zohar, j., fostick, l. (2004). comparison of mortality rates between
TRANSCRIPT
1
Page 1 of 23
Zohar, J., Fostick, L. (2014). Comparison of Mortality Rates between Israeli Veterans With and
Without Post Traumatic Stress Disorder. European Neuropsychopharmachology, 24, 117–124.
***This is a self-archiving copy and does not fully replicate the published version***
Mortality Rates between Treated Post Traumatic Stress Disorder Israeli
Male Veterans Compared to Non-Diagnosed Veterans
Joseph Zohar1 and Leah Fostick
2
on behalf of the Israeli Consortium on PTSD*
1Department of Psychiatry, Sheba Medical Center, Tel Hashomer, and Sackler School of
Medicine, Tel Aviv University, Israel
2Ariel University, Ariel, Israel
* The Israeli Consortium on PTSD includes:
J. Zohar (chair), A. Bleich, Z. Kaplan, I. Katz, E. Klein, M. Kotler, A. Ohri, AY. Shalev,
& Z. Weissman.
**Address for correspondence: Prof. J. Zohar, Department of Psychiatry, Sheba Medical
Center, Tel Hashomer 52621, Israel. Telephone: 972-3530-3300. Fax: 972-3535-2788.
Email: [email protected].
2
Page 2 of 23
Abstract
The literature suggests that post-traumatic stress disorder (PTSD) is associated with
increased mortality. However, to date, mortality rates amongst veterans diagnosed with
post-traumatic stress disorder have not been reported for Israeli veterans, who bear a
different profile than veterans from other countries. This study aims to evaluate age-
adjusted mortality rates amongst Israeli Defense Forces veterans with and without PTSD
diagnosis. The study was carried out in a paired sample design with 2,457 male veterans
with treated PTSD and 2,457 matched male veterans without a PTSD diagnosis. Data on
PTSD and non-PTSD veterans was collected from the Rehabilitation Division of the
Israeli Ministry of Defense (MOD) and the Israeli Defense Forces’ (IDF) special unit for
treatment of combat stress reaction. Mortality data were collected from the Ministry of the
Interior (MOI) computerized database. Comparison of mortality rates between PTSD and
non-PTSD veterans was done using paired observations survival analysis by applying a
proportional hazards regression model. Overall no statistically significant difference in
mortality rates was found between veterans with treated PTSD and veterans without
PTSD. These findings hold even when excluding veterans who died in battle and
including non-PTSD veterans who died before their matched PTSD veteran was
diagnosed. However, among pairs with similar military jobs PTSD group had
significantly less mortality. The results of this large national cohort suggest that treated
PTSD is not associated with increased mortality. We submit that the lack of this
association represents the “net“ pathophysiology of PTSD due to the unique
characteristics of the sample.
Keywords: PTSD; Mortality; Veterans
3
Page 3 of 23
Introduction
Post-traumatic stress disorder (PTSD) is a chronic and disabling disorder characterized by
re-experiencing trauma suffered, avoidance of trauma-related stimuli, restricted affect,
hypervigilance, and social isolation (American Psychiatric Association, 2013). PTSD has
also been found to be linked to considerable physical comorbidity (Schnurr et al., 1998;
Schnurr & Green, 2004) and increased all-cause mortality (Sareen et al., 2007) (see Table
1 for a review). Increased suicidal behaviors and attempts have been reported in anxiety
disorders (Sareen et al., 2005), including PTSD (Pfeiffer et al., 2009).
To date, the majority of studies investigating mortality rates amongst those diagnosed
with PTSD have been conducted on Vietnam combat veterans (Table 1) in comparison
either with the general U.S. population (Bullman & Kang, 1994; Johnson et al., 2004),
those stationed in Vietnam but serving in non-combat roles (Breslin et al., 1998), or U.S.
veterans who did not participate in the Vietnam War (Breslin et al., 1998; Bullman &
Kang, 1994; Centers for Disease Control, 1988; Fett et al., 1987; Lawrence et al., 1985;
Sareen et al., 2005, 2007; Schnurr & Green, 2004; Thomas et al., 1991; Watanabe et al.,
1991; Watanabe & Kang, 1995, 1996). These studies evince several confounding effects
which are appropriate to be taken into consideration: (1) a priori differences between the
veteran population in such a semi-mandatory conscription system as the U.S. Vietnam
draft and the general population; (2) a selection bias reflecting the “healthy veteran” effect
– i.e., comparison of mortality rates between a medically-selected group such as veterans
and an unselected group such as the general population (Macfarlane et al., 2000; Seltzer
& Jablon, 1974; Watanabe & Kang, 1995); (3) the impact of the high comorbidity of
alcohol and drug abuse amongst U.S. veterans with PTSD (Boscarino, 2008a).
The study aim is to evaluate age-adjusted mortality rates and causes of death amongst
Israeli Defense Forces (IDF) veterans diagnosed and treated for PTSD in comparison with
4
Page 4 of 23
IDF veterans never diagnosed with PTSD. This cohort potentially differs from those
reported in other studies in several ways. Firstly, since military service in Israel is both
mandatory and prestigious, Israeli veterans might more accurately represent the general
population than other countries where draftees do not come from all walks of life.
Secondly, the selection of cohorts of veterans for both the PTSD and comparison groups
should enable a better control of the selection bias deriving from the “healthy veteran”
effect. Finally, in contrast to the 11% alcohol abstinence and 35% drinkers reported in
Europe (Neumark et al., 2007), the Israel National Health Survey reports 40% alcohol
abstinence and only 10% drinkers reporting three or more drinking episodes weekly.
Since alcohol and drug use are less common in Israel – and thus also in Israeli veterans –
this might help to minimize the potentially-confounding effect of these factors on
mortality. We therefore propose that this unique sample is well suited for examining the
issue of PTSD-related mortality amongst veterans.
5
Page 5 of 23
Study procedure
Data collection
The study was part of a survey designed to analyze and characterize Israeli veterans with
PTSD who were referred either to the Rehabilitation Division of the Israeli Ministry of
Defense (MOD) or to the Israeli Defense Forces’ (IDF) special unit for treatment of
combat stress reaction. In the survey – conducted between 2000 and 2001 – all the charts
from the seven MOD rehabilitation branches and the single IDF center were screened. In
this sample, all index participants were veterans who died between 1957 and 2002. The
age of death ranged from 22 to 89. The PTSD veterans experienced a traumatic incident
either during their mandatory (age 18-21) or reserve service (age 21-45) between 1948
and 2000. Traumatic experiences included: combat action (81.2%), accidents during
routine work or training (5.2%), traffic accidents (6.6%), terror attacks (4.0%), and other
events (3.0%). The survey covered all the records relating to any psychiatric diagnoses –
5,871 in number, constituting 91% of the existing records of the entire population of
those diagnosed with any psychiatric disorder between 1948 and 2000. A diagnosis of
PTSD was found for 2,463 files. As the number of female veterans in the PTSD group
was very low (n=6) and their trauma history not combat related, we confined the sample
to males. The total number of PTSD files in the current study was 2,457, comprising all
the files surveyed for male veterans who met the DSM-IV criteria for PTSD. According to
power analysis, this sample size is sufficient to detect differences as small as 3% in
mortality rates at 93% power.
Although PTSD diagnosis exists in some of the files, it was reevaluated for all by
surveyors on the basis of the veterans’ charts using DSM-IV criteria (American
Psychiatric Association, 1994) – i.e., the person had been exposed to a traumatic event
6
Page 6 of 23
and evidenced at least one symptom of re-experiencing, three avoidance/numbing
symptoms, and two symptoms of hyper-arousal. The surveyors were 32 psychology
students in the final year of their undergraduate studies, who were given specific training
in the diagnosis of PTSD. The surveyors were overseen by senior, well-informed
supervisors who reevaluated 50% of the files randomly during the first two months for
each surveyor, and 25% at subsequent stages. Inter-rater reliability between the surveyors'
and the supervisors' diagnoses was found to be within the acceptable range (kappa=0.77).
The data were first coded into data sheets and then entered into a computerized database.
Non-PTSD veterans were matched with PTSD veterans in accordance with their draft
identification number. To each PTSD participant, a draftee with a sequential army
identification number who did not apply to the MOD or IDF centers for either recognition
or psychiatric treatment was assigned. This procedure ensured an identical draft period
and age/sex correspondence. To some extent it also ensured the matching of other
background variables – such as education and socioeconomic characteristics, since
soldiers of the same sex, age, and location being drafted in groups. Sequential draft
identification numbers also indicate matched physical characteristics, those with similar
abilities being placed in the same units.
Veterans with PTSD did not differ from non-PTSD veterans in age, sex, socio-
economic status, or draft-board assessment (Zohar et al., 2011). The study cohort was
comprised of a total of 2,457 male veterans with PTSD and an equal number of matched
non-PTSD veterans. The analysis thus included 4,914 participants – more cases than in a
previous publication (Zohar et al., 2011) due to the fact that no missing data limitations
applied to this study.
The analysis was conducted in 2005 and included a snap-shot collection of mortality
data obtained from the Ministry of the Interior’s (MOI) computerized database which, by
7
Page 7 of 23
law, covers the entire population. A total of 176 veterans were registered as deceased in
2005. Mortality causes for 120 of the 176 veterans who had died were obtained from the
Ministry of Health registry. Causes of death for the remaining 56 veterans were
unavailable. The study was approved by the Institutional Ethics Committee of Chaim
Sheba Medical Center, Israel. Since it was based solely on subjects’ records, no written
consent was required. ID numbers were encoded during the analysis and decoded for
retrieving data purposes only.
Data analysis
The comparison of mortality rates between treated PTSD and non-PTSD veterans was
performed six times: (1) analysis of the total sample; (2) analysis only of veterans who did
not die in battle (in order to control for incidents of combat-mortality amongst the non-
PTSD group, all the non-PTSD veterans who died during service and their matched PTSD
veterans were excluded); (3) analysis only of pairs in which the non-PTSD veterans died
after his matched PTSD veteran developed PTSD symptoms (in order to make PTSD and
non-PTSD groups more comparable in terms of length of follow-up, this analysis was
restricted to deaths occurring only after the trauma had taken place and PTSD had
developed); (4) a final analysis with data cleaned for all the above (i.e., only veterans who
did not die in battle and where the non-PTSD veteran did not die before his matched
PTSD veteran developed PTSD symptoms; (5) PTSD and non-PTSD veterans have the
same military job (combat/officer vs. non/service back/support); (6) same as 5 with
conditions of 2 and 3.
The survival analysis of the paired observations was conducted by applying a
proportional hazards regression model. Employing the marginal approach suggested by
Gharibvand and Liu (2009), the model was fitted by applying the SAS PROC PHREG,
with a robust sandwich estimate to account for the dependence within pairs. This model
8
Page 8 of 23
was found to be adequate in five out of the six cases (analyses 2-6), based on tests
regarding the proportionality assumption. The first proportionality assumption was based
on the correlation between the Schoenfeld residuals and the ranking of individual failure
times. The second ultilized the method proposed by Lin et al (1993), being based on the
cumulative sums of martingale residuals. In the single case where the model assumption
was found to be inadequate (analysis 1), the stratified log rank was applied.
Results
The demographic characteristics of year of birth, age upon death, draft year, number of
years of education, rank, and role during service for the two groups – deceased treated
PTSD and non-PTSD veterans – are described in Table 2. A difference of five years was
found between PTSD and non-PTSD veterans in year of death and age upon death, with
PTSD veterans being older. No other differences were found between the two groups.
As seen in Table 3 the results of the survival analysis show that for the entire cohort
the difference in survival between PTSD and non-PTSD groups was not statistically
signifcant (analysis 1). This remained true when removing death in battle (analysis 2),
pairs in which the non-PTSD veteran died prior to PTSD diagnosis (analysis 3), or both
(analysis 4). However, when examining PTSD and non-PTSD veterans with the same
military jobs, the PTSD group had significantly less mortalility rates as compared to the
non-PTSD (analysis 5), even after removing death in battle and pairs where the non-
PTSD veteran died prior to PTSD diagnosis (analysis 6).
Causes of death according to ICD-10 classification were coded and are described in
Table 4. Due to the small number of observations, no analysis was carried out on this
data.
9
Page 9 of 23
Discussion
No statistically-significant difference obtained between the age-adjusted mortality rate of
veterans with treated PTSD and that of non-PTSD veteran which were matched on age,
sex, and combat role during exposure to trauma . Moreover, when controlled for military
job, less mortality rates were found for treated PTSD than to non-PTSD veterans. This
finding is contrary to the majority of publications on the subject (see Table 1).
Conceivably, the unique population of this study might account for a portion of these
differences. The current sample represents the general population due to compulsory
conscription in Israel, while since most of the previous publications relate to voluntary
drafting cases (as obtains in the U.S. and the U.K.), it is subject to population bias.
Recent findings show no excess of mortality amongst U.K. veterans (not limited to
those diagnosed with PTSD) (Macfarlane et al., 2000, 2005), and lower rates of suicide
among depressed veterans with comorbid PTSD (Pfeiffer et al., 2009; Zivin et al., 2007).
The current study expanded these findings by focusing exclusively on mortality rates
veterans diagnosed with PTSD. To the best of our knowledge, it constitutes the first study
to examine the association between PTSD and mortality amongst Israeli veterans. In light
of findings of mortality excess amongst veterans with PTSD in other studies (Boscarino,
2008a, 2008b; Bullman et al., 1990; Drescher et al., 2003; Johnson et al., 2004; Thomas
et al., 1991), we submit that the variant findings of this study might be related to the
uniqueness of the Israeli veteran sample. In contrast to the U.S., service in Israel is
mandatory, holds a very broad consensus, and affords far fewer loopholes for avoiding the
draft. The sample thus reflects virtually all social layers within the population.
A further potential difference between this and other studies which should be
addressed relates to drugs and alcohol abuse. During the years covered by this study
10
Page 10 of 23
neither alcohol nor drug abuse were common in Israeli society (Neumark et al., 2007).
This study correspondingly found a relatively low rate of drug and alcohol usage amongst
veterans – in direct contrast to the widespread prevalence of drug and alcohol usage
amongst U.S. veterans, mostly Vietnam veterans (Boscarino, 2008a). The benefit of
studying veterans with low drug and alcohol comorbidity lies in the fact that the results
might reflect the “net” effect of PTSD on mortality without any "contamination" from
these confounders (Johnson et al., 2004), although it also might limits the generalizability
of the findings. We thus suggest that the above-mentioned advantages of the present
sample may account – at least partially – for the difference in the results obtained in this
rather “cleaner” sample as compared to other studies, wherein the PTSD sample may have
evinced a higher level of socio-economic and comorbidity confounds. Yet additional
explanation for these results could lay in the difference between Israel and other countries
in higher social benefit and standard healthcare provided for those who are recognized as
injured in battle. This also might lead to a situation in which not all the veterans in the
PTSD group are actually suffer from PTSD, as some of them might over-report their
PTSD symptoms.
An interesting finding repeatedly reported in previous studies of people with PTSD –
as well as in some cases comparing deployed vs. non-deployed veterans (Macfarlane et
al., 2000, 2005)– is the presence of excess mortality due to external causes (Boscarino,
2006a, 2006b; Breslin et al., 1998; Bullman et al., 1990; Bullman & Kang, 1994; Centers
for Disease Control, 1988; Drescher et al., 2003; Fett et al., 1987; Johnson et al., 2004;
Kang & Bullman, 1996; Watanabe et al., 1991; Watanabe & Kang, 1995, 1996),
especially suicide (Breslin et al., 1998; Bullman & Kang, 1994; Centers for Disease
Control, 1988; Drescher et al., 2003; Ferrada-Noli et al., 1998; Watanabe & Kang, 1996).
The results of the current study demonstrate a similar tendency towards excess mortality
11
Page 11 of 23
due to external causes (including injury and poisoning) amongst PTSD veterans in
comparison with non-PTSD veterans (18 vs. 8, respectively; Table 4). Mortality due to
external causes can point to a set of behavioural characteristics possibly associated with
suicide and suicidal behaviour. The numerical excess of deaths related to injury and
poisoning in the PTSD group corresponds to this hypothesis.
One of the limitations of this study is the apparent difficulty in distinguishing suicide
from other causes of death in the database due to cultural-religious confounds. According
to religious law (e.g., Jewish and Muslim), a person who commits suicide must be buried
on the outskirts of the cemetery (“near the fence”). The burial consequences of suicide
lead physicians responsible for recording cause of death to be extremely careful and to
avoid registering death as suicide, self-inflicted injury (code E950-E959, ICD-9), or
intentional self-harm (code X60-X84, ICD-10) unless left with no other recourse. Indeed,
only in the case of two veterans (one PTSD and one non-PTSD) was the mortality cause
classified as ”intentional self-harm” (code X60-X84, ICD-10).
Additional shortcomings are that the PTSD group was diagnosed based on chart
review, and that the non-PTSD group was defined on the basis of lack of referral rather
than via a systematically-structured diagnostic questionnaire, thereby hampering our
ability to assess whether veterans in the non-PTSD group actually suffer from PTSD.
Since referral to the MOD and IDF following combat-related trauma is not rare, however,
we suspect this group to be rather small. The complementing limitation is that – in similar
fashion to other studies in the field – we only tested PTSD veterans who applied for
treatment, the results consequently being limited to this group. Given the awareness of
PTSD in veterans in Israel and the substantial financial benefits and treatment provided by
the MOD, we opine that many (if not a majority) of the veterans with PTSD applied for
treatment.
12
Page 12 of 23
Other limitations could relate to other variables not measured in the study, such as
social support and family bonding, which, differing between Israel and the U.S. might
explain the variation in the results. Similarly, both the current study and previous studies
surveying military veterans exclusively observed males, whereas PTSD is more prevalent
amongst women. The implications of these studies are therefore limited to the male
population.
The finding of no mortality excess among treated PTSD veterans in this sample, and
even lower mortality rates when controlling for army job during the incident, appears not
only to differ from the literature on higher mortality rates amongst civilians and veterans
(Boscarino, 2006a, 2006b, 2008a, 2008b; Breslin et al., 1998; Bullman et al., 1990;
Bullman & Kang, 1994; Centers for Disease Control, 1988; Fett et al., 1987; Drescher et
al., 2003; Ferrada-Noli et al., 1998; Johnson et al., 2004; Mollica, et al., 2001; Thomas et
al., 1991; Watanabe et al., 1991; Watanabe & Kang, 1995, 1996), but also to be somewhat
counterintuitive. Individuals with PTSD are not only subject to considerable stress but
also evince a high comorbidity of depression, thus being expected to be more prone to
stress-related morbidity and increased mortality. A similar pattern of results was found
amongst Holocaust survivors, whereas no mortality excess was found in comparison with
age-matched individuals unexposed to the Holocaust (Collins et al., 2004; Stessman et al.,
2008). Previously it was suggested that lower rates of suicide among depressed veterans
with comorbid PTSD as compared to depressed veterans without PTSD can be explained
by having more psychotherapeutic attention for those who have PTSD (Pfeiffer et al.,
2009; Zivin et al., 2007), which might also be the case here. However, it is not clear
whether PTSD veterans are being followed more closely medically (since they are already
being monitored by their physicians for their PTSD) or, conversely, whether they are
more sensitive to threats (both internal and external) and therefore seek help more
13
Page 13 of 23
frequently. These and other factors are yet to be explored. For example, stress per se has
traditionally been associated with a higher occurrence of “diseases of the circulatory
system”. In this sample, however, a numerical excess of deaths in this system was
observed in the non-PTSD rather than the PTSD group.
If replicated, the principal finding of the study – namely, that treated PTSD veterans
evinced no age-adjusted mortality excess – might shed new light on the complex
interaction between stress and mortality. PTSD – a disorder closely linked to death and/or
fear of dying – was not associated with actual mortality rates. Although power analysis
found that the sample size of 2,457 in each group is sufficient to detect even a small
differences (as small as 3%) in mortality rates, this finding could still be reversed.
Therefore, this finding requires further exploration and re-examination in another 10-15
years, when more incidents of death will have occurred in both PTSD and non-PTSD
groups.
14
Page 14 of 23
References
American Psychiatric Association, 2013. Diagnostic and statistical manual of
mental disorders. 5th edition ed. Arlington, VAWashington, DC: American Psychiatric
Association.
Boscarino, J. A., 2006. External-cause mortality after psychologic trauma: the
effects of stress exposure and predisposition. Compr. Psychiat.47(b), 503-14.
Boscarino, J. A., 2006. Posttraumatic stress disorder and mortality among U.S.
Army veterans 30 years after military service. Ann. Epidemiol. 16(a), 248-56.
Boscarino, J. A., 2008. A prospective study of PTSD and early-age heart disease
mortality among Vietnam veterans: Implications for surveillance and prevention.
Psychosom. Med. 70b, 668-76.
Boscarino, J. A., 2008. Psychobiologic predictors of disease mortality after
psychological trauma. Implications for research and clinical surveillance. J. Nerv. Ment.
Dis. 196a, 100-107.
Breslin, P., Kang, H., Lee, Y., Burt, V., & Shepard, B., 1998. Proportionate
mortality study of US Army and US Marine Corps veterans of the Vietnam War. Occup.
Environ. Med. 30, 412-19.
Bullman, T., & Kang, H., 1994. Posttraumatic stress disorder and the risk of
traumatic deaths among Vietnam veterans. J. Nerv. Ment. Dis. 182, 604-10.
Bullman, T., Kang, H., & Watanabe, K.,1990. Proportionate mortality among US
Army Vietnam veterans who served in military region I. Am. J. Epidemiol. 132, 670-4.
Centers for Disease Control., 1988. Serum 2, 3,7,8-tetrachlorodibenzo-p-dioxin
levels in US Army Vietnam-era veterans. JAMA.
15
Page 15 of 23
Collins, C., Burazeri, G., Gofin, J., & Kark, J., 2004. Health status and mortality in
holocaust survivors living in Jerusalem 40–50 Years Later. J. Trauma. Stress. 17, 403-
411.
Drescher, K., Rosen, C., Burling, T., & Foy, D., 2003. Causes of death among male
veterans who received residential treatment for PTSD. J. Trauma. Stress. 16, 535-43.
Ferrada-Noli, M., Asberg, M., Ormstad, K., Lundin, T., & Sundbom, E., 1998.
Suicidal behavior after severe trauma. Part 1: PTSD diagnoses, psychiatric comorbidity,
and assessments of suicidal behavior. J. Trauma. Stress. 11, 103-12.
Fett, M., Nairn, J., Cobbin, D., & Adena, M., 1987. Mortality among Australian
conscripts of the Vietnam conflict era. II. Causes of death. Am. J. Epidemiol.125, 878-84.
Gharibvand, L., & Liu, L., 2009. Analysis of survival data with clustered events.
Paper 237-2009. Washington, DC: SAS Global Forum Proceedings.
Johnson, D., Fontana, A., Lubin, H., Corn, B., & Rosenheck, R., 2004. Long-term
course of treatment-seeking Vietnam veterans with posttraumatic stress disorder:
mortality, clinical condition, and life satisfactio. J. Nerv. Ment. Dis. 192, 35-41.
Kang, H., & Bullman, T., 1996. Mortality among U.S. veterans of the Persian Gulf
War. N. Engl. J. Med.335, 1498-504.
Lawrence, C., Reilly, A., Quickenton, P., Greenwald, P., Page, W., & Kuntz, A.
(1985). Mortality patterns of New York State Vietnam Veterans. Am. J .Public Health.
75, 277-79.
Lin, D., Wei, L., & Ying, Z., 1993. Checking the Cox model with cumulative sums
of martingale based residuals. Biometrika, 80, 557-572.
Macfarlane, G. J., Hotopf, M., Maconochie, N., Blatchley, N., Richards, A., & Lunt,
M., 2005. Long-term mortality amongst Gulf War Veterans: Is there a relationship with
16
Page 16 of 23
experiences during deployment and subsequent morbidity? Int. J. Epidemiol. 34, 1403-
1408.
Macfarlane, G. J., Thomas, E., & Cherry, N., 2000. Mortality among UK Gulf War
veterans. Lancet, 356, 17-21.
Mollica, R., Sarajlic, N., Chernoff, M., Lavelle, J., Vukovic, I., & Massagli, M.,
2001. Longitudinal study of psychiatric symptoms, disability, mortality, and emigration
among Bosnian refugees. JAMA, 286, 546-54.
Neumark, Y., Lopez-Quintero, C., Grinshpoon, A., & Levinson, D., 2007. Alcohol
drinking patterns and prevalence of alcohol-abuse and dependence in the Israel National
Health Survey. Isr. J. Psychiatry. Relat. Sci. 44, 126-35.
Pfeiffer, P., Ganoczy, D., Ilgen, M., Zivin, K., & Valenstein, M., 2009. Comorbid
anxiety as a suicide risk factor among depressed veterans. J. Depress. Anxiety. 26, 752-
757.
Sareen, J., Cox, B., Afifi, T., de Graaf, R., Asmundson, G., ten Have, M., & Stein,
M., 2005. Anxiety disorders and risk for suicidal ideation and suicide attempts: a
population-based longitudinal study of adults. Arch. Gen. Psychiatry. 62, 1249-5.
Sareen, J., Cox, B., Stein, M., Afifi, T., Fleet, C., & Asmundson, G., 2007. Physical
and mental comorbidity, disability, and suicidal behavior associated with posttraumatic
stress disorder in a large community sample. Psychosom. Med. 69, 242-8.
Schnurr, P., & Green, B., 2004. Understanding relationships among trauma, post-
traumatic stress disorder, and health outcomes. Adv. Mind. Body. Med. 20, 18-29.
Schnurr, P., Spiro, A. 3., Aldwin, C., & Stukel, T., 1998. Physical symptom
trajectories following trauma exposure: longitudinal findings from the normative aging
study. J. Nerv. Ment. Dis. 186, 522-8.
17
Page 17 of 23
Seltzer, C., & Jablon, S., 1974. Effects of selection on mortality. Am. J. Epidemiol,
100, 367-72.
Stessman, J., Cohen, A., Hammerman-Rozenberg, R., Bursztyn, M., Azoulay, D.,
Maaravi, Y., & Jacobs, J., 2008. Holocaust survivors in old age: The Jerusalem
longitudinal study. J. Am. Geriatr. Soc. 56, 470-477.
Thomas, T., Kang, H., & Dalager, N., 1991. Mortality among Vietnam veterans,
1973-1987. Am. J. Epidemiol. 134, 973-80.
Watanabe, K., & Kang, H., 1995. Military service in Vietnam and the risk of death
from trauma and selected cancers. Ann .Epidemiol. 5, 407-12.
Watanabe, K., & Kang, H., 1996. Mortality patterns among Vietnam veterans: a 24-
year retrospective analysis. Occup. Environ. Med. 38, 272-78.
Watanabe, K., Kang, H., & Thomas, T., 1991. Mortality among Vietnam veterans:
with methodological considerations. Occup. Environ. Med., 33, 780-85.
Zivin, K., Kim, M., McCarthy, J., Austin, K., Hoggatt, K., Walters, H., &
Valenstein, M., 2007. Suicide mortality among individuals receiving treatment for
depression in the Veterans Affairs health system: Associations with patient and treatment
setting characteristics. Am. J .Public Health. 97, 2193-2198.
Zohar, J., Yahalom, H., Koslovsky, N., Cwikel-Hamzany, S., Matar, M., Kaplan, Z.,
& Cohen, H., 2011. High dose hydrocortisone immediately after trauma may alter the
trajectory of PTSD: Interplay between clinical and animal studies.
Eur.Neuropsychopharmacol. 21(11), 796-809.
18
Page 18 of 23
19
Page 19 of 23
Table 1. Review of papers studying the relationship between PTSD and mortality
Paper Study group N Comparison
group
N Excess mortality
rates amongst
study group?
Cause of death
Bullman &
Kang,
1990 [6]
USA Vietnam
veterans
diagnosed
with PTSD
6,668 Non-Vietnam
veterans
27,917 Yes Accidental poisoning
Motor accidents
Bullman &
Kang,
1994 [7]
USA Vietnam
veterans
diagnosed
with PTSD
4,247 Non-PTSD
Vietnam
veterans
US males
12,010 Yes Suicide
Accidental poisoning
All causes
Mollica et
al., 2001
[8]
Bosnian
refugees
diagnosed
with PTSD
139 Bosnian
refugees not
diagnosed
with PTSD
389 No
Drescher et
al., 2003
[9]
USA dead
male
veterans
diagnosed
with PTSD
1,866 General
population
norms
General
population
norms
Yes Accidents
Motor vehicle accidents
Overdose
Injury
Intentional deaths
Suicide
Effects of chronic
substance use
Alcoholic liver cirrhosis
Alcohol dependence
HIV/hepatitis
Liver disease (unrelated to
alcohol)
20
Page 20 of 23
Johnson et
al., 2004
[10]
USA Vietnam
veterans
diagnosed
with PTSD
154 USA males
aged 45-54
USA males
aged 45-54
Yes
Boscarino
2006 [11,
12],
USA Vietnam
theater and era
veterans
diagnosed with
PTSD
1,050 Non-PTSD
Vietnam theater
and era veterans
14,238 Yes
Ferrada-
Noli et al.,
1998 [13]
Refugees
diagnosed
with PTSD
117 Refugees not
diagnosed with
PTSD
32 Yes Suicide
Boscarino,
2008a [14]
USA Vietnam
veterans
(theater and
era) diagnosed
with PTSD
323 Vietnam
veterans (theater
and era) not
diagnosed with
PTSD
4,139 Yes All causes
Cardiovascular-related
conditions
External causes
Cancer
Infectious diseases
Digestive conditions
Other disease-related
conditions
Boscarino,
2008b [15]
USA Vietnam
veterans
(theater and
era) diagnosed
with PTSD
311 Vietnam
veterans (theater
and era) not
diagnosed with
PTSD
4,017 Yes
21
Page 21 of 23
Table 2. Demographic characteristics of all cohort and of only deceased PTSD and non-
PTSD veterans
All Deceased
PTSD Non-
PTSD
Significance and
Effect Size
(95% CI) *
PTSD Non-
PTSD
Significance and
Effect Size (95%
CI) *
Year of birth Mean=
SD=
n=
1955.1
9.80
2457
1955.7
10.5
2457
t=1.27, p=.21
ES=-.06 (-.11;
.00)
1945
10.14
75
1940
14.73
101
t(173.1)=2.3, p<.02,
ES=.39 (.08; .68)
Age upon death Mean=
SD=
n=
(See Deceased
column)
49.88
10.63
75
45.5
16.81
101
t(170.1)=2.11,
p=.04, ES=.30
(.00; .60)
Draft year Mean=
SD=
n=
1974.6
9.03
2269
1974.5
9.18
2279
t=.60, p=.54
ES=.01 (-.05;
.07)
1969
8.29
15
1967
8.10
60
t(73)=.91, p=.37
ES=.25 (-.32;
.81)
No. of years of
education
Mean=
SD=
n=
10.1
2.1
2184
10.1
2.3
2110
t=.56, p=.58
ES=.00 (-.06;
.06)
9.15
1.68
13
8.81
3.25
58
t=.54, p=.59
ES=.11 (-.49;
.71)
Marital status χ2
(3) 14.1,
p=.003
χ2
(3)=.83, p=.84
Single 11.3%
(n=270)
14.5%
(n=349)
22.2%
(n=4)
26.6%
(n=25)
Married 81.3%
(n=1950)
77.2%
(n=1853)
61.1%
(n=11)
58.5%
(n=55)
Divorced 6.8%
(n=163)
7.7%
(n=185)
11.1%
(n=2)
12.8%
(n=12)
Widower 0.6%
(n=15)
0.5%
(n=13)
5.6%
(n=1)
2.1%
(n=2)
Rank during
incident
χ2
(1)= 4.1, p=.04 χ2
(3)=1.41, p=.77
Private 71.7%
(n=1471)
68.8%
(n=1413)
OR=1.15 (1.00;
1.31)
75%
(n=12)
69%
(n=43)
OR=1.33 (.38;
4.64)
Sergeant** 28.3%
(n=582)
31.2%
(n=642)
25%
(n=4)
31%
(n=19)
Role during
incident
χ2
(1)= 30.0,
p<.001
χ2=(3)7.12, p=.27
Combat 46.6%
(n=898)
37.7%
(n=691)
OR=1.44 (1.26;
1.64)
63%
(n=10)
45%
(n=27)
OR=2.04 (.66;
6.32)
Service 53.4%
(n=1031)
62.3%
(n=1141)
37%
(n=6)
55%
(n=33)
*Effect Size is mean difference/ pooled standard deviation or odd ratio for dichotomous variables.
**Includes 5 officers in the non-PTSD group
22
Page 22 of 23
Table 3. Number of deceased PTSD and non-PTSD veterans
Total
No.
No. of
deceased
amongst
PTSD
No. of deceased
amongst
non-PTSD
Hazards
ratio
(0=non-
PTSD;
1=PTSD)*
95% confidence
interval
Analysis 1 4,914 75/2382 101/2356 .73 0.44-1.02, p=.26
Analysis 2 4,904 75/2382 91/2356 .91 0.66-1.24, p=.19
Analysis 3 4,892 75/2382 79/2356 .94 0.77-1.46, p=.88
Analysis 4 4,889 75/2382 76/2356 .94 0.77-1.46, p=.88
Analysis 5 2,360 10/1170 33/1147 .30 0.14-0.63, p=.00
Analysis 6 2,348 10/1170 21/1147 .47 0.22-1.01, p=.05
Analyses: (1) All participants (2) persons who did not die in battle; (3) Only of pairs in
which the non-PTSD veterans died after his matched PTSD veteran developed PTSD
symptoms; (4) Conditions 1 & 2 (only veterans who did not die in battle and where the
non-PTSD veteran did not die before his matched PTSD veteran developed PTSD
symptoms); (5) PTSD and control veterans have the same military job (combat/officer vs.
non/service back/support); (6) same as 5 with conditions of 2 and 3.
*Survival anlaysis of matched pairs using proportional hazards regression model
23
Page 23 of 23
Table 4. Causes of death by ICD-10 amongst PTSD and non-PTSD veterans
Cause of death PTSD Control
Certain infectious and parasitic diseases 2 (3%) 2 (2%)
Neoplasms 18 (24%) 12 (12%)
Diseases of blood and blood-forming organs 0 1 (1%)
Endocrine, nutritional, and metabolic diseases 1 (1%) 1 (1%)
Mental disorders 0 3 (3%)
Diseases of the nervous system 1 (1%) 0
Diseases of the circulatory system 13 (17%) 23 (23%)
Diseases of the respiratory system 0 2 (2%)
Diseases of the digestive system 2 (3%) 1 (1%)
Diseases of the genitourinary system 2 (3%) 0
Symptoms, signs, and ill-defined conditions 5 (7%) 5 (5%)
Injury and poisoning 13 (17%) 6 (6%)
External causes of morbidity and mortality 5 (7%) 2 (2%)
Missing data 13 (17%) 43 (42%)