higher rates of these disordersthan the never pregnant group. More important, how-ever, are the prospective analyses reported, as thesecapitalize on the longitudinal strengths of the study.The authors conducted a prospective analysis using re-productive history prior to age 21 years to predicttotal number of mental health problems experiencedfrom 21–25 years (samples were too small to permitanalyses by disorder). Controlling for covariates, theabortion group had a significantly higher number ofdisorders than the other two groups, which did notdiffer significantly from each other.

This study is unusual in the quality of measurement ofthe mental health variables, range of outcomes as-sessed, and number of co-occurring risk factors con-trolled. However, several design features limitconclusions that can be drawn from this study. First,neither wantedness nor intentionality of pregnancywas controlled. Second, women with multiple abor-tions were not separated from women with one abor-tion (21.6% of the abortion group had more than oneabortion).2 Third, as with other survey studies of thistype, comparisons of reported abortions with popula-tion data suggest that abortion was underreported inthis sample, although not to a great extent. Finally,differing abortion regulations between the UnitedStates and NZ also mean that caution should be usedin generalizing from these studies to women in generalin the United States.

In order to obtain a legal abortion in NZ, a womanmust obtain the approval of two specialist consultants,the consultants must agree that either (1) the preg-nancy would seriously harm the life or the physical or


mental health of the woman, (2) the pregnancy is theresult of incest, (3) the woman is severely mentallyhandicapped, or (4) a fetal abnormality exists. Anabortion will also be considered on the basis of thepregnant woman’s young age or when the pregnancyis the result of rape.

Evaluation of record-based and secondary analysisstudies. In weighing the evidence regarding abortionand mental health derived from the record-based andsecondary analysis studies reviewed above, it must bekept in mind that the body of evidence is not as largeas it appears. The 10 studies based on medical recordsare based on two data sets, one from the United Statesand one from Finland. The 15 studies based on sec-ondary analyses of survey data are based on nine datasets, eight from the United States and one from NewZealand. Given that caution, what can be concludedfrom examination of these studies? An answer to thatquestion requires considering their methodologicalquality.

Problems of sampling. First, many of the above studiescannot be generalized to the majority of women inthe United States who seek abortions. Some are basedon specialized data sets not representative of womenin general (e.g., Coleman, Maxey, et al., 2005; Cole-man, Reardon, et al 2005), some used screening crite-ria that eliminated a huge proportion of the largersample (e.g., all of the Medi-Cal studies), some differ-entially excluded women from one outcome groupbut not the other (Reardon & Cougle, 2002a), andsome were based on samples of women who obtainedabortions under more restrictive regulations (Fergus-son et al., 2006). Only one of the above studies basedon survey data used sampling weights in its analyses(Coleman, 2006a). The study by Coleman (2006a),which did use sample weights, used a school-basedpopulation that did not include the most disadvan-taged adolescents—those who dropped out of schoolto care for a child.

Problems of comparison groups. Although it is necessaryto control for wantedness of pregnancy to assess apregnant woman’s mental health risks if she were tochoose abortion compared to its alternatives, onlythree data sets (the NSFG, ADD-Health, and NLSYdata sets) included questions about the intendednessor wantedness of pregnancy. Even when this informa-tion was available, it was not always used (Cougle et

al., 2003). In addition, interpretation of differencesobserved between the abortion and delivery groupswas often compromised by differential exclusionsfrom the delivery group.

Problems inmeasurement of independent variables. Otherthan the studies based on medical records, all of thestudies reviewed above established abortion historythrough retrospective self-reports, raising serious relia-bility concerns. Few of the above studies took ade-quate steps to enhance the accuracy of reports ofsensitive data. Thus, not surprisingly, abortion wasunderreported relative to national norms in all of thestudies based on survey data. Furthermore, becausenone of these public data sets was designed specificallyto identify the mental health effects of abortion com-pared with its alternatives, none provides adequate in-formation about the characteristics of the abortionexperience, such as the length of gestation at time ofthe abortion, age at which the abortion occurred, thereason for having the abortion (including medical rea-sons), and wantedness of the pregnancy. This informa-tion is not available for the medical record studieseither. Such data are essential to understand the psy-chological implications of abortion.

Problems inmeasurement of outcomes. Studies based onsecondary analysis of survey data typically did not usestandard measures of mental health. Some studieswere based on single-item measures of outcomes (e.g.,Coleman, 2006a); others used an unvalidated measureof a psychological problem (e.g., Cougle et al., 2005)or only one or two measures of general psychologicalwell-being (e.g., Russo & Zierk, 1992). Only two ofthe studies based on survey data (Fergusson et al.,2006; Harlow et al., 2004) used psychometricallystrong assessments of clinically significant outcomes(i.e. a diagnosis). Further, in some cases, it was impos-sible to determine whether the “outcome” variable oc-curred prior or subsequent to the abortion (Coleman,2006a; Cougle et al., 2005; Russo & Denious, 2001).Although less severe, there are problems with outcomemeasurement in the Medi-Cal data as well. Only onestudy (Gissler et al., 2004b) made an attempt to sepa-rate out therapeutic abortions from elective abortions,a distinction shown to be critical by the Finnish re-searchers.

Confounds and co-occurring risks. Researchers relying onsecondary analysis of both medical records and survey


data collected for other purposes only have access tovariables collected in those data sets. As a consequence,key variables that have documented relationships withboth pregnancy outcome and mental health and whichare thus potential confounders of any observed relation-ship between those variables may not be included in thedata set. These include, for example, measures of priorsubstance abuse, prior or ongoing exposure to sexualabuse or partner violence, poverty, number of currentchildren, number of prior unwanted pregnanciesand prior unwanted births (both of which are corre-lated with number of abortions), and, most importantly,adequate measures of mental health prior to pregnancy.Only one of the 23 studies reviewed above (Fergussonet al., 2006) contained adequate measures of mentalhealth prior to the pregnancy. In addition, with regardto the studies that focus on low-income populations(Medi-Cal studies, Washington study, Baltimore study),such populations are more likely to be in poor health,which itself is associated with psychological problems.Given that pregnant women who have serious illnessessuch as diabetes, AIDS, and heart disease may be ad-vised to have an abortion for health reasons, the corre-lation of abortion and physical and mental healthproblems might be expected to be higher in low-incomepopulations.

Problemswith statistical analyses. Large public data sets,particularly multiyear data sets, are complex and havean enormous number of variables from which to selectfor a particular analysis. As seen by the studies abovethat have published corrections of coding errors (e.g.,Reardon & Cougle, 2002b; Schmiege & Russo,2005), it is easy to make mistakes in the constructionof variables. Moreover, it is important to have a con-ceptual rationale for selecting among the large numberof potential variables. The variables researchers selectto include in reanalyses of the original data reflect theinterests (and sometimes the biases) of the researcherdoing the reanalysis. The approach to the data analy-ses reflected in these studies is also of concern. Largenumbers of statistical tests were often performed, in-creasing the probability of finding significant resultswhen there was in fact no effect. The large samplesizes mean that effect sizes that are a statistically sig-nificant may be clinically meaningless. On the otherhand, analyses were often based on small subgroupsor subgroups for which no sample size was provided.In addition, results were frequently overinterpreted,with one significant finding emphasized over a num-

ber that were not significant or were in the reversedirection.

The selection of covariates in these studies also raisedserious concerns. As noted above, the choice of co-variates to include in analyses can play a key role inhow much variance in the outcome variable is ex-plained by pregnancy outcome. Given the large num-ber of variables often assessed in these data sets, thereis considerable room for researcher discretion in se-lection of covariates. Inclusion of covariates wasoften based on atheoretical preliminary analyses andoften varied for unspecified reasons across analyses,even within the same study. In some studies, key co-variates known to be associated with the outcome inquestion were omitted from the analyses despite theirpresence in the data set. For example, Reardon et al.(2004) used NLSY data to compare alcohol and druguse of women who aborted a first pregnancy to thosewho delivered their first pregnancy or were not preg-nant. They did not control for history of drug useprior to the first pregnancy in their analyses, despitethe availability of this information in the data set anddespite prior published studies based on this samedata set showing that use of drugs and alcohol pre-dicted onset of early sexual activity (Rosenbaum &Kandel, 1990) and was uniquely predictive of subse-quent premarital teen pregnancy as well as the deci-sion to terminate a premarital teen pregnancy(Mensch & Kandel, 1992). As another example, intheir analysis of the NSFG, Cougle et al. (2005) didnot include items assessing rape history in their analy-sis, despite the presence of relevant items in the dataset and publication of other studies (e.g., Reardon etal., 2002; Russo & Denious, 2001) suggesting thatwomen who have abortions are at higher risk for rapeand other forms of violence in their lives.

Summary of medical-record and secondary analysesstudies. In sum, our careful evaluation of studiesbased on secondary analyses of medical records andexisting public data sets revealed that in generalthey were methodologically quite poor. Problems ofsampling, measurement, design, and analyses cloudinterpretation. Because of the absence of adequatecontrols for co-occurring risks and prior mentalhealth in these studies, it is impossible to determinewhether any observed differences between abortiongroups and comparison groups reflect consequencesof pregnancy resolution or preexisting differences


between groups or methodological artifact. Conse-quently, these studies do not provide a strong basisfor drawing conclusions regarding the relative risks ofabortion compared to its alternatives.

ComparisonGroup Studies Based on PrimaryDataSeventeen studies were conducted between 1990 and2007 with the primary purpose of comparing womenwho had a first-trimester abortion (or an abortion inwhich trimester was unspecified) to a comparisongroup of other women on a mental health related vari-able. These studies resulted in 19 publishedpapers. Details, key findings, and limitations of thesestudies are summarized in Tables 3a and 3b.

Description of findings:U.S. samples. Seven studieswere based on U.S. samples. These studies are summa-rized in Table 3a. Cohan et al. (1993) examined re-sponses of 33 women 1 month postpregnancy, 21 ofwhom had terminated their pregnancy and 12 ofwhom continued their pregnancy. Almost all had re-ported that their pregnancy was unintended. Therewere no significant differences between the 21 womenwho had terminated their pregnancy versus the 12 ofthose who continued their pregnancy on any of theoutcomes assessed (positive and negative affect anddecision satisfaction).

Lydon, Dunkel-Schetter, Cohan, and Pierce (1996) as-sessed initial commitment to a possible pregnancy aswell as positive affect and negative affect (Derogatis,1975) among women just prior to obtaining a preg-nancy test at health clinics in the United States andCanada. For the women who received a positive preg-nancy result, these variables were reassessed within 9days (T2) and again at 4–7 weeks (T3) after learning ofthe positive test result. By the T3 follow-up, 30 womenhad terminated their pregnancy, and 25 had decided tocontinue their pregnancy. Initial commitment to thepossible pregnancy (assessed at T1) interacted withoutcome decision (abort vs. deliver) to predict affect atT3. Among women continuing their pregnancy, thosehigh (N=11) and low (N=12) in initial commitment tothe pregnancy did not differ significantly in affect atT3. Both expressed more positive than negative affect.Among women who had aborted their pregnancy,those who had been initially less committed to the pos-sible pregnancy (N=13) did not differ significantly inaffect from those deciding to continue their pregnancy.They too expressed more positive than negative affect.

The women who had initially indicated somewhatmore commitment to the possible pregnancy but whodecided to terminate the pregnancy (N=14) reportedsignificantly less positive affect and significantly morenegative affect than the other three groups. A particu-lar strength of this study is its tracking of commitmentand affective state over the time course of first learningof a pregnancy and its resolution. Other strengths areits strong theoretical framework and good measure-ment of predictor variables. Limitations include thevery small sample sizes and absence of measures ofclinically significant mental health outcomes.

The remaining four U.S. studies measured abortion his-tory through retrospective self-reporting (see Table 3a).Felton, Parsons, and Hassell (1998) found no signifi-cant differences on overall health-promoting behaviors,appraisals of problem-solving effectiveness, or globalself-image between 26 adolescents attending a familyplanning clinic who reported a history of abortion and26 demographically matched adolescents who reportednever being pregnant. Williams (2001) found no signif-icant differences on any of the subscales of the GriefExperience Inventory between 45 women waiting tosee their health care provider who reported a history ofabortion and 48 demographically similar women whoreported no elective abortions. Medora et al. (1993)found that among a sample of 121 single, never mar-ried, pregnant teenagers, the 28 girls who reported aprior abortion had significantly higher self-esteem thanthe 93 girls who reported no abortion history. Medoraand von der Hellen (1997) reported that among a sam-ple of 94 teen mothers, teens who reported a priorabortion did not differ in self-esteem from teens whodid not report an abortion (number in each group wasnot specified). The only U.S. study to report thatan abortion group had a poorer outcome than a com-parison group was conducted by Reardon and Ney(2000). This study was based on a reproductive historyquestionnaire mailed to the homes of a large sample ofwomen, only 14.2% of whom responded. In analysesrestricted to White women, women who reported hav-ing had at least one induced abortion (N = 137) weremore likely than women who reported having had noabortions (N = 395) to also agree with a single yes/noquestion: “Have you ever abused drugs or alcohol?”

Description of findings:Non-U.S. samples. Ninestudies were based exclusively on non-U.S. samples.Most were methodologically quite poor (see Table


3b). The most methodologically sound papers werebased on a study conducted by Broen and colleaguesin Norway (Broen, Moum, Bodtker, Ekeberg, 2004,2005, 2006) and one conducted jointly by the RoyalCollege of General Practitioners and the Royal Col-lege of Obstetricians and Gynecologists in the UnitedKingdom (Gilchrist et al., 1995).

The study by Broen and colleagues followed twogroups of Norwegian women from 10 days to 5 yearsafter a first-trimester induced abortion (N = 80) orearly miscarriage (< 17 weeks; N = 40). Experiencesof anxiety and depression, avoidance, intrusion stressreactions (assessed with the Impact of Events scale),subjective well-being, and feelings about the preg-nancy termination were assessed at four intervals postabortion. Comparisons between the miscarriage andinduced abortion groups, controlling for potentialconfounders, revealed no significant differences be-tween the two groups in mean anxiety or depressionscores or subjective well-being scores at any timepoint. Women who had an induced abortion reportedfeeling more guilt, shame, and relief and also moreavoidance on the IES scale than women who miscar-ried. Women who miscarried reported more feelings ofgrief and loss than those who had an induced abortionin the short term, but this difference disappeared by5 years post event.

Strengths of this study included its repeated and long-term follow-up, attempt to control for prepregnancymental health (although this was assessed retrospec-tively via self-report and psychiatric evaluation postabortion), use of established and reliable outcomemeasures, and high retention rate (91%), althoughonly 47% of those initially approached agreed to par-ticipate in the study. This study is useful for compar-ing grief reactions among different forms of pregnancyloss. However, the comparison group used in thisstudy is inappropriate for drawing conclusions aboutthe relative risks of abortion versus its alternatives. Aspontaneous miscarriage of a (wanted) pregnancy isnot an alternative for women faced with a decisionabout how to resolve an unintended or unwantedpregnancy.

The strongest study reviewed (Gilchrist et al.,1995)was prospective and longitudinal and employed alarge sample size, appropriate comparison groupsof women with unplanned pregnancies, and a long

postpregnancy/abortion follow-up time. Importantly,this study also controlled for mental health prior tothe pregnancy as well as other covariates. Women’smedical, psychiatric, and obstetric history prior tothe pregnancy was recorded from their medicalrecords or the recruiting physicians’ case notes. Thefinal sample consisted of four pregnancy outcomecomparison groups: (a) 6,410 women who obtainedterminations (85% occurred before 12 weeks of ges-tation), (b) 6,151 women who did not seek termina-tion, (c) 379 who requested termination but weredenied, and (d) 321 who requested termination butchanged their mind.

Postdelivery/abortion psychiatric morbidity was as-sessed using established diagnoses and grouped intothree categories in order of severity: (a) psychosis,(b) nonpsychotic illness (e.g., depression, anxiety),and (c) deliberate self-harm (DSH) without otherpsychiatric illness (e.g., drug overdoses). Similarly,prepregnancy psychiatric history was classified intofour categories in order of severity: (a) psychoticepisode, (b) nonpsychotic illness, (c) DSH withoutother psychiatric illness, and (d) no psychiatric illness.The two largest subgroups of prepregnancy historywere women with no prepregnancy history of psychi-atric problems or DSH prior to the pregnancy (2476women) and women with a history of nonpsychoticillness (1100 women), followed by women with a his-tory of psychosis (N=106 ) and women with a historyof DSH alone (N=36). Differences between the deliv-ery reference group and each of the other three com-parison groups were examined within each of thefour categories of prepregnancy psychiatric history.Age, marital status, smoking, education level, gravid-ity, and prior history of abortion were controlled inanalyses that focused on the overall rate of postpreg-nancy psychiatric morbidity as well as the rate ofeach of the three postpregnancy diagnoses among thefour comparison groups.

Among women with equivalent past psychiatric his-tories, there were no significant differences betweenthe four comparison groups in overall rates of psy-chiatric illness. Rates of specific postpregnancy psy-chiatric illnesses, however, differed among thecomparison groups depending on prepregnancy diag-nostic history and diagnostic outcome as follows: (1)With respect to postpregnancy nonpsychotic illness,no significant differences were found between abor-


tion and delivery groups, irrespective of prepreg-nancy diagnostic history. (2) With respect to post-pregnancy psychoses, women who had an abortionwere significantly less likely to have a postpregnancypsychotic episode than those who delivered among

the subgroup of women with no prepregnancy his-tory of psychotic illness (1.1 vs. 4.1) and among thesubgroup of women with a history of nonpsychoticillness (4.9 vs. 11.8). A similar, but nonsignificantpattern was observed among the subgroup of women


Table 3A:Primary Data Comparison Group StudiesUNITES STATES SAMPLES

Citation

Cohan,C.L.,Dunkel-Schetter,Christine,&Lydon, J. (1993).Preg-nancy decisionmaking:Predictors of earlystress and adjustment.Psychology ofWomenQuarterly,17,223-239.

Felton,G.M.,Parsons,M.A.,& Hassell, J.S.(1998).Health behaviorand related factors inadolescents with a his-tory of abortion andnever pregnant adoles-cents. Health Care forWomen International,19,37-47.

Sample &Design

U.S.Recruited at healthclinic prior to preg-nancy testing (88% re-sponse rate).Pregnancyintendedness and out-come intentions as-sessed prior to learningoutcome.81% indi-cated pregnancy wasunintended.33 of the44 whowere pregnantcompleted question-naires at two points: (24hrs post-Pg test out-come & 1month post-Pg test outcome).Ofthe 33,21 had an abor-tion & 12 carried toterm.Criteria for partici-pation: 18 yrs or older &English speaking.

U.S.26 adolescents (age16-19) attending edu-cation classes at pub-licly supported familyplanning clinics who re-ported a history ofabortion on question-naires. Criteria for par-ticipation: nevermarried,not currentlypregnant., never gavebirth, and completionof 9th grade.

ComparisonGroup

Fifteen women whoinitially intended toabort and did so (de-cided aborters) and 6women who were ini-tially undecided andlater aborted (unde-cided aborters) werecompared to 10women who initiallyintended to carry toterm and did so.

Twenty-six never-pregnant adolescentsmatched to abortiongroup on age, race,education,& Medicaidstatus.Two groupsalso similar on age atfirst coitus and pat-terns of contraceptiveuse.

Primary Outcome

Positive and negativeaffect (Affect Balancescale).Decision satisfac-tion (single item).

Healthy lifestyle (HealthPromoting Lifestyle Pro-file). Perceived effec-tiveness of problemsolving (Problem Solv-ing Inventory). Adjust-ment (Offer Self-ImageQuestionnaire)

Key Findings

Onemonth post-test,there were no signifi-cant differences in ei-ther positive ornegative affect be-tweenwomenwhoaborted (both initiallydecided and unde-cided) vs. thosewhocontinued their preg-nancy.Women commit-ted to carrying theirpregnancy to termweremarginallymore satis-fiedwith their decisionthan both abortiongroups,who did not dif-fer from each other.Overall,womenwhoabortedwere satisfiedwith their decision.

No significant differ-ence between abortionand never-pregnantgroups on overallhealth-promoting be-haviors, appraisals ofproblem-solving effec-tiveness, and globalself-image.Both groups’scores on the Offer Self-ImageQuestionnairewere also compared tonormed referencegroup scores.Adoles-cents with history ofabortion scored belowthe norm on 10 out of12 areas of adjustment;never-pregnant adoles-cents scored below thenorm on 8 out of 12areas of adjustment.

Limitations

Extremely small samplesizes.Single-itemmeasure of decisionsatisfaction.Analysesdo not control forwhether pregnancywas intended or not.Nomeasures ofpre-pregnancymentalhealth.

Abortion history retro-spectively self-reported.No information aboutrecruitment strategy, re-sponse rate, samplerepresentativeness,orabortion context (e.g.,timing,gestation,age,etc). Extremely smallsample size. Compari-son group not appro-priate. Nomeasures ofpre-pregnancymentalhealth.


Table 3A:Primary Data Comparison Group StudiesUNITES STATES SAMPLES (continued)

Citation

Lydon, J.,Dunkel-Schet-ter, C.,Cohan,C.L.,Pierce,T. (1996).Preg-nancy decisionmakingas a significant lifeevent:A commitmentapproach. Journal ofPersonality and SocialPsychology,71,141-151.

Medora,N P.,Goldstein,A.,& von der Hellen,C.(1993).Variables relatedto romanticism andself-esteem in pregnantteenagers.Adolescence,28,159-170.

Medora,N.P.& von derHellen,C. (1997).Ro-manticism and self-es-teem among teenmothers.Adolesscense,32,811-814.

Sample &Design

U.S.and Canada.Re-cruited at health clinicsprior to pregnancy test-ing (90% response rate).Pregnancy intended-ness, wantedness,meaningfulness,com-mitment, concerns,andpositive and negative af-fect assessed prior tolearning Pg test out-come (T1).85womentested positive; 57 ofwhomcompleted inter-viewswithin 9 days oftest result (T2) andwithin 4-7wks of test re-sult (T3). 30 had abor-tion prior toT3;25continued Pg,2 hadabortion afterT3 follow-up.Criteria for participa-tion: 18 yrs or older,English speaking in U.S.Eng or French in Canada.

U.S.28 pregnantteenagers whoweresingle,never married,and enrolled in a preg-nant minor program orresiding in amaternityhome,who reported aprior abortion historyon a questionnaire.

U.S.Full sample con-sisted of 94 teenmoth-ers enrolled in a teenmother program affili-ated with a high schoolin Southern CA.Ages13-18 yrs.51 (54%)Latino, (23%) AfricanAmerican, (18%) Anglo,(4%) were Asian.Un-specified number ofgirls in sample reportedprior abortion.

ComparisonGroup

Thirty women whoaborted and 25women who carriedto term were dividedby high vs. low earlycommitment to preg-nancy at T1 and com-pared on affectbalance at T2 and T3.

Ninety-three preg-nant teenagers whowere single, nevermarried, and in samepregnant-minor pro-gram or maternityhome,who reportedno abortion history.

Unspecified numberof girls in sample whodid not report a priorabortion.

Primary Outcome

Negative affect (anxiety,guilt,depression,hostil-ity) and positive affectassessed with AffectBalance Scale (Dero-gatis, 1975).Affect Bal-ance (ave pos emominus ave neg emo) asmeasure of emotionaladjustment.

Self-esteem (BachmanSelf-Esteem scale )

Self-esteem (BachmanSelf-Esteem Scale )

Key Findings

Initial commitment atT1interactedwith outcomedecision (abort vs.de-liver) to predict affect atT3.Amongwomen con-tinuing Pg,those high(N=11) and low (N=12)in initial commitment toPghad equal pos affectatT3.Amongwomenwho aborted Pg,thoseless committed initiallyto Pg (N=13) did not dif-fer in pos affect fromthose continuing Pg.Those somewhatmorecommitted to Pg initially(N=14) had less sig posaffect andmore neg af-fect than those continu-ing Pg.

Pregnant teens who re-ported a prior abortionhad higher self-esteemthan pregnant teenswho reported no priorabortion

No significant differ-ence in self-esteembe-tween teenmotherswho reported an abor-tion and teenmotherswho did not.

Limitations

Strength of study is track-ing of commitment andaffect over timeduringcourse of pregnancydecision;good theoreticalframework,goodmeasurement of predic-tors. Limitations includesmall sample size,high at-trition. Outcomemeasurenot clinically significant.

Abortionhistory retrospec-tively self-reported.No in-formation about abortioncontext (e.g.,timing,gesta-tion). Small sample size.Samplenot representative.Comparisongroupnot ap-propriate. Nomeasures ofpre-pregnancymentalhealth.

No information aboutnumber of teenmotherswhodid anddid not abort;abortion history retrospec-tively self-reported.No in-formation about abortioncontext (e.g., timing,gesta-tion). Small sample size.Sample not representative.Comparison groupnot ap-propriate. Nomeasures ofprepregnancymentalhealth.


Table 3A:Primary Data Comparison Group StudiesUNITES STATES SAMPLES (continued)

Citation

Reardon,D.C.& Ney,P.G. (2000).Abortion andsubsequent substanceabuse.American Journalof Drug andAlcoholAbuse,26, 61-75.

Williams,G.B. (2001).Short term grief afteran elective abortion.Journal of Obstetric,Gynecologic,andNeonatal Nursing,30,174-183.

Sample &Design

U.S.Reproductive historyquestionnaire sent to anational sample of 4929womenbetweenagesof24 and44,selected ran-domly from“nationalmail-ing list housedatabase.”700 completed formsreturned (14.2%;94%ofrespondentsWhite).Onehundred and fifty-twowomen reportedhavingat least one inducedabortion.Analyses re-stricted toWhitewomenwhoaborted (N=137).

US.45 women (ave age23 years) waiting to seetheir health careprovider in a gynecol-ogical clinic who re-ported a history of oneor more abortions on aquestionnaire.Exclu-sion criteria included aperinatal loss of a non-voluntary nature withinthe past 5 years, a priorabortion for medicalreasons,or a docu-mented psychiatric his-tory.

ComparisonGroup

Comparison group of395White womenwho reported noabortions

Forty-eight womenwho completed samequestionnaire undersame circumstancesbut who reported noabortion history.There were no signifi-cant differences be-tween the two groupsin age, ethnicity,mari-tal status, education,income, or religion.

Primary Outcome

Single itemmeasure:"Have you ever abuseddrugs or alcohol?"yes/no

Grief. (Grief ExperienceInventory).

Key Findings

Significant positive associ-ationobservedbetweenself-reported abortionhis-tory and self-reportedsubstance abuse. Amongwhitewomen,65%whoreported ahistory of sub-stance abuse identifiedtheonset as occurringprior to age at first preg-nancy.

There were no signifi-cant differences be-tween the abortiongroups and no abortiongroups on any of the12 clinical scales ofthe Grief ExperienceInventory.

Limitations

Abortion history retro-spectively self-reported.Extremely low responserate.Sample not repre-sentative of U.S.women.Abortions un-derreported comparedto national statistics.Noinformation about con-text of abortion.Singleitem,dichotomous de-pendentmeasure not avalid indicator of sub-stance abuse.Responsebias likely, i.e.,womenwilling to report one so-cially sanctioned action(abortion) may bemorewilling to also reportanother (substanceabuse). Inappropriatecomparison group.Many tests of signifi-cance conducted, capi-talizing on chance.Analyses performed onextremely small subsetsof women (e.g.,N's <5).Nomeasures of pre-pregnancymentalhealth.

Abortion history retro-spectively self-reported.No information aboutresponse rate or repre-sentativeness of thesamples was provided.Small sample size.Comparison group notappropriate.Nomeas-ures of pre-pregnancymental health.


Table 3B:Comparison Group StudiesNON-U.S.SAMPLES

Citation

Bailey,P.E., Bruno,Z.V.,Bezerra,M.F.,Queiroz, I.,Oliveira,C.M.,& Chen-Mok,M. (2001).Adoles-cent pregnancy 1 yearlater:The effects ofabortion vs.mother-hood in northeastBrazil. Journal of Adoles-cent Health,29,223-232.

Barnett,W,Freuden-berg, N.,&Wille,R.(1992).Partnership afterinduced abortion:Aprospective controlledstudy.Archives of SexualBehavior,21, 443-455.

Sample &Design

Brazil.125 adolescentsadmitted to hospital forcomplications from ille-gal induced abortion in-terviewed beforedischarge.95 inter-viewed 1 year postabor-tion. Criteria forparticipation:18 oryounger,never gavebirth but not necessarilyfirst Pg,within 21weeksof gestation for abortersU.S.28 pregnantteenagers whoweresingle,never married,and enrolled in a preg-nant minor program orresiding in amaternityhome,who reported aprior abortion historyon a questionnaire.

Germany. Ninety-twowomen seeking abor-tion for socially indi-cated reasons (withoutmedical indication)were interviewed priorto and 1 year post abor-tion. All were referredto the study by their gy-necologists and were ina stable relationshipwith their partner.Nonehad an abortion duringthe previous year.

ComparisonGroup

Cohort of 367 preg-nant teens whosought prenatal careat the same hospital.

Comparison group of92 women drawn ran-domly from each gy-necological practicewho were in a stablerelationship,wereusing safe contracep-tives, had not hadabortion in prior year,and did not desire achild.They werematched to abortiongroup on martial sta-tus, age, number ofchildren, duration ofpartnership, and edu-cational background.They were inter-viewed at the sametwo time points.

Primary Outcome

Self-esteem (RosenbergSelf Esteem scale). Per-cent enrolled in schoolone year later.

Quality of relationshipwith partner prior toand 1 year post abor-tion: Affection, conflictbehavior, andmutualinterests (PartnershipQuestionnaire);Mutualtrust (Interpersonal Re-lationships scale); Per-cent separated frompartner at one year; Sat-isfaction with sex life.

Key Findings

Lower percent of teenswith high self-esteemamong induced abor-tion groupbothbeforedischarge andone yearlater than among teenswith intendedor unin-tendedpregnancies.Teens in abortion groupwere 6.9 timesmorelikely to be enrolled inschool 1 year later thanteenswith intendedpregnancies.

At Time 1 (preabortion),relationships of abor-tion groupwere ofpoorer quality (moreconflict, less affection,less trust) than controlgroup.At Time 2 (oneyear postabortion),there were no differ-ences between abor-tion and control groupin relationship quality,mutual trust,percentseparated,or satisfac-tionwith sex life.

Limitations

Sample not generalizabletoU.S. Abortion is illegal inBrazil unless pregnancy re-sults from rape or placeswoman’s life at risk.Samplewas recruited fromwomenexperiencingmedicalcomplications froman ille-gal abortion.Comparisongroup (teens carrying toterm) does not control forwantedness of pregnancy.Nomeasures of pre-preg-nancymental health.

Onlywomen in stable rela-tionships included in study.Nomeasures of prepreg-nancymental health.Someinitial differences betweenabortion and controlgroup (a higher percent ofabortion groupwerework-ing class and reportedmarital disharmony inchildhood). Comparisongroup (not pregnant) notappropriate.



Citation

(1) Broen,A.,Moum,T.Bodtker,A.S.,& Ekeberg,O. (2004).Psychologicalimpact on women ofmiscarriage versus in-duced abortion:A 2-year follow-up study.Psychosomatic Medi-cine, 66,265-271.

(2) Broen,A.,Moum,T.Bodtker,A.S.,& Ekeberg,O. (2005a).The course ofmental health after mis-carriage and inducedabortion:A longitudinalfive-years follow-upstudy.BioMedCentralMedicine, 3, 18.

(3) Broen,A.,Moum,T.Bodtker,A.S.,& Ekeberg,O.(2006).Predictors ofanxiety and depressionfollowing pregnancytermination:A longitu-dinal five-year follow-up study.ActaObstetricia et Gyneco-logica, 85, 317-323.

Bradshaw,Z.,& Slade,P.(2005).The relation-ships between inducedabortion,attitudes to-wards sexuality andsexual problems.Sexualand RelationshipTher-apy, 20,391-406.

Sample &Design

Norway. Recruitedwomen (age 18-45) inhospital for inducedabortion (< 13 weeks;none due to fetal ab-normality) (N=80) ormiscarriage (< 17weeks). (N=40).Womenin both groups were in-terviewed 10 days (T1),6 months (T2), 2 years,(T3) and 5 years (T4)post event.91% of sam-ple retained over 5years.Data are reportedin 3 papers.

UnitedKingdom.Ninety-eightwomenattending apre-abortionmeeting ata clinic for a first-trimesterabortion askedabout at-titudes toward sex andsexual problems in the 2monthsprior to theirpregnancy andafterlearningof their preg-nancy. 44 responded tothe samequestionnaires2-monthspost-abortion.

ComparisonGroup

Comparison group ofwomen in hospital formiscarriage (<17 wks);N=40General Norwegianpopulation norms foranxiety and depres-sion (HADS).AB group had morechildren,were lesslikely to be married,more likely to be stu-dents, and had poorermental health thanmiscarriage groupprior to abortion ormiscarriage.Women'spsychiatric healthprior to pregnancy as-sessed post-event bycombined self-reportand diagnostic evalu-ation by interviewer.

Comparison group of51 women attendinga health center whohad been in a sexualrelationship over thelast 3 months,whowere not pregnant,and who had not hadan abortion in the last5 years completedsame questionnairesonce.

Primary Outcome

Stress reactions (Intru-sion and avoidance,as-sessed with Impact ofEvent Scale). Feelingsabout pregnancy termi-nation (7 items); anxietyand depression (Hospi-tal Anxiety and Depres-sion Scale-HADS).Subjective well-being(Quality of Life Scale).

Attitudes toward sex(Sexual Opinion Sur-vey); sexual problems(Go Lombok Rust In-ventory of Sexual Satis-faction - GRISS).

Key Findings

Miscarriagegroup (MIS)reportedmore IES intru-sion thanabortiongroup(AB) atT1only.AB re-portedmore IES avoid-ance atT1,T2,T3 andT4.Quality of life scores didnot differ betweenMISandABgroups and im-provedover the courseof the study.MISgroupreportedmore feelingsof grief atT1,T2,andT3,andmore feelings of lossatT1 andT2 thanABgroup. ABgroup re-portedmore relief andshameat all timepoints,andmoreguilt atT2,T3,andT4.HADS scores didnot differ betweenMISandABgroups at anytimepointwhenpoten-tial confounderswerecontrolled.ABgrouphadhigher anxiety thangen-eral populationnormsatall timepoints.Bothgroups scoredhigherthangeneral populationindepression atT1butnot atT3orT4.Recent lifeevents and formerpsy-chiatric healthwere im-portant predictors ofanxiety anddepressionamongABgroup.

Abortion group andcomparison group didnot differ in attitudestoward sex or sexualproblems (assessed ret-rospectively for abor-tion group).

Limitations

Low participation rate(47%).Comparisongroup (miscarriage)does not control for in-tendedness of preg-nancy. Small samplesizes. "Pre-pregnancy":psychiatric health as-sessed post abortion ormiscarriage.Abortion history retro-spectively self-reported.No information aboutresponse rate or repre-sentativeness of thesamples was provided.Small sample size.Com-parison group not ap-propriate. Nomeasuresof pre-pregnancymen-tal health.

Low recruitment rate(45%) and retentionrate (46%) in abortionsample. Inappropriatecomparison group.Nocomparisonsmade onpost-abortionmeas-ures.Women's retro-spective reports of theirsexual attitudes andproblems“pre-preg-nancy” are unreliable.




Citation

Conklin,M.P.,& O’Con-nor, B.P. (1995).Beliefsabout the fetus as amoderator of postabor-tion psychological well-being. Journal of Socialand Clinical Psychol-ogy,14, 76-95

Sample &Design

Canada.Participants re-cruited fromwaitingrooms of physicians' of-fices and asked to com-plete a questionnaire.153 out of 817whocompleted question-naire reported at leastone abortion.

ComparisonGroup

Six hundred and sixty-four women who re-ported no abortionhistory on question-naire.

Primary Outcome

Self-esteem (Rosenbergself-esteem scale); posi-tive and negative Affect(Positive and NegativeAffect schedule); lifesatisfaction (Satisfac-tion with Life Scale).Be-liefs about thehumanness of the fetus(7-item scale reliabilitynot provided).

Key Findings

Therewere nodiffer-ences on any outcomevariable betweenwomenwho reportedhaving an abortion andwomenwho reportedno abortion oncemaritalstatuswas controlled.Belief in the humannessof the fetusmoderatedresponses.Womenwhohad an abortion and at-tributed humanness tothe fetus had lower self-esteem,more negativeaffect,and lower life sat-isfaction thanwomenwho reported no abor-tion.Womenwhohadan abortion butwhodidnot attribute humanqualities to the fetus didnot differ on any out-come variable fromwomenwhodid nothave an abortion.

Limitations

Abortion history retrospec-tively self-reported.No in-formation about abortioncontext.No informationabout response rate orsample representativeness.Comparison groupnot ap-propriate. Nomeasures ofpre-pregnancymentalhealth.

with a history of psychosis (28.2 vs. 35.2).3 (3) Find-ings with regard to the outcome of deliberate self-harm (DSH) were mixed. Rates of DSH did notsignificantly differ for abortion versus deliverygroups among the categories with the highest DSHrates—women with a past history of psychosis (18.2vs. 19.3) or past history of DSH (8.4 vs. 13.5).Among women with no previous psychiatric history,however, DSH was significantly higher amongwomen who were refused an abortion (5.1) or whohad an abortion (3.0) compared with those who de-livered (1.8). Most DSH episodes (89%) were drugoverdoses; none were fatal. In sum, the authors con-cluded that, “Rates of total reported psychiatric dis-order were no higher after termination of pregnancythan after childbirth.” Further, they noted thatwomen with a history of previous psychiatric illnesswere most at risk, irrespective of the pregnancy out-come.

Evaluation of primary data comparison group studies.Conclusions that can be drawn from these studies arelimited by the methodological problems that charac-terize the vast majority. Below, we briefly summarizethe nature of these problems.

Sampling problems. Most of the studies had one ormore sampling problems. Most were based on smallsample sizes (fewer than 100 women). Many providedlittle or no information about the sample recruitmentstrategy, response rates, or sample representativenessor were based on a sample that clearly is not represen-tative of the population of women who obtain abor-tions (e.g., Reardon & Ney, 2000). Only six of thesestudies were conducted in the United States, raisingconcerns about generalizability. The rest were con-ducted in Canada (3), the United Kingdom (3), Nor-way (1), Germany (1), Israel (1), and Brazil (1). Theabortion regulations and sociocultural context of



Citation

Gilchrist ,A.C.,Han-naford, P.C., Frank,P.,&Kay,C.R. (1995).Termi-nation of pregnancyand psychiatric morbid-ity. British Journal of Psy-chiatry, 167,243-248.

Houston,H.,& Jacob-son, L. (1996).Overdoseand termination ofpregnancy:an impor-tant association? BritishJournal of General Prac-tice, 46,737-738.

Sample &Design

United Kingdom.Prospective cohortstudy of 13,261 womenwith unplanned preg-nancies. One-thousandfive-hundred and ninevolunteer GPs asked torecruit all womenwhorequested a termina-tion of a pregnancy anda comparison group ofwomenwho did not re-quest termination butwhose pregnancy wasunplanned.Womenwere enrolled between1976 and 1979 andwere followed every 6months until they leftthe study or end ofstudy (1987).Final sam-ple consisted of 6410who obtained termina-tion.

United Kingdom.Au-thors examined all med-ical records of femalepatients aged 15-34years inclusivewithintheir practice in 1994 toexaminewhether therewas an association be-tween drug overdoseand induced termina-tion of a pregnancy (ex-cluding pregnancy forfetal abnormality orma-ternal illness).

ComparisonGroup

Comparison groupsincluded 6151 womenwho did not seek ter-mination, 379 whorequested termina-tion but were denied,and 321 who re-quested terminationbut changed mind.For purposes of analy-ses, each comparisongroup was dividedinto four subgroupsaccording to severityof previous psychi-atric history (assessedat study recruitment):psychosis, nonpsy-chotic illness, deliber-ate self-harm alone,and no psychiatric ill-ness or self-harm.Data also standard-ized (i.e. covariate ad-justment) for age,marital status, smok-ing, education level,gravidity and priorhistory of abortion.

Out of 1359 patients,163 (12%) had anabortion history, and47 (3.5%) had a his-tory of a deliberateoverdose. Fifteenwomen had a historyof both events.

Primary Outcome

Psychiatric morbiditycoded by GP using ICD-8 diagnostic categories:psychoses;nonpsy-chotic illnesses (depres-sion, anxiety), andepisodes of deliberateself-harm (DSH)

Drug overdose requir-ing hospital treatment(excluding accidentaloverdose).

Key Findings

Inwomenwith equiva-lent past psychiatric his-tories, therewerenosignificant differencesbetween the compari-songroups inoverallrates of psychiatric ill-ness. Risk of psychotic ill-ness and risk ofnonpsychotic illnessesdidnot differ betweentermination andnonter-minationgroups.RatesofDSHdidnotdiffer bypregnancyoutcomeamongwomenwith apast history of psychosisorDSH.Amongwomenwithnopreviouspsychi-atric history,DSHwashigher amongwomenwhohadanabortionorwhowere refused anabortion.Conclusion:“Rates of total reportedpsychiatric disorderwerenohigher after termina-tionof pregnancy thanafter childbirth.”Abortion group andcomparison group didnot differ in attitudestoward sex or sexualproblems (assessed ret-rospectively for abor-tion group).

The association be-tween overdose andterminationwas signifi-cant. More terminationstended to follow over-dose than the reverse.

Limitations

Analyses did not differ-entiate between termi-nations carried out at <12 weeks (85%) vs.over12 weeks (15%) gesta-tion. Sampling by GPrecruitmentmay haveled to nonrepresenta-tive sample.GPsmayunderrecognize or im-precisely diagnose psy-chiatric disorder.

No details known aboutcontext of abortion,reasons for termination,marital status or othercharacteristics ofwomen.Representa-tiveness of sample un-known. Presence ofsignificant associationdoes not establish cau-sation. Nomeasures ofpre-pregnancymentalhealth.


Table 3B:Primary Data Comparison Group StudiesNON-U.S.SAMPLES (continued)

Citation

Lauzon,P., Roger-Achim,D.,Achim,A.,& Boyer,R.(2000).Emotional dis-tress among couples in-volved in first-trimesterinduced abortions.Canadian Family Physi-cian, 46,2033-2040.

Ney,P.G., Fung,T.,Wick-ett, A.R.,& Beaman-Dodd,C. (1994).Theeffects of pregnancyloss on women’s health.Social Science & Medi-cine, 38,1193-1200.

Sample &Design

Canada.Recruitedwomenhaving a 1sttrimester abortion atone of 3 public abortionclinics. Excluded ifunder 15 years of age orpregnancy result of rapeor incest.197womencompleted question-naires prior to abortion.127 completed ques-tionnaires 1-3weekspostabortion.

Canada.Asked 238 fam-ily physicians to handout questionnaires tothe first 30women ofchild bearing agewhowalked into their officesin a givenweek,69physicans provided us-able questionnairesfrom1428women.Womenwere askedquestions about theirhealth, family life,enjoy-ment of being a parent,the supportiveness oftheir partner,and theoutcomes of up to ninepregnancies.

ComparisonGroup

Comparison group of728 women (aged 15-35 years) who hadtaken part in a previ-ous public health sur-vey and completedsame outcomemeas-ure. Compared tocontrol group, abor-tion group was signifi-cantly younger, lesseducated, less likely tobe living with aspouse, less likely tohave children,morelikely to be students,more likely to be di-vorced, separated orsingle, and more likelyto have had suicidalideation or suicide at-tempts prior to theabortion.Abortionhistory unspecified.

The number ofwomen who reportedvarious pregnancyoutcomes (e.g., thosewho reported abor-tions, still births, infantdeaths, full-termbirths, prematurebirths, etc) was notprovided.

Primary Outcome

Psychological distress.(Ilfeld Psychiatric Symp-tom).

Women’s reports that“My health is not good.”

Key Findings

Before the abortion,56.9%ofwomenweremore distressed thancomparison group.Threeweeks after abor-tion, 41.7%ofwomenmore distressed thancomparison group.Pre-dictors of distress priorto abortionwere pasthistory of suicidalideation,fear of negativeeffects on relationship,unsatisfactory relation-ship, andnopreviouschild.

Results of a number ofpoorly specified analy-ses appear to show thatperceptions of an un-supportive partner,number of abortionsandnumber ofmiscar-riageswere positivelycorrelatedwithwomen’sreports that“Mypresenthealth is not good.”Ofthese,perceptions of anunsupportive partnerweremost strongly re-lated to self-reportedhealth.The number ofstill births or infantdeathswas not relatedto self-reported health.

Limitations

Sample representative-ness unknown.One thirdof subjects lost to attrition.Very short follow-up pe-riod. Comparison groupinappropriate.Abortiongroup differed from com-parison group inways thatmay fully account for anydifferences observed postabortion. No significancetests reported for differ-ences between abortionand comparison group.Nomeasures of pre-preg-nancymental health.

Abortion history retro-spectively self-reported.No information providedabout response rate orrepresentativeness ofsample.Methods,meas-ures, and analyses wereparticularly poorly speci-fied, making it impossibleto tell exactly what wasmeasured.No reliabilitieswere reported for anymeasure.Single item de-pendentmeasure notvalid indicator of health.Nomeasures of prepreg-nancymental health.


Table 3B:Primary Data Comparison Group StudiesNON-U.S.SAMPLES (continued)

Citation

Teichman,Y., Shenhar, S.,& Segal, S. (1993).Emo-tional distress in Israeliwomen before andafter abortion.AmericanJournal of Orthopsychi-atry, 63,277-288.

Sample &Design

Israel.Seventy-sevenwomen requesting legalabortion compared topregnantwomen andnonpregnantwomenprior to their abortions.Only 17women in abor-tion group agreed toparticipate at 3-monthpostabortion follow-up

ComparisonGroup

Two comparisongroups: 32 womenwho were in the 40thweek of pregnancyand 45 nonpregnantwomen who be-longed to the samecommunity and wererecruited throughchild care center orworkplaces.

Primary Outcome

State and trait anxiety(STAI); depression (De-pression AdjectiveCheck List).

Key Findings

Prior to the abortion,abortion grouphadhigher anxiety andde-pression than compari-son groups.Nocomparisons betweengroups onpost-abortionmeasures.

Limitations

No comparisons on post-abortionmeasures.Verysmall (N=17) postabortionsample. Initial sample re-sponse rate and represen-tativeness unknown.Comparison groups donot control for unin-tended pregnancy.Differ-ent regulations forobtaining abortion in Is-rael make generalizationto US inappropriate. In Is-rael, womenmust go be-fore a committee to getapproval for abortion.Anxiety and depressionwere assessed just prior tothis (likely stressful) com-mittee appearance.Nomeasures of pre-preg-nancymental health.


abortion in some of these countries differ in importantways from those of the United States. For example,in some countries where abortion is legal, such asBritain, all abortions must be approved by two physi-cians, usually on grounds that continuation of a preg-nancy involves greater risk to the woman’s physical ormental health than does termination (although suchrequirements may be more of a formality than a bar-rier).4 Another example is Brazil, where induced abor-tion is illegal, except in cases where the pregnancy isdangerous to the mother’s health or resulted from rapeor incest. Caution must be exercised in drawing con-clusions about the responses of women in the UnitedStates based on data collected on non-U.S. samples.

Inappropriate comparison groups. With two exceptions(Cohan et al., 1993; Gilchrist et al., 1995), none ofthese studies used a comparison group that controlledfor the occurrence of an unintended or unwantedpregnancy, and hence was able to adequately addressthe question of relative risk. Comparison groups used

included women who reported never being pregnant(Felton, Parsons, Hassell, 1998), women who werecurrently pregnant (Bailey et al., 2001; Lydon et al.,1996; Medora et al., 1993; Teichman, Shenhar, &Segal, 1993), women who were not currently pregnant(Bradshaw & Slade, 2005; Teichman et al., 1993),women who reported no elective abortions (Conklin& O’Conner, 1995; Medora et al., 1993; Reardon &Ney, 2000; Williams, 2001), women who had miscar-ried (Bailey et al., 2001; Broen et al., 2004, 2005a,2006), women who had participated in a previouspublic health survey (Lauzon, Roger-Achim, Achim,& Boyer 2000), and women matched on demographicvariables (Barnett, Freundenburg, & Wille, 1992).

Co-occurring risk factors. Just as important as the lackof appropriate comparison groups in this set of stud-ies was the absence of measures of mental health andother variables prior to the pregnancy or abortionlikely to be related to the outcome studied (e.g., co-occurring risk factors such as prior engagement in

problem behaviors). Hence, any between-group dif-ferences observed post abortion may reflect between-group differences present prior to the pregnancyand/or abortion. With one exception (Gilchrist et al.,1995), none of the studies had adequate measures ofpreabortion mental health, and thus none could sepa-rate problems observed post abortion from thosepresent prepregnancy. Furthermore, few of the studiescontrolled for important covariates, such as age, mar-ital status, number of children, race, education, andduration of partnership that might be related to out-come variables independently of abortion history.

Measurement problems. In six of the papers, the keyevent—abortion—was determined from retrospectiveself-report, with no checks on accuracy of report-ing,and no information on how long since the abor-tion occurred, whether the pregnancy was wanted ornot, whether the abortion was first or secondtrimester, or what the age of the woman was at thetime of the abortion (Conklin & O’Conner, 1995; Fel-ton et al., 1998; Medora et al., 1993; Ney, Fung,Wickett, Beaman-Dodd, 1994; Reardon & Ney, 2000;Williams, 2001). As noted above, retrospective self-re-ports are notoriously unreliable and subject to bias,rendering conclusions of these six papers particularlyuntrustworthy. In studies where abortion was verified,mental health outcomes were often assessed withinonly a few weeks or months after the abortion. Onlytwo studies assessed mental health outcomes morethan a year post abortion (Broen et al., 2006; Gilchristet al., 1995).

In several cases a single item of unknown reliabilitywas used as a measure of mental health (Ney et al.,1994; Reardon & Ney, 2000). Only one study as-sessed clinically significant outcomes, that is, whetherparticipants met diagnostic levels for psychologicaldisorder or had sought psychiatric treatment (Gilchristet al., 1995). The remainder focused on a variety ofmental health-related outcomes, including self-esteem,positive and negative affect, decision satisfaction, lifesatisfaction, self-reported health-promoting behaviors,relationship quality, sexual attitudes and problems,grief, anxiety or depressive symptoms, and stress re-sponses.

Statistical problems. Some of the studies report numer-ous analyses capitalizing on chance (e.g., Reardon &Ney, 2000), some used small sample sizes lacking suf-

ficient power to detect potentially meaningful differ-ences (e.g., Cohan et al., 1993), some did not reportsample sizes at all (Ney et al., 1994), and some re-ported no statistical tests of comparisons on postabor-tion measures but discussed results as if they had (e.g.,Lauzon et al., 2000).

StudiesofAbortion forReasonsofFetalAbnormalityAll of the studies reviewed above either were re-stricted to samples of women undergoing first-trimester abortions or did not differentiatefirst-trimester from later-trimester abortions. Al-though the vast majority of abortions in the UnitedStates are of unplanned pregnancies that are eithermistimed or unwanted (Finer & Henshaw, 2006a),and they occur in the first trimester (Boonstra et al.,2006), the increasing accessibility and use of ultra-sound technology and other prenatal screening tech-niques has increased the likelihood of prenataldiagnosis of fetal anomalies, often in the second andsometimes even in the third trimester. Following sucha diagnosis, many couples elect to terminate theirpregnancy, especially when informed that the fetalanomaly is lethal or severely disabling (see Statham,2002, for a review of research in this area).

Abortion under these circumstances is a very differentphysical and psychological event than an abortion ofan unplanned or unwanted pregnancy. Not only doesabortion for reasons of fetal anomaly typically occurlater in pregnancy, but more importantly, it usuallyoccurs in the context of a pregnancy that was initiallyplanned and wanted. Consequently, the meaning andsignificance of the pregnancy and abortion are apt tobe quite different, as is the extent of loss experienced.Understanding women’s psychological experiencesfollowing an abortion for fetal anomaly is important.Some authors have speculated that women may feelmore responsible for the death of their child whenthey make an active decision to terminate theirpregnancy, leading to more negative long-term psy-chological sequelae compared with experiencingspontaneous miscarriage or perinatal loss (Salvesen,Oyen, Schmidt, Malt, & Eik-Nes, 1997). A full un-derstanding of this issue requires comparing re-sponses of women who undergo induced terminationof a pregnancy due to fetal anomaly to responses ofwomen who experience a miscarriage of a wantedpregnancy in the second or third trimester or experi-ence a neonatal loss (e.g., a stillbirth or death of a



newborn) or deliver a child with severe physical ormental disabilities.

Our literature search identified six studies in whichwomen who terminated an initially wanted pregnancybecause of fetal anomaly were compared with anothergroup of women. Five were based on non U.S. sam-ples. These studies are summarized in Table 4. We alsoidentified one U.S. study that examined psychologicalexperiences among women who terminated an initiallywanted pregnancy due to fetal anomaly, but the studydid not include a contrast group. Findings of thisstudy are summarized in Table 5.

Description of findings. Zeanah, Dailey, Rosenblatt,and Saller (1993) compared grief and depressionscores of 23 women in the United States who under-went induced termination of a wanted pregnancy be-cause of fetal anomalies to 23 demographicallymatched women who experienced spontaneous peri-natal losses (stillbirth or death of a newborn infant).Controlling for age, there were no significant differ-ences between the induced and spontaneous lossgroups in grief, difficulty coping, despair or depres-sion 2 months post abortion, or post spontaneousperinatal loss.

Lorenzen and Holzgreve (1995) compared grief reac-tions of 35 women in Germany who terminated apregnancy due to fetal anomalies and 15 women whoexperienced a spontaneous second- or third-trimestermiscarriage. Eight weeks post event, women who hadterminated their pregnancy expressed significantly lessgrief than those who had a spontaneous child loss.

Iles and Gath (1993) compared psychiatric distur-bance and grief among 71 women who underwent sec-ond-trimester abortion for reasons of fetal anomaly to26 women who had a second-trimester spontaneousmiscarriage. There were no significant differences inpsychiatric disturbance (determined by interviews witha trained psychiatrist) between the termination andmiscarriage groups or differences in grief between thetwo groups 4-6 weeks or 13 months post loss. Somesigns of normal grief persisted for a full year in somewomen in both groups.

Kersting et al. (2005) compared stress responses ofthree groups of women in Germany—83 women whohad had an induced late-trimester abortion for reasons

of fetal anomaly 2-7 years previously, 60 women whohad a late-trimester abortion for fetal anomaly 14days earlier, and 65 women who delivered a healthychild (time since delivery and abortion history unspec-ified). Women who delivered a healthy baby hadlower stress scores (assessed with the Impact of Eventsscale-IES) than women who had a late-term abortionfor fetal anomaly, regardless of whether the abortionoccurred 14 days or 2-7 years previously. The twoabortion groups did not differ in their grief responses.While 88% of the women in the abortion group be-lieved they had made the right decision, 9.6% ex-pressed doubts about their decision, and one womanfelt she had made the wrong decision.

Salvesen et al. (1997) compared depression, generalhealth, stress reactions, and anxiety of 24 women inNorway who terminated a pregnancy for fetal anom-aly to 29 Norwegian women who experienced a peri-natal death or late-trimester spontaneous miscarriage.Immediately after the event, both groups of womenreported high intrusion scores on the IES, but theperinatal loss group reported significantly higher de-pressed affect and had higher scores on the intrusionand avoidance scales of the IES than did the inducedtermination group. At later assessments, including at1 year post abortion, there were no significant differ-ences between the two groups. One woman out of 36exhibited symptoms of traumatic stress; she was inthe perinatal loss group.

Rona, Smeeton, Beech, Barnett, and Sharland (1998)compared depression and anxiety (assessed with theHospital Anxiety and Depression (HAD) scale) ofthree groups of women in the United Kingdom.One group consisted of 28 women who received aconfirmed diagnosis during their second trimester ofa severe fetal heart malformation and terminated thepregnancy. A second group consisted of 40 womenin whom a fetal heart malformation was initially di-agnosed but later disconfirmed by a specialist. Athird group consisted of 40 women whose fetal mal-formation was not identified and who had givenbirth to an infant with a severe heart malformation.The HAD scale was administered 6-10 months afterthe heart malformation was initially diagnosed orpost delivery in the latter group. Based on cutoffscores on the HAD (> 11), a significantly greaterproportion of mothers who had an infant with a se-vere heart malformation reported clinical levels of

Table 4:Abortion for Reasons of Fetal Anomaly

Citation

Iles, S.& Gath.D. (1993).Psychiatric outcome oftermination of preg-nancy for foetal abnor-mality. PsychologicalMedicine,23,407-413.

Sample andDesign

United Kingdom.Womenwith secondtrimester abortion forfetal abnormality (ABgroup) recruited fromhospitals (ave.age 30.7years).77%of pregnan-cies planned.86%par-ticipation rate.Interviewed by psychia-trist three times:4-6weeks post- (T1,N=71),6months post- (T2,N=65),and 13monthspost- (T3,N=61) termi-nation.

ComparisonGroup

Twenty-six womenwith second trimestermiscarriage (MISgroup; ave age 30.3years) interviewed atsame three timepoints (84% participa-tion rate) 77% ofpregnancies planned.Also compared ABand MIS groups to di-agnostic norms fornon-puerperalwomen and 12 monthpost-partum women

Primary Outcome

Intensity of psychiatricdisturbance (PSE Indexof Definition (ID)), es-tablished via interviewswith trained psychia-trist at three timepoints. ID levels of 5 orabove indicate a psychi-atric“case.“Grief alsoassessed via interview.

Key Findings

No significant differ-ences betweenAB andMIS groups in psychi-atric disturbance atT1,T2,orT3.AtT1 bothgroups showed consid-erable psychiatricmor-bidity and impairmentof social adjustment rel-ative to the normingsamples of the instru-ments. ByT2 andT3,psy-chiatricmorbiditywasnear norms in bothgroups.Nodifferences ingrief between theABandMIS groups atT1andT4.Some signs ofnormal grief persistedfor a full year in somewomen in both groups.

Limitations

Small sample sizes.Sam-ple representativenessunknown.Abortion forfetal abnormality nottypical of most abor-tions. Nomeasures ofprepregnancymentalhealth.


anxiety (43%) and depression (18%) compared towomen in the other two groups. Among those whohad terminated their pregnancy, 32% were catego-rized as anxious, and 4% as clinically depressed.Among mothers whose initial diagnosis of fetal ab-normality was later disconfirmed, the comparablepercentages were 15% (anxiety) and 5% (depres-sion). Women who had terminated their pregnancywere more anxious than this latter group of womenwho had delivered healthy infants. The authors at-tributed the higher anxiety in the termination groupthan the latter group to either the experience of ther-apeutic abortion or to a fear of a subsequent abnor-mal pregnancy. Younger age was associated withhigher anxiety.

Evaluation of fetal abnormality studies. All of theabove studies are limited by high attrition rates, typi-cally low response rates, and extremely small samplesizes. The small sample sizes restrict power, and, hence,the ability of these studies to detect significant differ-ences between groups. In most studies, the sample also

was of unknown representativeness. Despite thesemethodological limitations, these studies tell a fairlyconsistent story. Women’s levels of negative psychologi-cal experiences subsequent to a second-trimester abor-tion of a wanted pregnancy for fetal anomalies werehigher than those of women who delivered a healthychild (Kersting et al., 2005; Rona et al., 1998) andcomparable to that of women who experienced asecond-trimester miscarriage (Iles & Gath, 1993),stillbirth, or death of a newborn (Salveson et al., 1997;Zeanah et al., 1993). There was no evidence, however,that induced termination was associated with greaterdistress than spontaneous miscarriage or perinatal loss.Indeed, the one difference observed was that womenwho terminated a pregnancy because of fetal anomalyexperienced significantly less grief than women whomiscarried 8 weeks post loss (Lorenzen & Holzgreve,1995). Nonetheless, grief among both groups was highand appears to persist for some time. The one studythat compared the mental health of women who termi-nated a pregnancy for fetal abnormality and womenwho delivered an infant with a severe abnormality


Table 4:Abortion for Reasons of Fetal Anomaly (continued)

Citation

Kersting,A.,Dorsch,M.,Kreulich,C., Reutemann,M.,Ohrmann,P., Baez,E.,& Aroldt,V. (2005).Trauma and grief 2-7years after terminationof pregnancy becauseof fetal anomalies—apilot study. Journal ofPsychosomatic Obstet-rics &Gynecology,26,9-14.

Lorenzen & Holzgreve(1995),Helping parentsto grieve after secondtrimester termination ofpregnancy for feto-pathic reasons.FetalDiagnostic Therapy,10,147-156.

Sample andDesign

Germany.Recruited atDept of Gyn &Obstet-rics.Womenwho hadlate trimester abortions(15-33weeks gestation)for fetal abnormality.83responded tomailedquestionnaire 4 yearspost abortion (ave.age31 years),49% responserate.60women com-pleted questionnaires14 days post abortion(ave.age 34 years).Re-sponse rate not pro-vided.

Germany.Comparedgrief reactions of 35womenwho terminateda pregnancy for fetal ab-normality (65% re-sponse rate) to 15women after the spon-taneous loss of a childbetween the 12th and24thweek of gestation(60% response rate).Atthe time of the termina-tion ormiscarriage,allwomen had been en-couraged by hospitalpersonnel tomakethe lost baby a tangibleperson.

ComparisonGroup

Sixty-five women whohad delivered ahealthy child (timesince delivery notspecified) (averageage 32 years)

Fifteen women expe-riencing the sponta-neous loss of a childbetween the 12th and24th week of gesta-tion (60% responserate). There were nosig diff between thetwo groups in age,marital status, or pre-vious child losses.

Primary Outcome

Stress reactions (avoid-ance, intrusion,hyper-arousal, assessed withImpact of Events scale).Grief (Perinatal Griefscale) and Decision sat-isfaction (terminationgroups only).

Both groups completedthe Perinatal Grief scalein response to amailedquestionnaire an aver-age of 8 weeks after theloss of the child.

Key Findings

Womenwho had a late-term abortion for fetalabnormality scoredhigher than those whodelivered a healthybaby on the IES (bothoverall, and on all threesubscales), regardless ofwhether they had ter-minated their preg-nancy 14 days earlier or2-7 years earlier.Thetwo abortion groupsdid not differ in grief re-sponses, except thatthe womenwho hadthe abortionmore re-cently scored higher onfear of loss. 87.9% ofabortion group be-lieved (very strongly tofairly strongly) that theyhadmade the right de-cision; 9.6% expresseddoubts about their de-cision, and one womanfelt she hadmade thewrong decision.

Womenwho experi-enced a spontaneouschild loss expressed sig-nificantly more griefthan those having un-dergone termination 8weeks post child loss.Themajority of womenwho terminated due tofetal abnormality wereconvinced of the right-ness of their decisionand said they wouldagain vote for termina-tion in a similar situa-tion.

Limitations

Sample representative-ness unknown.Low re-sponse rate,or responserate unknown.Compari-son group (delivery ofhealthy child) not appro-priate. Abortion forfetal abnormality nottypical of most abor-tions. Nomeasures ofpre-pregnancymentalhealth.

Very small sample sizesof unknown representa-tiveness. Short follow-upinterval.


Citation

Rona,R. J., Smeeton,N.C.,Beech,R., Barnett,A.,& Sharland,G. (1998).Anxiety and depressioninmothers related toseveremalformation ofthe heart of the childand fetus. Acta Paedi-atrica, 87,201-205.

Sample andDesign

United Kingdom.Com-pared depression andanxiety 6-10monthspost termination ofthree groups of women.GroupA consisted of 28womenwho terminateda pregnancy during thesecond trimester due tosevere fetal heartmal-formation 6-10monthsposttermination.

ComparisonGroup

40 women referred tofetal cardiology inwhom a fetal heartmalformation wassuspected but laterdisconfirmed (GroupB), and 40 womenwhose fetal heartmalformation wasnot diagnosed duringpregnancy, andwho gave birth to achild with a severeheart malformation(Group C).

Primary Outcome

Anxiety and depressionassessed with theHospital Anxiety andDepression (HAD) scale.A score of 11 or moreindicates probablepresence of clinical anx-iety or depression.HADscale administered bymailed questionnaire6-10months after initialdiagnosis of a heartmalformation or 6-10months post deliveryof a child with severeheart malformation.67.5% response rate.

Key Findings

Greater percent ofmothers of infantswithsevere heartmalforma-tion (GroupC) had clini-cal levels of anxiety(43%) anddepression(18%) compared towomen inGroupAwhohad terminated for fetalanomaly (anxious = 32%;depressed=4%) orGroupBwhose initial di-agnosiswas later discon-firmed (anxious = 15%;depressed= 5%).Women inGroupsA andCwere significantlymore anxious thanwomen inGroupB.Younger agewas associ-atedwith higher anxiety.Authors attributed highanxiety inGroupA to ei-ther the experience oftherapeutic abortion orto fear of a subsequentabnormal pregnancy.

Limitations

Small sample sizes.Sam-ple representativenessunknown.Abortion forfetal abnormality nottypical of most abor-tions. Nomeasures ofpre-pregnancymentalhealth.


found that 6-10 months post event, a greater propor-tion of women in the delivery group reported clinicallysignificant anxiety and depression compared to womenin the abortion group.

REVIEWOFABORTION-ONLY STUDIES

In addition to the primary research reviewed above,our literature search also identified a set of papers thatmet all inclusion criteria except that they did not in-clude a comparison group. Studies without a compari-son group are not appropriate for addressingquestions of relative risk. However, studies focusedsolely on reactions and feelings of women who have

had an abortion can be useful for identifying factorsthat predict individual variation in women’s psycho-logical experiences following abortion. Furthermore,they can potentially address questions related to theprevalence of harm associated with abortion to theextent that their sample is representative of the popu-lation to which one wants to generalize. Becausedifferences between the United States and other coun-tries in cultural contexts surrounding abortion andabortion regulations make generalization from non-U.S. samples to U.S. women problematic, the TFMHAreviewed only those noncomparison group studies thatmet inclusion criteria that were based on U.S. samples.

The TFMHA identified 23 published papers thatwere based solely on samples of women who hadabortions in the United States, but that otherwise



Citation

Salveson,K.A.,Oyen,L.,Schmidt,N.,Malt,U.F.,&Eik-Nes,S.H. (1997).Comparison of long-term psychological re-sponses of women afterpregnancy terminationdue to fetal anomaliesand after perinatal loss.UltrasoundObstetrics &Gynecology,9,80-85.

Zeanah,C.H.,Dailey, J.,Rosenblatt,M.,& Saller,D.N. (1993).Do womengrieve after terminatingpregnancies because offetal anomalies? A con-trolled investigation.Obstetrics andGynecol-ogy, 82,270-275.

Sample andDesign

Norway.Compared de-pression, general health,stress reactions,andanxiety of 24womenwho terminated a preg-nancy for fetal anomaly(< 24wks gestation) to29womenwho experi-enced perinatal loss(82% response rate). In-terviewed day of or sev-eral days after event andsentmailed question-naires 7weeks,5months,and 1 year postevent.

U.S.23 of 36womenwho underwent in-duced termination ofwanted pregnancies forfetal anomalies (ave age31.4 years) interviewed2months post termina-tion. (64% responserate).

ComparisonGroup

Twenty-nine womenexperiencing latespontaneous miscar-riage (16-27 wks preg-nancy) or perinataldeath (death of a liveborn child within 7days after birth or stillbirth after 28 wkspregnancy). Abortionand perinatal lossgroups similar in par-ity, age, education,%nulliparous and psychhealth in 2 weeks pre-ceding event (as-sessed retrospectivelywith GHQ).

23 womenmatcheddemographically (so-cial class, education,number of children,age,gestational age atloss) who experiencedspontaneous perinatalloss (stillbirth or deathof newborn infant) in-terviewed 2monthspost loss.Comparisongroupwas signifi-cantly younger (aveage 27.2 years) thantermination group,and gestational agewas greater.Age wasinversely related togrief.

Primary Outcome

Depression (Mont-gomery & Ashberg De-pression Rating scale),anxiety (State-Trait Anx-iety Inventory), andstress responses (Im-pact of Events scale-IES-avoidance and intru-sion subscales).Gold-berg General HealthQuestionnaire (GHQ)used to retrospectivelyassess women’s psycho-logical health in the 2wks preceding event.Schedule for RecentLife Events used to con-trol for other life eventsthat might influencegrief response.Time 1measures given by in-terviewer; remainingmeasures sent bymailed questionnaire.Made diagnosis of post-traumatic stress disor-der based onmultiplecriteria.

Grief,difficulty coping,and despair (PerinatalGrief Inventory).De-pression (Beck Depres-sion Inventory). Clinicaldiagnosis by psychiatricevaluation (terminationgroup only).

Key Findings

Immediately post-event,both groups reportedhigh intrusion scores onIES,but abortion groupshowed less depression,andhad lower scores onintrusion and avoidancescales of IES thanperina-tal loss group.Therewere no significant dif-ferences betweenABandperinatal lossgroups on IES intrusionor avoidance scores,anx-iety, general health(GHQ),or depression atsubsequent assess-ments (7wks,5months,or 1 year post event).At1 year postevent onewoman (1/36 or 3%)met criteria for PTSD.Shewas in perinatal lossgroup.

Controlling for age,therewere no significant dif-ferences between thetermination and sponta-neous perinatal lossgroups in grief,difficultycoping,despair,or de-pression. Psychiatricevaluation of termina-tion group 2monthspost revealed that 74%reported theywere stillgrieving,17%met crite-ria formajor depression,and 23%had soughtpsychiatric help.Only 1regretted her decision.

Limitations

Strong aspects of studyinclude use of psycho-metrically valid meas-ures and comparabilityof AB and comparisongroups.Major limitationis extremely small sam-ple sizes.

Extremely small samplesizes.Short follow-up in-terval. No comparisonsof termination andspontaneous loss groupon psychiatric evalua-tion. Thirty-six percentnonparticipation rate intermination group.Nomeasures of pre-preg-nancymental health.


Table 5:U.S.Samples of Abortion Group(s) Only/No Comparison

Prospective Analyses UsingMajor et al.Multiple Site SampleGeneral Description:Women followed for 2 years after an elective first-trimester abortion for an unintended pregnancy, recruited from 3 sites in Buf-falo, NY in 1993. Four assessments: 1 hour before abortion, and 1 hour, 1 month, and 2 years after the abortion; 85% (N = 882) of eligible women agreed toparticipate, completing preabortion and 1 hour postabortion questionnaires; follow-up questionnaires were completed 1 month (N= 615) and 2 years(N= 442) post abortion.The age range was 14-60; 65% wereWhite/other.

Limitations Common to All Studies Based on this Data Set: Common to All Studies Based on this Data Set:No comparison group (not a lim-itation for majority of studies which examined risk factors,mediators, and moderators of post-abortion psychological distress). Sample may not be repre-sentative of women who obtain abortions in the U.S., although only sociodemographic difference from national comparison sample wasunderrepresentation of Hispanic women.High attrition: 30% at 1 month and 50% at 2-year follow-up, but women retained did not differ from women lostto follow-up at either time point on demographic or psychological measures.Does not include measures such as domestic violence and sexual abuse thatmay be related to post-abortion adjustment.

Citation

Major,B.,Cozzarelli,C.,Cooper M.L., et al.(2000).Psychological re-sponses of women afterfirst-trimester abortion.Archives of General Psy-chiatry, 557, 777-784.


Sample consisted of thetotal 442Women fol-lowed for 2 years afterabortion. This is theonly study whoseanalysis used data fromall 4 time points.

Controls/Covariates

In one simultaneousregression analysis,demographic charac-teristics, prior mentalhealth and self re-ports of physical com-plications werecontrolled. (Note:controls not requiredfor most analyses.)

Primary Outcome

Measures include BriefSymptom Inventory,modified DiagnosticInterview Schedule,4-item Rosenberg Self-Esteem Inventory,adapted PTSD scale,emotional reactions,satisfaction withdecision,appraisal ofabortion-related harm.

Results

Most womenwere sat-isfied with their deci-sion (78.7% at 1month)although decision satis-faction decreased overtime (72% satisfied at 2years).Most women feltmore benefit than harmfrom abortion decisionand this did not changeover time.Negativeemotions increased,and positive emotionsdecreased over timebutmost women feltmore relief than eitherpositive or negativeemotions.Depressionlower and self-esteemhigher 2 years post-abortion than pre-abor-tion. Depression ratewas similar to rates inthe general populationfor women in this agegroup.

AdditionalLimitations

Harm and regret arenon-standardizedmeasures, and difficultto interpret with nocomparison group.Cannot use findings toexamine prevalence ofpsychiatric outcomesassociated with abor-tion nationally.


met inclusion criteria. These studies are summarizedin Table 5. The studies were of two major types: (1)prospective or concurrent studies that usually in-cluded preabortion measures of psychological adjust-ment and risk factors and one or more postabortionassessments of adjustment, and (2) retrospectivestudies that assessed women’s perceived reactionsto the event and current level of psychological func-

tioning several years after the abortion. The formerprovide a wealth of information on predictors ofpostabortion psychological functioning. The retro-spective studies—although supporting many of theconclusions of research prior to 1990—have seriousmethodological problems that negate their ability toanswer questions about psychological experiencesfollowing abortion.


Table 5:U.S.Samples of Abortion Group(s) Only/No Comparison(continued)

Citation

Major B.,& Gramzow,R.(1999).Abortion asstigma:cognitive andemotional implicationsof concealment. Journalof Personality and SocialPsychology,77, 735-745.

Cozzarelli, C.,Major,B.,Karrasch,A.,& Fuegen,K. (2000).Women’s ex-periences of and reac-tions to anti-abortionpicketing.Basic and Ap-plied Social Psychology,25, 265-275.


442Women followed for2 years after an electivefirst trimester abortion.

442women followed for2 years after an electivefirst trimester abortion.

Controls/Covariates

Positive and negativeaffectivity, personalconflict over abortion,demographic vari-ables of age, race,number of prior livebirths,Medicaid sta-tus.

Correlations betweenmodel variables anddemographic vari-ables and negative af-fectivity (NA) wereexamined.Only ageand NA were relatedto more than one ofmodel variables.When model wasrerun with controlvariables added, re-sults were similar.

Primary Outcome

Pre-abortion and 2 yearpost abortion distressmeasured by the BriefSymptom inventory.

Depression assessedusing the 7-item de-pression subscale of theBrief Symptom Inven-tory about one hourpost abortion in thedelivery room and 2years postabortion atfollow-up.

Results

Average levels of psy-chological distress 2years post abortionwerelow,and lower than av-erage pre-abortion dis-tress. 2 years postabortion,47%ofwomenagreedor stronglyagreed that they felttheywould be stigma-tized if others knewabout the abortion.44.9% felt need to keepabortion a secret.Con-cealing stigmawas asso-ciatedwithmoreresidualized distress,viaincreased thought sup-pression anddecreasedemotional disclosure.

Feeling guilty in re-sponse to seeingpick-eters andhaving highpersonal conflict aboutabortion predicted im-mediate postabortiondepression,whereasfeeling angry was unre-lated to postabortiondepression.Althoughguilt andpersonal con-flict hadnodirect effectsondepression 2-yearpost abortion,depres-sion at the two timepointswas correlated.The authors concludethatwomen’s encoun-terswith picketers evokeshort-termnegativeemotional reactions butdonot have long-termnegative psychologicaleffects.


Non-standardizedmeas-ure of emotional reac-tions to picketing;noobjective (coders) re-ports of picketing activ-ity. Single measure ofpostabortion adjust-ment. No pre-abortionmeasure of depression.


Citation

QuintonW.J.,Major B.,&Richards C. (2001).Ado-lescents and adjust-ment to abortion:Areminors at greater risk?Psychology,Public Policyand Law,7,491-514.

Major,B., Richards,C.,Cooper, L.M.,& Zubek, J.(1998).Personal re-silience, cognitive ap-praisals and coping: Anintegrativemodel ofadjustment to abortion.Journal of Personalityand Social Psychology,74,735-752.

Cozzarelli,C., Sumer,N.,& Major,B. (1998).Men-tal models of attach-ment and coping withabortion. Journal of Per-sonality and Social Psy-chology, 74,453-467.


38minors and 402adults followed for 2years after an electivefirst trimester abortion.

527 women;all women(N=615) completedpreabortion and ap-proximately 1-monthpostabortion question-naires; analysis is lim-ited to 527 womenwhoprovided completedata on all relevantstudy variables.

615 womenwho com-pleted a preabortion,immediate postabor-tion and approximately1month follow-upquestionnaire.

Controls/Covariates

None.

All models tested con-trolling for measuresof prior adjustment.Neuroticism, age, edu-cation, religion, race,and whether it wasthe woman’s firstabortion.

Age,marital status,whether or not thiswas a first abortion.

Primary Outcome

Post-abortion adjust-ment (depression,deci-sion satisfaction,benefit-harm ap-praisals, abortion-spe-cific emotions,wouldmake the same deci-sion), at 1 month and 2years; risk factors as-sessed on day of abor-tion.

Post-abortion adjust-mentmeasured by theCoping Operation Pref-erence Enquiry, residu-alized distress (thedepression,hostility,and anxiety subscalesof the Brief SymptomInventory), the PositiveWell-Being scale anddecision satisfaction.

Psychological distress(42 items from the SCL-90) and psychologicalwell-being (18 itemindex developed byRyff ).

Results

No significant differ-ence between adultsandminors at 2 yearspost abortion; at 1month,adolescentsslightly less satisfiedand have less perceivedbenefit.

Preabortion personalresources (items takenfrom existingmeasuresof self-esteem,disposi-tional optimism andpersonal control) re-lated to postabortionadjustment throughpreabortion cognitiveappraisals and post-abortion coping.Cogni-tive appraisals’effectson adjustmentmedi-ated by postabortioncoping.Womenwhohadmore personal re-sources perceived theirabortions as less stress-ful and had better cop-ing skills.

Mental models of at-tachment were relatedto postabortion func-tioning. This relation-ship wasmediated byperceived social sup-port, perceived socialconflict, and self-effi-cacy. Models of self wasa stronger predictor ofadjustment thanmodelof others.


Small sample of womenunder age 18.

Non-standardizedmeasures of personalresources, cognitive ap-praisals, and decisionsatisfaction.

All measures of socialsupport based onwomen’s self-reports.Limited indirect globalmeasure of mentalmodels of attachment.Missing data onmentalmodels with sociode-mographic differencesbetweenmissing andnon-missing datagroups.




Citation

Major,B.,Zubek, J.M.,Cooper,M.L.,Cozzarelli,C.,& Richards,C. (1997).Mixedmessages: Impli-cations of social conflictand social supportwithin close relation-ships for adjustment toa stressful life event.Journal of Personalityand Social Psychology,72,1349-1363.


615 womenwho com-pleted preabortion and1-month follow-upquestionnaires.

Controls/Covariates

Positive and negativereactivity, lifetime his-tory of depression(from DIS), seekingprofessional mentalhealth counseling,demographic vari-ables related to oneor more criterionmeasures (includesage, race, education,marital status, religion,whehter this is firstabortion).

Primary Outcome

Separatemeasures ofdistress andwell-beingat 1-month follow-up.Psychological distressassessed using theSCL-90 subscales ofdepression,anxiety,hostility and somatiza-tion. Positive well-beingwasmeasured usingthe 18-item short ver-sion of the Ryff PositiveWell-Being scale.

Results

Perceived abortion-specific social supportand social conflict(measured preabortion)were related to1-month postabortionadjustment after poten-tial confounds werecontrolled.Perceivedsocial conflict frompartner predicted dis-tress but not well-being; social supportfrom partner predictedwell-being but not dis-tress. Perceived supportfrommother or friendwas associated withwell-being.Social con-flict withmother orfriends interacted withsocial support to pre-dict distress.Womenwho perceived highsupport from thesesources weremore dis-tressed if they also per-ceived high conflict.


All measures of socialsupport and social con-flict based onwomen’sself-reports.

Prospective StudiesThe majority of prospective studies were conducted byone group of investigators, Major and colleagues. Sevenpapers published since 1990 were based on data from amultisite sample of first-trimester abortion patients inthe Buffalo, NY, area (Sample 1). These papers are notindependent of each other because they are based onthe same sample. Four additional papers were based onthree separate samples of women from the same geo-graphic area obtaining first-trimester abortions (Sam-ples 2, 3, and 4). Four of the seven Sample 1 studiesanalyzed data of 442 women followed for 2 years aftera first- trimester abortion for an unintended pregnancyat one of three sites. Assessments took place at fourtime points: preabortion and 1-hour, 1-month, and 2-years post abortion. The three other papers based onSample 1 did not include the 2-year follow-up in their

analyses. The other studies by Major and colleagueswere based on smaller samples of 291 (Sample 2), 283(Sample 3), and 247 (Sample 4) women recruited froma single abortion facility who provided preabortion and30-minute- and 1-month postabortion follow-up data.

Although the lack of comparison groups of womenwith an unintended pregnancy who carry to term isa significant limitation for assessing relative risk ofabortion versus alternatives, as a group, the Sample 1studies have a number of methodological strengths,including use of standardized measures of psychologi-cal experiences, appropriate data collection andanalysis procedures, a large sample, reasonably longpostabortion follow-up, analyses of changes in abor-tion reactions over time, and sound social-psychologi-cal theory to direct analyses. One potential limitation

Table 5:U.S.Samples of Abortion Group(s) Only/No Comparison (continued)

Prospective Analysis UsingMajor & Colleagues 1-Site SampleGeneral Description: 291 English-speaking women who obtained a first trimester abortion at a private free-standing clinic in Buffalo,NY.Womencompleted a preabortion, immediate postabortion and 3-week follow-up questionnaire. From a larger sample of 336 women,but 45 eliminated fromanalysis because they did not complete the immediate post-questionnaire.Average age was 23.3 (range = 14-40); 66%White, 74% single.

Limitations Common to All Studies Based on this Data Set: No comparison group (not a limitation for majority of analyses which examinedrisk factors,mediators and moderators of postabortion psychological distress). Sample is limited to women from one clinic and is not nationally or re-gionally representative of women who obtain abortions. Short follow-up period; high attrition rate: 38% completed the 3-week follow-up questionnaire.Differences in sociodemographic characteristics of those who completed 3-week follow-up and those lost to follow-up.Does not include measures suchas domestic violence and sexual abuse that may be related to postabortion adjustment.

Citation

Cozzarelli, C.,& Major,B.(1994).The effects ofanti-abortion demon-strators and pro-choiceescorts on women’spsychological re-sponses to abortion.Journal of Social andClinical Psychology,13,404-427.

Cozzarelli, C. (1993).Personality and self-efficacy as predictors ofcoping with abortion.Journal of Personalityand Social Psychology,65,1224-1236.


291 womenwho re-ceived first trimesterabortions.

291 English-speakingwomenwho obtained afirst trimester abortion.

Controls/Covariates

None.

Preabortiondepression

Primary Outcome

Outcome immediatelypost abortion and at 3-week follow-up wasmeasured by the SCL-90 Depression subscale.

SCL-90 Depression sub-scale and 9-item scaleassessing current affec-tive state were com-bined to create apostabortion distressindex.

Results

Prior to the abortionwomenwere askedabout their perceptionsof anti-abortiondemonstrator and pro-choice escort activity.Pro-choice escortsbuffered the effects ofanti-abortion demon-strators but not the in-tensity of theirpicketing onwomen’spsychological adjust-ment. Themore womenfelt upset by thedemonstrators and themore intense the an-tiabortion activity, themore depression theyexperienced immedi-ately postabortion.

Self-efficacy regardingpost-abortion copingwas the strongest pre-dictor of psychologicaladjustment immedi-ately after and 3-weekspost-abortion.Self-efficacymediated theeffects of self-esteem,optimism,and per-ceived control on ad-justment at both timepoints. Initial depres-sion strongly predictedboth self-efficacy andadjustment.


Correlations aremod-est, although authorsstate that % of varianceexplained is more thanfor social support or forreligious/attitudinalconflict in this data set.




Other Prospective Studies

Citation

Major,B.,Cozzarelli,C.,Sciacchitano,A.M.,Cooper,M.L.,Testa,M.,&Mueller,P.M. (1990).Per-ceived social support,self-efficacy and adjust-ment to abortion. Jour-nal of Personality andSocial Psychology,59,452-463.

Major,B.,Cozzarelli,C.,Testa,M.,& Mueller,P.(1992).Male partners’appraisals of undesiredpregnancy and abor-tion: Implications forwomen’s adjustment toabortion. Journal of Ap-plied Social Psychology,22,599-614.


283women obtaining afirst trimester abortionat an abortion clinic inBuffalo,NY, in 1987 (91%participation rate).Ave.age =22,78%white,80% single (see alsoMueller &Major,1989).Perceived social supportand self-efficacy for cop-ingwith abortion as-sessed prior to theabortion.Adjustmentassessed 30min postabortion.

73 couples inwhichwoman received a firsttrimester abortion andmale partner accompa-nied her to the clinic.(Women’s ave age = 20,79%nevermarried,93%White).Theywere partof a larger sample of 247women obtaining abor-tions at a clinic in BuffaloNY in 1983;88%ofthosewhowere accom-panied by their partnerparticipated in thisstudy.Original sample(88%White,78% single)had 92%participationrate (seeMajor,Mueller&Hildebrandt (1985.)

Controls/Covariates

Demographic vari-ables related to crite-rion variables,included marital sta-tus, religion (Catholic,non-Catholic), andrace (White, other).

Women’s coping ex-pectancies for analy-ses of impact of men’sappraisal on partner’sadjustment.

Primary Outcome

1.For the primary pathanalyses 3 psychologi-cal outcomemeasures(mood,anticipation ofnegative consequencesand depression asmeasured by the shortform of the BDI) given30minutes postabor-tion were standardizedand summed to createa single adjustmentmeasure.2.For assessing the ef-fects of nondisclosureand disclosure on ad-justment, four separateoutcome variables weredepression,mood,an-ticipated negative con-sequences, andphysical complaints.

Women’s adjustmentmeasured 30minutespost abortion usingshort form of BDI.

Results

Highperceived socialsupport predicted in-creasedpreabortionself-efficacy for copingwith abortion andbetterpostabortion adjust-ment. Self-efficacymedi-ated the positive effectsof perceived social sup-port on adjustment.Also,womenwho told closeothers of their abortionand felt these otherswere not completelysupportive had lowerpostabortion adjust-ment than thosewhodid not tell others orthosewho told and feltcompletely supported.85% told partner;66%told friends;40% toldfamily of their abortion.

Coping expectanciesand attributions as-sessed immediately pre-abortion.Men’s copingexpectancies regardingthis abortion influencedtheir female partners’depression levels onlyforwomenwith lowcoping expectancies.Womenwith low copingexpectancieswhosepartners also had lowcoping expectancieswere themost de-pressed.Men’s attribu-tions about thepregnancywere unre-lated to their partners’adjustment.


Extremely shortpostabortion interval; noadditional follow-up.Nonstandardizedmeas-ures of coping self-effi-cacy and social support.No preabortion assess-ment of psychologicaloutcomes.e (coders) re-ports of picketing activ-ity. Single measure ofpostabortion adjust-ment. No pre-abortionmeasure of depression.

Sample unrepresenta-tive of larger sample ofwomen obtaining abor-tions at this particularclinic,most of whomwent to the clinic with-out a partner.Relativelysmall sample size.Ex-tremely short postabor-tion interval; noadditional follow-up.1-itemmeasure of copingexpectations;no pre-abortion assessment ofdepression.



Citation

Pope,L.M.,Adler,N.E.,&Tschann J.M. (2001).Postabortion psycho-logical adjustment:Areminors at increasedrisk? Journal of Adoles-cent Health,29,2-11.

Burgoine,G.A.,Van Kirk,S.D.,Romm,J., Edelman,A.B., Jacobson,S.,&Jensen, J.T. (2005).Com-parison of perinatalgrief after dilation andevacuation or labor in-duction in secondtrimester terminationsfor fetal anomalies.American Journal of Ob-stetrics andGynecology,192,1928-1932.

Phelps,R.H., Schaff, E.A.,& Fielding,S.L. (2001).Mifepristone abortioninminors.Contracep-tion, 64,339-343.


96women (23 under 18,40 aged 18-21) seekingpregnancy terminationat 6-12weeks gestationin four clinics in SanFrancisco,CA;63 com-pleted follow-up.Englishspeakers only,1/3wereAfrican American.

49womenwho termi-nated a desired secondtrimester pregnancy be-cause of a fetal abnor-mality through eitherdilation and evacuation(D & E) or induction oflabor (IOL).

35 adolescents 14-17years of age inRochester,NY,who hadmifepristone abortionsat < 56 days gestation.

Controls/Covariates

Did not control for de-mographic variablesbecause none wererelated to postabor-tion adjustment.

None.

None.

Primary Outcome

Follow-up 4 weeks postabortion,with assess-ment of Beck Depres-sion Inventory,“emotion”scale, Spiel-berger State Anxiety In-ventory, Rosenbergself-esteem scale, Im-pact of Events scale,Positive States of Mindscale.

Depression wasmeas-ured with the Edin-burgh PostnatalDepression scale at en-rollment, 4 month and12month follow-upand grief,using thePerinatal Grief scale at4-month and 12-monthfollow-up.

Rating scales assessedemotional responsevariables on question-naires at Day 1 (first visitwhenmifepristone wasadministered) and im-mediately post abor-tion (Days 4-8) andtelephone interview 4weeks post abortion.

Results

Nodifference betweenunder 18 andover 18group,except youngergroup scored slightlylower on“comfortablewith decision”; for com-bined agegroups pre-abortion emotional stateandperceivedpartnerpressure predictedpostabortion adjust-ment.

Cutoff scoreswere setfor clinical depressionandgrief.No significantdifferenceswere foundbetween the surgical(D&E) andmedical (IOL)groups in levels of griefor depression at anytimepoint.

Little emotional im-provement from firstvisit to immediate postabortion.Greater emo-tional improvement re-ported frompostabortion to fourweek follow-up,e.g.,stress (57% to 21%) andfeeling scared (43% to8%) decreased signifi-cantly from first visit to 4week follow-up.


Small sample size.Lim-ited representativenessof sample;urban popu-lation in state withoutparental requirement forabortion,6-12 weeksgestation only.Attrition:34% lost to follow-up;differences betweenthose retained and lostto follow-up, (e.g., on re-ligion and depression).Functional relevance notwell-established for allof themeasures used.

Small sample; very lim-ited statistical power.High attrition:57% com-pleted 4-month and58% completed 12-month follow-up;only28.5% completed both(use of mail back ques-tionnaires at 4 and 12months).No random as-signment to group.

Small sample.Limitedgeneralizability: Studylimited to teens withparental consent to par-ticipate but parentalconsent not required inNY for an abortion.Nocomparison groups suchas surgical abortionclients or adult women.Non-standardized singletimemeasures of emo-tional responses.Someadolescents still had in-complete abortionswhen they completedthe immediatepostabortion question-naire.





Citation

Miller,W.B. (1992).Anempirical study of thepsychological an-tecedents and conse-quences of inducedabortion. Journal of So-cial Issues,43,67-93.

Sit,D., Rothchild,A. J.,Creinin,M.D.,Hanusa,B.H.,&Wisner,K.L. (2007).Psychiatric outcomesfollowingmedical andsurgical abortion.HumanReproduction,22,878-884.


64womenwho had in-duced abortionswhowere part of a largerprospective longitudinalstudy of 987 nevermar-ried, recentlymarriedwomen,or recent first-timemotherswho deliv-ered living in the SanFrancisco Bay area in the1970s.Thewomenwereinterviewed 4 times atyearly intervals.

47womenwho ob-tained surgical abor-tions and 31womenwho obtained non-sur-gical abortions in Pitts-burgh andWesternPennsylvania at < 9weeks gestation.

Controls/Covariates

None.

Age and race initiallyincluded.No differ-ences between groupsin other demographiccharacteristics,pastreproductive history,or psychiatric history.

Primary Outcome

Postabortion“regret”assessed by a one-itemquestion that asked ifthe womanwouldchoose to have anabortion again.Emo-tional upset assessed atfinal interview by a one-itemmeasure thatasked if the woman hadexperienced emotionalupset from the abor-tion after first fewweeks.

Depression assessedimmediately pre-abor-tion and approximatelyone-month (range =14-60 days) post abor-tion using the Edin-burgh PostnatalDepression Scale.

Results

Womenwith a Protes-tant religious back-groundhad less regretand thosewith a tradi-tional gender role orien-tation reportedmoreregret.Emotional upsetafter first fewweeks ofabortion associatedwithnot beingmarried atttimeof the abortionandbeing low in tradi-tional gender-role orien-tation .

Nodifferences indepression betweengroups.Both groups ex-perienced a significantdecline in depressionfrompre- to postabortion (35-36%atincreased risk pre-abortion vs.17-21%at risk post abortiondefined as EPDS>10)Womenwith a pasthistory of psychiatricproblems at a higherrisk of post abortiondepression.


Single-itemmeasures ofthe negative psycholog-ical reactions to abor-tion. Retrospectivereporting of the emo-tional impact of theabortion.Lack of specifi-cation of abortion his-tory. Probableunder-reporting of abor-tions. Sample limited toWhite English speakingwomen.Only small sub-set of representativesample (64 of 987) are inthe abortion group.

Small sample; limitedmeasures of pre-abor-tion characteristics; lackof differences betweenin participant character-istics between groupsmay be due to smallsample size and limitedpower.

is the high attrition rate; the 442 women for whomdata were available 2 years post abortion represent50% of the original sample. However, the researchersconducted detailed analyses to show that women whocompleted the follow-up and those lost to follow-upnot did not significantly differ on any demographic orpsychological characteristic. A second limitation isthe lack of measures of mental health prior to thepregnancy. Strengths and limitations of Samples 2, 3,and 4 are similar to those of Sample 1 with the addedcaveat that these were smaller samples from a singlesite followed for a shorter time period.

Analyses based on the Sample 1 data set examinedchanges over time in women’s psychological experi-ences. Most women reported that they had benefitedfrom their abortion more than they had been harmedby it, and these appraisals did not change from 1month to 2 years post abortion (Major et al., 2000).Most women also reported that they were satisfiedwith their decision, although the percentage satisfieddecreased from 1 month (79%) to 2 years (72%).Women also reported feeling more relief than positiveor negative emotions both immediately and 2 yearsafter their abortion. Over the 2 years, however, relief


Retrospective Studies (all these studies lacked a preabortionmeasure of psychological functioning)

Citation

Coleman,P.K.,& Nelson,E. I. (1998).The quality ofabortion decisions andcollege students’ re-ports of post-abortionemotional sequelaeand abortion attitudes.Journal of Social andClinical Psychology,17,425-442.

Franz,W.,& Reardon,D.(1992).Differential im-pact of abortion onadolescents and adults.Adolescence,105,161-172.


31 female and 32malecollege students at amidsized southeasternuniversity who reporteda previous abortion;asubsample of a largerstudy of abortion atti-tudes.

252women aged 16-64who have had an abor-tionwere divided intoadolescent vs.adultgroups based on age attime of abortion (114younger than 20 and138,20 or older).Re-spondents recruited bysending survey forms toall identifiedWomen Ex-ploited by Abortiongroups in the U.S.

Controls/Covariates

Time since theabortion.

None.

Primary Outcome

Single-item nonstan-dardizedmeasures ofpostabortion depres-sion and depressionand anxiety.

Apparently single-itemassessed self-report of"severe psychologicalreactions" to the abor-tion. Item/scale not ad-equately described.

Results

Dimensions of abortiondecisions (ambivalence,regret,comfort) andemotional connection tothe fetuswere not asso-ciatedwith self-reportedanxiety anddepressionforwomenwith the ex-ception that comfortwas related to anxiety.

Adolescent participantsreported significantlygreater severity of psy-chological stress thanadult participants andweremore likely to feelforced to have the abor-tion andmisinformed atthe timeof abortion.Pre-dictors of severe psycho-logical stresswerefeeling forced to abort,being dissatisfiedwithabortion services andhaving a very negativeviewof abortion.


Small sample; abortionhistory retrospectivelyself-reported.Single-item non-standardizedoutcomemeasures.Un-warranted conclusions,e.g., state that“morethan one-half of thewomen and over one-quarter of themen ex-perience post-abortionincrease in depression”based on responses toan item stating,“I haveexperienced some de-pression since the timeof my abortion.”

Unrepresentative con-venience sample ofwomen already in a sup-port group.Abortion his-tory retrospectivelyself-reported.Time sinceabortion varied greatly(1-15 years).Differencesbetween groups in so-ciodemographic charac-teristics and pregnancyhistory are unknownand not controlled.Noinformation on ethnicityof (total) sample.Lessthan half of surveysmailed to groups (47%)were returned.


and positive emotions declined, whereas negativeemotions increased. Depression scores were lower,and self-esteem was higher 2 years after the abortioncompared with just prior to the abortion.

Collectively, these findings add to knowledge of pre-dictors and mediators of psychological outcomes overa longer follow-up period than earlier abortion-only

studies. These studies showed that women at higherrisk for negative emotions 2 years post abortionincluded those with a prior history of mental healthproblems (Major et al., 2000), younger age at thetime of the abortion (Major et al., 2000), low per-ceived or anticipated social support for their decision(Cozzarelli, Sumer, & Major, 1998; Major, Zubek,Cooper, Cozzarelli, & Richards, 1997), greater




Citation

Rue,V.M.,Coleman,P.K.,Rue, J. J.,& Reardon,D.C.(2004). Induced abor-tion and traumaticstress:Preliminary com-parison of Americanand Russian women.Medical ScienceMonitor,10,SR5-16.

Lemkau, J.P. (1991).Postabortion adjustment ofhealth care profession-als in training.AmericanJournal of Orthopsychia-try, 6,102.


217 Americanwomenand 331 Russianwomenages 18-40who had hadone ormore inducedabortions and had notexperienced other preg-nancy losses; recruitedin 1994 from a hospitaland two outpatient clin-ics in the U.S.and a hos-pital in Russia.

63women studentswhowere enrolled indegree programs innursing,professionalpsychology,ormedicineat aMid-westernmetro-politan university andacknowledged havinghad an abortion; theyrepresented 12%of allwomen students sur-veyed.

Controls/Covariates

Different sets of co-variates for differentanalyses.

Age, age at abortion,ethnicity,marital sta-tus, religion, sexualabuse, gestation time,total number of abor-tions, etc. entered intoregression equation.

Primary Outcome

Traumawasmeasuredusing the 14-item PTSDscale of the PregnancyLoss Questionnaire.Thisscale’s items corre-spond to the 14 symp-toms of PTSD describedin the DSM-IV.The Trau-matic Stress Institute’s(TSI) Belief scale wasused tomeasure dis-ruptions in beliefsabout self and othersthat arise form expo-sure to trauma.

Short-term adjustment(STA) wasmeasured assummed ratings (1 =not at all; 4 = moder-ately; 7 = extremely)of assessed relief,guilt,anger, anxiety, concernabout future relation-ships and concernabout future pregnan-cies threemonths postabortion.Long-termadjustment (LTA) con-sisted of the sum ofparallel items for thepresent time.equatelydescribed.

Results

Americanwomen re-portedmore PTSDsymptoms than theirRussian counterparts;14.3%of American and0.9%of Russianwomenmet full diagnostic crite-ria for PTSD.Russianwomen reportedmoredisruption of cognitiveschemas.For U.S.women,predictors ofpoorer psychological ad-justment (greater stressrelated-symptoms) onceprior stress and abusewere controlled in-cludedbeing younger,more years of education,havingbonded to thefetus,not believing inwomen’s right to havean abortion,feeling pres-sured tomake the deci-sion.

Current and 3-monthpostabortion distresswere low,means of allitems<4with the excep-tion of relief ($5).Per-ceivedpreparation forthe abortion and confi-dence in thewisdomoftheir choicewere predic-tors of STA and LTA.Womenwho recalledbeingpressured re-portedpoor STA and LTAandwere less confidentabout the decision theyhadmade.


Abortion history retro-spectively self-reported.Two groups of womenwere dissimilar in age,mean number of weekspregnant etc.Translationproblems led to use ofdifferent data collectionmethods (questionnairein U.S.vs. interview inRussia).Greater rates ofbehavioral and psycho-logical symptoms in U.S.womenmay be associ-ated with an environ-mentmore conflictedabout abortion.

Abortion history andsomemeasures ofpostabortion distressretrospectively self-re-ported. Abortion oc-curred an average of 9years previously.




Citation

Congleton,G.K.,& Cal-houn, L.G. (1993).Post-abortion perceptions:Acomparison of self-identified distressedand nondistressed pop-ulations. InternationalJournal of Social Psychi-atry, 39,255-265.

Tamburrino,M.B.,Franco,K.N.,Campbell,N.B., Pentz, J. E., Evans,C.L.,& Jurs, S.G. (1990).Postabortion dysphoriaand religion.SouthernMedical Journal,83,736-738.


25womenwhoreported responding toabortionwith emotionaldistress comparedwith25 non- distressedwomen.Participantsrecruited nationallyfromposted notices andvolunteers fromNOW,post-abortion supportgroups,etc.

71women frompatient-led support groups forwomenwith post abor-tion dysphoria.

Controls/Covariates

None.

None.

Primary Outcome

Mental health assessedvia two indices fromthe Brief Symptom In-ventory: the GlobalSeverity Index and thePositive SymptomDis-tress Index.The Impactof Event scale was usedtomeasure traumaticstress.

Mental health (dyspho-ria) measured by sub-scales of theMillonClinical Multiaxial In-ventory.

Results

Thedistressedgroup re-called higher past trau-matic stress levels andcurrently hadhighertraumatic stress.Neithergroup showeddistressonGSI,and their PSDIscores did not differ.

46%of total groupchanged their religion toEvangelical and Funda-mentalist Protestant de-nominations.Thosewhoweremembers of thesedenominations scoredlower onpassive-aggres-sive, ethanol abuse,andavoidance subscales.


Small unrepresentativeconvenience samples.Abortion history retro-spectively self-reported.Retrospective self-re-ports of stress that oc-curredmany years ago.Two groups differed oncurrent religious affilia-tion.

Unrepresentative con-venience sample limitedtowomenwho feel ex-ploited by abortion.Abortion history retro-spectively self-reported;psychological reactionsafter abortion retrospec-tively reported; someparticipants had anabortion decades earlier.Non-standardized singleitem primary outcomemeasure;age and agerange at time of abor-tion unclear; assumeadolescents evidenceimmature decisionmak-ing but no evidence tosupport assumption.

personal conflict about abortion (Cozzarelli, Major,Karrasch, & Fueger, 2000), and low self-efficacyabout their ability to cope with the abortion (Coz-zarelli, Sumer, & Major, 1998; Cozzarelli, 1993;Major et al., 1990).

This research also provided new insight into the roleof cognitive mediators, coping, and stigma inpostabortion functioning. Two studies investigatedthe effects of antiabortion picketing on women’spostabortion responses. Cozzarelli and Major (1994)found that the greater the number of antiabortionpicketers and the more aggressive the picketing that

women encountered when entering an abortion clinic(as coded by observers), and the more the womenreported feeling upset by the demonstrators, themore depressed affect they reported right after theirabortion. These effects were partially mitigated bythe presence of prochoice escorts outside the clinic,suggesting that prochoice escorts altered not only thesocial context, but also the meaning of that context.A later study that included 2-year follow-up assess-ments concluded the women’s encounters with pick-eters evoke short-term negative psychologicalreactions but do not appear to have long-term nega-tive psychological effects (Cozzarelli et al., 2000).


Other



Citation

Layer, S.D., Roberts,C.,Wild,K.,&Walters, J.(2004).Post abortiongrief: Evaluating thepossible efficacy of aspiritual group inter-vention. Research on So-cial Group Practice,14,344-350.


35womenwith“post-abortion grief”recruitedfrom three faith-basedorganizations in Florida.

Controls/Covariates

None.

Primary Outcome

Postabortion grief com-posed on shame andpost-traumatic stress.Shame is assessedusing Cook’s Internal-ized Shame scale;post-traumatic stressmeasured by the Im-pact of Events scale- Re-vised.

Results

Womenparticipated apsychoeducational spiri-tual-basedgroup inter-vention forwomenwithpostabortion grief of-fered in an 8-week orweekend format.Shameandpost traumaticstress showed significantreductions frompre-in-tervention to immedi-ately post-intervention.


Small unrepresentativeconvenience sample.Nocontrol/comparisongroup.No sociodemo-graphic or pregnancyhistory informationother than age.No infor-mation on length oftime since abortion.Nomental health history.

Examination of perceived stigma revealed that almosthalf of the 442 women in the multisite sample (Sample1) felt that they would be stigmatized if others knewabout the abortion, and over 45% felt a need to keepit secret from family and friends (Major & Gramzow,1999). Secrecy was associated with increases in psy-chological distress (anxiety and depression) over time,via the mediators of increased thought suppressionand decreased emotional disclosure. In particular,Major and Gramzow (1999) found that the morewomen felt that others would look down on them ifthey knew about the abortion, the more they felt thatthey had to keep the abortion a secret from theirfriends or family. Perceived need for secrecy, in turn,was associated with less disclosure of feelings to fam-ily and friends, increased thought suppression and in-trusion, and increased psychological distress 2 yearspost abortion (controlling for initial distress). Thus,feelings of stigmatization led women to engage in cop-ing strategies that were associated with poorer adapta-tion over time.

This research group also extended earlier knowledgeabout the role of social support in abortion. One studyshowed that perceived social support mediated the re-lationship between cognitive models of attachment andadjustment (Cozzarelli et al., 1998). Another study in-vestigated the joint and interactive effects of perceivedsocial conflict and perceived social support from others

surrounding the abortion on negative psychological re-actions and well-being (Major et al., 1997). Greaterperceived social conflict with the partner predicted in-creased distress (but not decreased well-being), whereasgreater perceived support from partner predicted in-creased well-being (but not decreased distress). More-over, for mothers and friends, perceived conflict andsupport interacted to predict distress, whereas supportwas a direct predictor of well-being.

Three studies established the importance of cognitiveappraisals and self-efficacy as proximal predictors ofpostabortion adjustment. One study showed that therelationship between social support and adjustmentwas mediated by coping appraisals and self-efficacy.Women who perceived more social support from oth-ers for their decision felt more able to cope with theirabortion prior to the procedure, and these appraisalsmediated the positive relationship between perceivedsocial support and postabortion well-being (Major etal., 1990). Two other studies showed that self-efficacyand cognitive appraisals mediated the effects of pre-abortion personal resources on postabortion copingand adjustment (Cozzarelli, 1993; Major et al.,1998). Women with more resilient personalities (highself-esteem, internal locus of control, and an opti-mistic outlook on life) felt more capable of copingwith their abortion and appraised it more benignlyprior to the procedure. Their more positive cognitive

appraisals, in turn, were associated with more adap-tive forms of coping in the month following the abor-tion (more acceptance, less avoidance), which in turnwere associated with reductions in psychological dis-tress (depression, anxiety) and increases in positivewell-being over time.

Two studies specifically compared the responses ofminor adolescents and adult abortion patients. Theyreported very similar findings. Using data from Sam-ple 1 of Major et al. (2000), Quinton, Major, andRichards (2001) found no differences between minors(N = 38) and adults (N = 404) in psychological dis-tress and well-being 2 years after an abortion, al-though the adolescents were slightly less satisfied withtheir decision and perceived less personal benefit fromit. In a different sample of 96 women (23 adoles-cents), Pope, Adler, and Tschann (2001) reported thatat 4 weeks post abortion, there were no differencesin depression, anxiety, self-esteem, or posttraumaticstress between the younger and older groups, al-though the adolescents scored slightly lower on“comfort with decision.” Both of these studies arelimited by small samples of adolescents. These resultsappear to conflict with Major et al. (2000), whichidentified younger age at time of abortion as a riskfactor for negative postabortion emotional experi-ences. However, the latter study examined the associ-ation of mental health outcomes with the continuousvariable of age among a larger sample.

Miller (1992) examined psychological experiencessubsequent to abortion among 64 women who hadparticipated in a larger longitudinal study on the psy-chology of reproduction in the San Francisco Bay areain the 1970s. All of the 967 women in the larger studywere White, English speaking, and between ages 18and 27 years. At the final interview, the 64 womenwho reported an abortion during the study were askeda series of one-item questions about how their abor-tion had affected them. Prospective analyses using re-sponses from earlier interview periods examinedpredictors of “regret” (the extent to which womensaid they would choose the abortion again (1 = no, 2 =not sure, 3 = yes)) and “upset” (how emotionallyupset the women recalled being in the first few weeksafter the abortion). Having a Protestant religiousbackground was associated with less regret, whereashaving a traditional gender role orientation was asso-ciated with greater regret. Not being married at the

time of the abortion was related to greater postabor-tion upset, whereas a traditional gender-role orienta-tion was associated with less upset. Other single itemsmeasuring reasons for having and not having an abor-tion (measured at the final interview) were also relatedto the two outcome variables. Despite its prospectivedesign, this study is severely limited by the single-itemmeasures of the negative psychological reactions toabortion, retrospective reporting of the emotional im-pact of the abortion, lack of specification of abortionhistory, probable underreporting of abortions, smallsample, and nonrepresentative sample.

Two other prospective studies examined emotionalimprovement after mifepristone abortions in minors(Phelps, Schaff, & Fielding, 2001) and depression riskafter surgical and nonsurgical abortion (Sit et al.,2007). Phelps et al. assessed emotional responses (e.g.,perceived stress, fear) of adolescents aged 14-17 yearsat three time points: when mifepristone was first ad-ministered, 4-8 days later, and 4 weeks later. The re-searchers found little emotional improvement fromfirst visit to 4-7 days later, but greater emotional im-provement (e.g., lower perceived stress, lower fear) at4-week follow-up. This study was limited by smallsamples (N=35), high attrition rates, and othermethodological problems.

Sit et al. (2007) compared depression scores preabor-tion and 1 month post abortion among women ob-taining surgical (N = 47) versus nonsurgical(mifepristone-misoprostol) abortions (N = 31) at lessthan 9 weeks’ gestation. One month post abortion,17% (7/42) of surgical and 21% (5/24) of medical pa-tients had an EPDS depression score equal to orgreater than 10. Both groups experienced a significantdecline in depression from pre- to post abortion, andthe difference in depression between the two groupswas not significant either before or after the abortion.As observed in other studies, women with a historyof past psychiatric problems were at higher risk forpostabortion depression, irrespective of procedure.Findings of this study are consistent with severalothers based on non-U.S. samples in suggesting thatmethod of termination during the first trimester doesnot affect emotional adjustment or psychological ex-periences after the procedure among women, given achoice of procedure (Ashok et al., 2005; Howie, Hen-shaw, Naji, Russell, & Templeton, 1997; Lowensteinet al., 2006).



A final U.S. study (Burgoine et al., 2005) examineddepression and grief among 49 women who termi-nated a desired pregnancy during the second trimester.They examined whether responses differed as a func-tion of the abortion procedure they underwent: dila-tion and evacuation (D&E) or induction of labor(IOL). Levels of depression were relatively high inboth groups 4 months and 12 months post abortion,but incidence of clinically significant depression didnot differ as a function of abortion procedure. Griefscores did not differ at 4 or 12 months betweenwomen choosing either of the two abortion methods.

Retrospective StudiesMost of the half dozen retrospective studies of abor-tion samples had serious methodological flaws and donot warrant further discussion except as examples ofpoor study designs. In these studies women’s currentor recalled past mental health or distress often was at-tributed to an abortion that occurred many years pre-viously (e.g., Franz & Reardon, 1992; Lemkau, 1991;Tamburrino et al., 1990). For instance, Lemkau(1991) queried women about their level of distress ex-perienced 3 months post abortion although the targetabortion had occurred an average of 9 years previ-ously. Other limitations include use of one-item un-standardized outcome measures (Coleman & Nelson,1998; Franz & Reardon, 1992) and small samplesizes (Coleman & Nelson, 1998; Congleton & Cal-houn, 1993; Tamburrino et al., 1990). Finally, authorsof several papers drew conclusions about prevalenceof postabortion mental health problems in the generalpopulation from samples of women who had self-identified as having postabortion mental health prob-lems, attributed their psychological problems tohaving had an abortion, and were members of supportgroups that foster such attributions (Congleton &Calhoun, 1993; Franz & Reardon, 1992; Tamburrinoet al., 1990).

SummaryandEvaluationofAbortion-Only StudiesProspective studies of U.S. abortion-only samples haveadded to knowledge about predictors, mediators, andmoderators of psychological experiences subsequentto abortion. The most methodologically strong studiesin this group identified personal and social factors thatinfluence how women cognitively appraise and copewith abortion and demonstrated how appraisals andcoping processes predict postabortion psychologicalexperiences, both positive and negative. The retrospec-

tive studies in this group suffered from methodologicallimitations that decreased confidence in the results andlimited conclusions that can be drawn from them.

SUMMARYANDCONCLUSIONS

As noted at the beginning of this report, the empiricalliterature on the association between abortion andmental health has been asked to address four primaryquestions: (1) Does abortion cause harm to women’smental health? (2) How prevalent are mental healthproblems among women in the United States whohave had an abortion? (3) What is the relative risk ofmental health problems associated with abortion com-pared to its alternatives (other courses of action thatmight be taken by a pregnant woman in similar cir-cumstances)? and (4) What predicts individual varia-tion in women’s psychological experiences followingabortion? As discussed above, the first question is notscientifically testable from an ethical or practical per-spective. The second and third questions obscure theimportant point that abortion is not a unitary event,but encompasses a diversity of experiences. That said,in the following section we address what the literaturereviewed has to say with respect to the last three ques-tions.

The Relative Risks of AbortionCompared to its AlternativesThe TFMHA identified 50 papers published in peer-re-viewed journals between 1990 and 2007 that analyzedempirical data of a quantitative nature on psychologi-cal experiences associated with induced abortion, com-pared to an alternative. These included 10 papersbased on secondary analyses of two medical recorddata sets, 15 papers based on secondary analyses ofnine public data sets, 19 papers based on 17 studiesconducted for the primary purpose of comparingwomen who had first-trimester abortions (or an abor-tion in which the trimester was unspecified) with acomparison group, and 6 studies that comparedwomen’s responses following an induced abortion forfetal abnormality to women’s responses followingother reproductive events. These studies were evaluatedwith respect to their ability to draw sound conclusionsabout the relative mental health risks associated withabortion compared to alternative courses of action that

can be pursued by a woman facing a similar circum-stance (e.g., an unwanted or unintended pregnancy).

A careful evaluation of these studies revealed that themajority suffered from methodological problems,sometimes severely so. Problems of sampling, meas-urement, design, and analyses cloud interpretation.Abortion was often underreported and underspecifiedand in the majority of studies, wantedness of preg-nancy was not considered. Rarely did research designsinclude a comparison group that was otherwise equiv-alent to women who had an elective abortion, impair-ing the ability to draw conclusions about relativerisks. Furthermore, because of the absence of adequatecontrols for co-occurring risks, including systemic fac-tors (e.g., violence exposure, poverty), prior mentalhealth (including prior substance abuse), and personal-ity (e.g., avoidance coping style), in almost all of thesestudies, it was impossible to determine whether anyobserved differences between abortion groups andcomparison groups reflected consequences of preg-nancy resolution, preexisting differences betweengroups, or artifacts of methodology. Given this state ofthe literature, what can be concluded about relativerisks from this body of research?

One approach would be to simply calculate effect sizesor count the number of published papers that suggestadverse effects of abortion and those that show no ad-verse effects (or even positive effects) of abortionwhen compared to an alternative course of action(e.g., delivery). Although tempting, such approacheswould be misleading and irresponsible, given the nu-merous methodological problems that characterizethis literature, the many papers that were based on thesame data sets, and the inadequacy of the comparisongroups typically used. Given this state of the literature,the TFMHA judged that the best course of action wasto base conclusions on the findings of the studies iden-tified as most methodologically rigorous and sound.

Of the studies based on medical records, the mostmethodologically rigorous studies were conducted inFinland. The largest and strongest of these examinedthe relative risk of death within a year of end of preg-nancy associated with abortion versus delivery(Gissler et al., 2004b). It demonstrated that the rela-tive risk differs depending on how cause of death iscoded. Compared to women who delivered, womenwho had an abortion had lower rates of direct preg-

nancy-related deaths (cause of death was directly re-lated to or aggravated by the pregnancy or its man-agement, but not from accidental or incidentalcauses) but higher rates of pregnancy-associateddeaths (deaths occurring within one year from end ofpregnancy, regardless of whether deaths are preg-nancy-related). When therapeutic abortions wereexcluded from the category of pregnancy-associateddeaths, however, this latter difference was notsignificant. Across both the Medi-Cal and Finlandrecord-based studies, a higher rate of violent death(including accidents, homicide, and suicide) was ob-served among women who had an abortion comparedto women who delivered. This correlational findingis consistent with other evidence indicating that riskfor violence is higher in the lives of women who haveabortions and underscores the importance of control-ling for violence exposure in studies of mental healthassociated with pregnancy outcome.

With respect to the studies based on secondary analy-ses of survey data, the conclusions regarding relativerisk varied depending on the data set, the approach tothe design of the study, the covariates used in analyses,the comparison group selected, and the outcome vari-ables assessed. Analyses of the same data set (theNLSY) with respect to the same outcome variable (de-pression) revealed that conclusions regarding relativerisk differed dramatically depending on the samplingand exclusion criteria applied.

The strongest of the secondary analyses studies wasconducted by Fergusson et al. (2006). This study wasbased on a representative sample of young women inChristchurch, NZ, was longitudinal (although Fergus-son also reported concurrent analyses), measuredpostpregnancy/abortion psychiatric morbidity usingestablished diagnostic categories, and controlled formental health prior to the pregnancy in prospectiveanalyses. Fergusson et al. compared women who ter-minated a pregnancy to women who delivered or hadnot been pregnant. The prospective analyses reportedby Fergusson et al. are most informative. These analy-ses compared number of total psychiatric disordersamong women who had an abortion prior to age21 to number of total psychiatric disorders amongwomen who had delivered a child by age 21 oramong women who had never been pregnant by age21, controlling for prepregnancy mental health andother variables that differed initially among the three



groups. In these analyses, women who had one ormore abortions prior to age 21 had a significantlyhigher number of total psychiatric disorders by age 25than women who had delivered or had never beenpregnant by age 21. This study thus suggeststhat women who have one or more abortions at ayoung age (<21) are at greater relative risk for psychi-atric disorder compared to women who deliver a childat a young age or women who do not get pregnant ata young age.

There are several reasons why caution should be usedin drawing the above conclusion from this study. Firstand most importantly, Fergusson et al. (2006) did notassess the intendedness or wantedness of the preg-nancy. As noted earlier, approximately 90% of preg-nancies that are aborted are unintended, compared toonly 31% of those that are delivered (Henshaw,1998). Thus, although these were young women, it isreasonable to assume that at least some of the womenin the delivery group were delivering a planned andwanted child. Delivery of a planned and wanted childwould be expected to be associated with positive out-comes and is not a viable option for women facing anunintended pregnancy. Second, the other comparisongroup used by Fergusson et al.—women who hadnever been pregnant—is not a viable option forwomen already facing an unintended pregnancy.Third, the prospective analyses were based on only 48women who had abortions, an extremely small sam-ple. Fourth, the study did not control for number ofprior abortions or births. Fifth, the study focused onwomen who had one or more abortions at a youngage (< 21 years), limiting its generalizability toyounger women; younger age has been linked in somestudies to more negative psychological experiences fol-lowing abortion (e.g., Major et al., 2000). Finally, thisstudy was conducted in New Zealand, a country withmore restrictive abortion regulations than those in theUnited States. Because the focus of APA is on mentalhealth in the United States, it may thus be less usefulas a basis for drawing conclusions about relative risksof abortion for U.S. women.

The TFMHA also reviewed and evaluated 19 papersbased on 17 studies conducted for the primary purposeof comparing women who had first-trimester abortions(or an abortion in which trimester was unspecified)with a comparison group on a mental health relevantvariable. These studies varied widely in methodological

quality and cultural context. Although most of thestudies showed no significant differences between thepsychological experiences of women who had an in-duced first-trimester abortion and women in a varietyof comparison groups once important covariates (e.g.,marital status, age) were controlled, most also werecharacterized by methodological deficiencies. These in-cluded problems of sampling, measurement, design,analyses, and inappropriate comparison groups. Thus,as a group, these studies also do not provide good an-swers to questions of relative risk or prevalence.

One study, however, stood out from the rest in termsof its methodological rigor. This study was conductedin the United Kingdom by the Royal College of Gen-eral Practitioners and the Royal College of Obstetri-cians and Gynecologists (Gilchrist et al., 1995). It waslongitudinal, based on a representative sample, meas-ured postpregnancy/abortion psychiatric morbidityusing established diagnostic categories, controlledfor mental health prior to the pregnancy as well asother relevant covariates, and compared women whoterminated an unplanned pregnancy to women whopursued alternative courses of action. In prospectiveanalyses, Gilchrist et al. compared postpregnancy psy-chiatric morbidity (stratified by prepregnancy psychi-atric status) of four groups of women, all of whomwere faced with an unplanned pregnancy: women whoobtained abortions, who did not seek abortion, whorequested abortion but were denied, and who initiallyrequested abortion but changed their mind. The re-searchers concluded that once psychiatric disordersprior to the pregnancy were taken into account, therate of total reported psychiatric disorder was nohigher after termination of an unplanned pregnancythan after childbirth.

This study provides high-quality evidence that amongwomen faced with an unplanned pregnancy, the rela-tive risks of psychiatric disorder among women whoterminate the pregnancy are no greater than the risksamong women who pursue alternative courses of ac-tion. What appears to be a discrepancy between theconclusions of this study and those of Fergusson et al.(2006) is likely due to differences in sampling andstudy design. First and most importantly, Gilchrist etal. (1995) restricted their study to women identified bytheir family doctor as having an “unplanned” preg-nancy, whereas Fergusson et al. did not assess the in-tendedness of the pregnancy, as noted above.

Consequently, the comparison groups used by Gilchristet al. are more appropriate for addressing the questionof relative risk of negative psychological experiencesfollowing elective abortion compared to other coursesof action women in similar circumstances (i.e., facingan unplanned pregnancy) might take. Second, theGilchrist et al. study was not restricted to women whobecame pregnant at a young age; hence the sample ismore representative of women who seek abortion.Third, differences in abortion sample size were dra-matic. The prospective analyses by Gilchrist et al. werebased on an abortion sample of 6,410 women, as com-pared to 48 in the Fergusson et al. study. Fourth, unlikethe study by Fergusson et al., the Gilchrist et al. studycontrolled for number of prior abortions and births.For these reasons, the TFMHA had more confidence inarriving at conclusions about relative risk based on thefindings of Gilchrist et al. Nonetheless, it should benoted that the abortion context in the United Kingdommay differ from that in the United States, weakeninggeneralization to the U.S. context.

The TFMHA reviewed six studies that comparedwomen’s responses following an induced abortionfor fetal abnormality to women’s responses followingother reproductive events. These studies were basedon extremely small samples often characterized byhigh attrition rates and low response rates. Nonethe-less, these studies suggest that terminating a wantedpregnancy, especially late in pregnancy, can be asso-ciated with negative psychological experiences com-parable to those experienced by women whomiscarry a wanted pregnancy or experience a still-birth or death of a newborn, but less severe thanthose experienced by women who deliver a childwith a severe abnormality. At least one study alsosuggests that the majority of women who make thisdifficult choice do not regret their decision (e.g.,Kersting et al., 2005). As a group, these studies ofresponses to termination of a wanted pregnancy forfetal abnormality underscore the importance of con-sidering the wantedness of the pregnancy, as well asthe reason for and timing of the abortion, in studyingits psychological implications. Interpretation ofprevalence of psychological distress and relative riskis clouded when researchers lump together under thecategory of “abortion” women who abort a wantedpregnancy for reasons of fetal anomaly with womenwho have an elective abortion of an unplanned andunwanted pregnancy.

In summary, although numerous methodologicalflaws prevent the published literature from provid-ing unequivocal evidence regarding the relative mentalhealth risks associated with abortion per se comparedto its alternatives (childbirth of an unplanned preg-nancy), in the view of the TFMHA, the best scientificevidence indicates that the relative risk of mentalhealth problems among adult women who have an un-planned pregnancy is no greater if they have an elec-tive first-trimester abortion than if they deliver thatpregnancy (Gilchrist et al., 1995).

The evidence regarding the relative mental health risksassociated with multiple abortions is more equivocal.One source of inconsistencies in the literature may bemethodological, such as differences in sample size orage ranges among samples. Positive associations ob-served between multiple abortions and poorer mentalhealth (e.g., Harlow et al., 2004) also may be due toco-occurring risks that predispose a woman to bothunwanted pregnancies and mental health problems.

Terminating a wanted pregnancy late in pregnancydue to fetal abnormality appears to be associated withnegative psychological experiences equivalent to thoseexperienced by women who miscarry a wanted preg-nancy or experience a stillbirth or the death of a new-born.

Prevalence ofMental Health Problems AmongU.S.WomenWhoHave anAbortionA second question this literature has been used to ad-dress concerns the prevalence of mental health prob-lems among women in the United States who have hadan abortion. As noted at the outset of this report, re-search capable of adequately addressing this questionrequires at minimum: (1) a clearly defined, agreedupon, and appropriately measured mental healthproblem (e.g., a clinically significant disorder, assessedvia validated criteria); (2) a sample representative ofthe population to which one wants to generalize (e.g.,women in the United States); and (3) knowledge of theprevalence of the same mental health problemin the general population, equated with the abortiongroup with respect to potentially confounding fac-tors. None of the studies reviewed met all these crite-ria and hence provided sound evidence regardingprevalence. Few of the U.S studies assessed clinicallysignificant disorders with valid and reliable measuresor physician diagnosis. In those studies that did use


clinically relevant outcome measures, sampling strate-gies were inadequate to address the question of preva-lence in the larger U.S. population either because thesamples were biased, highly selected, geographicallyrestricted, or failed to use appropriate samplingweights. Furthermore, because of the lack of adequatecontrol for co-occurring risks, the extent to which theincidence of mental health problems associated withabortion was due to the procedure versus to poten-tially confounding factors such as poverty, poorerprior mental health, etc., was impossible to establish.

Given these caveats, however, the prevalence of mentalhealth problems observed among women in the UnitedStates who had a single, legal, first- trimester abortionfor nontherapeutic reasons appeared to be consistentwith normative rates of comparable mental healthproblems in the general population of women in theUnited States. Consider, for example, the overallprevalence of depression among women in the NLSY,a longitudinal national survey of a cohort of men andwomen aged 14–21 years in 1979. Among all womenin the NLSY, irrespective of reproductive history andwithout controlling for any covariates, 22% met crite-ria for depression in 1992 (i.e., scored above the clini-cal cutoff on the CES-D). Among women whoreported one abortion, the corresponding percentagewas 23%. Among women who reported multipleabortions, however, the percentage was higher; 31%met criteria for depression (see Table 6).5 A similarpattern was reported by Harlow et al. (2004) in theirstudy of a representative sample of women in theBoston metropolitan area.

To say that women in general do not show an in-creased incidence of mental health problems followinga single abortion, however, does not mean that nowomen experience such problems. Abortion is an ex-perience often hallmarked by ambivalence, and a mixof positive and negative emotions is to be expected(Adler et al., 1990; Dagg, 1991). Some women experi-ence beneficial outcomes, whereas others experiencesadness, grief, and feelings of loss following the elec-tive termination of a pregnancy. Some women experi-ence clinically significant outcomes, such as depressionor anxiety. However, the TFMHA reviewed no evi-dence sufficient to support the claim that an observedassociation between abortion history and a mentalhealth problem was caused by the abortion per se, asopposed to other factors. As observed throughout this

report, unwanted pregnancy and abortion are corre-lated with preexisting conditions (e.g., poverty), lifecircumstances (e.g., exposure to violence, sexualabuse), problem behaviors (e.g., drug use), and per-sonality characteristics (e.g., avoidance style of copingwith negative emotion) that can have profound andlong-lasting negative effects on mental health. Differ-ences in prevalence of mental health problems orproblem behaviors observed between women whohave had an abortion and women who have not maybe primarily accounted for by these preexisting andongoing differences among groups.

Predictors of Individual Variationin Responses Following AbortionA third issue addressed in the literature on abortionand mental health concerns individual variation inwomen’s psychological experiences following abor-tion. The TFMHA reviewed 23 papers based on 15data sets that were based solely on samples of womenwho had abortions in the United States, but that oth-erwise met inclusion criteria. These noncomparisongroup studies typically focused on predictors of indi-vidual variation in response. They were of two major


Table 6Population estimates of proportion of all women and womenidentified as having been pregnant exceeding CES-D clinicalcutoff score,National Longitudinal Survey of Youth: 1992.

Group (N) CES-D> 15

All women(unweighted N= 4401) 22 %

No abortion ever 21 %

Ever abortion 25 %

One abortion 23 %

Multiple abortions 31 %

All women ever pregnant+(unweighted N=3503) 23 %

No abortion ever 23 %

Ever abortion 25 %

One abortion 22 %

Multiple abortions 31 %

Notes: +Includes pregnancies ending in miscarriages.No covariates are controlled.

types: (1) prospective or concurrent studies that usu-ally included preabortion measures of psychologicaladjustment and risk factors and one or more postabor-tion assessments of adjustment, and (2) retrospectivestudies that assessed women’s perceived reactions tothe event and current level of psychological function-ing several years after the abortion. The retrospectivestudies had serious methodological problems thatmade interpretation of their findings difficult. Theprospective studies, despite limitations of high attri-tion, geographically limited samples, and potentialconfounds that were not measured, provided valuableinformation about sources of variation in individualwomen’s psychological experiences and, to a morelimited extent, mental health problems subsequent toabortion.

The most methodologically strong studies in thisgroup showed that interpersonal concerns, includingfeelings of stigma, perceived need for secrecy, exposureto antiabortion picketing, and low perceived or antici-pated social support for the abortion decision, nega-tively affected women’s postabortion psychologicalexperiences. Characteristics of the woman also pre-dicted more negative psychological experiences afterfirst-trimester abortion, including a prior history ofmental health problems, personality factors such aslow self-esteem and low perceived control over herlife, and use of avoidance and denial coping strategies.Feelings of commitment to the pregnancy, ambivalenceabout the abortion decision, and low perceived abilityto cope with the abortion prior to its occurrence alsopredicted more negative postabortion responses.Across studies, prior mental health emerged as thestrongest predictor of postabortion mental health(Major et al., 2000). Type of abortion procedures, atleast those used in the first trimester, did not appear tobe related to postabortion psychological well-being ormental health.

In considering these risk factors, it is important torecognize that many of the same factors shown to beassociated with more negative postabortion psycho-logical experiences also predict more negative reac-tions to other types of stressful life events, includingchildbirth (e.g., low perceived social support, low self-esteem, low self-efficacy, avoidance coping). For in-stance, low perceived social support and lowself-esteem also are risk factors for postpartum depres-sion (Beck, 2001; Logsdon & Usui, 2001). Most risk

factors are not uniquely predictive of psychologicalexperiences following abortion. Women characterizedby one or more such risk factors might be equally (ormore) likely to experience negative psychological reac-tions if they pursued an alternative course of action(motherhood or adoption).

Conclusions and Future ResearchBased on our comprehensive review and evaluation ofthe empirical literature published in peer-reviewedjournals since 1989, this Task Force on Mental Healthand Abortion concludes that the most methodologi-cally sound research indicates that among women whohave a single, legal, first-trimester abortion of an un-planned pregnancy for nontherapeutic reasons, the rel-ative risks of mental health problems are no greaterthan the risks among women who deliver an un-planned pregnancy. This conclusion is generally con-sistent with that reached by the first APA task force(Adler et al., 1990).

This report has highlighted the methodological failingsthat are pervasive in the literature on abortion andmental health. This focus on methodological limita-tions raises the question of whether empirical scienceis capable of informing understanding of the mentalhealth implications of and public policy related toabortion. Some policy questions cannot be definitivelyanswered through empirical research because they arenot pragmatically or ethically possible.

Other questions, however, are amenable to the meth-ods of well-designed, rigorously conducted scientificresearch. For example, empirical research can identifythose women who might be more or less likely thanothers to show adverse or positive psychological out-comes following an abortion. Well-designed researchcan also answer questions of relative risk and preva-lence. What would this research look like?

Such research would use methods that are prospectiveand longitudinal and employ exacting sampling meth-ods (including the use of sampling weights that allowproper generalization back to the populations to whomthe conclusions are being applied). Careful attentionwould be paid to adequately assessing preexisting andco-occurring conditions such as marital status, domes-tic violence, age, socioeconomic status, parity, priormental health, and prior problem behaviors, as well asother situations that are known to be associated with



both differential utilization of abortion and mentalhealth problems. Importantly, comparison groupswould be selected so as to be equivalent to the abortiongroup on all variables other than abortion history. Crit-ical variables such as intendedness and wantedness ofthe pregnancy would be assessed, and abortion statusverified objectively (not only through self-report).Careful use of covariance or similar adjustment tech-niques (applied to pre-defined covariates) would beemployed. Precision of measurement (both in terms ofspecification of outcome measure and psychometricadequacy of the measurements) would also be guaran-teed. Positive psychological responses and experiencesas well as negative mental health would be assessed.Repeated assessment of responses over time would bemade to assess relevant changes, positive and negative,in the trajectory of responses following abortion. Sam-ples sufficiently large to guarantee adequate power todetect effects that are present would be used, and at-tention would be paid to effect-size estimation in addi-tion to the simple reliance of null hypothesis statisticaltesting.

Research that met the above scientific standardswould help to disentangle confounding factors and es-tablish relative risks of abortion compared to its alter-natives. Even so, there is unlikely to be a singledefinitive research study that will determine the men-tal health implications of abortion “once and for all”as there is no “all,” given the diversity and complexityof women and their circumstances. Important agendasfor future research are to further understand and alle-viate the conditions that lead to unwanted pregnancyand abortion and to understand the conditions thatshape how women respond to these life events, withthe ultimate goal of improving women’s lives andwell-being.

ENDNOTES

1. In an attempt to assess whether underreporting ofabortion might have biased findings in the NLSY,Russo and Dabul (1997) also undertook a reanaly-sis of the NLSY data to examine whether the rela-tionship between reproductive outcomes andself-esteem held across racial and religious groupsknown to vary in underreporting, specifically Blackversus White and Catholic versus non-Catholicgroups. They again found that neither having oneabortion nor having repeat abortions was signifi-cantly related to RSE when contextual variableswere controlled. They also found that the pattern ofrelationships did not vary by race or religion. Thissuggests that differential underreporting by somegroups did not introduce systematic bias into the re-sults.

2. Personal communication to NFR from David Fer-gusson, e-mail, 8/8/2007.

3. Although no women in the subgroup with a previ-ous history of DSH were identified as havinga postpregnancy psychotic episode, the number ofwomen in that category (N = 36) was too small forreliable analysis by reproductive outcome.

4. Personal communication from Ellie Lee.

5. The TFMHA would like to thank K. C. Blackwellfor providing these analyses.



REFERENCES

Adler, N. E. (1976). Sample attrition in studies ofpsychological sequelae of abortion: How great aproblem? Journal of Applied Social Psychology, 6,240-259.

Adler, N. E., David, H. P., Major, B. N., Roth, S. H.,Russo, N. F., & Wyatt, G. E. (1990). Psychological re-sponses after abortion. Science, 248, 41-44.

Adler, N. E., David, H. P., Major, B. N., Roth, S. H.,Russo, N. F., & Wyatt, G. E. (1992). Psychologicalfactors in abortion. American Psychologist, 47, 1194-1204.

Aldwin, C., & Revenson, T. A. (1987). Does copinghelp? A reexamination of the relation between copingand mental health. Journal of Personality and SocialPsychology, 53, 337-348.

American Psychiatric Association. (2002). Diagnosticand statistical manual of mental disorders (4th ed.,text revision) (DSM-IV-TR). Arlington, VA: Author.

Ashok, P. W., Hamoda, H., Flett, G. M., Kidd, A.,Fitzmaurice, A., & Templeton, A. (2005). Psychologi-cal sequelae of medical and surgical abortion at 10–13weeks’ gestation. Acta Obstetricia et GynecologicaScandinavica, 84, 761-766.

Bailey, P. E., Bruno, Z. V., Bezerra, M. F., Queiroz, I.,Oliveira, C. M., & Chen-Mok, M. (2001). Adolescentpregnancy 1 year later: The effect of abortion vs.motherhood in Northeast Brazil. Journal of Adoles-cent Health, 29, 223-232.

Barber, J. S., Axinn, W. G., & Thornton, A. (1999).Unwanted childbearing, health, and mother–childrelationships. Journal of Health and Social Behavior,40, 231-257.

Barnett, W., Freudenberg, N., & Wille, R. (1992).Partnership after induced abortion: A prospectivecontrolled study. Archives of Sexual Behavior, 21,443-455.

Bazelon, E. (2007, January 21). Is there a post-abortion syndrome? New York Times Magazine,pp. 40-47, 62, 66, 70.

Beck, C. T. (2001). Predictors of postpartum depres-sion: An update. Nursing Research, 50, 275-285.

Billings, A. G., & Moos, R. H. (1981). The role ofcoping responses and social resources in attenuatingthe impact of stressful life events. Journal of BehavioralMedicine, 4, 139-157.

Blascovich, J., Spencer, S. J., Quinn, D., & Steele, C.M. (2001). African Americans and high blood pres-sure: The role of stereotype threat. Psychological Sci-ence, 12, 225-229.

Bolzendahl, C., & Brooks, C. (2005). Polarization,secularization, or differences as usual? The denomina-tional cleavage in U.S. social attitudes since the 1970s.The Sociological Quarterly, 46, 47-78.

Boonstra, H., Gold, R., Richards, C., & Finer, L.(2006). Abortion in women’s lives. New York:Guttmacher Institute.

Boyer, D., & Fine, D. (1992). Sexual abuse as a factorin adolescent pregnancy and child maltreatment. Fam-ily Planning Perspectives, 24, 4-11.

Bradshaw, Z., & Slade, P. (2003). The effects of in-duced abortion on emotional experiences and relation-ships: A critical review of the literature. ClinicalPsychology Review, 23, 929-958.

Bradshaw, Z., & Slade, P. (2005). The relationshipbetween induced abortion, attitudes toward sexuality,and sexual problems. Sexual and RelationshipTherapy, 20, 390-406.

Breslau, N., Kessler, R. C., Chilcoat, H. D., Schultz, L.R., Davis, G. C., & Andreski, P. (1998). Trauma andposttraumatic stress disorder in the community: The1996 Detroit Area Survey of Trauma. Archives ofGeneral Psychiatry, 55, 626-632.

Broen, A. N., Moum, T., Bodtker, A. S., & Ekeberg,O. (2004). Psychological impact on women of miscar-riage versus induced abortion: A 2-year follow-upstudy. Psychosomatic Medicine, 66, 265-271.

Broen, A. N., Moum, T., Bodtker, A. S., & Ekeberg,O. (2005a). The course of mental health after miscar-riage and induced abortion: A longitudinal, five-yearfollow-up study. BMC Medicine, 3, 18.

Broen, A. N., Moum, T., Bodtker, A. S., & Ekeberg,O. (2005b). Reasons for induced abortion and theirrelation to women’s emotional distress: A prospective,two-year follow-up study. General Hospital Psychia-try, 27, 36-43.

Broen, A. N., Moum, T., Bodtker, A. S., & Ekeberg,O. (2006). Predictors of anxiety and depression fol-lowing pregnancy termination: A longitudinal five-year follow-up study. Acta Obstetricia etGynecologica Scandinavica, 85, 317-323.

Brown, G. W., & Harris, T. (1978). Social origins ofdepression. New York: The Free Press.

Burgoine, G. A., Van Kirk, S. D., Romm, J., Edelman,A. B., Jacobson, S. L., & Jensen, J. T. (2005). Compar-ison of perinatal grief after dilation and evacuation orlabor induction in second-trimester terminations forfetal anomalies. American Journal of Obstetrics andGynecology, 192, 1928-1932.

Chouinard, E., & Walter, S. (1994). Recall bias incase-control studies: An empirical analysis and theo-retical framework. Journal of Clinical Epidemiology,48, 245-254.

Cohan, C. L., Dunkel-Schetter, C., & Lydon, J.(1993). Pregnancy decision making: Predictors of earlystress and adjustment. Psychology of Women Quar-terly, 17, 223-239.

Cohen, S. A. (2006). Abortion and mental health:Myths and reality. Guttmacher Policy Review, 9,8-11, 16.

Coker, A. L. (2007). Does physical intimate partner vi-olence affect sexual health? A systematic review.Trauma, Violence, and Abuse, 8, 149-177.

Coleman, P. K. (2006a). Resolution of unwanted preg-nancy during adolescence through abortion versuschildbirth: Individual and family predictors and psy-chological consequences. Journal of Youth and Ado-lescence, 35, 903-911.

Coleman, P. K. (2006b). [Testimony in South DakotaPlanned Parenthood Minnesota vs. Rounds], No. Civ.05-4077-KES, 2006 U.S. Dist. LEXIS 72778 (D.S.D.Oct. 4, 2006).

Coleman, P. K., Maxey, C. D., Rue, V. M., & Coyle,C. T. (2005). Associations between voluntary and in-voluntary forms of perinatal loss and child maltreat-ment among low-income mothers. Acta Paediatrica,94, 1476-1483.

Coleman, P. K., & Nelson, E. S. (1998). The quality ofabortion decisions and college students’ reports ofpost-abortion emotional sequelae and abortion atti-tudes. Journal of Social and Clinical Psychology, 17,425-442.

Coleman, P. K., Reardon, D. C., & Cougle, J. (2002).The quality of the caregiving environment and childdevelopmental outcomes associated with maternal his-tory of abortion using the NLSY data. Journal ofChild Psychology and Psychiatry, 43, 743-757.

Coleman, P. K., Reardon, D. C., & Cougle, J. R.(2005). Substance use among pregnant women in thecontext of previous reproductive loss and desire forcurrent pregnancy. British Journal of Health Psychol-ogy, 10, 255-268.

Coleman, P. K., Reardon, D. C., Rue, V. M., &Cougle, J. (2002a). A history of induced abortion inrelation to substance use during subsequent pregnan-cies carried to term. American Journal of Obstetricsand Gynecology, 187, 1673-1678.

Coleman, P. K., Reardon, D. C, Rue, V. M., &Cougle, J. (2002b). State-funded abortions versus de-liveries: A comparison of outpatient mental healthclaims over 4 years. American Journal of Orthopsychi-atry, 72, 141-152.

Coleman, P. K., Reardon, D. C., Strahan, T., &Cougle, J. R. (2005). The psychology of abortion: Areview and suggestions for future research. Psychologyand Health, 20, 237-271.

Congleton, G. K., & Calhoun, L. G. (1993). Post-abortion perceptions: A comparison of self-identifieddistressed and non-distressed populations. Interna-tional Journal of Social Psychiatry, 39, 255-265.

Conklin, M. P., & O’Connor, B. P. (1995). Beliefsabout the fetus as a moderator of post-abortion psy-chological well-being. Journal of Social and ClinicalPsychology, 14, 76-95.



Cook, E. A., Jelen, T. G., & Wilcox, C. (1992). Be-tween two absolutes: Public opinion and the politicsof abortion. Boulder, CO: Westview Press.

Cooper, M. L., Wood, P. K., Orcutt, H. K., & Albino,A. (2003). Personality and the predisposition to en-gage in risky or problem behaviors during adoles-cence. Journal of Personality and Social Psychology,84, 390-410.

Costa, F., Jessor, R., & Donovan, J. E. (1987).Psychosocial correlates and antecedents of abortion:An exploratory study. Population and Environment:A Journal of Interdisciplinary Studies, 9, 3-22.

Cougle, J. R., Reardon, D. C., & Coleman, P. K.(2003). Depression associated with abortion andchildbirth: A long-term analysis of the NLSY cohort.Medical Science Monitor, 9, CR105-112.

Cougle, J. R., Reardon, D. C., & Coleman, P. K.(2005). Generalized anxiety following unintendedpregnancies resolved through childbirth and abortion:A cohort study of the 1995 National Survey of FamilyGrowth. Journal of Anxiety Disorders, 19, 137-142.

Cozzarelli, C. (1993). Personality and self-efficacy aspredictors of coping with abortion. Journal of Person-ality and Social Psychology, 65, 1224-1236.

Cozzarelli, C., & Major, B. (1994). The effects of anti-abortion demonstrators and pro-choice escorts onwomen’s psychological responses to abortion. Journalof Social and Clinical Psychology, 13, 404-427.

Cozzarelli, C., Major, B., Karrasch, A., & Fuegen, K.(2000). Women’s experiences of and reactions to an-tiabortion picketing. Basic and Applied Social Psy-chology, 22, 265-275.

Cozzarelli, C., Sumer, N., & Major, B. (1998). Men-tal models of attachment and coping with abortion.Journal of Personality and Social Psychology, 74,453-467.

Crocker, J., Major, B., & Steele, C. (1998). Socialstigma. In S. Fiske, D. Gilbert, & G. Lindzey (Eds.),Handbook of social psychology (Vol. 2, pp. 504-553).Boston: McGraw-Hill.

David, H. D., Dytrych, Z., & Matejcek, Z. (2003).Born unwanted: Observations from the Prague Study.American Psychologist, 58, 224-229.

Dagg, P. K. B. (1991). The psychological sequelae oftherapeutic abortion—denied and completed. Ameri-can Journal of Psychiatry, 148, 578-585.

Derogatis, R. L. (1975). Affect balance scale. Balti-more: Clinical Psychometrics Research.

Dietz, P. M., Spitz, A. M., Anda, R. F., Williamson,D. F., McMahon, P. M., Santelli, J. S., Nordenberg,D. F., Felitti, V. J., & Kendrick, J.S. (1999). Unin-tended pregnancy among adult women exposed toabuse or household dysfunction during their child-hood. Journal of the American Medical Association,282, 1359-1364.

Dryfoos, J. D. (1990). Adolescents at risk: Prevalenceand prevention. New York: Oxford University Press.

Dugger, K. (1998). Black women and the question ofabortion. In L. J. Beckman & S. M. Harvey (Eds.),The new civil war: The psychology, culture, and poli-tics of abortion (pp. 107-132). Washington, DC:American Psychological Association.

Elashoff, J. D. (1969). Analysis of covariance: A deli-cate instrument. American Educational Research Jour-nal, 6, 383-401.

Erickson, P. I., & Kaplan, C. P. (1998). Latinas andabortion. In L. J. Beckman & S. M. Harvey (Eds.),The new civil war: The psychology, culture, and poli-tics of abortion (pp. 133-156). Washington, DC:American Psychological Association.

Felton, G. M., Parsons, M. A., & Hassell, J. S. (1998).Health behavior and related factors in adolescentswith a history of abortion and never-pregnantadolescents. Health Care for Women International,19, 37-47.

Fergusson, D. M., Boden, J. M., & Horwood, L. J.(2007). Abortion among young women and subse-quent life outcomes. Perspectives on Sexual andReproductive Health, 39, 6-12.

Fergusson, D. M., Horwood, L. J., & Ridder, E. M.(2006). Abortion in young women and subsequentmental health. Journal of Child Psychology and Psy-chiatry, 47, 16-24.

Finer, L. B., Frohwirth, L. F., Dauphinee, L. A., Singh,S., & Moore, A. M. (2005). Reasons U.S. womenhave abortions: Quantitative and qualitative perspec-tives. Perspectives on Sexual and ReproductiveHealth, 37, 110-118.

Finer, L. B., & Henshaw, S. K. (2006a). Disparities inrates of unintended pregnancy in the United States,1994 and 2001. Perspectives on Sexual and Reproduc-tive Health, 38, 90-95.

Finer, L. B., & Henshaw, S. K. (2006b). Estimates ofU.S. abortion incidence, 2001-2003. New York:Guttmacher Institute. Retrieved October 26, 2007,from http://www.guttmacher.org/pubs/2006/08/03/ab_incidence.pdf

Franz, W., & Reardon, D. (1992). Differential impactof abortion on adolescents and adults. Adolescence,27, 161-172.

Fu, H., Darroch, J. E., Henshaw, S. K., & Kolb, E.(1998). Measuring the extent of abortion underreport-ing in the 1995 National Survey of Family Growth.Family Planning Perspectives, 30, 128-133.

Gazmararian, J. A., Lazorick, S., Spitz, A. M., Ballard,T. J., Saltzman, L. E., & Marks, J. S. (1996). Preva-lence of violence against pregnant women. Journal ofthe American Medical Association, 275, 1915-1920.

Gilchrist, A. C., Hannaford, P. C., Frank, P., & Kay,C. R. (1995). Termination of pregnancy and psychi-atric morbidity. British Journal of Psychiatry, 167,243-248.

Gissler, M., Berg, C., Bouvier-Colle, M. H., &Buekens, P. (2004a). Methods for identifying preg-nancy-associated deaths: Population-based data fromFinland 1987-2000. Paediatric and Perinatal Epidemi-ology, 18, 448-455.

Gissler, M., Berg, C., Bouvier-Colle, M. H., &Buekens, P. (2004b). Pregnancy-associated mortalityafter birth, spontaneous abortion, or induced abortionin Finland, 1980-2000. American Journal of Obstet-rics and Gynecology, 190, 422-427.

Gissler, M., Hemminki, E., & Lonnqvist, J. (1996).Suicides after pregnancy in Finland, 1987-94: Registerlinkage study. British Medical Journal, 313, 1431-1434.

Gissler, M., Kauppila, R., Merilainen, J., Toukomaa,H., & Hemminki, E. (1997). Pregnancy-associateddeaths in Finland 1987-1994—Definition problemsand benefits of record linkage. Acta Obstetricia et Gy-necologica Scandinavica, 76, 651-657.

Golding, J. M. (1999). Intimate partner violence as arisk factor for mental disorders: A meta-analysis. Jour-nal of Family Violence, 14, 99-132.

Greenberg, B. G., Kuebler, R. R., Abernathy, J. R., &Horvitz., D. G. (1971). Applications of randomizedresponse technique in obtaining quantitative data.Journal of the American Statistical Association, 66,243-256.

Grote, N. K., & Bledsoe, S. (2007). Predicting post-partum depressive symptoms in new mothers: The roleof optimism and stress frequency during pregnancy.Health and Social Work, 32, 107-118.

Harlow, B. L., Cohen, L. S., Otto, M. W., Spiegelman,D., & Cramer, D. W., (2004). Early life menstrualcharacteristics and pregnancy experiences amongwomen with and without major depression: The Har-vard Study of Moods and Cycles. Journal of AffectiveDisorders, 79, 167-176.

Henshaw, S. K. (1998). Unintended pregnancy inthe United States. Family Planning Perspectives, 30,24-29, 46.

Hope, T. L., Wilder, E. I., & Terling Watt, T. (2003).The relationships among adolescent pregnancy, preg-nancy resolution, and juvenile delinquency. Sociologi-cal Quarterly, 44, 555-576.


Howie, F. L., Henshaw, R. C., Naji, S. A., Russell,I. T., & Templeton, A. (1997). Medical abortion orvacuum aspiration? Two-year follow up of a patientpreference trial. British Journal of Obstetrics andGynaecology, 104, 829-833.

Iles, S., & Gath, D. (1993). Psychiatric outcome of ter-mination of pregnancy for fetal abnormality. Psycho-logical Medicine, 23, 407-413.

Jessor, S. L., & Jessor, R. (1977). Problem behaviorand psychosocial development. New York: AcademicPress.

Jones E. F., & Forrest J. D. (1992) Contraceptive fail-ure rates based on the 1988 NSFG. Family PlanningPerspectives, 4, 12-19.

Jones, R. K., Darroch, J. E., & Henshaw, S.. K.(2002a). Contraceptive use among U.S. women havingabortions in 2000-2001. Perspectives on Sexual andReproductive Health, 35, 294-303.

Jones, R. K., Darroch, J. E., & Henshaw, S. K.(2002b). Patterns in the socioeconomic characteristicsof women obtaining abortions in 2000-2001. Perspec-tives on Sexual and Reproductive Health, 34, 226-235.

Jones, R. K., & Kost, K. (2007). Underreporting of in-duced and spontaneous abortion in the United States:An analysis of the 2002 National Survey of FamilyGrowth. Studies in Family Planning, 38, 187-197.

Jones, R. K., Zolna, M. R. S., Henshaw, S. K., &Finer, L. B. (2008). Abortion in the United States: Inci-dence and access to services, 2005. Perspectives onSexual and Reproductive Health, 40, 6-16.

Kandel, D. B. (1989). Issues of sequencing of adoles-cent drug use and other problem behaviors. Drugs andSociety, 3, 55-76.

Kersting, A., Dorsch, M., Kreulich, C., Reutemann,M., Ohrmann, P., Baez, E., & Arolt, V. (2005).Trauma and grief 2-7 years after termination of preg-nancy because of fetal anomalies—A pilot study. Jour-nal of Psychosomatic Obstetrics and Gynecology, 26,9-14.

Kessler, R. C., Avenevoli, S., Merikangas, K. R.(2001). Mood disorders in children and adolescents:An epidemiologic perspective. Biological Psychiatry,49, 1002-1014.

Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C.B., Hughes, M., Eshleman, S., Wittchen, H. U., &Kendler, K. S. (1994). Lifetime and 12-month preva-lence of DSM-III-R psychiatric disorders in the UnitedStates. Archives of General Psychiatry, 51, 8-18.

Koop, C. E. (1989). The federal role in determiningthe medical and psychological impact of abortions onwomen (HR No. 101-392, p.14). Testimony given tothe Committee on Government Operations, U.S.House of Representatives, 101st Congress, 2nd ses-sion, December 11, 1989.

Lauzon, P., Roger-Achim, D., Achim, A., & Boyer, R.(2000). Emotional distress among couples involved infirst-trimester induced abortions. Canadian FamilyPhysician, 46, 2033-2040.

Lazarus, R. S., & Folkman, S. (1984). Coping andadaptation. In W. D. Gentry (Ed.). The handbook ofbehavioral medicine (pp. 282-325). New York: Guil-ford.

Lee, E. (2003). Abortion, motherhood and mentalhealth: Medicalizing reproduction in the United Statesand Great Britain. New York: Aldine de Gruyter.

Lemkau, J. P. (1991). Post-abortion adjustment ofhealth care professionals in training. American Journalof Orthopsychiatry, 61, 92-102.

Link, B. G., Struening, E. L., Rahav, M., Phelan, J. C.,& Nuttbrock, L. (1997). On stigma and its conse-quences: Evidence from a longitudinal study of menwith dual diagnoses of mental illness and substanceabuse. Journal of Health and Social Behavior, 38,177-190.

Logsdon, M. C., & Usui, W. (2001). Psychosocial pre-dictors of postpartum depression in diverse groups ofwomen. Western Journal of Nursing Research, 23,563-574.


Lorenzen, J., & Holzgreve, W. (1995). Helping par-ents to grieve after secondtrimester termination ofpregnancy for fetopathic reasons. Fetal Diagnosis andTherapy, 10, 147-156.

Lowenstein, L., Deutcsh, M., Gruberg, R., Solt, I.,Yagil, Y., Nevo, O., & Bloch, M. (2006). Psychologi-cal distress symptoms in women undergoing medicalvs. surgical termination of pregnancy. General Hospi-tal Psychiatry, 28, 43-47.

Lydon, J., Dunkel-Schetter, C., Cohan, C. L., &Pierce, T. (1996). Pregnancy decision-making as a sig-nificant life event: A commitment approach. Journalof Personality and Social Psychology, 71, 141-151.

MacNair, R. M. (2005). Perpetration-induced trau-matic stress: The psychological consequences ofkilling. New York: Authors Choice Press.

Major, B., & Cozzarelli, C. (1992). Psychosocial pre-dictors of adjustment to abortion. Journal of Social Is-sues, 48, 121-142.

Major, B., Cozzarelli, C., Cooper, M. L., Zubek, J.,Richards, C., Wilhite, M., et al. (2000). Psychologicalresponses of women after first-trimester abortion.Archives of General Psychiatry, 57, 777-784.

Major, B., Cozzarelli, C., Sciacchitano, A. M., Cooper,M. L., Testa, M., & Mueller, P. M. (1990). Perceivedsocial support, self-efficacy, and adjustment to abor-tion. Journal of Personality and Social Psychology, 59,452-463.

Major, B., Cozzarelli, C., Testa, M., & Mueller, P.(1992). Male partners’ appraisals of undesired preg-nancy and abortion: Implications for women’s adjust-ment to abortion. Journal of Applied SocialPsychology, 22, 599-614.

Major, B., & Gramzow, R. H. (1999). Abortion asstigma: Cognitive and emotional implications of con-cealment. Journal of Personality and Social Psychol-ogy, 77, 735-745.

Major, B. N., Mueller, P. M., & Hildebrandt, K.(1985). Attributions, expectations, and coping withabortion. Journal of Personality and Social Psychol-ogy, 48, 585-599.

Major, B., & O’Brien, L. T. (2005). The social psy-chology of stigma. Annual Review of Psychology, 56,393-421.

Major, B., Richards, C., Cooper, M., Cozzarelli, C., &Zubek, J. (1998). Personal resilience, cognitive ap-praisals, and coping: An integrative model of adjust-ment to abortion. Journal of Personality and SocialPsychology, 74, 735-752.

Major, B., Zubek, J. M., Cooper, M., Cozzarelli, C.,& Richards, C. (1997). Mixed messages: Implicationsof social conflict and social support within close rela-tionships for adjustment to a stressful life event. Jour-nal of Personality and Social Psychology, 72,1349-1363.

Martino, S. C., Collins, R. L., Ellickson, P. L., &Klein, D. J. (2006). Exploring the link between sub-stance use and abortion: The roles of unconventional-ity and unplanned pregnancy. Perspectives on Sexualand Reproductive Health, 38, 66–75.

Mather, M., & Rivers, K. L. (2006). City profiles ofchild well-being: Results from the American Commu-nity Survey. Washington, DC: Annie E. Casey Founda-tion.

McCall, R. B., & Appelbaum, M. I. (1991). Some is-sues of conducting secondary analyses. DevelopmentalPsychology, 27, 911-917.

Medora, N. P., Goldstein, A., & von der Hellen, C.(1993). Variables related to romanticism and self-esteem in pregnant teenagers. Adolescence, 28, 159-170.

Mensch, B., & Kandel, D. B. (1992). Drug use as arisk factor for premarital teen pregnancy and abortionin a national sample of young White women. Demog-raphy, 29, 409-429.

Messer, L. C., Kaufman, J. S., Dole, N., Savitz, D. A.,& Laraia, B. A. (2006). Neighborhood crime, depriva-tion, and preterm birth. Annals of Epidemiology, 16,455-462.

Miller, W. B. (1992). An empirical study of the psy-chological antecedents and consequences of inducedabortion. Journal of Social Issues, 48, 67-93.


Moore, K. A. (1995). Executive summary: Nonmaritalchildbearing in the United States. In U.S Departmentof Health and Human Services Working Group onNonmarital Childbearing (Eds.), Report to Congresson out-of-wedlock childbearing (DHHS PublicationNo. PHS 95-1257) (pp. ii-xxii). Washington, DC: De-partment of Health and Human Services. RetrievedOctober 20, 2007, from http://www.cdc.gov/nchs/data/misc/wedlock.pdf

Mueller, P., & Major, B. (1989). Self-blame, self-efficacy, and adjustment to abortion. Journal ofPersonality and Social Psychology, 57, 1059-1068.

Neugebauer, R., & Ng, S. (1990). Differential recall asa source of bias in epidemiologic research. Journal ofClinical Epidemiology, 43, 1337-1341.

Ney, P. G., Fung, T., Wickett, A. R., & Beaman-Dodd,C. (1994). The effects of pregnancy loss on women’shealth. Social Science and Medicine, 38, 1193-1200.

Pallitto, C. C., & O’Campo, P. (2005). Communitylevel effects of gender inequality on intimate partnerviolence and unintended pregnancy in Colombia: Test-ing the feminist perspective. Social Science and Medi-cine, 60, 2205-2216.

Phelps, R. H., Schaff, E. A., & Fielding, S. L. (2001).Mifepristone abortion in minors. Contraception, 64,339-343.

Pope, L. M., Adler, N. E., & Tschann, J. M. (2001).Postabortion psychological adjustment: Are minorsat increased risk? Journal of Adolescent Health, 29,2-11.

Posavac, E., & Miller, T. (1990). Some problemscaused by not having a conceptual foundation forhealth research: An illustration from studies of thepsychological effects of abortion. Psychology andHealth, 5, 13-23.

Quinton, W. J., Major, B., & Richards, C. (2001).Adolescents and adjustment to abortion: Are minorsat greater risk? Psychology, Public Policy, and Law, 7,491-514.

Radloff, L. S. (1977). The CES-D scale: A self-reportdepression scale for research in the general popula-tion. Applied Psychological Measurement, 1, 385-401.

Reardon, D. C. (2007). A new strategy for endingabortion: Learning the truth—telling the truth.Downloaded on January 2008 from http://www.afterabortion.org

Reardon, D. C., & Coleman, P. K. (2006). Relativetreatment for sleep disorders following abortion andchild delivery: A prospective record-based study. Sleep,29, 105-106.

Reardon, D. C., Coleman, P. K., & Cougle, J. R.(2004). Substance use associated with unintendedpregnancy outcomes in the National Longitudinal Sur-vey of Youth. American Journal of Drug and AlcoholAbuse, 30, 369-383.

Reardon, D. C., & Cougle, J. R. (2002a). Depressionand unintended pregnancy in the National Longitudi-nal Survey of Youth: A cohort study. British MedicalJournal, 324, 151-152.

Reardon, D. C., & Cougle, J. R. (2002b). Depressionand unintended pregnancy in the National Longitudi-nal Survey of Youth: A cohort study: Reply. BritishMedical Journal, 324, 1097-1098.

Reardon, D. C., Cougle, J. R., Rue, V. M., Shuping,M. W., Coleman, P. K., & Ney, P. G. (2003). Psychi-atric admissions of low-income women followingabortion and childbirth. Canadian Medical Associa-tion Journal, 168, 1253-1256.

Reardon, D. C., & Ney, P. G. (2000). Abortion andsubsequent substance abuse. American Journal ofDrug and Alcohol Abuse, 26, 61-75.

Reardon, D. C., Ney, P. G., Scheuren, F., Cougle, J.,Coleman, P. K., & Strahan, T. W. (2002). Deaths asso-ciated with pregnancy outcome: A record linkagestudy of low income women. Southern Medical Jour-nal, 95, 834-841.

Robins, L., & Regier, D. (Eds.). (1991). Psychiatricdisorders in America: The Epidemiological CatchmentArea study. New York: Free Press.


Rona, R. J., Smeeton, N. C., Beech , R., Barnett, A.,and Sharland , G. (1998). Anxiety and depression inmothers related to heart of the child and foetus. ActaPædiatr 87, 201–205.

Rosenbaum, E., & Kandel, D. B. (1990). Early onsetof adolescent sexual behavior and drug involvement.Journal of Marriage and Family, 52, 783-798.

Rosenberg, M. (1965). Society and adolescent self-image. Princeton, NJ: Princeton University Press.

Rotter, J. B. (1966). Generalized expectancies for in-ternal versus external control of reinforcement. Psy-chological Monographs: General and Applied, 80,Whole No. 609).

Rue, V. M., Coleman, P. K., Rue, J. J., & Reardon,D. C. (2004). Induced abortion and traumatic stress:Preliminary comparison of American and Russianwomen. Medical Science Monitor, 10, SR5-16.

Russo, N. F., & Dabul, A. J. (1997). The relationshipof abortion to well-being: Do race and religion make adifference? Professional Psychology: Research andPractice, 28, 23-31.

Russo, N. F., & Denious, J. E. (1998a). Understandingthe relationship of violence against women to un-wanted pregnancy and its resolution. In L. J. Beckman& S. M. Harvey (Eds.), The new civil war: The psy-chology, culture, and politics of abortion (pp. 211-234). Washington, DC: American PsychologicalAssociation.

Russo, N., & Denious, J. (1998b). Why is abortionsuch a controversial issue in the United States? In L. J.Beckman & S. M. Harvey (Eds.), The new civil war:The psychology, culture, and politics of abortion (pp.25-60). Washington, DC: American Psychological As-sociation.

Russo, N. F., & Denious, J. E. (2001). Violence in thelives of women having abortions: Implications forpractice and public policy. Professional Psychology:Research and Practice, 32, 142-150.

Russo, N. F., & Zierk, K. L. (1992). Abortion, child-bearing, and women’s well- being. Professional Psy-chology: Research and Practice, 23, 269-280.

Salvesen, K. A., Oyen, L., Schmidt, N., Malt, U. F., &Eik-Nes, S. H. (1997). Comparison of long-term psy-chological responses of women after pregnancy termi-nation due to fetal anomalies and after perinatal loss.Ultrasound Obstetrics and Gynecology, 9, 80-85.

Schmiege, S., & Russo, N. F. (2005). Depression andunwanted first pregnancy: Longitudinal cohort study.British Medical Journal, 331, 1303.

Sit, D., Rothschild, A. J., Creinin, M. D., Hanusa, B.H., & Wisner, K. L. (2007). Psychiatric outcomes fol-lowing medical and surgical abortion. Human Repro-duction, 22, 878-884.

Speckhard, A. C., & Rue, V. M. (1992). Postabortionsyndrome: An emerging public health concern. Journalof Social Issues, 48, 95-119.

Steele, C. M., & Aronson, J. (1995). Stereotype threatand the intellectual test performance of African Ameri-cans. Journal of Personality and Social Psychology,69, 797-811.

Strahan T. W. (2001). Detrimental effects of abortion:An annotated bibliography with commentary. Spring-field, IL: Acorn Books.

Tamburrino, M. B., Franco, K. N., Campbell, N. B.,Pentz, J. E., Evans, C. L., & Jurs, S. G. (1990).Postabortion dysphoria and religion. Southern Med-ical Journal, 83, 736-738.

Taylor, J., Henderson, D., & Jackson, B. B. (1991). Aholistic model for understanding and predicting de-pressive symptoms in African-American women. Jour-nal of Community Psychology, 19, 306-320.

Taylor, J., & Jackson, B. (1990). Factors affecting al-cohol consumption in Black women, Part II. The In-ternational Journal of Addictions, 25, 1415-1427.

Teichman, Y., Shenhar, S., & Segal, S. (1993). Emo-tional distress in Israeli women before and after abor-tion. American Journal of Orthopsychiatry, 63,277-288.


Thorp, J. M., Hartmann, K. E., & Shadigin, E.(2003). Long-term physical and psychological healthconsequences of induced abortion: Review of theevidence. Obstetrical and Gynecological Survey, 58,67-79.

Torres, A., & Forrest, J. (1988). Why do women haveabortions? Family Planning Perspectives 20, 169-176.

Trybulski, J. (2006). The long-term phenomena ofwomen’s postabortion experiences: Reply to the letterto the editor. Western Journal of Nursing Research,28, 354-356.

Vestal, C. (2006, June 22). States probe limits of abor-tion policy (Updated October 25, 2006). RetrievedJanuary 3, 2007, from http://www.stateline.org/live/ViewPage.action?siteNodeId=136&languageId=1&contentId=121780

Williams, G. B. (2001). Short-term grief after an elec-tive abortion. Journal of Obstetric, Gynecologic, andNeonatal Nursing, 30, 174-183.

Willoughby, T., Chalmers, H., & Busseri, M. A.(2004). Where is the syndrome? Examining co-occur-rence among multiple problem behaviors in adoles-cence. Journal of Consulting and Clinical Psychology,72, 1022-1037.

Wilmoth, G. H., de Alteriis, M., & Bussell, D. (1992).Prevalence of psychological risks following legal abor-tion in the U.S.: Limits of evidence. Journal of SocialIssues, 48, 37-65.

Zeanah, C. H., Dailey, J. V., Rosenblatt, M. J., &Saller, D. N., Jr. (1993). Do women grieve after termi-nating pregnancies because of fetal anomalies? A con-trolled investigation. Obstetrics and Gynecology, 82,270-275.


LIST OFTABLES

Table 1: Medical Record Studies: U.S. Samples

Table 1B: Medical Record Studies: InternationalSamples

Table 2: Secondary Analyses of Survey Data:U.S. Samples and International Samples

Table 3A: Primary Data Comparison Group Studies:U.S. Samples

Table 3B: Primary Data Comparison Group Studies:International Samples

Table 4: Abortion for Reasons of Fetal Anomaly

Table 5: U.S. Studies of Abortion Only:No Comparison Groups

Table 6: Population estimates of proportion of allwomen and women identified as havingbeen pregnant exceeding CES-D clinicalcutoff score, National Longitudinal Surveyof Youth: 1992.


ACKNOWLEDGMENTS

Brenda Major’s contributions to this report were sup-ported in part by grants from the American Philosoph-ical Society and the James McKeen CattellFoundation.

Thanks are extended to Julia Cleaver, Rennie Georgieva,and Yelena Suprunova for library assistance.

Task force members would like to express their appre-ciation to the following individuals for their thought-ful reviews and comments on earlier versions of thisreport: Nancy E. Adler, PhD; Toni C. Antonucci,PhD; Bonita Cade, PhD, JD; Priscilla Coleman, PhD;M. Lynne Cooper, PhD, MPH; Henry P. David, PhD;Patricia Dietz, DrPh; David Fergusson; Barbara Fiese,PhD; Irene Frieze, PhD; Mika Gissler, PhD; Ellie Lee,PhD; Marcia Lobel, PhD; Bernice Lott, PhD; RachelM. MacNair, PhD; Debra Mollen, PhD; Carol C.Nadelson, MD; Robert Post, PhD, JD; Jaquie Resnick,PhD; Gail Erlick Robinson MD, DPsych, FRCPC;Elizabeth Shadigian, MD; Reva Siegal, JD; Nada LStotland, MD, MPH; John M. Thorp, Jr., MD; BrianWilcox, PhD; and Greg Wilmoth, PhD.

We’d also like to thank the staff of the APA Women’sPrograms Office for their support: Tanya Burrwell,Shari Miles-Cohen, Leslie Cameron, Gabe Twose,Liapeng Matsau, and Ashlee Edwards.


mentalhealth andabortionreportofthe apataskforceon mentalhealth andabortion...

Documents