Psychological Bulletin Copyright 2003 by the American Psychological Association, Inc.

2003, Vol. 129, No. 1, 52–73 0033-2909/03/$12.00 DOI: 10.1037/0033-2909.129.1.52

Predictors of Posttraumatic Stress Disorder and Symptoms in Adults:

A Meta-Analysis

Emily J. Ozer Suzanne R. Best and Tami L. Lipsey

University of California, San Francisco Veterans Affairs Medical Center, San Francisco, California

Daniel S. Weiss
University of California, San Francisco

A review of 2,647 studies of posttraumatic stress disorder (PTSD) yielded 476 potential candidates for
a meta-analysis of predictors of PTSD or of its symptoms. From these, 68 studies met criteria for
inclusion in a meta-analysis of 7 predictors: (a) prior trauma, (b) prior psychological adjustment, (c)
family history of psychopathology, (d) perceived life threat during the trauma, (e) posttrauma social
support, (f) peritraumatic emotional responses, and (g) peritraumatic dissociation. All yielded significant
effect sizes, with family history, prior trauma, and prior adjustment the smallest (weighted r ⫽ .17) and
peritraumatic dissociation the largest (weighted r ⫽ .35). The results suggest that peritraumatic psycho-
logical processes, not prior characteristics, are the strongest predictors of PTSD.

Throughout the 20th century, interest in the psychological im- coherence of chronic, long-term posttraumatic stress reactions had
pact of trauma has peaked during and after wartime, with the first not materialized, even though these earlier nomenclatures were
major study of combat-related psychological sequelae (then called explicit about the phenomenology of the short-term acute effects
physioneurosis) published in 1941 by A. Kardiner (see Kolb, 1993, of exposure to traumatic stressors.
for a more detailed accounting). Dealing with survivors of World In the early and mid 1970s, veterans of the Vietnam War were
War II death and prisoner of war camps brought some insight (but being hospitalized in Department of Veterans Affairs hospital
less than might have been expected) into the effects of extreme psychiatry units (Haley, 1974) and receiving diagnoses of schizo-
trauma on psychological functioning. In 1968, the Diagnostic and phrenia or other psychotic disorders, even though combat-related
Statistical Manual of Mental Disorders (2nd ed.; DSM–II; Amer- problems had been seen in World War II and Korean War veterans.
ican Psychiatric Association, 1968, p. 49) was published, and a Moreover, at least two prescient scholars (Horowitz & Solomon,
year later, the Manual of the International Statistical Classification 1975) had predicted that the U.S. government would have to
of Diseases, Injuries, and Causes of Death (8th ed.; ICD– 8; World respond to widespread stress-related problems in the Vietnam
Health Organization, 1969, p. 158) appeared, with reference to veteran population, a prediction that was unfortunately all too
“combat fatigue” in the latter and “fear associated with military accurate. Among the phenomena that were most commonly re-
combat and manifested by trembling, running, and hiding” in the ported and observed in combat veterans were intrusive thoughts
former. Both of these clinical phenomena, however, were placed in and images, nightmares, social withdrawal, numbed feelings, hy-
the category of “transient situational disturbances,” even though a pervigilance, and even frank paranoia, especially regarding the
volume entitled The Traumatic Neurosis (Keiser, 1968) was pub- government. Vivid dissociative phenomena, such as flashbacks,
lished the same year that suggested a more profound and chronic occasionally characterized the symptom picture, and these symp-
effect of exposure to traumatic stress. A clear recognition of the toms may have centrally contributed to the misdiagnoses in the
psychotic realm.
Contemporaneously, clinicians began to recognize and write
Emily J. Ozer and Daniel S. Weiss, Department of Psychiatry, Univer- about common patterns in the psychological sequelae of women
sity of California, San Francisco; Suzanne R. Best and Tami L. Lipsey, who had been sexually assaulted, and the term rape trauma syn-
Mental Health Service, Veterans Affairs Medical Center, San Francisco, drome entered the literature (Becker, 1982; Burgess & Holmstrom,
California. 1974). The psychological suffering described was surprisingly
This article is based on a paper presented at the 14th Annual Meeting of similar in terms of the processes, if not the vocabulary. These
the International Society for Traumatic Stress Studies, Washington, DC, in women were observed to be avoidant, on guard, easily startled, and
November of 1998.
flooded with memories and images of the assault that could not be
Correspondence concerning this article should be addressed to either
Emily J. Ozer, who is now at the Department of Psychology and Social
easily dispelled.
Behavior, University of California, Irvine, California 92697-7085; or to It required close to a decade, however, for investigators in both
Daniel S. Weiss, Department of Psychiatry, Box 0984, University of fields of inquiry to recognize that these two separate syndromes
California, San Francisco, California 94143-0984. E-mail: [email protected] were more similar than different despite arising, for the most part,
or [email protected] in different genders and in different contexts. Increased interest in

52
 PREDICTORS OF PTSD: A META-ANALYSIS 53

the study of posttraumatic symptoms and the consequent increase itself need to examined from a more neurobiological perspective.
in empirical investigations culminated in the introduction of post- As well, such an understanding would help shed light on the thorny
traumatic stress disorder (PTSD) into the American diagnostic issue of the wide heterogeneity of traumatic events that give rise to
nomenclature in 1980 in the DSM–III (3rd ed.; American Psychi- PTSD. For example, a broadening of the types of events that some
atric Association, 1980); this introduction was somewhat contro- have considered to be traumatic has led to inclusion in the PTSD
versial, even though the publication of the ninth edition of the literature of studies of highly distressing events that carry infor-
International Classification of Diseases (ICD–9; World Health mation about life threats (e.g., receiving a diagnosis of cancer) that
Organization, 1978) had indicated that organized psychiatry had may or may not invoke the same adrenergic arousal that acute
recognized the basic nucleus of the symptoms. In something of an life-threatening situations do. The presence or absence of adren-
irony, the DSM–III also ushered in the elimination of any diagnosis ergic arousal may well become a key phenomenon that has impli-
that encompassed the acute short-term effects of exposure to a cations for how symptoms of PTSD present and whether an event
traumatic event. It has been only recently (4th ed.; DSM–IV; is deemed “traumatic.”1 If the subjective emotional and physio-
American Psychiatric Association, 1994) that this disorder reap- logical response to the event is overlooked, research that catego-
peared in the American nomenclature as acute stress disorder rizes events may not yield consistent findings that would perhaps
(ASD), despite the knowledge of the short-term pattern of psycho- emerge if arousal were required. Indeed, the criteria for consider-
logical reactions to traumatic stressors that had been acknowl- ing the presence of PTSD itself imply likely adrenergic arousal in
edged and described for at least 50 years prior. the phrase “intense fear, helplessness, or horror” in describing the
Since the introduction of DSM–III, the criteria for PTSD have Criterion A response to exposure to the traumatic event.
been revised twice, but neither revision has drastically modified Despite some suggestions that PTSD be moved from the anxiety
the fundamental set of symptomatic criteria. The core DSM–IV disorders of the DSM to the dissociative disorders, such a change
criteria for PTSD consist of distressing symptoms of (a) reexpe- has little empirical basis because dissociative phenomena are not
riencing of the trauma (e.g., nightmares, intrusive thoughts), (b) routinely present in symptomatic presentations. As we shall dem-
avoidance and numbing (e.g., avoiding reminders, not being able onstrate, however, dissociative phenomena may play an important
to have loving feelings), and (c) increased arousal (e.g., difficulty role in predicting who develops PTSD, and we argue that the
sleeping, hypervigilance, exaggerated startle response; American dissociative phenomena are more closely related to activation of
Psychiatric Association, 1994). Besides the unusual feature of the the HPA axis than are predictors such as a family history of
requirement of a traumatic (life-threatening) event in Criterion A, psychopathology.
a cardinal feature of PTSD is that the disabling aspects of the One obvious predictor of having PTSD symptoms is having met
phenomenology are linked to the event. Thus, intrusive images or the criteria for ASD during the month after exposure (the only
thoughts are typically of some aspect of the actual event, not just window during which it can be diagnosed), as a severe immediate
random content that comes to mind with a distressing and intrusive reaction might logically be implicated in having symptoms further
quality and cannot easily be dispelled once it has entered conscious out from the trauma. It is clear that many who meet criteria for
awareness. PTSD some time after the exposure to the traumatic stressor do not
Because the traumatic event is typically one of immediate life evidence symptoms of ASD; thus, ASD is not two-way pathogno-
threat and horror (e.g., sexual assault at knifepoint, torture, com- monic. There is, however, less evidence and certainty about
bat, being in a flooded shelter during a hurricane and having the whether having ASD is typically followed by chronic PTSD.
water level quickly rise to one’s shoulders), there is also typically Because the diagnostic criteria of ASD include dissociative phe-
very high adrenergic arousal that accompanies the experience. nomena, there may well be an important link between peritrau-
Current neurobiological models of the acute stress response (e.g., matic dissociation or other peritraumatic emotional responses and
Davis, 1992; LeDoux, 2000; McEwen, 1995) implicate the amyg- the formation or manifestation of symptoms. Nonetheless, ASD is
dala and hippocampus as key brain areas that are involved in the beyond the scope of our efforts here.
registration of potentially dangerous situations and in the later The study of traumatic sequelae across cultures, types of trau-
formation of the memories of such events, and give the matic exposure, and populations has increased dramatically in
hypothalamic–pituitary–adrenal (HPA) axis a central role in both recent years, with a burgeoning published literature (i.e., more
the development of PTSD and its maintenance (e.g., Yehuda, than 4,200 PsycINFO citations can be found using PTSD as a key
1998). The underlying phenomenon that these elegant brain mod- word). Of particular interest has been the search for factors that
els are trying to explain is related to the long-known phenomenon explain why some people who are exposed to traumatic stress
that memories formed under emotionally arousing situations be- develop PTSD whereas others similarly exposed do not (see
have differently from those that are not (Bower, 1981). There is
Brewin, Andrews, & Valentine, 2000), as the epidemiologic liter-
some evidence from preclinical and human studies (McGaugh &
ature has made clear that PTSD is not an inevitable result of
Cahill, 1997) that memory formation under circumstances of emo-
exposure. The particular focus of our meta-analysis is a compari-
tional arousal can be altered by blocking the effects of adrenalin.
son of more static predictors to predictors more likely to be
This finding suggests that the degree of arousal during or imme-
implicated in the psychological and neurobiological processes
diately after the traumatic event may have fundamental importance
consequent to exposure to a traumatic stressor.
for the development of the intrusive and hyperarousal symptoms of
PTSD through the mechanisms involved in both the formation and
recall of episodic memories. If this is so, an understanding of 1
The conceptual discussion of the relationship between the event and
which variables are good predictors of the development of symp- the normative response to the event as definitory of traumatic is beyond the
toms as well as an understanding of some aspects of the disorder scope of this discussion but merits serious consideration elsewhere.
54 OZER, BEST, LIPSEY, AND WEISS

Prevalence of PTSD there are, how strong are the effects? An important clue may be
found in the observation that there is considerable individual
In the past decade, several large studies assessed the prevalence variability in psychological response to traumatic stress.
of PTSD in adults. The National Vietnam Veterans Readjustment
Study (NVVRS; Kulka et al., 1990; Weiss et al., 1992), an exten-
Prior Reviews and Focus of the Present Meta-Analysis
sive national probability survey of male and female Vietnam
theater veterans (i.e., military service personnel, including health Qualitative reviews of the PTSD literature highlighted the press-
professionals, who served in the Vietnam War), estimated ing need for attention to the role of other personal and environ-
that 30.9% of men and 26.0% of women met the diagnostic criteria mental variables in predicting PTSD (Emery, Emery, Shama,
for PTSD at some point since their service in Vietnam. The Quiana, & Jassani, 1991; Fontana & Rosenheck, 1994; Green,
estimate of current prevalence was 15.2% for men and 8.5% for 1994). Those reviews were neither comprehensive nor quantitative
women (Schlenger et al., 1992). The percentage of veterans with (using measures of effect size [ES]) in their efforts to analyze the
current clinically significant PTSD symptoms (i.e., partial PTSD) adult PTSD literature.
was 11.1% for men and 7.8% for women. Combining estimates for An initial quantitative review of the predictors of PTSD in
all current full and partial PTSD resulted in an estimate of roughly adults was published during the period in which our work was
830,000 Vietnam theater veterans who continued to experience being conducted (Brewin et al., 2000). That meta-analysis studied
significant posttraumatic distress or impairment approximately 20 five demographic predictors (age, gender, socioeconomic status,
years after their exposure to one or more traumatic stressors education, and race) and nine other variables. Though the authors
(Weiss et al., 1992). did not do so, these nine can be categorized into three groups: (a)
Epidemiological research with civilian populations has found historical or static person characteristics such as family psychiatric
lifetime PTSD prevalence rates to be 9.2% for adults belonging to history, intelligence, childhood adversity and trauma, and other
an urban health maintenance organization (Breslau, Davis, An- previous trauma; (b) trauma severity; and (c) social support and
dreski, & Peterson, 1991), 12.3% for a nationally representative intercurrent life stress in the interval between traumatic exposure
sample of women in the National Women’s Study (NWS; Resnick, and measurement of PTSD symptoms or the presence of the
Kilpatrick, Dansky, Saunders, & Best, 1993), and 7.8% in the best disorder. The literature searching, study identification, and selec-
epidemiological study to date—the National Comorbidity Study tion procedures in the Brewin et al. study yielded a set of 77
(NCS; Kessler, Sonnega, Bromet, Hughes, & Nelson, 1995)—a studies. Their main finding was that every predictor produced a
nationally representative study of 5,877 people ages 15 to 45 years. weighted average ES (r) that was statistically significant, but that
Of particular note is that the lifetime prevalence for women was the effect of the predictors was quite heterogeneous: Race, the
twice that for men in the NCS (10.4% vs. 5.0%). The prevalence weakest, was associated with an average ES of .05, whereas the
of PTSD is high among immigrants and refugees, particularly strongest, lack of social support, was associated with an average
those who immigrated because of armed conflict or political re- weighted ES of .40. Most of the other ESs were in the range of .10
pression, with a community study indicating that 52.0% of Central to .19. There was no particular conceptualization offered to ac-
American refugees from war or political persecution met criteria count for the different ESs associated with the different predictors.
for PTSD (Cervantes, Salgado de Snyder, & Padilla, 1989); Another key finding of Brewin et al. (2000) was evidence of
smaller studies with Cambodian refugees also indicate high levels considerable heterogeneity in effects for specific predictors, de-
of PTSD (e.g., Carlson & Rosser-Hogan, 1993, 1994). pending on which trauma group was being analyzed; predictors did
Another important finding from these epidemiological studies is not demonstrate consistent degrees of magnitude in every trauma
that the PTSD criterion in DSM–III and DSM–III–R (3rd ed., rev.; group in which they were studied. Moreover, the heterogeneity of
American Psychiatric Association, 1987) referring to the charac- ES estimates within the studies for any individual predictor also
teristics of the stressor (Criterion A, a necessary but not sufficient showed wide variability across the set of predictors, leading to
condition for the diagnosis of PTSD) was empirically inaccurate in even less capability to draw generalizable conclusions. Brewin et
that it required the event to be outside the range of normal human al. also examined the effects of sample and study characteristics on
experience. The NWS found that 69.0% of respondents reported ESs. As with the ESs themselves, these variables produced a
having been exposed to a traumatic event at some point in their conflicting pattern of results. As an example, consider the com-
lives, a finding that was replicated in the NCS, which found a parison of studies with a civilian sample and studies with a military
lifetime prevalence of 60.7% for men and 51.2% for women for sample. The average ES for social support was significantly dif-
exposure to an event that would qualify as a Criterion A event. As ferent for the two types of samples, but the average ES for life
a consequence, in the DSM–IV the phrase, “outside the range of stress was not different in civilian and military samples. Similarly,
normal human experience,” was deleted and replaced with Crite- whether PTSD was indexed by a dichotomous diagnosis or a
rion A2, which requires that the person’s response to the stressor continuous measure of symptoms made a significant difference on
be one of intense fear, helplessness, or horror. the average ES for trauma severity but not for social support.
The contrast between the lifetime prevalence of exposure to a The main conclusion of Brewin et al. (2000) was that the set of
traumatic event of over 50% and the lifetime prevalence of PTSD studies displayed remarkable heterogeneity. This heterogeneity
at roughly 7% was the disparity that led directly to the key aspect was both in the range of ESs across studies but within individual
of our investigation. Given that roughly 50%– 60% of the U.S. predictors as well as within the sets of studies of individual
population is exposed to traumatic stress but only 5%–10% de- predictors, as these sets were moderated by characteristics of the
velop PTSD, are there any systematic risk factors or correlates that studies such as the type of sample or the method of measuring
identify who will or will not develop the disorder? Moreover, if PTSD symptoms or diagnosis. As a consequence, Brewin et al.
 PREDICTORS OF PTSD: A META-ANALYSIS 55

warned against attempting “to build a general vulnerability model The meta-analysis did not include as a predictor the exposure to
for all cases of PTSD” (p. 756), an approach that has been taken the index traumatic event about which PTSD symptoms were
with other mental disorders such as schizophrenia or bipolar dis- measured because exposure to such an event is a necessary crite-
order. Instead, they made the following suggestion: rion for diagnosis of PTSD.
Our decision to use the conceptual term predictor rather than the
The data may be regarded as consistent with a model in which the more operational term correlate, despite the largely retrospective
impact of pretrauma factors on later PTSD is mediated by responses nature of the PTSD literature, was based on our desire to distin-
to the trauma or, alternatively, with a model in which pretrauma
guish between factors that, if measured prospectively, could in-
factors interact with trauma severity or trauma responses to increase
the risk of PTSD. In either case, [these data] suggest that it may be
deed serve as predictors from those factors that could not be
more productive to investigate more proximal links in the causal conceptualized as predictors even in prospective designs. Two
chain, such as the association between pretrauma risk factors and examples of the latter are comorbid diagnoses and survivor guilt.
immediate trauma responses. (Brewin et al., 2000, p. 756) In using the term predictor, we make no strong claim about
causality; instead, our conceptual framework is more akin to a risk
Quite independently of and prior to the conclusions of Brewin et factor than to a causal factor.
al. (2000), our focus was on only two types of predictors: (a)
person characteristics salient for psychological processing and
Method
functioning and (b) aspects of the traumatic event or its sequelae,
just the factors that Brewin et al. suggested required greater em- Sample of Studies
phasis and exploration. We omitted consideration of demographic
factors such as gender, education, and ethnicity, as none of these Relevant studies were initially identified through broad, computerized
factors is plausibly implicated in the psychological processes of database searches of MEDLINE, PsycINFO, and PILOTS for publications
trauma response. As it turned out, Brewin et al.’s review showed between 1980 (the year PTSD was first included in the DSM) and 2000.
that the effects of these factors were relatively small and did not Key words entered were posttraumatic stress disorder, PTSD, trauma
(including combat trauma, rape trauma, and emotional trauma) and stress
lead to a significantly greater understanding of the development of
(including stress reaction, stress response), and traumatic neurosis. We
PTSD. used a range of key words to reflect the changing terminology in the
Whereas there are numerous individual and environmental fac- empirical literature over the course of the 2 decades covered here. In
tors that could be conceptually linked to the development of PTSD addition, each issue of the Journal of Traumatic Stress (published by the
symptoms, there has been sufficient empirical research on only a International Society for Traumatic Stress Studies)— beginning with Vol-
small number of these predictors, and between the current study ume 1, No. 1, in January 1988 through Volume 13, No. 4, in October
and the work of Brewin et al. (2000), the entire universe of 2000 —was reviewed for potentially pertinent research. We also reviewed
predictors with even minimal empirical research has been covered. the reference sections of articles selected for study inclusion to identify any
This meta-analysis focused only on those nondemographic vari- additional relevant studies. Studies included for review were those in which
ables other than exposure to trauma itself that (a) could be legit- quantitative methods were used to examine the predictors of PTSD or
PTSD symptoms in adults (ages 18 and over). A predictor was defined as
imately conceptualized as predictors of PTSD symptoms and de-
any variable examined as a potential contributor to variability in PTSD
velopment of the disorder, or that could influence the development symptom levels or diagnostic status. Studies focusing on conditions co-
of PTSD symptoms before they are expressed; and (b) have been morbid with PTSD (e.g., substance abuse, depression) that were not as-
studied adequately to generate sufficient data for meta-analytic sessed prior to the onset of trauma were not included. Also excluded were
purposes (i.e., at least two empirical articles). Comorbid disorders examinations of the longitudinal course of PTSD, as well as studies of
or associated symptoms occurring simultaneously with or after the acute trauma response occurring prior to 1 month posttrauma. Single case
onset of PTSD were not considered as predictors (further criteria studies were also eliminated. Finally, we excluded doctoral dissertations,
for inclusion and exclusion are discussed in the Method section). reasoning that those dissertations whose positive findings would be rele-
Because of the almost exclusively retrospective assessment meth- vant would have had a high probability of being published. To address the
odology of the PTSD literature, the large majority of predictors concern that unpublished data from dissertations or other research studies
might have findings that could alter the conclusions of our meta-analysis,
studied in the meta-analysis were not assessed prior to the devel-
we conducted a “file drawer” calculation (Rosenthal, 1979, 1991) to test
opment of PTSD symptoms (although this would have been far the robustness of our findings for each ES (see below).
preferable), though there are several exceptions to this rule The search identified 2,647 articles in the English language. Our review
(Schnurr, Friedman, & Rosenberg, 1993; Shalev, Sahar, et al., of the abstracts using the criteria outlined above identified 476 studies
1998). eligible for study inclusion. In the second stage of review, additional
The approach to the literature we have described led us to focus studies were eliminated because they did not specifically assess PTSD
on seven predictors of PTSD symptoms or of PTSD diagnosis that symptoms (e.g., studies that reported only general symptoms) or because
had been sufficiently studied to include in this meta-analysis: (a) a they assessed some symptoms of PTSD but not all of the clusters as defined
history of at least one other trauma prior to the index traumatic by DSM–IV criteria (e.g., studies that assessed intrusion and avoidance and
event, (b) psychological adjustment prior to the traumatic event, numbing symptoms using the original Impact of Event Scale, Horowitz,
Wilner, & Alvarez, 1979, but did not assess hyperarousal). A subset of 47
(c) family history of psychopathology, (d) perceived life threat
studies was not included in the meta-analysis because it contained studies
during the traumatic event, (e) perceived social support following that assessed the relationship between severity of exposure to traumatic
the traumatic event, (f) peritraumatic emotionality— high levels of stress and PTSD symptoms only and did not assess any of the correlates we
emotion during or in the immediate aftermath of the traumatic considered. Decisions to exclude studies from the meta-analysis were
event, and (g) peritraumatic dissociation— dissociative experi- independently reviewed by Emily J. Ozer and Suzanne R. Best to ensure
ences during or in the immediate aftermath of the traumatic event. that decisions were consistent with study inclusion criteria. These proce-
56 OZER, BEST, LIPSEY, AND WEISS

dures yielded 68 empirical studies that we included in the meta-analysis. multiple ESs for the same general construct (e.g., separate ESs for family
Because of our reduced number of predictors and slightly different decision history of phobia, generalized anxiety disorder, and depression), the ESs
rules for inclusion, our total number of studies was eight fewer than Brewin were averaged (weighting for dfs) to calculate one overall ES for the
et al. (2000). Our meta-analysis, however, included 21 studies that they did construct. Fifth, only single-df comparisons were appropriate for selection
not include. These nonoverlapping studies did not solely report on peri- in the meta-analysis (Rosenthal, 1994). Sixth, articles were reviewed
traumatic dissociation and emotional responses, the two novel predictors in closely to ensure that the same study published in more than one article
our meta-analysis. Nineteen of these studies were published before 2000, would be counted as one study; authors, study design, and study sampling
and 2 studies were published during 2000 (although Brewin et al., 2000, were examined to ensure that no study contributed more than one ES per
included studies published in 2000, it is likely that these two latter studies construct.
were not possible to include in their meta-analysis). This number of We converted ES estimates for single-df analyses to the common metric
nonoverlapping studies excludes instances in which the two meta-analyses of Fisher’s z transformation of r (Rosenthal, 1994). ESs with artificial
report results for predictors from the same dataset but from published dichotomization of PTSD symptoms were corrected using the adjustment
articles with differing publication dates or authorship. We evaluated for procedure recommended by Hunter and Schmidt (1990a, 1990b, 1994).
inclusion all studies included in the Brewin et al. meta-analysis. This correction takes into account the proportion of subjects in each half of
the dichotomy and increases the ES accordingly to make up for artificial
attenuation. Once appropriate adjustments were made, the mean of z
Coding of Studies (weighted by the df associated with each ES) and 95% confidence limits
were calculated for each set of predictors using the more conservative
All eligible studies were thoroughly reviewed by Emily J. Ozer and
simple random effects approach (k ⫽ number of studies; Rosenthal, 1994).
Suzanne R. Best and were coded for the date of publication, country of
Mean ESs and confidence limits were converted back to r for ease of
origin, size and demographics of study sample (i.e., age, ethnicity, gender),
interpretation. If the 95% confidence interval did not include zero, the null
type of trauma experienced, length of time elapsed between the trauma and
hypothesis that the relationship between the specific predictor and PTSD
assessment of PTSD, clinical status of sample (i.e. community sample,
symptoms was zero could be rejected at the p ⫽ .05 level.
medical patients, individuals seeking mental health services), method and
Cohen’s (1988) guidelines for interpreting the size of sample-weighted
measure for assessing PTSD, and a measure of ES for the relationship
average correlations were used: A small ES is r ⫽ .10; a medium ES is r ⫽
between specific correlates and PTSD symptoms. In addition, we coded
.30; and r ⫽ .50 is a large ES.
methodological aspects of each study, including the following: (a) Was the
issue of missing data addressed by the study? and (b) Was PTSD assessed
as a categorical (i.e., met diagnostic criteria for PTSD) or continuous (i.e., Investigation of Within-Predictor Variability
symptom severity) variable? To reduce the possibility of coding errors,
We investigated the extent to which the ESs for each predictor varied
Suzanne R. Best entered the codes, and then Emily J. Ozer rechecked them.
according to (a) the type of sample studied (i.e., community, medical
When studies neglected to provide a single degree of freedom (single-df)
patients, or individuals seeking mental health services); (b) the length of
ES, we used a combination of strategies consistent with established guide-
time that had elapsed between exposure to the traumatic event and assess-
lines for meta-analytic research (Cooper & Hedges, 1994; Lipsey & Wil-
ment of PTSD symptoms; (c) the type of trauma studied as the target
son, 2000) to derive ES estimates. When sufficient raw data were available
incident; and (d) the method used to assess either PTSD symptoms or
from the published study, we conducted t tests and chi-square analyses as
diagnosis. To facilitate between-groups comparisons, we categorized stud-
appropriate on the study data to obtain ESs. If the study did not provide
ies into three groups on the basis of the amount of time elapsed between
sufficient raw data or ESs but indicated the degrees of freedom and
trauma exposure and PTSD assessment: 1 to 6 months, 6 months to 3 years,
significance level for a particular analysis, we estimated the ES as outlined
and beyond 3 years. The selection of these time frames followed a rough
by Rosenthal (1994). If the study indicated that an analysis was not
categorization to distinguish among acute reactions (1 to 6 months), clearly
statistically significant but provided no significance level or data (this
chronic reactions (beyond 3 years), and intermediate-term reactions. With
occurred for three ESs), we imputed an ES of zero for use in subsequent
regard to the type of trauma, events were broadly characterized as combat
analyses (Rosenthal & DiMatteo, 2000). Studies that did not provide
exposure, interpersonal violence (i.e., human-perpetrated violence that
single-df ESs nor sufficient data to calculate a single-df ES were excluded
occurred in a civilian or nonmilitary context), accidents, or disasters.
from the quantitative calculations; the findings of several epidemiological
Because of the small number of studies of disaster as the target event (three
studies that were excluded for this reason are reviewed in the Discussion
in total and no more than one appearing in any one predictor category),
section with respect to the consistency of those studies’ results with our
comparative analyses focused on the first three types of trauma only.
overall findings.
Method of assessment was dichotomized into self-report versus interviewer
or clinician assessment (in a few studies, the PTSD decision included both
Selection and Calculation of ES Estimates self-report and interview data, and for these few studies, method of assess-
ment was assigned as missing). The significance of the difference between
The studies reviewed for this meta-analysis were conducted using a wide the z-transformed correlation coefficients for each subgroup (divided on
range of research designs and ESs. Because only one ES per construct per the basis of type of sample, time elapsed since trauma, and type of trauma)
study could be included in the meta-analysis (Rosenthal, 1994), we estab- was then tested within each predictor (Cohen & Cohen, 1983).
lished a set of decision rules consistent with standards for meta-analytic To investigate how these moderator variables were related to each other
analyses (Cooper & Hedges, 1994) for this diverse literature. First, ESs for as well as to the ES they moderated, we conducted two sets of analyses.
current rather than past PTSD symptoms were selected when both were The first was a set of chi-square analyses crossing each moderator with
presented. Second, for longitudinal studies that presented multiple assess- every other. The second set was a series of hierarchical multiple regression
ments of symptoms over time, the ES representing the assessment of PTSD analyses focused on looking at the increment in variance accounted for by
symptoms closest in time to the trauma (but after 1 month—the minimum a moderator in the presence of the other moderators.
time frame at which PTSD can be diagnosed— had elapsed) was selected.
Third, when researchers conducted analyses using both a diagnosis of Results
PTSD and symptom severity as outcome measures, we used the ES for
PTSD symptom severity because of the statistical advantages of continu- In this section we present the evidence for each predictor of
ously measured variables in predictive research. Fourth, if a study reported PTSD symptoms. The tables contain specific information regard-
 PREDICTORS OF PTSD: A META-ANALYSIS 57

ing each study included in the analysis for each predictor as well sonality Inventory scores predicted PTSD symptoms among Viet-
as the combined ES r (unweighted and weighted for sample size) nam combat veterans (Schnurr et al., 1993) and precombat neu-
and 95% confidence intervals for the overall unweighted and roticism predicted PTSD among WWII veterans (Lee, Vaillant,
weighted ESs. Torrey, & Elder, 1995). Preservice psychiatric problems, however,
were found to be negatively related to PTSD symptoms in one
History of Prior Trauma study of WWII prisoners of war (Speed, Engdahl, Schwartz, &
Eberly, 1989).
The weighted average correlation from 23 studies (combined Because a small cluster of studies assessed prior depression as
n ⫽ 5,308) of the relationship between a history of prior trauma the predictor representing prior adjustment, it was possible to
and PTSD symptoms or diagnosis was .17, a statistically signifi- examine whether studies in which depression was the prior psy-
cant but small ES (see Table 1). On average, somewhat higher chological problem had a different average ES from the remainder
levels of PTSD symptoms were reported by those who reported of the studies that had examined the impact of other prior adjust-
experiencing a traumatic event prior to the target stressor than ment problems. The results of this comparison were informative.
those who indicated that they had not been previously exposed. The weighted r for the four studies of depression was .32, and the
Individual ESs ranged from .00 to .46. Slightly more than half of comparison of this value to the remaining studies (weighted r ⫽
the studies (13) investigated the role of at least one prior trauma .15) was statistically significant (z ⫽ 3.78, p ⬍ .01). This was a
that had occurred in childhood. Calculation of the average conservative comparison because studies that examined any prior
weighted ES for this subgroup of studies yielded the identical ES affective or Axis I disorder (including depression) were included
(r ⫽ .17) as the overall analysis produced, implying that, in in the remainder.
general, prior childhood trauma imparts no greater risk than prior The strength of the relationship between prior adjustment prob-
adult trauma. lems and PTSD symptoms or diagnosis did not vary according to
The strength of the relationship between prior trauma and PTSD the type of sample studied. The relationship between prior adjust-
did not differ according to the type of sample studied (i.e., com- ment problems and PTSD did differ as a function of the type of
munity, medical patients, or individuals seeking mental health traumatic experience that constituted the target event, the amount
services) nor the time elapsed since the trauma, nor the method by of time elapsed, and the method of assessing PTSD. Having prior
which PTSD was assessed. The relationship between prior trauma adjustment problems was more strongly related to PTSD when the
and PTSD did vary by the type of traumatic experience studied as traumatic experience involved noncombat interpersonal violence
the target event: Having had a prior trauma was more strongly (weighted r ⫽ .31) or accident (weighted r ⫽ .28) than when the
related to PTSD when the traumatic experience involved noncom- traumatic experience resulted from combat exposure (weighted
bat interpersonal violence (e.g., civilian assault, rape, domestic r ⫽ .06; for interpersonal violence vs. combat, z ⫽ 8.70, p ⬍ .01;
violence; weighted r ⫽ .27) than when the traumatic experience for accident vs. combat, z ⫽ 4.72, p ⬍ .01). Stronger relationships
resulted from combat exposure (weighted r ⫽ .18; z ⫽ 3.02, p ⬍ were found between prior adjustment problems and PTSD among
.01) or an accident (weighted r ⫽ .12; z ⫽ 2.10, p ⬍ .05). studies in which less time had elapsed between the traumatic event
and assessment of PTSD (1 to 6 months [weighted r ⫽ .24] vs. 6
Psychological Problems Prior to Target Stressor months to 3 years [weighted r ⫽ ⫺.02]; z ⫽ 5.14, p ⬍ .01, or 3
years and longer [weighted r ⫽ .11]; z ⫽ 3.35, p ⬍ .01). The
Twenty-three studies (combined n ⫽ 6,797) contributed to the weighted ES for prior adjustment was higher for interview studies
ES for prior adjustment problems. The weighted correlation be- (r ⫽ .29) than for self-report studies (weighted r ⫽ .15; z ⫽ 5.31,
tween prior adjustment problems and PTSD symptoms or diagno- p ⬍ .01).
sis was also .17, a statistically significant but small ES (see Table
2). This positive correlation shows that individuals who reported Psychopathology in Family of Origin
problems in psychological adjustment prior to experiencing the
target stressor reported higher PTSD symptoms, on average, than The weighted average correlation for nine studies (combined
those who disavowed prior adjustment problems. The ESs ranged n ⫽ 667) of the relationship between psychopathology in the
from ⫺.13 to .47. family of origin and PTSD symptoms or diagnosis was also
The types of prior adjustment problems associated with in- significantly different from zero at .17 (see Table 3). This was a
creased PTSD symptoms included previous mental health treat- small ES, indicating that individuals who reported a family history
ment (Carlier, Lamberts, & Gersons, 1997; Jeavons, Greenwood, of psychopathology reported higher PTSD symptoms or higher
& Horne, 2000), pretrauma emotional problems (Ehlers, Mayou, & rates of PTSD than those without such a family history. The ESs
Bryant, 1998), pretrauma anxiety or affective disorders (Blan- for this predictor ranged from ⫺.06 to .43.
chard, Hickling, Taylor, & Loos, 1995; Breslau et al., 1991; The strength of the relationship between family history of psy-
Fauerbach et al., 1997; North, Smith, & Spitznagel, 1994; Resnick, chopathology and PTSD symptoms or diagnosis did not vary
Kilpatrick, Best, & Kramer, 1992; Shalev, Freedman, et al., 1998; according to the type of sample or the time elapsed between the
Solomon, Oppenheimer, Elizur, & Waysman, 1990; Tedstone & traumatic event and the assessment of PTSD. The relationship
Tarrier, 1997; Ursano et al., 1999), and antisocial personality between family history of psychopathology and PTSD, however,
disorder prior to military service (Cottler, Compton, Mager, did vary by the type of traumatic experience that constituted the
Spitznagel, & Janca, 1992). In two of the genuinely prospective target event and the method by which PTSD was assessed. Having
studies in the field (i.e., the predictor actually was assessed prior to a family history of psychopathology was more strongly related to
the target event), elevated precombat Minnesota Multiphasic Per- (text continues on page 61)
Table 1 58
Weighted Correlations of Prior Trauma With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

Study Sample Broad trauma type Clinical status Trauma recency Type of prior trauma PTSD measure df ra

Andrews et al. (2000) Crime victims Interpersonal violence Medical 6 months Childhood abuse PSS 157 .21
Astin et al. (1995) Domestic violence victims Interpersonal violence Medical Variable Childhood abuse SCID (DSM–III–R) 87 .17
Blanchard et al. (1995) Motor vehicle accident Accident Medical 1–4 months Prior trauma CAPS 158 .02
victims
Bremner et al. (1993) Vietnam veterans Combat Clinical 19 years Prior trauma SCID 66 .31b
Breslau et al. (1998) Community Variable Community Variable Mixed DIS (DSM–IV) 1922 .14
Dunmore et al. (1999) Assault survivors Interpersonal violence Community Variable Any prior trauma not including PSS–SR 92 .30
child abuse
Engel et al. (1993) Desert Storm veterans Combat Clinical 7–12 months Precombat abuse MCS for Desert Storm 292 .19
Fontana et al. (1997) Female Vietnam veterans Combat Community 20⫹ years Adult sexual abuse MCS, CAPS 396 .27
(NVVRS sample)
Jeavons et al. (2000) Motor vehicle accident Accident Medical 3 months Not specified PTSD–I 62 .03
victims
Kemp et al. (1995) Female victims of Interpersonal violence Variable Variable Childhood abuse MCS–CV 227 .14
domestic violence and
verbal abuse
Koopman et al. (1994) Fire survivors Accident Community 7–9 months Prior trauma and highly MCS–CV 154 .26
stressful life events
Kramer & Green (1991) Women treated for sexual Interpersonal violence Medical Variable Previous sexual assault Semistructured interview 30 .40
assault based on DSM–III
Michaels et al. (1999) Motor vehicle accident Accident Medical 6 months Prior abuse MCS–CV 176 .18
victims
North & Smith (1992) Homeless Varied Community Variable Child physical abuse DIS 522 .21
Perrin et al. (1996) Female domestic violence Interpersonal violence Clinical Not specified Prior battering or rape as adult MMPI–PK 69 .11
victims
Resnick et al. (1995) Rape victims Interpersonal violence Medical 17–57 days Previous assault SCID (DSM–III–R) 37 .43
Riggs et al. (1992) Female crime victims Interpersonal violence Mixed 1 month Prior sexual assault PTSD Symptom Scale 86 .15
Roberts et al. (1998) Female domestic violence Interpersonal violence Medical Variable Child abuse PTSD Checklist 161 .24
victims
OZER, BEST, LIPSEY, AND WEISS

Solomon (1995) Israeli combat veterans Combat Community 18 years Childhood exposure to war IES, PTSD–I 348 .03
Speed et al. (1989) WWII prisoners of war Combat Community 40⫹ years Child trauma Rating scale based on 62 .16
DSM–III
Ursano, Fullerton, Motor vehicle accident Accident Medical 1 month Prior trauma of any kind SCID (DSM–III–R, DSM– 122 .00c
Epstein, Crowley, victims IV)
Kao, et al. (1999)
Van Velsen et al. (1996) Torture victims Interpersonal violence Medical Variable History of other persecution Clinical diagnosis using 60 .37b
DSM–III
Zaidi & Foy (1994) Vietnam combat veterans Combat Clinical Variable Child physical abuse MCS 22 .46
Unweighted average r ⫽ .21 (CI ⫽ .16, .27)
Weighted average r ⫽ .17 (CI ⫽ .11, .22)

Note. Average r calculated from sample-weighted correlation for each effect size. PSS ⫽ Posttraumatic Stress Disorder Symptom Scale (Foa, Riggs, Dancu, & Rothbaum, 1993); SCID ⫽ Structured
Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders ([3rd ed.; DSM–III; American Psychiatric Association, 1980]; Spitzer & Williams, 1983); DSM–III–R ⫽ Diagnostic and
Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); CAPS ⫽ Clinician Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); DIS ⫽
Diagnostic Interview Schedule (Robins, Helzer, Croughan, & Ratcliff, 1981); DSM–IV ⫽ Diagnostic and Statistical Manual of Mental Disorders (4th ed., American Psychiatric Association, 1994);
PSS–SR ⫽ PSS—Self-Report; MCS ⫽ Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); NVVRS ⫽ National Vietnam Veterans Readjustment
Study (Kulka et al., 1990); PTSD–I ⫽ PTSD Inventory (Solomon et al., 1993); MCS–CV ⫽ MCS—Civilian Version (Lauterback, Vrana, King, & King, 1997); MMPI–PK ⫽ Minnesota Multiphasic
Personality Inventory PTSD—Keane Scale (Keane, Malloy, & Fairbank, 1984); IES ⫽ Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979); CI ⫽ 95% confidence interval.
a
Positive correlation indicates presence of prior trauma associated with PTSD symptoms. b Effect size adjusted for dichotomization of PTSD variable. c Effect size of zero imputed because study
indicated only that analysis was “not statistically significant.”
Table 2
Weighted Correlations of Psychological Adjustment Prior to the Traumatic Event With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms
Study Sample Broad trauma type Clinical status Trauma recency Prior adjustment assessment method PTSD measure df ra
Basoglu et al. (1994) Survivors of torture Interpersonal violence Community Variable Prior psychopathology SCID 55 .09b
Blanchard et al. (1995) Motor vehicle accident Accident Medical 1–6 months Depression prior to trauma CAPS 158 .35c
victims
Boscarino (1995) Vietnam combat and era Combat Community 3⫹ years Self-reported child/adolescent delinquency DIS 2490 .06c
veterans
Carlier et al. (1997) Police officers with and Interpersonal violence Community 1–6 months Previous psychiatric care SI–PTSD 262 .26c
without PTSD
Cottler et al. (1992) Community survey Mixed Community Variable History of antisocial personality disorder DIS 430 .12c
respondents
Dunmore et al. (1999) Physical assault victims Interpersonal violence Community Variable Preassault psychological problems PTSD Scale 92 .31c
Ehde et al. (2000) Burn victims Accident Medical 1–6 months Prior positive mental health Checklist based on 68 .13
DSM–III–R
Ehlers et al. (1998) Motor vehicle accident Interpersonal violence Medical 1–6 months Preaccident emotional problems PSS 888 .20
victims
Fauerbarch et al. (1997) Burn patients Accident Medical 1–6 months Preaccident anxiety or affective disorder SCID (DSM–III–R) 95 .34c
Fontana et al. (1997) Female Vietnam veterans Combat Community 3⫹ years Premilitary conduct disorder MCS, CAPS 396 .05
(NVVRS sample)
Foy et al. (1984) Vietnam combat veterans Combat Clinical 3⫹ years Premilitary mental health contacts Self-report checklist 43 .07
Jeavons et al. (2000) Motor vehicle accident Accident Medical 1–6 months Any prior psychiatric treatment PTSD–I 62 .31
victims
Lee et al. (1995) Longitudinal study of Combat Community 6 months to 3 years Neuroticism assessed prior to military Checklist based on 107 .20
WWII veterans service DSM–III
Madakasira & O’Brien Tornado victims Disaster Community 1–6 months History of mental illness Adapted Hopkins 116 .03c
(1987) Checklist 90
North & Smith (1992) Homeless Mixed Community Variable Any prior psychiatric diagnosis DIS 522 .33c
North et al. (1994) Survivors of shooting spree Interpersonal violence Community 1–6 months Predisaster depression DIS 124 .41c
Resnick et al. (1992) Crime victims Interpersonal violence Community 3⫹ years Prior Axis I disorder Adapted DIS using 295 .47c
DSM–III criteria
Schnurr et al. (1993) Vietnam era veterans Combat Community 3⫹ years MMPI Paranoia Scale assessed before SCID (DSM–III–R) 91 .31c
military service
Shalev, Freedman, et al. Medical patients Mixed Medical 1–6 months Depression prior to trauma MCS–CV 211 .28
PREDICTORS OF PTSD: A META-ANALYSIS

(1998)
Solomon et al. (1990) Combat veterans Combat Clinical 6 months to 3 years Prior combat stress reaction PTSD–I 63 .09
Speed et al. (1989) WWII prisoners of war Combat Community 3⫹ years Preservice psychiatric problem Rating scale based 62 ⫺.13
on DSM–III
Tedstone & Tarrier (1997) Burn victims Accident Medical 1–6 months Prior depression Penn Inventory 45 .17c
Ursano, Fullerton, Epstein, Motor vehicle accident Accident Medical 1–6 months Prior anxiety or depressive disorder SCID (DSM–III–R, 122 .25c
Crowley, Kao, et al. victims DSM–IV)
(1999)
Unweighted average r ⫽ .21 (CI ⫽ .13, .26)
Weighted average r ⫽ .17 (CI ⫽ .10, .23)

Note. Average r calculated from sample-weighted correlation for each effect size. SCID ⫽ Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders ([3rd ed.; DSM–III; American
Psychiatric Association, 1980]; Spitzer & Williams, 1983); CAPS ⫽ Clinician Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); DIS ⫽ Diagnostic Interview Schedule (Robins, Helzer,
Croughan, & Ratcliff, 1981); SI–PTSD ⫽ Structured Interview for Posttraumatic Stress Disorder (Davidson, Kudler, & Smith, 1990); DSM–III–R ⫽ Diagnostic and Statistical Manual of Mental Disorders (3rd
ed., rev.; American Psychiatric Association, 1987); PSS ⫽ Posttraumatic Stress Disorder Symptom Scale (Foa, Riggs, Dancu, & Rothbaum, 1993); NVVRS ⫽ National Vietnam Veterans Readjustment Study
(Kulka et al., 1990); MCS ⫽ Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); PTSD–I ⫽ PTSD Inventory (Solomon et al., 1993); WWII ⫽ World War
II; MMPI ⫽ Minnesota Multiphasic Personality Inventory (Hathaway & McKinley, 1951); MCS–CV ⫽ MCS—Civilian Version (Lauterbach, Vrana, King, & King, 1997); DSM–IV ⫽ Diagnostic and Statistical
Manual of Mental Disorders (4th ed., American Psychiatric Association, 1994); CI ⫽ 95% confidence interval.
a
Positive correlation indicates poorer prior psychological adjustment associated with PTSD symptoms. b Effect size of zero imputed because study indicated only that analysis was “not statistically
59

significant.” c Effect size adjusted for dichotomization of PTSD variable.

 60

Table 3
Weighted Correlations of Family History of Psychopathology With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

Broad Trauma Family psychopathology

Study Sample trauma type Clinical status recency assessment method PTSD measure df ra

Basoglu et al. (1994) Turkish torture survivors Interpersonal Community Variable Family history of SCID 55 .36
violence psychopathology
Emery et al. (1991) Vietnam veterans Combat Clinical 20⫹ years Alcohol-dependent parent Semistructured interview, 40 .34b
MMPI and IES
Jeavons et al. (2000) Motor vehicle accident Accident Medical 3 months Psychiatric or psychological PTSD–I 62 .05
victims treatment in family
McFarlane (1988) Firefighters who responded Accident Community 4 months Family history of Clinical interview using 45 .41b
to Australian bush fire psychopathology DSM–III criteria
Reich et al. (1996) VA patients Combat Medical Variable Family history of GAD or SCID (DSM–III–R) 82 .12b
depression
Skre et al. (1993) Twins (one proband had Not Clinical Not Twin diagnosed with SCID (DSM–III–R) 81 .43b
received treatment for specified specified anxiety disorder
mental disorder)
Speed et al. (1989) WWII prisoners of war Combat Community 40⫹ years Family history of Rating scale based on 62 ⫺.06
psychopathology DSM–III
Ursano, Fullerton, Motor vehicle accident Accident Medical 1 month Family history of SCID (DSM–III–R, 122 .00c
Epstein, Crowley, victims psychopathology DSM–IV)
Kao, et al. (1999)
OZER, BEST, LIPSEY, AND WEISS

Watson et al. (1996) Vietnam veterans Combat Mixed 20⫹ years Any psychiatric diagnosis PTSD–I 118 .14b
in family
Unweighted average r ⫽ .21 (CI ⫽ .08, .33)
Weighted average r ⫽ .17 (CI ⫽ .04, .29)

Note. Average r calculated from sample-weighted correlation for each effect size. SCID ⫽ Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders ([3rd ed.; DSM–III;
American Psychiatric Association, 1980]; Spitzer & Williams, 1983); MMPI ⫽ Minnesota Multiphasic Personality Inventory (Hathaway & McKinley, 1951); IES ⫽ Impact of Event Scale (Horowitz,
Wilner, & Alvarez, 1979); PTSD–I ⫽ PTSD Inventory (Solomon et al., 1993); DSM–III–R ⫽ Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association,
1987); VA ⫽ Veterans Administration; DSM–IV ⫽ Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Association, 1994); CI ⫽ 95% confidence interval.
a
Positive correlation indicates history of family psychopathology associated with PTSD symptoms. b Effect size adjusted for dichotomization of PTSD variable. c Effect size of zero imputed because
study indicated only that analysis was “not statistically significant.”
 PREDICTORS OF PTSD: A META-ANALYSIS 61

PTSD when the traumatic experience involved noncombat inter- traumatic experience was an accident (weighted r ⫽ .20; z ⫽ 2.44,
personal violence (weighted r ⫽ .31) than when the traumatic p ⬍ .05).
experience was combat exposure (weighted r ⫽ .12; z ⫽ 3.40, p ⬍
.01) or accident (weighted r ⫽ .08; z ⫽ 3.00, p ⬍ .01). The Perceived Support Following Trauma
weighted r of .28 for interview assessment was much stronger than
the weighted r of .04 for self-report (z ⫽ 3.42, p ⬍ .01). The weighted average correlation for 11 studies (combined
n ⫽ 3,537) of the relationship between perceived social support
following the trauma and PTSD symptoms was ⫺.28, a statisti-
Perceived Life Threat
cally significant ES in the small-to-medium range (see Table 5).
The weighted average correlation for 12 studies (combined Individuals reporting lower levels of perceived social support after
n ⫽ 3,524) of the relationship between perceived life threat and the traumatic event reported higher levels of PTSD symptoms or
PTSD symptoms or diagnosis was .26, a statistically significant rates of current PTSD. The ESs ranged from ⫺.57 to .07.
effect in the small-to-medium range (see Table 4). Individuals who The strength of the relationship between perceived social sup-
perceived that their life was in danger during their index traumatic port after the traumatic event and PTSD did not vary by the type
event reported higher levels of PTSD symptoms or higher rates of of sample studied or by the method used to assess PTSD symptoms
current PTSD. The ESs ranged from .13 to .49. or diagnosis. The inverse relationship between perceived support
The strength of the relationship between perceived life threat and PTSD (i.e., those reporting more support showed lower rates
during the traumatic event and PTSD diagnosis or symptoms did of meeting diagnostic criteria or lower symptom levels), however,
not vary by the type of sample studied nor by the method used to was strongest among studies with more time elapsed between the
assess PTSD. The strength of the relationship between perceived traumatic event and the assessment of PTSD. That is, an incre-
life threat and PTSD, however, was higher among studies with mental pattern was found in which the inverse relationship was
more time elapsed between the traumatic event and the assessment strongest in studies in which more than 3 years had elapsed
of PTSD (i.e., 6 months to 3 years [weighted r ⫽ .44] vs. 1 to 6 between the traumatic event and assessment of PTSD (weighted
months [weighted r ⫽ .24]; z ⫽ 2.23, p ⬍ .05). Furthermore, r ⫽ ⫺.42), followed by studies in which 6 months to 3 years had
perceived life threat during the traumatic event was more strongly elapsed (weighted r ⫽ ⫺.16), followed by studies in which 1 to 6
associated with PTSD when the traumatic experience was non- months had elapsed (weighted r ⫽ .01; z ⫽ 2.64, p ⬍ .01, 1 to 6
combat interpersonal violence (weighted r ⫽ .36) than when the months vs. 6 months to 3 years; z ⫽ 7.50, p ⬍ .01, 1 to 6 months

Table 4
Weighted Correlations of Perceived Life Threat With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

Study Sample Broad trauma type Clinical status Trauma recency PTSD measure df ra

Bernat et al. (1998) College students Variable Community Variable IES–R 349 .21
Blanchard et al. Motor vehicle accident victims Accident Medical 1–4 months CAPS 98 .23
(1995)
David et al. (1996) Hurricane victims Accident Medical 6–12 months SCID (DSM–III–R) 61 .29b
Dunmore et al. Survivors of sexual and Interpersonal violence Community Variable PSS 91 .49b
(1999) physical assault
Ehlers et al. (1998) Motor vehicle accident victims Accident Medical 3 months PSS 888 .23b
Jeavons et al. (2000) Motor vehicle accident victims Accident Medical 3 months PTSD–I 62 .46
Kemp et al. (1995) Female victims of domestic Interpersonal violence Variable Variable MCS–CV 168 .15
violence and verbal abuse
Kilpatrick et al. Female crime victims Interpersonal violence Community Variable DIS 294 .40b
(1989)
King et al. (1998) Vietnam veterans (NVVRS Combat Community 20⫹ years MCS, DIS 1224 .26
sample)
McFarlane (1988) Firefighters of Australian bush Accident Community 4 months Structured interview 44 .13b
fire based on DSM–III
Michaels et al. Accident victims Accident Medical 6 months MCS–CV 176 .18b
(1999)
Perrin et al. (1996) Domestic violence victims Interpersonal violence Clinical Not specified MMPI–PK 69 .34b
Unweighted average r ⫽ .29 (CI ⫽ .21, .36)
Weighted average r ⫽ .26 (CI ⫽ .18, .34)

Note. Average r calculated from sample-weighted correlation for each effect size. IES–R ⫽ Impact of Event Scale—Revised (Weiss & Marmar, 1997);
CAPS ⫽ Clinician Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); SCID ⫽ Structured Clinical Interview for Diagnostic and
Statistical Manual of Mental Disorders [(3rd ed.; DSM–III; American Psychiatric Association, 1980]; Spitzer & Williams, 1983); DSM–III–R ⫽ Diagnostic
and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); PSS ⫽ Posttraumatic Stress Disorder Symptom Scale
(Foa, Riggs, Dancu, & Rothbaum, 1993); PTSD–I ⫽ PTSD Inventory (Solomon et al., 1993); MCS–CV ⫽ Mississippi Scale for Combat-Related
Posttraumatic Stress Disorder—Civilian Version (Lauterbach, Vrana, King, & King, 1997); DIS ⫽ Diagnostic Interview Schedule (Robins, Helzer,
Croughan, & Ratcliff, 1981); MCS ⫽ Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); NVVRS ⫽
National Vietnam Veterans Readjustment Study (Kulka et al., 1990); MMPI–PK ⫽ Minnesota Multiphasic Personality Inventory PTSD—Keane Scale
(Keane, Malloy, & Fairbank, 1984); CI ⫽ 95% confidence interval.
a
Positive correlation indicates perceived life threat associated with PTSD symptoms. b Effect size adjusted for dichotomization of PTSD variable.
 62

Table 5
Weighted Correlations of Social Support With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

Study Sample Broad trauma type Clinical status Trauma recency Support assessmenta PTSD measure df rb

Astin et al. (1993) Battered women Interpersonal violence Medical Mean of 8 Perceived support PTSD Symptom Checklist 53 ⫺.18
months
Green & Berlin (1987) Vietnam veterans Combat Clinical 30 years If spoke with others about Figley Rating Scale 60 ⫺.26
Vietnam during first
year postwar
Hough et al. (1990) Community where sniper Interpersonal violence Clinical 6 months Availability of a confidant DIS 290 .07
massacre occurred
King et al. (1998) Vietnam veterans Combat Community 20⫹ years Functional support after MCS, DIS 1224 ⫺.43
(NVVRS theater) return from Vietnam
Kramer & Green Women treated for Interpersonal violence Medical Not specified Use of informal support Semistructured interview 30 ⫺.42
(1991) sexual assault network to discuss the based on DSM–III
rape criteria and IES
Perrin et al. (1996) Women who had Interpersonal violence Clinical Not specified Interpersonal Support MMPI–PK 69 ⫺.57
experienced domestic Evaluation List
violence
Perry et al. (1992) Burn victims Accident Medical 2 months Interpersonal Support SCID 50 ⫺.44c
Evaluation List
Solomon et al. (1987) Israeli combat veterans Combat Community 12 months Perceived support during PTSD–I–SR 683 ⫺.18
war
Sutker et al. (1995) Persian Gulf War Combat Community Within 1 year Number of support MCS for Desert Storm 580 ⫺.18
veterans providers
OZER, BEST, LIPSEY, AND WEISS

Weiss et al. (1995) EMS personnel Mixed Community Variable Current Social Support MCS–CV 367 ⫺.25
Scale (NVVRS)
Wolfe et al. (1998) Female veterans of Combat Community 18–24 months Leader support during war MCS 131 ⫺.25
Persian Gulf War
Unweighted average r ⫽ ⫺.29 (CI ⫽ ⫺.41, ⫺.16)
Weighted average r ⫽ ⫺.28 (CI ⫽ ⫺.40, ⫺.15)

Note. Average r calculated from sample-weighted correlation for each effect size. DIS ⫽ Diagnostic Interview Schedule (Robins, Helzer, Croughan, & Ratcliff, 1981); NVVRS ⫽ National Vietnam
Veterans Readjustment Study (Kulka et al., 1990); MCS ⫽ Mississippi Scale for Combat-Related Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); DSM–III ⫽ Diagnostic and Statistical
Manual of Mental Disorders (3rd ed.; American Psychiatric Association, 1980); IES ⫽ Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979); MMPI–PK ⫽ Minnesota Multiphasic Personality
Inventory PTSD—Keane Scale (Keane, Malloy, & Fairbank, 1984); SCID ⫽ Structured Clinical Interview for DSM–III (Spitzer & Williams, 1983); PTSD–I–SR ⫽ PTSD Inventory—Self-Report
(Solomon et al., 1993); EMS ⫽ emergency medical service; MCS–CV ⫽ MCS—Civilian Version (Lauterback, Vrana, King, & King, 1997); CI ⫽ 95% confidence interval.
a
Support measures all assessed self-reported, perceived support. b Negative correlation indicates high support associated with lower PTSD symptoms. c Effect size adjusted for dichotomization of
PTSD variable.
 PREDICTORS OF PTSD: A META-ANALYSIS 63

vs. 3 years or more; z ⫽ 6.58, p ⬍.01, 6 months to 3 years vs. 3 described having dissociative experiences during or immediately
years or more). Furthermore, a stronger inverse relationship be- after the traumatic event reported appreciably higher levels of
tween perceived support and PTSD was found for studies of PTSD symptoms or rates of current disorder. The ESs ranged from
combat trauma (weighted r ⫽ ⫺.26) than for studies of noncombat .14 to .94. The coefficient of .94 was a conversion from a t-test
interpersonal violence (weighted r ⫽ ⫺.11; z ⫽ 2.85, p ⬍ .01). comparison in which the means of a modified dissociation scale
were 11.5 (SD ⫽ 1.6) and 1.8 (SD ⫽ 2.1), respectively (Bremner
Peritraumatic Emotional Responses et al., 1992). To assess the impact of this admittedly extreme ES,
we recalculated the weighted ES omitting the Bremner et al.
The weighted average correlation for five studies (combined (1992) study. The result was a weighted r of .34; thus, we chose to
n ⫽ 1,755) of the relationship between self-reports of emotional retain this study in the analyses because it did not unduly influence
responses at the time of the traumatic event and PTSD symptoms the overall ES.
was .26, statistically significant even given the small number of The strength of the relationship between peritraumatic dissoci-
studies (see Table 6). These studies showed that individuals who ation and PTSD symptoms or diagnosis varied as a function of the
described having intensely negative emotional responses during or
time elapsed between (a) the trauma and measurement of symp-
immediately after the index traumatic event reported appreciably
toms, (b) the type of sample, and (c) the method for assessment of
higher levels of PTSD symptoms or rates of current PTSD. The
symptoms, but it did not vary by the type of event. Peritraumatic
kinds of emotional responses studied in the literature included fear,
dissociation was most strongly related to PTSD among individuals
helplessness, horror, guilt, and shame. The ESs ranged from .15 to
.55. No within-group comparisons by sample, type of trauma, or seeking mental health services (weighted r ⫽ .60) as compared
time elapsed between trauma and PTSD assessment were possible with medical (weighted r ⫽ .33; z ⫽ 3.81, p ⬍ .01) or community
for this correlate because of insufficient quantity of studies in each samples (weighted r ⫽ .35; z ⫽ 3.96, p ⬍ .01). With respect to the
subgroup category. Assessment method was not analyzable be- time frame between the trauma and assessment of PTSD, the
cause no study used an interview exclusively. relationship between peritraumatic dissociation and PTSD symp-
toms or diagnosis was highest among studies in which 6 months
to 3 years had elapsed between the traumatic event and the assess-
Peritraumatic Dissociation ment of PTSD (weighted r ⫽ .45) compared with the time frames
The weighted average correlation for 16 studies (combined of 1 to 6 months (weighted r ⫽ .31; z ⫽ 2.51, p ⬍ .05) and more
n ⫽ 3,534) of the relationship between reports of dissociative than 3 years (weighted r ⫽ .30; z ⫽ 2.61, p ⬍ .01). For studies in
experiences (either self-report or interviewer assessment) during which PTSD was indexed by self-report measures, the weighted r
and immediately after the traumatic event and PTSD symptoms or was .48 as compared with a weighted r of .21 for studies using
diagnosis was .35, a statistically significant ES in the medium interview methods. This difference was statistically significant
range (see Table 7). These studies showed that individuals who (z ⫽ 7.81, p ⬍ .01).

Table 6
Weighted Correlations of Peritraumatic Emotions With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

Broad Trauma Peritraumatic emotion

Study Sample trauma type Clinical status recency assessment method PTSD measure df ra

Bernat et al. College students Mixed Community sample Variable Initial Subjective Reaction IES–R 349 .41
(1998) Emotional Scale
Brewin et al. Crime victims Interpersonal Combination of 6 months Any intense fear, PSS 138 .55b
(2000) violence medical, helplessness, or horror
community
Ehlers et al. Motor vehicle accident Accident Medical patient 3 months Rating of how frightening PSS 888 .28
(1998) victims recruited from the accident was
ERs
Epstein et al. Military health personnel Disaster Community sample 6–18 months Initial anxiety/fright coded Items from 306 .20b
(1998) from qualitative reports SCL–90–R,
DSM–III–R
Roemer et al. College students Mixed Community sample Not specified Single items assessing PTSD Checklist 74 .15
(1998) fear, helplessness, and
horror
Unweighted average r ⫽ .28 (CI ⫽ .10, .42)
Weighted average r ⫽ .26 (CI ⫽ .08, .42)

Note. Average r calculated from sample-weighted correlation for each effect size. IES–R ⫽ Impact of Event Scale—Revised (Weiss & Marmar, 1997);
PSS ⫽ Posttraumatic Stress Disorder Symptom Scale (Foa, Riggs, Dancu, & Rothbaum, 1993); ERs ⫽ emergency rooms; SCL–90 –R ⫽ Symptom
Checklist—90 —Revised (Derogatis, 1992); DSM–III–R ⫽ Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric
Association, 1987); CI ⫽ 95% confidence interval.
a
Positive correlation indicates positive association between peritraumatic emotional response and PTSD symptoms. b Effect size adjusted for dichoto-
mization of PTSD variable.
 64
Table 7
Weighted Correlations of Peritraumatic Dissociation With Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

Broad Trauma Peritraumatic dissociation

Study Sample trauma type Clinical status recency assessment method PTSD measure df ra

Bernat et al. (1998) College students Mixed Community Variable PDEQ–RV IES–R 349 .35
Bremner & Brett (1997) Vietnam combat veterans Combat Medical 20⫹ years Modified DES MCS 50 .52
Bremner et al. (1992) Vietnam combat veterans Combat Psychiatric 20⫹ years Modified DES MCS 60 .94
Ehlers et al. (1998) Motor vehicle accident Accident Medical 3 months Feeling numbed or dazed PTSD Symptom Scale 888 .33
victims recruited from ERs during accident
Koopman et al. (1994) Fire survivors Disaster Community 7–9 months Stanford Acute Stress MCS–CV 154 .59
Reaction Questionnaire
Lee et al. (1995) Longitudinal study of WWII Combat Community 2 years Peritraumatic symptoms Checklist based on DSM–III 107 .22
veterans during combat
Marmar et al. (1994) Male era and theater Vietnam Combat Community 20 years PDEQ–RV MCS 251 .51
veterans from NVVRS
Michaels et al. (1999) ER injury victims Accident Medical 6 months Michigan Critical Events MCS–CV 176 .24c
Perception Scale
O’Toole et al. (1999) Australian Vietnam veterans Combat Community 20⫹ years PDEQ–RV Adapted version of SCID 636 .14b,c
Roemer et al. (1998) College students Mixed Community Variable Single item assessing PTSD Checklist 74 .35
numbing during
trauma
Shalev et al. (1996) ER patients (mainly accidents; Mixed Medical 20⫹ years PDEQ–RV MCS–CV 51 .54
some terrorist or other
physical assault)
Shalev, Sahar, et al. ER patients (mainly accidents; Mixed Medical 4 months PDEQ–RV CAPS 86 .22c
(1998) some terrorist or other
physical assault)
Tichenor et al. (1996) Female Vietnam veterans Combat Community 20⫹ years PDEQ–RV MCS 77 .18
(NVVRS theater sample)
Tucker et al. (1997) Terrorist bombing victims Interpersonal Psychiatric 6 months Item assessing being on Posttraumatic Stress Symptom 86 .32
OZER, BEST, LIPSEY, AND WEISS

violence “automatic pilot” score (IES ⫹ 7 symptom

items)
Ursano, Fullerton, Medical patients from motor Accident Medical 1 month PDEQ–RV SCID (DSM–III–R) 122 .27c
Epstein, Crowley, vehicle accidents
Vance, et al. (1999)
Weiss et al. (1995) Two samples of emergency Disaster Community Variable PDEQ–RV MCS–CV ⫹ IES–R 367 .52
services workers (disaster
exposed and variable)
Unweighted average r ⫽ .43 (CI ⫽ .25, .58)
Weighted average r ⫽ .35 (CI ⫽ .16, .52)

Note. Average r calculated from sample-weighted correlation for each effect size. PDEQ–RV ⫽ Peritraumatic Dissociative Experiences Questionnaire—Rater Version (Marmar, Weiss, & Metzler,
1997); IES–R ⫽ Impact of Event Scale—Revised (Weiss & Marmar, 1997); DES ⫽ Dissociative Experiences Scale (Bernstein & Putnam, 1986); MCS ⫽ Mississippi Scale for Combat-Related
Posttraumatic Stress Disorder (Keane, Caddell, & Taylor, 1988); MCS–CV ⫽ MCS—Civilian Version (Lauterback, Vrana, King, & King, 1997); ERs ⫽ emergency rooms; DSM–III ⫽ Diagnostic
and Statistical Manual of Mental Disorders (3rd ed.; American Psychiatric Association, 1980); NVVRS ⫽ National Vietnam Veterans Readjustment Study (Kulka et al., 1990); CAPS ⫽ Clinician
Administered Posttraumatic Stress Disorder Scale (Blake et al., 1995); IES ⫽ Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979); SCID ⫽ Structured Clinical Interview for DSM–III (Spitzer
& Williams, 1983); DSM–III–R ⫽ Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; American Psychiatric Association, 1987); CI ⫽ 95% confidence interval.
a
Positive correlation indicates positive association between peritraumatic dissociation and PTSD symptoms. b Effect size estimated from df and p value. c Effect size adjusted for dichotomization
of PTSD variable.
 PREDICTORS OF PTSD: A META-ANALYSIS 65

File Drawer Problem of predictor. We conducted four hierarchical linear multiple re-
gression analyses, with each moderator entered last in a rotating
To assess the robustness of our findings and address the poten-
fashion. These results did not support the presence of meaningful
tial problem of selective publication of studies that show statisti-
confounding among the moderators. Overall, when the whole set
cally significant ESs for the relationship between predictors and
of 68 studies was analyzed, none of the moderator variables
PTSD, we followed the recommendation of Rosenthal (1979,
accounted for a significant increment in variance at the first step of
1991) to estimate the number of unknown, unpublished studies
the analysis nor at any subsequent steps (with the effects of the
with null results that would be sufficient to reduce our overall ESs
other moderator variables partialed out). The overall R2 with all
to below the point of statistical significance (i.e., such that confi-
effects in the model was only .135 (shrunken R2 ⫽ ⫺.002), and
dence intervals would include zero). The threat to the robustness of
increments at the last step ranged from .001 for assessment method
the meta-analysis findings can then be assessed, taking into con-
to .090 for time elapsed. From these results, it is not clear that any
sideration the breadth of the field and the likelihood that such a
of the moderators makes a difference by themselves, and certainly
quantity of unpublished studies exists.
not uniquely in the presence of the others for the PTSD predictor
The results of these calculations supported the robustness of the
literature as a whole.
overall weighted ESs we have reported for five of the seven
An argument can be made, however, that an analysis that pools
predictors. The most robust ES was found for prior trauma, for
across all ESs and assumes that the moderators would have the
which 68 unknown studies with a zero ES would be needed to
same effect regardless of what predictor the ES was associated
generate a nonsignificant ES. Prior adjustment problems would
with is not a meaningful examination of the effects of the moder-
require 43 zero-ES studies; perceived life threat would necessi-
ators. Consequently, we also conducted hierarchical multiple lin-
tate 30 zero-ES studies. Less robust but still requiring a relatively
ear regression analyses for prior psychological adjustment and
implausible number of zero-ES studies were peritraumatic disso-
peritraumatic dissociation, the only two predictors for which (a)
ciation, needing 17 zero-ES studies, and perceived social support,
the z comparisons had yielded more than one significant moderator
needing 14 zero-ES studies. In contrast, family history of psycho-
and (b) there was a sufficient number of studies to conduct a
pathology would require only 4 zero-ES studies in addition to the
relatively stable and interpretable regression analysis given the
nine studies we have identified, and peritraumatic emotional re-
degrees of freedom used for the coding of the moderators. Because
sponse would require only 2 zero-ES studies, fewer than half of the
of the very low statistical power associated with these analyses,
five studies we used for this predictor. It is worth noting that
this approach was exploratory and intended to help describe the
interest in family history as a predictor has a much longer history
relative independence of the predictors rather than to test for
than does peritraumatic emotionality, although this latter variable
significance per se. We included all three moderators for both
is currently receiving more attention (e.g., Brunet et al., 2001).
analyses.
The results for the predictor of prior adjustment revealed that the
Interrelationship of Moderators and Unique Effects only statistically significant finding was for method of assessment.
Analyses of the effects of moderators (type of sample, type of It explained a significant proportion of variance in prior adjust-
trauma, method of assessment, and time elapsed since the trauma) ment whether entered first, R2(1, 20) ⫽ .190, p ⫽ .05, or second,
described above indicated that more than one moderator had a R2 increment(1, 17) ⫽ .207, p ⫽ .04, but not last. Neither the set
significant impact on the weighted ES for four of the six predictors of variables representing time elapsed since the trauma nor the set
with a sufficient number of ESs to conduct these analyses. To what representing type of event yielded significant effects at any step.
extent were these moderators related or independent?2 To explore The results for peritraumatic dissociation did not yield any statis-
this question, we analyzed the data using two strategies. tically significant effects, probably partly as a result of the sample
Our first set of analyses systematically crossed the categories of size of only 16 studies. The shrunken R2 values at the first step
each moderator with those of the remaining three and tested the were .244 for type of sample, ⫺.069 for time elapsed, and .044 for
contingency between the two moderators. Our chi-square results method of assessment.
showed that three of the six were significant: type of trauma
crossed with time elapsed, ␹2(9, N ⫽ 68) ⫽ 57.4, p ⬍ .01; type of Discussion
trauma crossed with sample, ␹2(9, N ⫽ 68) ⫽ 37.2, p ⬍ .01; and
time elapsed crossed with sample, ␹2(9, N ⫽ 68) ⫽ 18.1, p ⫽ .05. Replication and Extension of Salient Predictors
All three chi-squares of assessment method crossed with the other
There are a number of seemingly obvious but nonetheless im-
three were not statistically significant. Thus, we concluded that
portant observations that flow from the results of our meta-
any effect of assessment method on prior psychological adjust-
analysis. First, understanding who does and does not respond to
ment, family history of psychopathology, and peritraumatic disso-
traumatic exposure with PTSD symptoms is not purely a random
ciation was not likely to be explained away by looking at any of
phenomenon. This general conclusion is consistent with Brewin et
the other three moderators. For type of trauma, time elapsed, and
al.’s (2000) initial analyses. Systematic and relatively consistent
type of sample, however, any effect of any one of these on prior
findings emerged despite considerations that could be understood
psychological adjustment, social support, and peritraumatic disso-
as working against predictability. Even though nearly half of adults
ciation could not be ruled out as being associated with either of the
other two. Thus, confounding as a general effect appeared
possible. 2
We thank several anonymous reviewers and Scott Monroe for raising
Our next step was to assess the overall impact of these moder- the question of the interrelationship of the moderators and their impact on
ators on the entire sample of ESs in the meta-analysis, regardless our conclusions.
66 OZER, BEST, LIPSEY, AND WEISS

report experiencing at least a single traumatic event at some point variability in response to traumatic stressors is at best less than
in their lives, and despite the (a) relatively low prevalence (7.9%) 20% of the total variance is consistent with the possibility that
of lifetime PTSD, and (b) differing classes of events, time frames, factors unique to the combination of the person exposed and the
measurement methods, and other “noise” (cf. Meehl, 1978) that nature of the exposure are the determining factors in understanding
accompany the nonexperimental psychosocial field research we who becomes symptomatic and who does not. We return to this
subjected to meta-analysis, replicable and relatively stable predic- point below as well.
tive relationships were identified (see Table 8). This state of affairs
is encouraging and implies that there is indeed a phenomenon to be Social Support as a Predictor
understood and explained.
A second observation is that the sizes of the effects for the The main substantive difference between Brewin et al.’s (2000)
predictors studied here are respectable and certainly not trivial. meta-analysis and ours, aside from our inclusion of peritraumatic
Furthermore, in another replication and extension of Brewin et emotionality and peritraumatic dissociation, is that social support
al.’s (2000) findings, there appear to be two classes of predictors was their strongest predictor (weighted r ⫽ .40), stronger both than
based on the weighted ESs and their temporal proximity to the our estimate of its strength (weighted r ⫽ |.28|) and peritraumatic
traumatic event: (a) characteristics of the individual or his/her life dissociation (weighted r ⫽ .35), our strongest predictor. The set of
history that were more distal to the traumatic event and produced studies that constituted each analysis were obviously somewhat
average coefficients smaller than .20 (i.e., prior adjustment, prior different; in particular, Brewin et al. included several studies that
history of trauma, and family history of psychopathology in our we did not because they did not meet our inclusion criteria (e.g.,
results; gender, age, education, socioeconomic status, IQ, and race not having published a single-df ES). Even though this difference
in Brewin et al., 2000); and (b) stronger predictive factors yielding in raw data produced the different results, both analyses suggest
coefficients greater than .20 that were more proximal to the trau- that social support is a predictor needing further exploration.
matic event (i.e., perceived life threat, perceived support, peritrau- Our analyses indicated that the strength of the relationship
matic emotionality, and peritraumatic dissociation from our results between social support and PTSD symptoms and diagnosis dif-
and intercurrent life stress from Brewin et al., 2000). These latter fered according to the length of time that had elapsed since the
ESs are strong enough to warrant a serious investigation of the trauma. Social support served as a stronger predictor in studies
mechanism or mechanisms that gave rise to these results. It is where the event had occurred more than 3 years prior than it was
worth noting that, in most cases, for comparable predictors, the for studies with less time elapsed. This result tends to lend some
weighted ESs we report are similar to those reported by Brewin et credence to the idea that social support may function as a kind of
al. For example, family history of psychopathology produced a secondary prevention that is seen more clearly when symptoms are
coefficient of .13 in Brewin et al.’s data compared with our most clearly symptoms of PTSD rather than common, shorter-term
coefficient of .17, and perceived life threat had coefficients of .23 reactions following traumatic exposure. Alternatively, these find-
and .26, respectively. The single finding where our results diverge ings may also suggest that the effects of social support are cumu-
from those of Brewin et al. is for the predictor of social support: lative over time and thus may be seen most strongly in studies that
Brewin et al. reported a weighted ES of |.40| (their strongest assessed PTSD symptoms after several years had elapsed since the
predictor) compared with our coefficient of |.28|, a statistically index trauma.
significant difference (z ⫽ 5.89, p ⬍ .01). We discuss this dis- The methodological problems associated with the retrospective
crepancy in more detail below. nature of the measurement of social support in most of the current
The results also indicate, nonetheless, that a large proportion of PTSD literature can be easily solved in prospective longitudinal
the variability in response is not explained by any of the predictors studies. More importantly, when we examined the measures used
that have been examined in both meta-analyses, despite the fact to generate the 11 ESs for social support, we found that in most
that the more proximal set of predictors had been suggested in the studies, the emphasis was on emotional support. This suggests that
prior literature as potentially able to explain substantial variability the kinds of phenomena for which individuals are receiving sup-
in PTSD symptoms or diagnosis. The fact that the prediction of the port likely have more to do with the psychological processing of

Table 8
Meta-Analyses of Predictors of Posttraumatic Stress Disorder (PTSD) Diagnosis or Symptoms

CI CI
Predictor k N ru r (unweighted) (weighted)

Prior trauma 23 5,308 .21 .17 .16, .27 .11, .22

Prior adjustment 23 6,797 .21 .17 .13, .26 .10, .23
Family history of psychopathology 9 667 .21 .17 .08, .33 .04, .29
Perceived life threat 12 3,524 .29 .26 .21, .36 .18, .34
Perceived support 11 3,537 ⫺.29 ⫺.28 ⫺.41, ⫺.16 ⫺.40, ⫺.15
Peritraumatic emotions 5 1,755 .28 .26 .10, .42 .08, .42
Peritraumatic dissociation 16 3,534 .43 .35 .25, .58 .16, .52

Note. k ⫽ number of effect sizes; N ⫽ total number of participants in the k samples; ru ⫽ unweighted effect
size estimate; r ⫽ effect size estimate adjusted for sample size (positive values reflect that the correlate is
positively associated with increased PTSD symptoms); CI ⫽ 95% confidence interval.
 PREDICTORS OF PTSD: A META-ANALYSIS 67

the meaning of the event or the management of the psychological predictor-specific regressions (i.e., 23 at most), definitive, multi-
distress and pain experienced during intrusive memories or night- variate analyses of the role of moderators for specific predictors of
mares, rather than with needs such as financial assistance, mobi- PTSD will only be possible in the future with a larger body of
lization of the criminal justice system, or restoration of lost pos- research from which to draw.
sessions. Further investigation into whether this conjecture can be
empirically supported would be of interest, and could also produce Strengths and Weaknesses of the Current Literature
knowledge that could inform intervention to help ameliorate the
impact of exposure to traumatic stressors. There are a number of issues in the PTSD literature that warrant
discussion and that lend some understanding to the variety of
Impact of Moderators on Prediction results within predictor categories. First, it is important to note the
heavy reliance on self-report measures and retrospective designs
The impact of our moderators of type of event, time elapsed that characterizes the PTSD literature on which this and Brewin et
since the event, type of sample, and method of assessment on al.’s (2000) meta-analysis are based. This naturalistic, retrospec-
prediction was not uniform across predictors. In this respect, our tive emphasis is certainly understandable given the unpredictabil-
results also replicate Brewin et al.’s (2000), whose moderator ity of traumatic exposure in many of the civilian populations
results were also mixed and could not be simply summarized. studied here (e.g., the unexpected nature of most natural and
Nonetheless, when approached from the perspective of compari- human disasters), and the obvious ethical prohibitions of inflicting
sons of average weighted ES within predictor, type of event was exposure to highly stressful events on populations in an experi-
the most salient moderator of prediction, having an impact on five mental or quasiexperimental design. Longitudinal research with
of the six predictors for which its effect could be analyzed. Time special, high-risk populations, such as military personnel, emer-
elapsed since the event was a significant moderator for four of the gency service workers, and police officers, however, can provide
six predictors, and method of assessment was a significant mod- opportunities for prospective studies of the relationship of factors
erator of three of the six. studied here with expression of PTSD symptoms in the context of
For two predictors, interviews produced stronger average ESs traumatic exposure. In fact, a small number of studies in the
than self-report surveys, but it is not clear that this finding is an literature reviewed here were able to serendipitously provide such
important lead. The effect of time elapsed since the trauma also prospective data by linking existing data (i.e., routine, premilitary
failed to produce a clear result across predictors. For two predic- service psychological assessments) to later functioning. As well,
tors (life threat and peritraumatic dissociation), the average ES was creative designs in which real-time psychophysiological data are
greater for studies in which 6 months to 3 years had elapsed since collected remotely from such samples during their on-duty service
the trauma. For two other predictors, however, the pattern was may also be an important improvement on the bulk of the extant
different and opposite of each other—prior adjustment was more set of studies, especially with respect to neurobiological measures.
predictive when the shortest period of time had elapsed, and social The studies that constitute this meta-analysis used as dependent
support was most predictive when the longest period of time (3 variables either (a) PTSD diagnosis or (b) severity of PTSD
years or more) had elapsed. symptoms. For those studies that diagnosed PTSD, the criterion of
The manner in which the type of event affected prediction, by significant impairment in the affected individuals was necessarily
contrast, was nearly uniform—average ESs were stronger for prior met; that cannot be said uniformly for those studies that assessed
trauma, prior adjustment, family history of psychopathology, and PTSD symptoms as a continuous measure, although high mean
life threat if the index trauma was noncombat interpersonal vio- symptom levels in many studies certainly suggest evidence of
lence (i.e., interpersonal violence that did not occur in a combat or impairment. Both our and Brewin et al.’s (2000) findings further
military context). The one exception to that pattern was the finding indicate that the use of PTSD diagnosis versus the “continuous”
that lower social support was in fact a stronger predictor of PTSD level of PTSD symptoms as outcomes does not generally have a
when the index trauma was combat than when it was other types significant impact on the ESs in this field. Thus, although it is
of trauma. Although it is possible to construct a number of expla- possible, we believe that heterogeneity in the degree of impairment
nations to account for these results, all would be ad hoc and not among research participants is not a cause for serious concern in
supported by ancillary data. The relative uniformity of the finding this literature.
does suggest that further research into the impact of the type of More specific (and fixable) shortcomings of the research liter-
event could be fruitful. ature on predictors of PTSD symptoms are these: (a) virtually total
A less optimistic perspective on the influence of moderators lack of information on strategies used to address the problem of
might conclude that moderation of prediction need not be included missing data (fewer than five studies explicitly mentioned this
as a major phenomenon in investigating the prediction of PTSD issue and how it was treated) and (b) insufficient information
symptoms or diagnostic status on the basis of our results from (a) regarding single-df ESs and degrees of freedom for findings in
chi-square analyses conducted across all studies indicating that the multiple studies (e.g., reporting only that a finding was not statis-
moderators of type of event, type of sample, and time elapsed were tically significant). Both of these factors, as well as others such as
not independent; (b) overall multiple regression equations con- the routine presentation of summary data for all variables tested
ducted across all studies that failed to demonstrate a significant rather than just those that were statistically significant, are reme-
effect for any moderator; and (c) predictor-specific regression diable editorially on a policy basis. Our review suggests such
analyses that found a statistically significant effect only for method policy changes would help advance knowledge by providing com-
of assessment. Because the appropriate literature currently avail- plete information. The article published by the American Psycho-
able for this meta-analysis yielded a relatively small set of ESs for logical Association’s Task Force on Statistical Inference (Wilkin-
68 OZER, BEST, LIPSEY, AND WEISS

son & Task Force on Statistical Inference, 1999) has many “no history of PTSD.” Though not unique to PTSD, this issue is
important suggestions in this regard. especially salient given the requirement that the symptoms be
Nonetheless, findings from several studies not included in the linked to a traumatic event, an event which can have occurred
meta-analysis because of the lack of published information neces- years or decades previously.
sary to calculate single-df ESs generally support the results of the
meta-analysis and further indicate that the average ESs calculated Implications of the Findings for Models of the
for prior adjustment and family history of psychopathology repre- Development of PTSD
sent relatively conservative estimates. Prior affective and anxiety
disorders were strongly associated with higher risk of PTSD in the There is a common thread between exposure to prior traumatic
NCS (i.e., affective disorders associated with at least a threefold stressors, family history of psychopathology, and the exposed
increased risk of PTSD; Bromet, Sonnega, & Kessler, 1998) and person’s own psychological difficulties as predictors of PTSD
among Breslau et al.’s (1991) community sample. In addition, symptoms. This thread is that psychological difficulties, perhaps
family history of psychopathology was associated with a higher also manifested in or as a consequence of poorer social support,
risk for PTSD in the same two studies. play some role in conferring risk of developing PTSD symptoms
after exposure to a traumatic stressor. It is tempting to make an
The Fiction of Homogeneity and Fungibility analogy to the flu or infectious disease: Those whose immune
systems are compromised are at greater risk of contracting a
Our meta-analysis, Brewin et al.’s (2000), and almost every subsequent illness. Similarly, this cluster of variables may all be
other study in our meta-analysis dataset make the assumption that pointing to a single source of vulnerability for the development of
different events are interchangeable with each other for the pur- PTSD or enduring symptoms of PTSD—a lack of psychological
pose of categorizing and conducting analyses to determine rela- resilience. The notion of resilience has a long history in measure-
tionships. What is often overlooked, however, is that there may be ment (e.g., Block, 1965), and several researchers invoke it (Funk
substantial heterogeneity in each construct that is typically viewed & Houston, 1987; Garmezy & Rutter, 1983; Haggerty, Sherrod,
more homogeneously. For example, our predictor of prior trauma Garmezy, & Rutter, 1996; Kobasa, Maddi, & Kahn, 1982; Kobasa,
makes the assumption that all the prior traumas are equal in their Maddi, Puccetti, & Zola, 1985; Maddi & Kobasa, 1991), though
effects (which they likely are not given that type of event appeared none of those researchers has closely examined resilience after
to be one of the most robust moderators), that a single exposure is exposure to truly traumatic events. It is possible that our results,
no different from multiple exposures, and that the variety of other taken in conjunction with these other large sets of findings, are yet
characteristics by which they differ will compensate for each other another view of the same core vulnerability, but seen from differ-
or in some other way not have a systematic impact. The hetero- ent perspectives and manifested in different ways. It would be
geneity of the index trauma for many of the studies in our dataset reasonable to suspect, however, that in circumstances where the
is well described, but then conceptually overlooked with respect to traumatic event is particularly horrific (i.e., on the extreme end
the presence of PTSD symptoms or disorder. We know that events even for traumatic events), resilience may play less of a role. At
differ with respect to potential lethality, time elapsed since the this time, there is little data with which to evaluate that
event, repetition of the event, developmental phase of the individ- proposition.
ual being subjected to the event, as well as the nature of the event How this resiliency, or lack of it, is related to both the psycho-
(e.g., natural vs. human, accidental vs. purposeful, etc.). This logical issues that working through exposure to a traumatic stres-
heterogeneity may be as responsible as any other factor for the sor require and to peritraumatic dissociation is of particular inter-
limits on predictability. est. Our finding that peritraumatic dissociation and, to a lesser
What makes this issue so vexing is that the main homogeneity degree, peritraumatic emotionality are as salient predictors as any
in the field is in the consequences of exposure to the heterogeneity yet identified raises the possibility that Brewin et al.’s (2000)
of traumatic events—the very signs and symptoms of PTSD. From conclusion that a general vulnerability model for PTSD was not
this perspective, more attention needs to be paid to the sequelae of warranted might require some amendment. With respect to the
exposure than to either the preexisting conditions prior to the former issue, the psychological issues involved in treating PTSD
exposure or the aspects of the exposure itself. One way in which (see, e.g., Goodman & Weiss, 1998; Resick, 2001) revolve, in
such attention might manifest itself is in the recognition that large part, around issues of loss. The impact of prior loss on later
because the course of PTSD is typically chronic (Bromet et al., risk for depression, for example, has been well documented (Ten-
1998) and is a waxing and waning disorder more like diabetes than nant, 1988; Tennant, Smith, Bebbington, & Hurry, 1981). It is thus
like myocardial infarction, some of the heterogeneity in terms of not surprising that prior trauma was found to be related to symp-
symptom status versus diagnostic status is that individuals who toms or diagnosis of PTSD as the overlap in symptoms between
have met criteria for PTSD prior to being studied are in partial depression and PTSD is well known. Roughly the same argument
remission when they are interviewed. This would systematically can be made for psychological adjustment and family history of
bias the results of studies using diagnosis in terms of underesti- psychopathology. A kind of convergence appears to characterize
mating the impact of exposure, whereas studies using continuous the finding that depression was the aspect of prior adjustment that
symptom measures would be less subject to these issues. Similarly, was most strongly related to PTSD and the observation that one of
there are individuals (see Weiss et al., 1992) who can best be the key psychological aspects of dealing with the sequelae of
described as having lifetime partial PTSD for whom symptom exposure to traumatic stress and PTSD is coming to terms with
measures would make them virtually indistinguishable from those loss. The degree to which depression is frequently comorbid with
who meet current criteria, but whose diagnostic status would be PTSD may be related to one, several, or all of these factors. An
 PREDICTORS OF PTSD: A META-ANALYSIS 69

important question for future work is the relationship between the nomena. Such a position would be similar to contending that the
presence of peritraumatic dissociation and the salience of loss, as absence of psychophysiological arousal in response to trauma cues
opposed to fear or horror in dealing with exposure to traumatic (one method of assessing a part of the criteria for the diagnosis;
stress. Orr & Roth, 2000) in a subset of those who meet criteria (Keane
It is worth noting that resiliency may also be manifest in the et al., 1998) means that hyperarousal criteria (Criterion D) should
individual differences in the functioning of the HPA axis, specif- be eliminated from the diagnostic criteria for PTSD. Until such
ically the manner in which high levels of physiological and emo- time as the definitory (Meehl, 1995) criteria for PTSD are identi-
tional arousal are managed in the brain structures that have been fied and understood, the approach of the DSM to use evidentiary
revealed to become increasingly important in animal models (e.g., criteria as definitory seems to be the only viable option.
LeDoux, 2000) and imaging findings (e.g., Rauch et al., 2000; Although it is relatively easy to quantify peritraumatic dissocia-
Schuff et al., 2001)—the amygdala and hippocampus have now tive experiences retrospectively, whether months or years after the
been implicated in PTSD. It remains to be clarified whether these event (Weiss et al., 1995) or hours or days after the event (Shalev
effects in the brain are consequences of exposure to traumatic et al., 1996), it is much more difficult to quantify peritraumatic
stress, consequences of intrusive and hyperarousal symptoms of distress as it actually is occurring in the context of a traumatic
PTSD, or in some way reflect preexisting vulnerabilities that event. Leaving aside the logistic issues of being present to observe
increase the likelihood of the development of PTSD when an an individual’s in-the-moment response to an ongoing traumatic
individual experiences severe stress. event (though accompanying emergency personnel on their runs
One argument about the predictive power of the static historical could be one approach), the case could be made that the subjective
variables that we and Brewin et al. (2000) illuminated is that they experience of peritraumatic dissociation may not be accessible in
are merely a consequence of the shared variability between general the moment and may only be understood or recognized after the
psychiatric symptoms and PTSD-specific symptoms. From the immediate danger or threat is passed. Although brain imaging is
analyses we conducted, we cannot make an empirical statement one possible method, it would require the experimental induction
about what would have occurred if each bivariate relationship of high arousal, something that may not be possible. There are,
between predictor and PTSD symptoms had been residualized however, a few studies using ketamine (e.g., Chambers et al.,
against general psychiatric symptoms. In most of the studies, such 1999) in volunteers that suggest this may be a possible avenue.
data were not gathered. The findings regarding peritraumatic dis- Ketamine is a synthesized anesthetic used primarily in veterinary
sociation, however, do shed some light on this thesis. A number of medicine and belongs to the class of dissociative anesthetics like
researchers (e.g., Shalev, Peri, Canetti, & Schreiber, 1996; Weiss, PCP and nitrous oxide. Its action is to block nerve paths without
Marmar, Metzler, & Ronfeldt, 1995) did examine the effect of the depressing respiratory or circulatory functioning. At lower doses in
predictors after other direct or indirect measures of psychological humans it often produces dissociative experiences, and its illicit
adjustment were included in the multivariate model. The results use has documented “trips” that may include out of body experi-
showed that after accounting for general adjustment and general ences when used at higher doses.
dissociative experiences, peritraumatic dissociative experiences Because one view of peritraumatic dissociation suggests that it
remained a significant predictor of PTSD symptoms. Future stud- occurs when the traumatic event is so severe as to feel intolerable
ies would benefit from an examination of the shared variance (Spiegel & Cardena, 1991), it would be important to take into
between general symptoms typically present in PTSD (e.g., de- account the severity of the traumatic event when evaluating the
pression, anxiety), specific symptoms of PTSD (e.g., intrusive impact of peritraumatic dissociation on PTSD symptoms. Concep-
images, emotional numbing, hypervigilance), and the predictors tually, there is no true objective assessment of severity that is
included in this meta-analysis (or other sets of predictors). Such totally divorced from response, because a rough assessment of the
analyses could go a long way toward clarifying more specific modal response to any particular event is typically understood to
aspects of PTSD, including the conundrum of the comorbidity be a rough index of its severity (see Footnote 1). Nonetheless,
between PTSD and depression. given a stance of “all other things being equal” with regard to the
Given these results for peritraumatic dissociation, it is difficult type of event, it appears that the peritraumatic response to the
not to entertain the notion that the psychological aspects of expo- event is the most robust factor contributing to the prediction of
sure may be the most important, and that the in-the-moment PTSD symptoms.
appraisal and meaning of the traumatic stressor may have as much That peritraumatic dissociation was the strongest predictor of
to do with explaining who develops PTSD as do the more static those we examined is probably the most important finding from a
factors such as adjustment, prior exposure, or concurrent psycho- conceptual perspective as well. If we assume that the results of the
pathology. As well, the salience of peritraumatic dissociation may meta-analysis are not confounded by some unknown artifact in the
also lend some empirical support to the conjecture that ASD may same way that we presume that each individual study is not
indeed be an important precursor to PTSD in some people, or after confounded by some unknown artifact, the strong implication is
some very horrifying event, or both. Nonetheless, the strength of that if one could bet on only one variable, the subjective psycho-
the relationship of r ⫽ .35 clearly does not support thinking about logical response to traumatic exposure is the variable on which to
dissociative phenomena as an integral part of the phenomenology place bets. It remains an open question as to how much of reported
of posttraumatic response. Many individuals who have high levels peritraumatic dissociation is related to the level of psychophysio-
of symptoms of PTSD and meet current diagnostic criteria do not logical arousal the individual endured during the event, and how
experience dissociative phenomena; thus, it would be a stretch to contemporaneous versus retrospective measurement differentially
argue that peritraumatic dissociation is by definition a part of impacts the reported degree of peritraumatic dissociation. These
PTSD because some individuals do experience dissociative phe- are important areas for further inquiry.
70 OZER, BEST, LIPSEY, AND WEISS

Implications and Future Directions torture in Turkey. Journal of the American Medical Association, 272,
357–363.
Although this meta-analysis and the results of Brewin et al. Becker, J. V. (1982). The effects of sexual assault on rape and attempted
(2000) provide evidence for statistically significant predictive re- rape victims. Victimology, 7, 106 –113.
lationships between multiple characteristics of individuals and *Bernat, J. A., Ronfeldt, H. M., Calhoun, K. S., & Arias, I. (1998).
their experiences and PTSD symptoms, the specific processes by Prevalence of traumatic events and peritraumatic predictors of posttrau-
which these factors may serve to influence the development of matic stress symptoms in a nonclinical sample of college students.
PTSD remain largely unexamined. Further specification of the Journal of Traumatic Stress, 11, 645– 665.
intervening as well as directly explanatory variables would point to Bernstein, E. M., & Putnam, F. W. (1986). Development, reliability, and
areas of opportunity for intervention and possible attenuation or validity of a dissociation scale. Journal of Nervous and Mental Diseases,
174, 727–735.
prevention of the development of PTSD. Theory-based develop-
Blake, D. D., Weathers, F. W., Nagy, L. M., Kaloupek, D. G., Gusman,
ment and evaluation of such interventions with good operational-
F. D., Charney, D. S., & Keane, T. M. (1995). The development of a
ization of critical variables could then provide data with which to
clinician-administered PTSD scale. Journal of Traumatic Stress, 8,
evaluate existing theoretical approaches, an important area of 75–90.
inquiry given the ethical limitations of experimental research in *Blanchard, E. B., Hickling, E. J., Taylor, A. E., & Loos, W. (1995).
this area. Psychiatric morbidity associated with motor vehicle accidents. Journal
The thrust of future research needs to be focused on the more of Nervous and Mental Disease, 183, 495–504.
proximal mechanisms or processes—in psychological and/or phys- Block, J. (1965). The challenge of response sets: Unconfounding meaning,
iological terms—that explain the statistical relationship between acquiescence, and social desirability in the MMPI. New York: Appleton
PTSD symptoms and the more distal, static predictors such as Century Crofts.
family history of psychopathology and prior trauma. Our exami- *Boscarino, J. A. (1995). Post-traumatic stress and associated disorders
nation of the PTSD literature sought to illuminate noncontingent, among Vietnam veterans: The significance of combat exposure and
simple, linear relationships. As described above, with respect to social support. Journal of Traumatic Stress, 8, 317–336.
peritraumatic dissociation, emotional and psychophysiological Bower, G. H. (1981). Mood and memory. American Psychologist, 36,
arousal, and the underlying brain activities in the HPA axis and the 129 –148.
*Bremner, J. D., & Brett, E. (1997). Trauma-related dissociative states and
limbic system, it may well be that some variables operate in a
long-term psychopathology in posttraumatic stress disorder. Journal of
contingent fashion— heightened arousal is a necessary but not
Traumatic Stress, 10, 37– 49.
sufficient condition for the experience of peritraumatic dissocia-
*Bremner, J. D., Southwick, S., Brett, E., Fontana, A., Rosenheck, R., &
tion, so that not all those who experience heightened arousal go on Charney, D. S. (1992). Dissociation and posttraumatic stress disorder in
to experience peritraumatic dissociation. The mechanism by which Vietnam combat veterans. American Journal of Psychiatry, 149, 328 –
peritraumatic dissociation occurs may be influenced not only by 332.
arousal, but also by temperament, prior experience, and other *Bremner, J. D., Southwick, S. M., Johnson, D. R., Yehuda, R., &
genetic or environmental factors, including factors at the level of Charney, D. S. (1993). Childhood physical abuse and combat-related
plasticity and learning in individual neurons (Malenka, 2000). Our posttraumatic stress disorder in Vietnam veterans. American Journal of
examination of the literature has been able to offer only tentative Psychiatry, 150, 235–239.
glimpses into explanatory mechanisms. Breslau, N., Davis, G. C., Andreski, P., & Peterson, E. (1991). Traumatic
events and posttraumatic stress disorder in an urban population of young
adults. Archives of General Psychiatry, 48, 216 –222.
References *Breslau, N., Kessler, R. C., Chilcoat, H. D., Schultz, L. R., Davis, G. C.,
References marked with an asterisk indicate studies included in the & Andreski, P. (1998). Trauma and posttraumatic stress disorder in the
meta-analysis. community: The Detroit area survey of trauma. Archives of General
Psychiatry, 55, 626 – 632.
American Psychiatric Association. (1968). Diagnostic and statistical man-
Brewin, C. R., Andrews, B., & Valentine, J. D. (2000). Meta-analysis of
ual of mental disorders (2nd ed.). Washington, DC: Author.
risk factors for posttraumatic stress disorder in trauma-exposed adults.
American Psychiatric Association. (1980). Diagnostic and statistical man-
Journal of Consulting and Clinical Psychology, 68, 748 –766.
ual of mental disorders (3rd ed.). Washington, DC: Author.
Bromet, E., Sonnega, A., & Kessler, R. C. (1998). Risk factors for DSM–
American Psychiatric Association. (1987). Diagnostic and statistical man-
III–R post-traumatic stress disorder: Findings from the National Comor-
ual of mental disorders (3rd ed., rev.). Washington, DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical man- bidity Survey. American Journal of Epidemiology, 147, 353–361.
ual of mental disorders (4th ed.). Washington, DC: Author. Brunet, A., Weiss, D. S., Metzler, T. J., Best, S. R., Neylan, T. C., Rogers,
*Andrews, B., Brewin, C. R., Rose, S., & Kirk, M. (2000). Predicting C., et al. (2001). The Peritraumatic Distress Inventory: A proposed
PTSD symptoms in victims of violent crime: The role of shame, anger, measure of PTSD Criterion A2. American Journal of Psychiatry, 158,
and childhood abuse. Journal of Abnormal Psychology, 109, 69 –73. 1480 –1485.
*Astin, M. C., Lawrence, K. J., & Foy, D. W. (1993). Posttraumatic stress Burgess, A. W., & Holmstrom, L. L. (1974). Rape trauma syndrome.
disorder among battered women: Risk and resiliency factors. Violence American Journal of Psychiatry, 131, 981–986.
and Victims, 8, 17–28. *Carlier, I. V. E., Lamberts, R. D., & Gersons, B. P. R. (1997). Risk factors
*Astin, M. C., Ogland-Hand, S. M., Coleman, E. M., & Foy, D. S. (1995). for posttraumatic stress symptomatology in police officers: A prospec-
Posttraumatic stress disorder and childhood abuse in battered women: tive analysis. Journal of Nervous and Mental Disease, 185, 498 –506.
Comparisons with maritally distressed women. Journal of Consulting Carlson, E. B., & Rosser-Hogan, R. (1993). Mental health status of
and Clinical Psychology, 63, 308 –312. Cambodian refugees ten years after leaving their homes. American
*Basoglu, M., Paker, M., Ozmen, E., Tasdemir, O., & Sahin, D. (1994). Journal of Orthopsychiatry, 63, 223-231.
Factors related to long-term traumatic stress responses in survivors of Carlson, E. B., & Rosser-Hogan, R. (1994). Cross-cultural response to
 PREDICTORS OF PTSD: A META-ANALYSIS 71

trauma: A study of traumatic experiences and posttraumatic symptoms of premilitary, military, and combat exposure influences. Journal of
in Cambodian refugees. Journal of Traumatic Stress, 7, 43–58. Consulting and Clinical Psychology, 52, 79 – 87.
Cervantes, R. C., Salgado de Snyder, V. N., & Padilla, A. M. (1989). Funk, S. C., & Houston, B. K. (1987). A critical analysis of the Hardiness
Posttraumatic stress in immigrants from Central America and Mexico. Scale’s validity and utility. Journal of Personality and Social Psychol-
Hospital & Community Psychiatry, 40, 615– 619. ogy, 53, 572–578.
Chambers, R. A., Bremner, J. D., Moghaddam, B., Southwick, S. M., Garmezy, N., & Rutter, M. (Eds.). (1983). Stress, coping, and development
Charney, D. S., & Krystal, J. H. (1999). Glutamate and post-traumatic in children. Baltimore: Johns Hopkins University Press.
stress disorder: Toward a psychobiology of dissociation. Seminars in Green, B. L. (1994). Psychosocial research in traumatic stress: An update.
Clinical Neuropsychiatry, 4, 274 –281. Journal of Traumatic Stress, 7, 341–362.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences *Green, M. A., & Berlin, M. A. (1987). Five psychosocial variables related
(2nd ed.). Hillsdale, NJ: Erlbaum. to the existence of post-traumatic stress disorder symptoms. Journal of
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation Clinical Psychology, 43, 643– 649.
analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Goodman, M., & Weiss, D. S. (1998). Double trauma: A group therapy
Cooper, H., & Hedges, L. V. (Eds.). (1994). The handbook of research approach for Vietnam veterans suffering from war and childhood
synthesis. New York: Russell Sage Foundation. trauma. International Journal of Group Psychotherapy, 48, 39 –54.
*Cottler, L. B., Compton, W. M., III, Mager, D., Spitznagel, E. L., & Janca, Haggerty, R. J., Sherrod, L. R., Garmezy, N., & Rutter, M. (Eds.). (1996).
A. (1992). Posttraumatic stress disorder among substance users from the Stress, risk, and resilience in children and adolescents: Processes,
general population. American Journal of Psychiatry, 149, 664 – 670. mechanisms, and interventions. New York: Cambridge University Press.
*David, D., Mellman, T., Mendoza, L., Kulick-Bell, R., Ironson, G., & Haley, S. A. (1974). When the patient reports atrocities: Specific treatment
Schneiderman, N. (1996). Psychiatric morbidity following Hurricane considerations of the Vietnam veteran. Archives of General Psychia-
Andrew. Journal of Traumatic Stress, 9, 607– 612. try, 30, 191–196.
Davidson, J. R. T., Kudler, H. S., & Smith, R. (1990). Assessment and Hathaway, S. R., & McKinley, J. C. (1951). Manual for the Minnesota
pharmacotherapy of posttraumatic stress disorder. In J. E. L. Giller (Ed.), Multiphasic Personality Inventory (Rev. ed.). New York: Psychological
Biological assessment and treatment of posttraumatic stress disorder Corporation.
(pp. 205–221). Washington, DC: American Psychiatric Press. Horowitz, M. J., & Solomon, G. F. (1975). A prediction of delayed stress
Davis, M. (1992). The role of the amygdala in fear and anxiety. Annual response syndromes in Vietnam veterans. Journal of Social Issues, 31,
Review of Neuroscience, 15, 353–375. 67– 80.
Derogatis, L. R. (1992). SCL–90R administration, scoring & procedures Horowitz, M. J., Wilner, N., & Alvarez, W. (1979). Impact of Event Scale:
manual—II. Towson, MD: Clinical Psychometric Research. A measure of subjective stress. Psychosomatic Medicine, 41, 209 –218.
*Dunmore, E., Clark, D. M., & Ehlers, A. (1999). Cognitive factors *Hough, R. L., Vega, W., Valle, R., Kolody, B., Griswald del Castillo, R.,
involved in the onset and maintenance of posttraumatic stress disorder & Tarke, H. (1990). Mental health consequences of the San Ysidro
(PTSD) after physical or sexual assault. Behaviour Research and Ther- McDonald’s massacre: A community study. Journal of Traumatic
apy, 37, 809 – 829. Stress, 3, 71–92.
*Ehde, D. M., Patterson, D. R., Wiechman, S. A., & Wilson, L. G. (2000). Hunter, J. E., & Schmidt, F. L. (1990a). Dichotomization of continuous
Post-traumatic stress symptoms and distress 1 year after burn injury. variables: The implications for meta-analysis. Journal of Applied Psy-
Journal of Burn Care and Rehabilitation, 21, 105–111. chology, 75, 334 –349.
*Ehlers, A., Mayou, R. A., & Bryant, B. (1998). Psychological predictors Hunter, J. E., & Schmidt, F. L. (1990b). Methods of meta-analysis: Cor-
of chronic posttraumatic stress disorder after motor vehicle accidents. recting error and bias in research findings. Newbury Park, CA: Sage.
Journal of Abnormal Psychology, 107, 508 –519. Hunter, J. E., & Schmidt, F. L. (1994). Estimation of sampling error
*Emery, V. O., Emery, P. E., Shama, D. K., Quiana, N. A., & Jassani, variance in the meta-analysis of correlations: Use of average correlation
A. K. (1991). Predisposing variables in PTSD patients. Journal of in the homogeneous case. Journal of Applied Psychology, 79, 171–177.
Traumatic Stress, 4, 325–343. *Jeavons, S., Greenwood, K. M., & Horne, D. J. (2000). Accident cogni-
*Engel, C. C., Jr., Engel, A. L., Campbell, S. J., McFall, M. E., Russo, J., tions and subsequent psychological trauma. Journal of Traumatic
& Katon, W. (1993). Posttraumatic stress disorder symptoms and pre- Stress, 13, 359 –365.
combat sexual and physical abuse in Desert Storm veterans. Journal of Keane, T. M., Caddell, J. M., & Taylor, K. L. (1988). Mississippi Scale for
Nervous and Mental Disease, 181, 683– 688. Combat-Related Posttraumatic Stress Disorder: Three studies in reliabil-
*Epstein, R. S., Fullerton, C. S., & Ursano, R. J. (1998). Posttraumatic ity and validity. Journal of Consulting and Clinical Psychology, 56,
stress disorder following an air disaster: A prospective study. American 85–90.
Journal of Psychiatry, 155, 934 –938. Keane, T. M., Kolb, L. C., Kaloupek, D. G., Orr, S. P., Blanchard, E. B.,
*Fauerbach, J. A., Lawrence, J., Haythornthwaite, J., Richter, D., McGuire, Thomas, R. G., et al. (1998). Utility of psychophysiological measure-
J., Schmidt, C., & Munster, A. (1997). Preburn psychiatric history ment in the diagnosis of posttraumatic stress disorder: Results from a
affects posttrauma morbidity. Psychosomatics, 38, 374 –385. Department of Veterans Affairs Cooperative Study. Journal of Consult-
Foa, E. B., Riggs, D. S., Dancu, C. V., & Rothbaum, B. O. (1993). ing and Clinical Psychology, 66, 914 –923.
Reliability and validity of a brief instrument for assessing post-traumatic Keane, T. M., Malloy, P. S., & Fairbank, J. A. (1984). The empirical
stress disorder. Journal of Traumatic Stress, 6, 459 – 473. development of an MMPI subscale for the assessment of combat-related
Fontana, A., & Rosenheck, R. (1994). Posttraumatic stress disorder among posttraumatic stress disorder. Journal of Consulting and Clinical Psy-
Vietnam theater veterans: A causal model of etiology in a community chology, 52, 888 – 891.
sample. Journal of Nervous and Mental Disease, 182, 677– 684. Keiser, L. (1968). The traumatic neurosis. Philadelphia: Lippincott.
*Fontana, A., Schwartz, L. S., & Rosenheck, R. (1997). Posttraumatic *Kemp, A., Green, B. L., Hovanitz, C., & Rawlings, E. I. (1995). Incidence
stress disorder among female Vietnam veterans: A causal model of and correlates of posttraumatic stress disorder among battered women:
etiology. American Journal of Public Health, 87, 169 –175. Shelter and community samples. Journal of Interpersonal Violence, 10,
*Foy, D. W., Sipprelle, R. C., Rueger, D. B., & Carroll, E. M. (1984). 43–55.
Etiology of posttraumatic stress disorder in Vietnam veterans: Analysis Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C. B.
72 OZER, BEST, LIPSEY, AND WEISS

(1995). Posttraumatic stress disorder in the National Comorbidity Sur- McGaugh, J. L., & Cahill, L. (1997). Interaction of neuromodulatory
vey. Archives of General Psychiatry, 52, 1048 –1060. systems in modulating memory storage. Behavioural Brain Re-
*Kilpatrick, D. G., Saunders, B. E., Amick-McMullan, A., Best, C. L., search, 83, 31–38.
Veronen, L. J., & Resnick, H. S. (1989). Victim and crime factors Meehl, P. E. (1978). Theoretical risks and tabular asterisks: Sir Karl, Sir
associated with the development of crime-related post-traumatic stress Ronald, and the slow progress of soft psychology. Journal of Consulting
disorder. Behavior Therapy, 20, 199 –214. and Clinical Psychology, 46, 806 – 834.
*King, L. A., King, D. W., Fairbank, J. A., Keane, T. M., & Adams, G. A. Meehl, P. E. (1995). Bootstraps taxometrics: Solving the classification
(1998). Resilience–recovery factors in post-traumatic stress disorder problem in psychopathology. American Psychologist, 50, 266 –275.
among female and male Vietnam veterans: Hardiness, postwar social *Michaels, A. J., Michaels, C. E., Moon, C. H., Smith, J. S., Zimmerman,
support, and additional stressful life events. Journal of Personality and M. A., Taheri, P. A., & Peterson, C. (1999). Posttraumatic stress disorder
Social Psychology, 74, 420 – 434. after injury: Impact on general health outcome and early risk assessment.
Kobasa, S. C., Maddi, S. R., & Kahn, S. (1982). Hardiness and health: A Journal of Trauma, 47, 460 – 467.
prospective study. Journal of Personality and Social Psychology, 42, *North, C. S., & Smith, E. M. (1992). Posttraumatic stress disorder among
168 –177. homeless men and women. Hospital and Community Psychiatry, 43,
Kobasa, S. C. O., Maddi, S. R., Puccetti, M. C., & Zola, M. A. (1985). 1010 –1016.
Effectiveness of hardiness, exercise and social support as resources *North, C. S., Smith, E. M., & Spitznagel, E. L. (1994). Posttraumatic
against illness. Journal of Psychosomatic Research, 29, 525–533. stress disorder in survivors of a mass shooting. American Journal of
Kolb, L. C. (1993). The psychobiology of PTSD: Perspectives and reflec- Psychiatry, 151, 82– 88.
tions on the past, present, and future. Journal of Traumatic Stress, 6, Orr, S. P., & Roth, W. T. (2000). Psychophysiological assessment: Clinical
293–304. applications for PTSD. Journal of Affective Disorders, 61, 225–240.
*Koopman, C., Classen, C., & Spiegel, D. (1994). Predictors of posttrau- *O’Toole, B. I., Marshall, R. P., Schureck, R. J., & Dobson, M. (1999).
matic stress symptoms among survivors of the Oakland/Berkeley, Cal- Combat, dissociation, and posttraumatic stress disorder in Australian
ifornia, firestorm. American Journal of Psychiatry, 151, 888 – 894. Vietnam veterans. Journal of Traumatic Stress, 12, 625– 640.
*Kramer, T. L., & Green, B. L. (1991). Posttraumatic stress disorder as an *Perrin, S., Van Hasselt, V. B., Basilio, I., & Hersen, M. (1996). Assessing
early response to sexual assault. Journal of Interpersonal Violence, 6, the effects of violence on women in battering relationships with the
160 –173.
Keanne MMPI–PTSD Scale. Journal of Traumatic Stress, 9, 805– 816.
Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan,
*Perry, S., Difede, J., Musngi, G., Frances, A. J., & Jacobsberg, L. (1992).
B. K., Marmar, C. R., & Weiss, D. S. (1990). Trauma and the Vietnam
Predictors of posttraumatic stress disorder after burn injury. American
War generation: Report of the findings from the National Vietnam
Journal of Psychiatry, 149, 931–935.
Veterans Readjustment Study. New York: Brunner/Mazel.
Rauch, S. L., Whalen, P. J., Shin, L. M., McInerney, S. C., Macklin, M. L.,
Lauterback, D., Vrana, S., King, D. W., & King, L. A. (1997). Psycho-
Lasko, N. B., et al. (2000). Exaggerated amygdala response to masked
metric properties of the civilian version of the Mississippi PTSD scale.
facial stimuli in posttraumatic stress disorder: A functional MRI study.
Journal of Traumatic Stress, 10, 499 –513.
Biological Psychiatry, 47, 769 –776.
LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of
*Reich, J., Lyons, M., & Cai, B. (1996). Familial vulnerability factors to
Neuroscience, 23, 155–184.
post-traumatic stress disorder in male military veterans. Acta Psychiat-
*Lee, K. A., Vaillant, G. E., Torrey, W. C., & Elder, G. H. (1995). A
rica Scandinavica, 93, 105–112.
50-year prospective study of the psychological sequelae of World War II
Resick, P. A. (2001). Cognitive therapy for posttraumatic stress disorder.
combat. American Journal of Psychiatry, 152, 516 –522.
Journal of Cognitive Psychotherapy, 15, 321–329.
Lipsey, M. W., & Wilson, D. B. (2000). Practical meta-analysis. Thousand
Oaks, CA: Sage. *Resnick, H. S., Kilpatrick, D. G., Best, C. L., & Kramer, T. L. (1992).
*Madakasira, S., & O’Brien, K. F. (1987). Acute posttraumatic stress Vulnerability-stress factors in development of posttraumatic stress dis-
disorder in victims of a natural disaster. Journal of Nervous and Mental order. Journal of Nervous and Mental Disease, 180, 424 – 430.
Disease, 175, 286 –290. Resnick, H. S., Kilpatrick, D. G., Dansky, B. S., Saunders, B. E., & Best,
Maddi, S. R., & Kobasa, S. C. (1991). The development of hardiness. In A. C. L. (1993). Prevalence of civilian trauma and posttraumatic stress
Monat & R. S. Lazarus (Eds.), Stress and coping: An anthology (3rd ed., disorder in a representative national sample of women. Journal of
pp. 245–257). New York: Columbia University Press. Consulting and Clinical Psychology, 61, 984 –991.
Malenka, R. (2000). Basic science in psychiatry. Neurobiology of Dis- *Resnick, H. S., Yehuda, R., Pitman, R. K., & Foy, D. W. (1995). Effect
ease, 7, 537–539. of previous trauma on acute plasma cortisol level following rape. Amer-
Marmar, C. R., Weiss, D. S., & Metzler, T. J. (1997). The Peritraumatic ican Journal of Psychiatry, 152, 1675–1677.
Dissociative Experiences Questionnaire. In J. P. Wilson & T. M. Keane *Riggs, D. S., Dancu, C. V., Gershuny, B. S., & Greenberg, D. (1992).
(Eds.), Assessing psychological trauma and PTSD: A handbook for Anger and post-traumatic stress disorder in female crime victims. Jour-
practitioners (pp. 412– 428). New York: Guilford Press. nal of Traumatic Stress, 5, 613– 625.
*Marmar, C. R., Weiss, D. S., Schlenger, W. E., Fairbank, J. A., Jordan, *Roberts, G., Lawrence, J. M., Williams, G. M., & Raphael, B. (1998). The
B. K., Kulka, R. A., & Hough, R. L. (1994). Peritraumatic dissociation impact of domestic violence on women’s mental health. Australian and
and posttraumatic stress in male Vietnam theater veterans. American New Zealand Journal of Public Health, 22, 796 – 801.
Journal of Psychiatry, 151, 902–907. Robins, L. N., Helzer, J. E., Croughan, J. L., & Ratcliff, K. S. (1981).
McEwen, B. S. (1995). Adrenal steroid actions on brain: Dissecting the National Institute of Mental Health Diagnostic Interview Schedule: Its
fine line between protection and damage. In M. J. Friedman, D. S. history, characteristics, and validity. Archives of General Psychiatry, 38,
Charney, & A. Y. Deutch (Eds.), Neurobiological and clinical conse- 381–389.
quences of stress: from normal adaptation to post-traumatic stress *Roemer, L., Orsillo, S. M., Borkovec, T. D., & Litz, B. T. (1998).
disorder (pp. 135–147). Philadelphia: Lippincott-Raven. Emotional response at the time of a potentially traumatizing event and
*McFarlane, A. C. (1988). The aetiology of post-traumatic stress disorders PTSD symptomatology: A preliminary retrospective analysis of the
following a natural disaster. British Journal of Psychiatry, 152, 116 – DSM–IV Criterion A-2. Journal of Behavior Therapy and Experimental
121. Psychiatry, 29, 123–130.
 PREDICTORS OF PTSD: A META-ANALYSIS 73

Rosenthal, R. (1979). The file drawer problem and tolerance for null Tennant, C. (1988). Parental loss in childhood: Its effect in adult life.
results. Psychological Bulletin, 86, 638 – 641. Archives of General Psychiatry, 45, 1045–1050.
Rosenthal, R. (1991). Meta-analytic procedures for social research (Rev. Tennant, C., Smith, A., Bebbington, P., & Hurry, J. (1981). Parental loss in
ed.). Newbury Park, CA: Sage. childhood: Relationship to adult psychiatric impairment and contact with
Rosenthal, R. (1994). Parametric measures of effect size. In H. M. Cooper psychiatric services. Archives of General Psychiatry, 38, 309 –314.
& L. V. Hedges (Eds.), The handbook of research synthesis (pp. 231– *Tichenor, V., Marmar, C. R., Weiss, D. S., Metzler, T. J., & Ronfeldt,
244). New York: Russell Sage Foundation. H. M. (1996). The relationship of peritraumatic dissociation and post-
Rosenthal, R., & DiMatteo, M. R. (2000). Meta analysis: Recent develop- traumatic stress: Findings in female Vietnam theater veterans. Journal of
ments in quantitative methods for literature reviews. Annual Review of Consulting and Clinical Psychology, 64, 1054 –1059.
Psychology, 52, 59 – 82. *Tucker, P., Dickson, W., Pfefferbaum, B., McDonald, N. B., & Allen, G.
Schlenger, W. E., Kulka, R. A., Fairbank, J. A., Hough, R. L., Jordan, (1997). Traumatic reactions as predictors of posttraumatic stress six
B. K., Marmar, C. R., & Weiss, D. S. (1992). The prevalence of months after the Oklahoma City bombing. Psychiatric Services, 48,
post-traumatic stress disorder in the Vietnam generation: A multi- 1191–1194.
method, multisource assessment of psychiatric disorder. Journal of *Ursano, R. J., Fullerton, C. S., Epstein, R. S., Crowley, B., Kao, T.-C.,
Traumatic Stress, 5, 333–363. Vance, K., et al. (1999). Acute and chronic posttraumatic stress disorder
*Schnurr, P. P., Friedman, M. J., & Rosenberg, S. D. (1993). Preliminary in motor vehicle accident victims. American Journal of Psychiatry, 156,
MMPI scores as predictors of combat-related PTSD symptoms. Ameri- 589 –595.
can Journal of Psychiatry, 150, 479 – 483. *Ursano, R. J., Fullerton, C. S., Epstein, R. S., Crowley, B., Vance, K.,
Schuff, N., Neylan, T. C., Lenoci, M. A., Du, A. T., Weiss, D. S., Marmar, Kao, T.-C., & Baum, A. S. (1999). Peritraumatic dissociation and
C. R., & Weiner, M. W. (2001). Decreased hippocampal N-acetyl- posttraumatic stress disorder following motor vehicle accidents. Amer-
aspartate in the absence of atrophy in posttraumatic stress disorder. ican Journal of Psychiatry, 156, 1808 –1810.
Biological Psychiatry, 50, 952–959. *Van Velsen, C., Gorst-Unsworth, C., & Turner, S. (1996). Survivors of
*Shalev, A. Y., Freedman, S., Peri, T., Brandes, D., Sahar, T., Orr, S. P., torture and organized violence: Demography and diagnosis. Journal of
& Pitman, R. K. (1998). Prospective study of posttraumatic stress Traumatic Stress, 9, 181–193.
disorder and depression following trauma. American Journal of Psychi- *Watson, C. G., Anderson, P. E. D., & Gearhart, L. P. (1996). Social
atry, 155, 630 – 637. maladjustment indicators in PTSD patients’ families of origin. Journal
*Shalev, A. Y., Peri, T., Canetti, L., & Schreiber, S. (1996). Predictors of
of Clinical Psychology, 52, 21–25.
PTSD in injured trauma survivors: A prospective study. American Jour-
Weiss, D. S., & Marmar, C. R. (1997). The Impact of Event Scale—
nal of Psychiatry, 153, 219 –225.
Revised. In J. P. Wilson & T. M. Keane (Eds.), Assessing psychological
*Shalev, A. Y., Sahar, T., Freedman, S., Peri, T., Glick, N., Brandes, D.,
trauma and PTSD: A practitioner’s handbook (pp. 399 – 411). New
et al. (1998). A prospective study of heart rate response following
York: Guilford Press.
trauma and the subsequent development of posttraumatic stress disorder.
*Weiss, D. S., Marmar, C. R., Metzler, T. J., & Ronfeldt, H. M. (1995).
Archives of General Psychiatry, 55, 553–559.
Predicting symptomatic distress in emergency services personnel. Jour-
*Skre, I., Onstad, S., Torgersen, S., Lygren, S., & Kringlen, E. (1993). A
nal of Consulting and Clinical Psychology, 63, 361–368.
twin study of DSM–III–R anxiety disorders. Acta Psychiatrica Scandi-
Weiss, D. S., Marmar, C. R., Schlenger, W. E., Fairbank, J. A., Jordan,
navica, 88, 85–92.
B. K., Hough, R. L., & Kulka, R. A. (1992). The prevalence of lifetime
*Solomon, Z. (1995). The effect of prior stressful experience on coping
and partial post-traumatic stress disorder in Vietnam theater veterans.
with war trauma and captivity. Psychological Medicine, 25, 1289 –1294.
Journal of Traumatic Stress, 5, 365–376.
Solomon, Z., Benbenishty, R., Neria, Y., Abramowitz, M., Ginzburg, K., &
Abraham, O. (1993). Assessment of PTSD: Validation of the revised Wilkinson, L., & Task Force on Statistical Inference. (1999). Statistical
PTSD Inventory. Israel Journal of Psychiatry & Related Sciences, 30, methods in psychology journals. American Psychologist, 54, 594 – 604.
110 –115. *Wolfe, J., Sharkansky, E. J., Read, J. P., Dawson, R., Martin, J. A., &
*Solomon, Z., Mikulincer, M., & Hobfall, S. (1987). Objective versus Ouimette, P. C. (1998). Sexual harassment and assault as predictors of
subjective measurement of stress and social support: Combat-related PTSD symptomatology among U.S. female Persian Gulf War military
reactions. Journal of Consulting and Clinical Psychology, 55, 577–583. personnel. Journal of Interpersonal Violence, 13, 40 –57.
*Solomon, Z., Oppenheimer, B., Elizur, Y., & Waysman, M. (1990). World Health Organization. (1969). Manual of the international statistical
Trauma deepens trauma: The consequences of recurrent combat stress classification of diseases, injuries, and causes of death (8th ed.). Ge-
reaction. Israel Journal of Psychiatry & Related Sciences, 27, 233–241. neva, Switzerland: Author.
*Speed, N., Engdahl, B., Schwartz, J., & Eberly, R. (1989). Posttraumatic World Health Organization. (1978). Clinical modification of the World
stress disorder as a consequence of the POW experience. Journal of Health Organization’s manual of the international statistical classifica-
Nervous and Mental Disease, 177, 147–153. tion of diseases, injuries, and causes of death (9th ed.). Geneva, Swit-
Spiegel, D., & Cardena, E. (1991). Disintegrated experience: The disso- zerland: Author.
ciative disorders revisited. Journal of Abnormal Psychology, 100, 366 – Yehuda, R. (1998). Psychoneuroendocrinology of post-traumatic stress
378. disorder. Psychiatric Clinics of North America, 21, 359 –379.
Spitzer, R. L., & Williams, J. B. W. (1983). Structured Clinical Interview *Zaidi, L. Y., & Foy, D. W. (1994). Childhood abuse experiences and
for DSM–III (SCID). New York: New York State Psychiatric Institute. combat-related PTSD. Journal of Traumatic Stress, 7, 33– 42.
*Sutker, P. B., Davis, J. M., Uddo, M., & Ditta, S. R. (1995). War zone
stress, personal resources, and PTSD in Persian Gulf War returnees.
Journal of Abnormal Psychology, 104, 444 – 452. Received August 2, 2000
*Tedstone, J. E., & Tarrier, N. (1997). An investigation of the prevalence Revision received August 8, 2002
of psychological morbidity in burn-injured patients. Burns, 23, 550 –554. Accepted August 9, 2002 䡲