Psychiatry and Primary CareRecent epidemiologic studies have found that most patients with mental illness are seen exclusively inprimary care medicine. These patients often present with medically unexplained somatic symptoms andutilize at least twice as many health care visits as controls. There has been an exponential growth in studiesin this interface between primary care and psychiatry in the last 10 years. This special section, edited byWayne J. Katon, M.D., will publish informative research articles that address primary care-psychiatric issues.

Factors Associated with Symptomatic Improvementand Recovery From Major Depression in PrimaryCare Patients

Charlotte Brown, Ph.D., Herbert C. Schulberg, Ph.D., andHolly G. Prigerson, Ph.D.

Abstract: This article describes a post-hoc analysis of clinicaland psychosocial factors and beliefs about health associatedwith treatment outcome in a sample of depressed primary carepatients (N5181) randomly assigned to a standardized treat-ment or physician’s usual care (UC). Different factors werefound to predict clinical outcomes for treatment modality [UCvs. interpersonal psychotherapy (IPT) or nortriptyline (NT)]and the type of outcome evaluated (i.e., depressive symptoms at8 months or symptomatic and functional recovery at 8months). Factors associated with treatment-specific outcomesare also described. Consistent with prior studies, lower depres-sive symptom severity at 8 months was associated with higherbaseline functioning, minimal medical co-morbidity, race, andstandardized pharmacologic or psychotherapeutic treatment.Additionally, an interaction between treatment modality andhealth locus of control indicated that individuals perceivingmore self-control of their health and who received a standard-ized treatment experienced greater depressive symptom reduc-tion at 8 months. Factors associated with symptomatic andfunctional recovery from the depressive episode were also ex-amined. Patients who received a standardized treatment (IPTor NT) perceived greater control of their health and lacked alifetime generalized anxiety disorder or panic disorder weremore likely to recover by month 8 than those who receivedusual care. While clinical severity and treatment adequacy play

an important role in both symptomatic improvement and fullrecovery from a depressive episode, other key factors such ashealth beliefs and non-depressive psychopathology also influ-ence recovery. © 2000 Elsevier Science Inc.

Introduction

The efficacy of treatments for major depression ishigh, with 50–70% of patients recovering when pro-vided a validated intervention. Conversely, 30–50%of patients treated for this disorder do not recover[1]. Numerous investigators, therefore, have triedto identify factors that can distinguish the formerfrom the latter patient subgroups. Psychiatric stud-ies of both a prospective and post-hoc nature haveidentified numerous clinical factors associated withrecovery from a depressive episode. These include:severity of depressive symptoms [2–4] functionaldisability [5], medical comorbidity [3,6], anxietydisorder comorbidity [7–9], and personality pathol-ogy [10–12]. Psychosocial factors such as the occur-rence of stressful life events [13,14] and social sup-port [3,15] also have been found to be associatedwith recovery in unipolar depression. Additionally,the importance of an individual’s belief in the con-trollability of their health has been emphasized indepressive outcomes. Using meta-analytic tech-niques, Benassi et al. [16] found a moderate rela-tionship between internal health locus of controland reduced depressive symptoms. Reynaert et al.[17] further demonstrated that depressed patients

Western Psychiatric Institute and Clinic, University of Pitts-burgh School of Medicine, Pittsburgh, Pennsylvania 15213 (C.B.,H.C.S.), Connecticut Mental Health Center, Department of Psy-chiatry, Yale University School of Medicine, 34 Park Street, NewHaven, Connecticut 06519 (H.G.P.).

Address reprint requests to: C. Brown, Western PsychiatricInstitute and Clinic, University of Pittsburgh School of Medicine,3811 O’Hara Street, Pittsburgh, PA 15213.

General Hospital Psychiatry 22, 242–250, 2000242© 2000 Elsevier Science Inc. All rights reserved.ISSN 0163-8343/00/$–see front matter

655 Avenue of the Americas, New York, NY 10010PII S0163-8343(00)00086-4

with internal health locus of control beliefs showedgreater improvement following antidepressanttreatment than those with an external health locusof control.

The majority of studies examining recovery fromunipolar depression have been conducted with psy-chiatric samples. It is also important to consider,however, whether these factors influence recoveryfrom depression in primary care patients given pos-sible differences in the medical comorbidity andsymptom presentation of these patients when com-pared to psychiatric samples [18]. Given this con-cern, it is striking that while several recent studieshave demonstrated the effectiveness of pharmaco-logic and psychotherapeutic treatments provideddepressed primary care patients [19–22], few inves-tigators have examined clinical characteristics, psy-chosocial factors, and health beliefs associated withtreatment outcome in this population.

Among studies of primary care patients, Schul-berg et al. [23], in a prospective observational study,found lifetime psychiatric symptoms and medicalcomorbidity associated with recovery from a majordepressive episode. In a post-hoc analysis, Brown etal. [24] found that depressed patients with lifetimepanic disorder randomized to pharmacotherapy orpsychotherapy had significantly poorer recoveryrates. Personality pathology has also been linkedwith clinical outcomes for depressed primary carepatients. Katon et al. [25] found that high neuroti-cism was the only significant predictor of non-remission of symptoms among patients with majordepression who were treated with antidepressantsby their primary care physician. Additionally, Pa-tience and colleagues [12] found that patients witha personality disorder were less likely to have re-covered at the end of acute phase treatment fordepression than those with no such disorder. Incontrast, Mynors-Wallis et al. [21] found no signif-icant association between clinical, demographic, orpsychosocial variables and change in depressiveseverity in patients receiving psychotherapy, phar-macotherapy, or placebo.

These earlier studies suggest that similar clinicalfactors such as depressive severity, disability, andmedical and psychiatric comorbidity, are associatedwith treatment outcome in both depressed primarycare and psychiatric patients. However, this litera-ture provides little guidance regarding the influ-ence of perceived control of health or psychosocialfactors such as life stress and social support onclinical course. Furthermore, few studies have ex-amined whether clinical and psychosocial charac-

teristics of patients are associated with treatment-specific response. The NIMH Treatment ofDepression Collaborative Research Program [26]addressed this by analyzing the relationship be-tween patient factors and treatment responseamong patients randomly assigned to one of fourtreatment types. The investigators found less socialimpairment associated with better response to in-terpersonal psychotherapy; less perfectionist andsocially dependent attitudes (cognitive dysfunc-tion) associated with superior response to cognitivebehavior therapy and to imipramine; and more im-pairment of work, school, and home functioning topredict better response to imipramine. To date, nopublished studies of primary care patients haveexamined whether particular patient factors are as-sociated with treatment-specific response.

An understanding of patient factors associatedwith or even predicting good outcomes for specifictreatments would permit clinicians to customizedepression interventions to particular patient pro-files and to minimize relapse or recurrence. Withthis in mind, the present post-hoc analyses wereundertaken to determine whether: demographicfactors; clinical characteristics such as depressiveseverity, anxiety comobidity, personality disorder,or medical comorbidity; psychosocial factors suchas stressful life events and social support; andhealth beliefs are significantly associated withgreater symptomatic improvement or recoveryfrom a major depressive episode in a sample ofprimary care patients randomized to interpersonalpsychotherapy (IPT), nortriptyline (NT), or physi-cian’s usual care (UC). We anticipated that patients’perceived control of health (internal vs. external)and treatment type (standardized vs. usual care)might differentially influence depression outcomes.Therefore, the interaction of health locus of controland treatment type was also tested for significance.Finally, we sought to determine which of thesepatient variables are significantly associated withdifferential response to IPT, NT, or UC.

Method

Participants and Procedures

The sample was composed of patients recruited fora randomized control trial of treatments for majordepression in two internal medicine clinics and twofamily health centers in Pittsburgh, Pennsylvania.Potential participants were patients ages 18–64 pre-senting in the waiting rooms of the four study sites.

Improvement and Recovery from Major Depression

243

Each patient was informed by a Research Associatethat investigators from the University of PittsburghSchool of Medicine were conducting an Institu-tional Research Board-approved study of treat-ments for major depression among ambulatorymedical patients, and informed consent was soughtfor a three-phase assessment of eligibility for theclinical trial [22].

Patients meeting DSM-III-R criteria for a currentmajor depression on the Diagnostic InterviewSchedule (DIS) [27] and scoring $13 on the 17-itemHamilton Rating Scale-Depression (HRSD) [28]were randomized to IPT provided by a mentalhealth specialist within treatment manual stan-dards [29]; NT provided by a primary care physi-cian within manualized pharmacotherapy stan-dards [30]; or UC, i.e., any intervention deemedappropriate by the treating physician, including nodepression-specific treatment, antidepressant med-ication, counseling, referral for psychotherapy, etc.[31]. Treatment protocols for IPT and NT were con-sistent with Agency for Health Care Policy andResearch (AHCPR) Depression Guidelines. Thus,both pharmacotherapy and psychotherapy in-cluded acute and continuation-phase treatment,and the entire trial lasted 8 months. Upon comple-tion of acute phase treatment, patients receiving NTentered the continuation phase and were seen for 6monthly visits. The acute phase of IPT consisted ofapproximately 16 weekly sessions, followed by 4monthly continuation-phase sessions. The meannumber of weeks in treatment was 34.8 for phar-macotherapy patients and 37 for psychotherapy pa-tients. This report focuses on the intent-to-treatsample of 181 patients for whom all clinical andpsychosocial assessments were available at base-line.

Measures

Following randomization, study participants wereevaluated with additional sections of the DiagnosticInterview Schedule and with the Structured Clini-cal Interview for DSM-III-R Personality Disorders(SCID-II) [32] to determine the presence of otherAxis I and any Axis II psychiatric disorders. Clinicaland psychosocial variables measured at baselinealso included medical severity with the Duke Se-verity of Illness Scale (DUSOI) [33]; functional dis-ability with the Global Assessment Scale (GAS)[34]; perceived control of health with the HealthLocus of Control Scale (HCL) [35]; stressful lifeevents with the Psychiatric Epidemiology Research

Interview (PERI) [36]; and social support with theInterpersonal Support Evaluation List (ISEL) [37].The HRS-D was administered at baseline andmonth 8 by clinical evaluators blind to a patient’streatment assignment and whose interrater reliabil-ity exceeded 0.90 when computed by rank-ordercorrelation. Sociodemographic characteristics as-sessed at baseline included the patient’s age, gen-der, marital status, educational attainment, employ-ment status, and self-reported race.

Statistical Analyses

T-tests for continuous variables and x2 tests forcategorical variables were used to compare patientswho had and had not recovered at month 8. Inorder to identify the most powerful and parsimo-nious set of predictors of clinical response two setsof stepwise regression analyses were completed.The first set of analyses evaluated variables associ-ated with symptomatic improvement in the fullsample, and those associated with symptomatic im-provement in each treatment group (i.e., IPT, NT,UC). Factors associated with symptomatic im-provement in the full sample were evaluated with astepwise linear regression analysis, using themonth 8 HRS-D total score as the dependent vari-able. Given the strong association between baselineseverity of depression and treatment outcome, ini-tial depressive status as measured by the HRS-Dserved as a covariate in all models. Independentvariables included gender, self-reported race (Afri-can American, white), marital status (married, un-married), employment status (employed full orpart-time, unemployed), education (high school ed-ucation or greater, less than high school education),medical comorbidity, functional disability, socialsupport, and stressful life events. Dichotomousvariables were created to assess presence/absenceof lifetime generalized anxiety disorder (GAD) orpanic disorder (PD), presence or absence of Axis IIpersonality disorder and whether a patient hadbeen randomized to a standardized treatment (IPTor NT) or UC. The above independent variableswere entered in the model in a stepwise fashion,after controlling for baseline depressive severity.The interaction term and variables included in theinteraction were entered in the last step of themodel.

Three separate stepwise linear regression analy-ses evaluated factors associated with symptomaticchange in each treatment condition (i.e., IPT, NT,and UC), also using month 8 HRD-D total score as

C. Brown et al.

244

the dependent variable. Independent variables in-cluded gender, self-reported race, marital status,employment status, education, medical comorbid-ity, functional disability, lifetime GAD or PD, AxisII personality disorder, social support, stressful lifeevents, and health locus of control. All independentvariables were entered into the model in a stepwisefashion after controlling for baseline severity of de-pression.

The second set of analyses evaluated treatmentrecovery using stepwise logistic regression analy-ses. Treatment recovery was defined as a HRS-Dscore #7 and GAS $71 at month 8. All participantswere classified as recovered or not recovered ac-cording to this criterion, and this dichotomous clas-sification served as the dependent variable in allstepwise logistic regression models. In the analysisassessing treatment recovery in the full sample,independent variables (gender, race, marital status,employment status, education, medical comorbid-ity, functional disability, lifetime GAD or PD, AxisII personality disorder, social support, stressful lifeevents, health locus of control, and type of treat-ment, i.e., standardized or usual care) were enteredin the analysis in a stepwise fashion after control-ling for baseline depressive severity. The interac-tion of health locus of control by treatment type andvariables included in the interaction were enteredin the last step of the model.

Three separate logistic regression analyses eval-uated factors associated with recovery in each treat-ment group (i.e., IPT, NT, and UC), also usingtreatment recovery at month 8 as the dependentvariable. Independent variables (i.e., gender, self-reported race, marital status, employment status,education, medical comorbidity, functional disabil-ity, lifetime GAD or PD, Axis II personality disor-der social support, stressful life events and healthlocus of control) were entered into the model in astepwise fashion after controlling for baseline se-verity of depression.

Results

Patient Characteristics

As shown in Table 1, this sample of 181 patientswas predominantly unmarried, female, had a meanage of 39 years, and was approximately 50% white.Most participants had at least a high school educa-tion, and less than half were employed full or part-time. Patients typically reported experiencing mod-erate to severe depressive symptoms, significant

psychosocial impairment, and moderate physicalillness. Other psychiatric diagnoses were common;more than two-thirds of the sample had co-occurring lifetime generalized anxiety disorder orpanic disorder, or personality disorder.

Depressive Symptomatic Improvement

Stepwise linear regression analyses indicated thatparticipants who were white, had higher initialfunctioning, less severe medical problems, and whoreceived a standardized treatment had significantlylower depressive symptoms at month 8 (Table 2).The significant interaction between perceived con-trol of health and treatment type indicates that pa-tients who believed that they had more control oftheir health had better outcomes when provided astandardized treatment (NT or IPT) compared withusual care. These factors accounted for 23% of thevariance.

Given the significant interaction between treat-ment type and health locus of control, we per-formed separate stepwise linear regressions for NT,IPT, and UC to identify variables associated withtreatment-specific clinical course. For patients re-ceiving NT, baseline level of depression (HRS-Dscore) and severity of medical illness (DUSOI) werethe only factors significantly associated with de-pressive symptoms at eight months (R5.51, Ad-justed R25.24, F(2,62)510.9, P5.0001) (data notshown). Less severe baseline depressive symptomsand better physical health were significantly asso-ciated with lower depressive symptoms at 8months. Perceived control of health was the onlyvariable significantly associated with improvedoutcome among patients randomized to IPT. Those

Table 1. Demographic and clinical characteristicsof sample (N5181)

Demographic characteristicsMean (SD) age 39.3 (11.5)Percent female 84.5Percent married 33.7Percent $ HS education 82.3Percent employed full/parttime 40.9Percent white 54.7

Clinical characteristicsMean (SD) HRSD 23.4 (5.1)Mean (SD) GAS 49.9 (7.1)Mean (SD) DUSOI 33.8 (15.9)Percent any GAD or panic disorder 70.7Percent any Axis II personality disorder 69.1

Improvement and Recovery from Major Depression

245

perceiving more internal control over their healthhad less severe depressive symptoms at montheight (R5.43, Adjusted R25.16, F(2, 62)56.9, P5.002) (data not shown). Among participants receiv-ing their physician’s ‘usual care,’ those with lesssevere medical illness and functional impairment atbaseline, and higher levels of social support hadless severe depressive symptoms at month 8 (R5.52, Adjusted R25.21, F(4, 46)54.2, P5.005) (datanot shown). Baseline depressive severity was notfound to be a significant covariate in the final re-gression models for IPT or UC.

Symptomatic and Functional Recovery

Additional analyses were undertaken with the fullsample to determine factors associated with recov-ery in general, as well as with specific treatments.Univariate analyses compared the baseline demo-graphic, clinical, psychosocial, and treatment as-signment of recovered and non-recovered partici-pants. Compared to those failing to recover,patients meeting recovery criteria at month 8(HRS-D # 7 and GAS $ 71) were more likely to beemployed full or part-time at baseline (x256.5,df51, P5.01); to have less severe depressive symp-

toms (t52.6, df5179, P5.01), functional impair-ment (t523.1, df5179, P5.002), and medical prob-lems (t52.3, df5179, P5.02); and were less likely tohave a lifetime comorbid anxiety disorder (x259.0,df51, P5.003) or personality disorder (x258.2,df51, P5.004). The two groups did not differ onlife stress or social support at baseline, but recov-ered patients perceived more internal locus of con-trol over their health status (t53.2, df5179,P5.002). Finally, a significantly greater proportionof patients randomized to standardized treatment(IPT or NT) recovered compared to those assignedto usual care (x259.4, df5 1, P5.002).

Stepwise logistic regression analyses, controllingfor baseline HRS-D, determined that patients withno history of GAD or PD, with greater perceivedinternal control of health, and randomized to astandardized treatment were more likely to recoverfrom the depressive episode by month 8. Depres-sive severity at baseline was not a significant co-variate in the final model (see Table 3).

Separate stepwise logistic regression analyseswere completed to examine treatment-specificpredictors of recovery. Among patients random-ized to NT, those who were younger, married,

Table 2. Summary of stepwise linear regression analysis for predictors of symptomatic outcomes at 8months

Variable Beta SE B t P R Adjusted R2 F P

Step 1Baseline HRS-D .43 .11 4.12 .0001 .29 .08 17.0 .0001

Step 2Baseline HRS-D .33 .11 2.98 .0030 .36 .12 13.5 .0001Functional impairment 2.24 .08 23.05 .0030

Step 3Baseline HRS-D .29 .11 2.66 .0080 .42 .16 12.5 .0001Functional impairment 2.24 .08 23.12 .0020Medical comorbidity .10 .03 3.04 .0030

Step 4Baseline HRS-D .22 .11 1.95 .0530 .45 .18 10.9 .0001Functional impairment 2.26 .08 23.43 .0010Medical comorbidity .09 .03 2.74 .0070Race 22.41 1.08 22.24 .0270

Step 5Baseline HRS-D .16 .11 1.48 .1410 .51 .23 8.6 .0001Functional impairment 2.22 .08 22.88 .0040Medical comorbidity .08 .03 2.51 .0130Race 22.09 1.05 22.00 .0470Health locus of control 3 type of treatment .45 .19 2.34 .0210Type of treatment 213.11 4.74 22.76 .0060Health locus of control 2.14 .17 2.84 .4010

C. Brown et al.

246

and who reported milder depressive symptomsat baseline were more likely to recover by month8 (Wald x2518.3, df53, P5.0004). Among patientsrandomized to IPT, those with no history of GADor PD, and no current Axis II personality disorderwere significantly more likely to recover from thedepressive episode at month 8 (Wald x2518.1,df54, P5.001).

There were no variables significantly associatedwith recovery for patients randomized to usualcare, a finding likely due to the fact that only 20% ofUC patients achieved full recovery. Therefore, wecombined them with the UC patients displaying apartial recovery at month 8 (HRS-D score 8–12 andGAS score 61–70), resulting in a subgroup of 29patients or 59% of those randomized to UC. Astepwise logistic regression analysis, controlling forinitial severity of depressive symptoms, indicatedthat initial functional disability was the sole vari-able significantly associated with full or partial re-covery (x259.7, df52, P5.008). Baseline depressiveseverity was not a significant covariate in themodel.

Discussion

We conducted a post-hoc analysis of the associationbetween clinical and psychosocial factors, and be-liefs about health and treatment outcome in a sam-ple of depressed primary care patients randomlyassigned to a standardized treatment or UC. Differ-ent factors were found to predict clinical outcomesfor treatment modality (physician’s usual care ver-sus IPT or NT), and the type of outcome evaluated

(i.e., depressive symptoms at 8 months or symp-tomatic and functional recovery at 8 months). Inkeeping with prior studies, lower depressive symp-tom severity at 8-months was associated withhigher baseline functioning [5], minimal medicalco-morbidity [6,23,38], and standardized pharma-cologic or psychotherapeutic treatment [19,21]. Aspredicted, health beliefs were significantly associatedwith clinical outcomes, although only partial supportwas found for the role of psychosocial factors.

Of note is the interaction between treatment mo-dality and health locus of control. Our finding that,in general, individuals perceiving more self-controlof their health experienced greater depressivesymptom reduction is supported by earlier reportsthat patient expectancies and attitudes are signifi-cant correlates of successful outcomes in psycho-therapy [39,40]. The fact that health beliefs were theonly significant predictor of symptom reduction forpatients randomized to IPT lends further support toSpanier et al.’s [39] conclusion that the ability toeffectively engage in IPT is likely linked to a pa-tient’s appraisal of the degree of control that she/hecan exercise over the emergence and suppression ofdepressive symptoms.

Among patients treated with NT, milder depres-sive symptoms and less severe medical illness atbaseline were significantly associated with reduceddepressive symptomatology at 8 months. This find-ing is inconsistent with the AHCPR DepressionGuideline Panel’s [1] recommendation that phar-macotherapy be considered as the first-line treat-ment for more severe depression in primary carepatients. However, our research group [41] deter-

Table 3. Summary of stepwise logistic regression analysis for predictors of recovery at 8 months

Variable Beta SE B Wald x2 P Odds Ratio 95% CI

Step 1Baseline HRS-D 2.08 .03 6.38 .0115 .92 0.87–0.98

Step 2Baseline HRS-D 2.07 .03 5.27 .0217 .93 0.87–0.99Type of treatment 1.07 .38 7.93 .0049 2.90 1.38–6.10

Step 3Baseline HRS-D 2.07 .03 4.52 .0335 .93 0.88–0.99Lifetime panic or GAD 2.89 .35 6.54 .0105 .41 0.21–0.81Type of treatment 1.00 .38 6.82 .0090 2.72 1.28–5.77

Step 4Baseline HRS-D 2.05 .03 2.36 .1249 .95 0.89–1.01Lifetime panic or GAD 2.82 .35 5.31 .0212 .44 0.22–0.89Type of treatment 1.00 .39 6.59 .0103 2.71 1.27–5.80Health locus of control 2.07 .03 4.81 .0283 .93 0.88–0.99

Improvement and Recovery from Major Depression

247

mined that less severely depressed primary carepatients (baseline HRS-D #19) prescribed NT im-proved more rapidly during the initial 3 months oftreatment than clinically similar patients providedIPT. The current analyses clarify this earlier findingby identifying medical status as a key factor possi-bly influencing symptom remission among patientstreated with antidepressants. It should be noted,however, that both IPT and NT were comparablyeffective in treating major depression among moreseverely depressed patients (HRS-D $20) in thissample [22].

We have reported earlier that only 43% of pa-tients assigned to their physician’s usual care wereprescribed antidepressant medications for dosagesand durations consistent with AHCPR guidelines[31]. It is not surprising then that patients in thistreatment arm achieved better symptomatic out-comes when presenting with fewer psychosocialand medical vulnerabilities. Indeed, prospective,observational studies have documented that pa-tients with greater medical comorbidity [6,23],poorer functioning [5], and less social support[3,15,42] experience poorer depressive outcomes.

Given the increasing emphasis on the assessmentof functioning as an outcome of depression-specifictreatment [43] and studies demonstrating that per-sistent functional disability following treatment in-creases the risk for recurrent depression [44], ourdefinition of recovery included both symptomaticand functional improvement. In the full sample,univariate analyses identified standardized treat-ment, demographic characteristics (i.e., married,employed), clinical status (i.e., milder depressivesymptoms, less severe functional impairment andmedical problems, no comorbid anxiety or person-ality disorder) and health beliefs as associated withrecovery at 8-months. In contrast to factors associ-ated with symptomatic outcomes, multivariate lo-gistic regression analyses examining predictors ofrecovery at 8 months found no interaction betweentreatment type and health locus of control. Patientswho received a standardized treatment (IPT or NT),perceived greater control of their health, and lackeda lifetime GAD or PD were more likely to recoverby month 8. This suggests that when consideringboth functional and symptomatic recovery, per-ceived control of health remains significantly asso-ciated with outcome regardless of treatment modal-ity. The absence of lifetime panic disorder or GADis also positively associated with recovery frommajor depression. These findings highlight the needto adequately assess symptoms of anxiety disorder

and a patient’s beliefs in the controllability of de-pressive symptoms and functioning in order totreat depression effectively, and to minimize therisk of relapse and recurrence.

When treatment-specific patterns were exam-ined, patients receiving NT, and who were marriedand younger were found more likely to recover.Our data suggest that in this predominantly femalesample, marriage can serve as a buffer against de-pression. This impression is supported by researchrelating perceived quality of social support withassociated with depression outcomes [2,3]. The pos-itive association between younger age and recoveryfor NT patients may be indirectly linked to medicalcomorbidity in that younger patients had signifi-cantly fewer comorbid medical conditions.

It is not surprising that IPT produced the bestoutcomes for patients without lifetime GAD, panic,or personality disorder. This psychotherapy focuseson the interpersonal nature of depression’s etiologyand course, and explicitly avoids longstanding dis-positional tendencies. Indeed, Pilkonis and Frank[45] have documented a longer time to recovery fordepressed patients with a personality disorder,when treated with IPT.

Finally, our finding that race was a significantpredictor of depressive symptom severity but notrecovery status at 8 months is consistent with ourearlier report that African American primary carepatients present with more severe somatic symp-toms, but have similar rates of recovery when com-pared to whites [46].

In summary, we have identified health beliefs,and several clinical, demographic, and psychosocialcharacteristics that are associated with treatmentoutcomes for depressed primary care patients. Ourfindings distinguished factors associated withsymptomatic improvements, and those specific tosymptomatic and functional recovery. In both in-stances, clinical severity (i.e., depressive severity,medical comorbidity, and functional impairment)and adequacy of treatment play an important role.However, when stringent recovery criteria are ap-plied, the key factors are more limited but do in-clude non-depressive psychopathology (i.e., anxi-ety or personality disorders) and health beliefs. Thissuggests that more trait-like factors can also influ-ence recovery from major depression over time.

This research was supported by National Institutes of Mental HealthGrants MH01328 (C.B.), MH45815 (H.C.S.), MH56529, and TheAmerican Foundation for Suicide Prevention (H.G.P.).

C. Brown et al.

248

References

1. AHCPR Depression Guideline Panel: Depression inprimary care: Vol. 2. Treatment of major depression:Clinical practice guideline, No. 5 (AHCPR) Publica-tion No. 93-0551). Washington, DC, U.S. GovernmentPrinting Office, 1993

2. Brugha T, Bebbington P, MacCarthy B, Sturt E,Wykes T, Potter J: Gender, social support and recov-ery from depressive disorders: a prospective clinicalstudy. Psychol Med 20:147–156, 1990

3. George LK, Blazer BG, Hughes DC, Fowler N: Socialsupport and the outcome of major depression. Br JPsychiatry 154:478–485, 1989

4. Ramana R, Paykel E, Cooper Z, Hayhurst H, Saxty M,Surtees P: Remission and relapse in major depression:a two-year prospective follow-up study. Psychol Med25:1161–1170, 1995

5. Coryell W, Endicott J, Keller M: Outcome of patientswith chronic affective disorder: a five-year follow-up.Am J Psychiatry 147:1627–1633, 1990

6. Keitner GJ, Ryan CE, Miller IW, Kohn R, Epstein NB:12-month outcome of patients with depression andcomorbid psychiatric or medical illness (CompoundDepression). Am J Psychiatry 148:345–350, 1991

7. Clayton P: The comorbidity factor: establishing theprimary diagnosis in patients with mixed symptomsof anxiety and depression. J Clin Psychiatry 51:35–39,1990

8. Coryell W, Endicott J, Winokur G: Anxiety syndromesas epiphenomena of primary major depression: out-come and familial psychopathology. Am J Psychiatry149:100–107, 1992

9. Grunhaus L, Harel Y, Krugler T, Pande A, Haskett R:Major depressive disorder and panic disorder. Effectsof comorbidity on treatment outcome with antide-pressant medications. Clin Neuropharmacol 11:454–461, 1988

10. Shea MT, Widiger TA, Klein MH: Comorbidity ofpersonality disorders and depression: implicationsfor treatment. J Consulting Clin Psychol 60:857–868,1993

11. Ezquiaga E, Garcia A, Bravo F, Pallares T: Factorsassociated with outcome in major depression: a6-month prospective study. Social Psychiatry Psychi-atric Epidemiol 33:552–557, 1998

12. Patience D, McGuire R, Scott A, Freeman C: TheEdinburgh primary care depression study: personal-ity disorder and outcome. Br J Psychiatry 167:324–330, 1995

13. Monroe SM, Kupfer DJ, Frank E: Life stress and treat-ment course of recurrent depression: 1. Responseduring index episode. J Consulting Clin Psychol 60:718–724, 1992

14. Vallejo J, Gasto C, Catalan R, Bulbena A, Menchon J:Predictors of antidepressant treatment outcome inmelancholia: psychosocial, clinical and biological in-dicators. J Affective Disorders 21:151–162, 1991

15. Veiel HO, Kuhner C, Brill G, Ihle W: Psychosocialcorrelates of clinical depression after psychiatric in-patient treatment: methodological issues and baselinedifferences between recovered and non-recoveredpatients. Psychol Med 22:415–427, 1992

16. Benassi VA, Sweeney PD, Dufour CL: Is there a rela-tion between locus of control orientation and depres-sion? J Abnormal Psychol 97:357–367, 1988

17. Reynaert C, Janne P, Vause M, Zdanowicz N, LejeuneD: Clinical trials of antidepressants: the hidden face:where locus of control appears to play a key role indepression outcome. Psychopharmacology 119:449–454, 1995

18. Schulberg HC, Block MR, Coulehan JL: Treating de-pression in primary care practice. An application ofdecision analysis. General Hospital Psychiatry 11:208–215, 1989

19. Katon W, VonKorff M, Lin E, Walker E, Simon G,Bush T, Robinson P, Russo J: Collaborative manage-ment to achieve treatment guidelines. Journal of theAm Med Assoc 273:1026–1031, 1995

20. Katon W, Robinson P, VonKorff M, Lin E, Bush T,Ludman E, Simon G, Walker E: A multifaceted inter-vention to improve treatment of depression in pri-mary care. Arch General Psychiatry 53:924–932, 1996

21. Mynors-Wallis LM, Lloyd-Thomas AR: Randomisedcontrolled trial comparing problem-solving treat-ment with amitriptyline and placebo for major de-pression in primary care. Br Med J 310:441–445, 1995

22. Schulberg HC, Block MR, Madonia MJ, Scott CP,Rodriguez R, Imber SD, Perel J, Lave J, Coulehan J:Treating major depression in primary care practice:eight-month clinical outcomes. Arch General Psychi-atry 53:913–919, 1996

23. Schulberg HC, McClelland M, Gooding W: Six-month outcomes for medical patients with major de-pressive disorders. J General Intern Med 2:312–317,1987

24. Brown C, Schulberg HC, Madonia MJ, Shear MK,Houck PR: Treatment outcomes for primary care pa-tients with major depression and lifetime anxiety dis-orders. Am J Psychiatry 153:1293–1300, 1996

25. Katon W, Lin E, Korff MV, Bush T, Walker E, SimonG, Robinson P: The predictors of persistence of de-pression in primary care. J Affective Disorders 31:81–90, 1994

26. Elkin I, Shea T, Watkins JT, Imber SD, Sotsky SM,Collins JF, Glass DR, Pilkonis PA, Leber WR, Do-cherty JP, Feister SJ, Parloff MB: National Institute ofMental Health Treatment of Depression Collabora-tive Research Program. Arch General Psychiatry 46:971–982, 1989

27. Robins L, Helzer J, Croughan J, Ratcliff K: NationalInstitute of Mental Health Diagnostic InterviewSchedule. Its history, characteristics, and validity.Arch General Psychiatry 38:381–389, 1981

28. Hamilton M: The Hamilton rating scale for depres-sion, In N Sartorius, T Banto (ed): Assessment ofDepression. New York, Springer Verlag, 1986

29. Klerman G, Weissman M, Rounsaville B, Chevron E:Interpersonal Psychotherapy of Depression. NewYork, Basic Books, 1984

30. Fawcett J, Epstein P, Fiester SJ, Elkin I, Autry JH:Clinical management of imipramine/placebo admin-istration manual. NIMH Treatment of DepressionCollaborative Research Program. Psychopharmacol-ogy Bulletin 23:309–324, 1987

31. Schulberg HC, Block MR, Madonia MJ, Scott CP,

Improvement and Recovery from Major Depression

249

Lave JR, Rodriquez E, Coulehan JL: The ‘usual care’of major depression in primary care practice. ArchFamily Med 6:334–339, 1997

32. Spitzer R, Williams J, Gibbon M, First M: The Struc-tured Clinical Interview for DSM-III-R PersonalityDisorders (SCID-II). New York, Biometrics Research,New York State Psychiatric Institute, 1989

33. Parkerson GJ, Broadhead W, Tse C: The Duke Sever-ity of Illness Checklist (DUSOI) for the measurementof severity and comorbity. J Clin Epidemiol 46:379–393, 1993

34. Endicott J, Spitzer RL, Fleiss JL, Cohen J: The globalassessment scale: a procedure for measuring overallseverity of psychiatric disturbance. Arch GeneralPsychiatry 33:766–771, 1976

35. Wallston B, Wallston K, Kaplan G, Maides S: Develop-ment and validation of the health locus of control (HLC)scale. J Consulting Clin Psychol 44:580–585, 1976

36. Dohrenwend BS, Krasnoff L, Askenasy AR, Dohren-wend BP: Exemplification of a method of scaling lifeevents: the PERI life events scale. J Health SocialBehav 19:205–229, 1978

37. Cohen S, Mermelstein R, Kamarck T, Hoberman H:Measuring the functional components of social sup-port, In I Sarason, B Sarason (ed): Social Support:Theory, Research and Applications. The Hague, Mar-tinus Nijhoff, 1985

38. Krantz SE, Moos RH: Risk factors at intake predictnonremission among depressed patients. J Consult-ing Clin Psychol 56:863–869, 1988

39. Spanier C, Frank E, McEachran A, Grochocinski V,

Kupfer D: The prophylaxis of depressive episodes inrecurrent depression following discontinuation ofdrug therapy: integrating psychosocial and biologicalfactors. Psychol Med 26:461–475, 1996

40. Stiles WB, Shapiro D, Elliott R: Are all psychothera-pies equivalent? Am Psychol 7:165–180, 1986

41. Schulberg HC, Pilkonis PA, Houck P: The severity ofmajor depression and choice of treatment in primarycare practice. J Consulting Clin Psychol 66:932–938,1998

42. Paykel E, Cooper Z: Life events and social stress. In EPaykel (ed): Handbook of Affective Disorders, 2ndEdition. Edinburgh, Churchill Livingstone, 1992, pp149–170

43. Magruder KM, Schulberg HC, Oxman TE: Measure-ment and meaning of disablement in primary care.Int J Psychiatry Med 26:355–362, 1996

44. Kupfer D, Frank E: The minimum length of treatmentfor recovery. In S Montgomery, F Rouillon (eds):Long-term Treatment of Depression. London, JohnWiley & Sons Ltd, 1992

45. Pilkonis PA, Frank E: Personality pathology in recur-rent depression: nature, prevalence, and relationshipto treatment response. Am J Psychiatry 145:435–441,1988

46. Brown C, Schulberg H, Sacco D, Perel J, Houck P:Effectiveness of treatments for major depression inprimary medical care practice: a post hoc analysis ofoutcomes for African American and white patients. JAffective Disorders 53:185–192, 1999

C. Brown et al.

250