Turner RJ, Lloyd DA. Stress Burden and the Lifetime Incidence of Psychiatric Disorder inYoung AdultsRacial and Ethnic Contrasts. Arch Gen Psychiatry. 2004;61(5):481-488. doi:10.1001/archpsyc.61.5.481
Copyright 2004 American Medical Association. All Rights Reserved.Applicable FARS/DFARS Restrictions Apply to Government Use.2004
With the exception of studies of individual traumatic events, the significance
of stress exposure in psychiatric disorder previously has not been effectively
To address the hypothesis that accumulated adversity represents an important
risk factor for the subsequent onset of depressive and anxiety disorders.
A community-based study of psychiatric and substance use disorders among
a large, ethnically diverse cohort representative of young adults in South
Florida. Adversity was estimated with a count of major and potentially traumatic
events experienced during one's lifetime and prior to the onset of disorder.
Most interviews took place in the homes of participants, with 30% conducted
We obtained a random sample of individuals aged 18 to 23 years from
a previously studied representative sample of young adolescents. Because participants
in the prior study were predominantly boys, a supplementary sample of girls
was randomly obtained from the early-adolescence school class rosters. A total
of 1803 interviews were completed, representing a success rate of 70.1%.
Main Outcome Measures
Depressive and anxiety disorders were assessed through computer-assisted
personal interviews using the DSM-IV version of the
Michigan Composite International Diagnostic Interview.
Level of lifetime exposure to adversity was found to be associated with
an increased risk of subsequent onset of depressive and/or anxiety disorder.
This association remained clearly observable when childhood conduct disorder,
attention-deficit/hyperactivity disorder, prior substance dependence, and
posttraumatic stress disorder were held constant and when the possibility
of state dependence effects was considered.
Evidence suggests that high levels of lifetime exposure to adversity
are causally implicated in the onset of depressive and anxiety disorders.
The idea that stress matters for health and well-being is now widelyaccepted by the public and among many researchers and physicians. Consistentwith this view, a large body of evidence has accumulated during the past quartercentury linking exposure to recent stressful events with psychological distress,most typically indexed in terms of depressive symptoms.1- 3 Instudies in which social stressors have been assessed more comprehensively,differences in such exposure have been found to account for a substantialportion of observed variation in psychological distress, both across individualsand population subgroups.4- 6
Although this evidence appears incontrovertible, a question remainswith respect to the role and significance of social stress in the occurrenceof psychiatric disorders in contrast to psychological distress or depressivesymptoms. Although certain studies, including some of our own,5,6 havereported associations of exposure to stressful events and chronic stressorswith depressive disorder, significant interpretive problems attach to suchfindings. The virtually insurmountable difficulty is that a substantial majorityof 1-year, 6-month, and 1-month prevalent cases involve recurrent rather thaninitial episodes. The first onset of disorder will generally have precededthe occurrence of the stressful experiences assessed, so these findings cannotprovide convincing support for the hypothesized significance of social stress.Any causation that may be involved in demonstrated associations may go frompsychiatric disorder to stress exposure, and reported stressful events andcircumstances may include both contributors to and consequences of psychiatricdisorder.
Thus, meaningful evaluation of the stress hypothesis requires that considerationbe limited to stressful experiences that are temporally prior to the firstonset of disorder. This article assesses the role and significance of lifetimeexposure to major and potentially traumatic events (cumulative adversity)in depressive and anxiety disorders. For these analyses, depressive and anxietydisorders are defined in terms of DSM-IV criteriaas estimated by the Michigan Composite International Diagnostic Interview7 (CIDI), and cumulative adversity is based on a countof the reported lifetime experience of 33 distinct events.
It is hardly novel to suggest that traumatic experiences can have significantadverse mental health consequences. Such experiences have represented prominentexplanatory factors since the early days of psychoanalytic theory.8 Moreover, a considerable body of research has accumulated,suggesting the relevance of specific forms of early trauma for adult mentalhealth.9 Principal among these are sexual abuse,10- 12 physical violenceand abuse,13- 15 parentaldeaths,16,17 and parental psychopathologiccharacteristics and substance abuse.18,19 Thehypothesis that major adverse experiences have long-term psychiatric significanceis clearly tenable. However, studies that have gone beyond the considerationof individual and particular events to assess the significance of an arrayof major events, either singly or cumulatively, have been extremely rare.20
Our approach to assessing lifetime exposure to major events treats suchexperiences as different from typically assessed life events primarily interms of their severity and presumably the duration of their consequences.The widespread practice of limiting consideration of stressful events to a1-year time frame has been based largely on evidence that the effects of theevents considered tend to be limited to less than a year,21,22 aswell as the wish to avoid the falloff observed in the ability of respondentsto recall many events beyond a 1-year time frame.23 However,the evidence regarding childhood victimization and parental death and divorce16,17,21,24 suggeststhat some events can and do have significant mental health consequences despiteoccurring years or even decades earlier. Moreover, the problem of reliabilityof reporting or recall does not apply equally to all events. For example,it would seem rare for subjects to forget or fail to respond to specific questionsasking if their parents had divorced, if they had been in an accident in whichsomeone was killed or badly injured, or if they had been shot with a gun orbadly injured with another weapon.
It appears that a range of severe events can be measured with reasonableaccuracy and, singly or in combination, may constitute important mental healthrisk factors. These considerations argue for the inclusion of lifetime experienceof such events within efforts to understand variations in the occurrence ofmental health problems. Our article reports the individual and cumulativesignificance of the lifetime experience of what is, to our knowledge, thewidest range of major and potentially traumatic experiences studied thus farfor the occurrence of depressive and anxiety disorders. Measures of exposureto recent life events, chronic stressors, and discrimination stress, all ofwhich assess conditions relatively proximate to the time of the interview,are not considered in this study because of the associated time-ordering problemsdescribed previously. However, racial and ethnic differences in stress exposure,estimated in terms of these dimensions, and the significance of such exposurefor psychological distress have been reported elsewhere.25,26
This article is based on a study of the prevalence and social distributionsof psychiatric and substance use disorders and of factors that increase anddecrease risk of such disorders among a representative South Florida communitycohort of 1803 young adults. Most (93%) were aged 19 to 21 years when interviewedbetween 1998 and 2000. The study possesses unique potential for contributionin several respects. First, these data are from perhaps the largest samplewithin this age range studied thus far in the United States. Second, thisis one of the first large-scale community studies to estimate the occurrenceof disorders based on DSM-IV criteria. Third, ourstudy population is ethnically diverse, allowing consideration of ethnic variationsin both stress exposure and the consequences of exposure. Specifically, approximately25% of the sample were of Cuban origin, 25% were other Hispanic individualsfrom the Caribbean Basin, 25% were African American, and 25% were non-Hispanicwhite.
Our approach in obtaining this sample was in accord with a growing consensusin the field that race is more a social than a biological categorization akinto ethnic status27 and that there are importantcultural variations within ethnic status. In an effort to minimize the effectsof such variations on results, we have distinguished Cuban from other Hispanicindividuals and have limited inclusion within this latter category to Hispanicpeople from countries in the Caribbean Basin. For the same reason, Haitianand other Caribbean black individuals were not studied and are not includedin the African American subsample.
This study was based on data from a representative sample of young adults,most of whom had been studied 5 to 7 years earlier.28 Between1998 and 2000, we interviewed 1803 respondents aged 18 to 23 years (92% wereaged 19 to 21 years), and all analyses presented in this article are basedon data from these interviews. Overall, 70.1% of those sampled were successfullyrecruited to the study. Most interviews were carried out face to face in thehomes of respondents, with 30% conducted by telephone and aided by mailedresponse booklets. Consistent with the bulk of evidence that in-person andtelephone interviews yield comparable data,29- 31 ouranalyses revealed no association between interviewing mode and the presencevs absence of an affective or anxiety disorder diagnosis (prevalence, 0.25and 0.27, respectively, for in-person and telephone interviews). Althougha slight difference in number of reported adversities was observed acrossinterviewing mode (8.4 vs 7.8, respectively, for in-person and telephone interviews),the fact that the higher stress exposure corresponds with a lower prevalenceof disorder suggests an absence of bias associated with interviewing mode.A more detailed description of the sample and study field procedures has beenpublished previously.32
Informed consent was obtained from all study participants. The institutionalreview board of Florida State University, Tallahassee, approved the proceduresfor obtaining consent and protecting the rights and welfare of participants.
Data on the lifetime and 1-year occurrence of psychiatric disorderswere obtained through computer-assisted personal interviews that allowed theestimation of DSM-IV diagnoses. Our basic instrumentwas the Michigan CIDI, which was used in the National Comorbidity Survey (NCS).7 The CIDI is a fully structured interview based substantiallyon the Diagnostic Interview Schedule33 anddesigned to be administered by lay interviewers trained in its use.34- 36 Using the MichiganCIDI as updated by NCS researchers to cover DSM-IV criteria,we assessed major depression, dysthymia, generalized anxiety disorder, socialphobia, panic disorder, alcohol abuse and dependence, drug abuse and dependence,posttraumatic stress disorder, and antisocial personality disorder. Theselatter 2 modules had been borrowed from the Diagnostic Interview Schedulefor the NCS. Evidence for the validity of Michigan CIDI diagnostic estimates,evaluated against the Structured Clinical Interview for DSM-III-R,36 has been reported for mostNCS disorders including affective disorders,37 anxietydisorders,38,39 addictive disorders,40,41 and posttraumatic stress disorder.42
Along with the Michigan CIDI, our assessment instrument43 includeda reliable module44 obtained from the revisedDiagnostic Interview Schedule45 to assess attention-deficit/hyperactivitydisorder and contained items to allow the assessment of childhood conductdisorder. The dependent variable for the analyses to be presented was thelifetime occurrence of major depressive disorder, dysthymia, social phobia,generalized anxiety disorder, and/or panic disorder.
A retrospective procedure is necessarily required to estimate the timingand sequencing during the life course of major events and psychiatric disorders.The validity of retrospective reports is uncertain, and opinions of theirusefulness vary widely. However, much of the information we have on age cohortdifferences in substance use and psychiatric disorders,41,46 lifetimecomorbidity within and across these domains,47 socialconsequences of early-onset psychiatric disorders,48- 51 andrisk significance of early traumas for psychiatric and substance use disorders9,20 is based on retrospective reportsobtained in the Epidemiologic Catchment Area Study,46 NCS,and an array of other studies. The potential significance of such relationshipsand the need to better understand their correctness and meanings emphasizesthe importance of maximizing the accuracy of these retrospective reports.Specifically, the central need is to effectively order within the life coursethe first onset of psychiatric disorder and the occurrence of major and potentiallytraumatic life events.
We used a life history calendar based on that developed by Freedmanet al52 as an aid in achieving the most accuraterecall of significant life course experiences. This calendar traced 5 categoriesof experience. The first 3 involved a process in which respondents describeddivisions in their lives in terms of where they lived (country, city, or streetas appropriate), landmark events (eg, birth of a sibling, getting a driver'slicense, or finishing school), and the teachers or best friends they had duringvarious years. These dimensions were completed at the beginning of the interview1 at a time, each building on the information already obtained. The calendarwas used at 3 points in the interview: (1) for questions about the age ofoccurrence of major or traumatic lifetime events; (2) for questions aboutthe onset and last occurrence of substance abuse and related problems; and(3) for questions regarding the age at first and last occurrence of psychiatricdisorder episodes. In each of these sections, the question of temporal orderwas established using all information available through scanning both upwardand across the life history calendar. Thus, a reported first onset of majordepression, for example, would be placed on the calendar in relation to otherpsychiatric disorders (if any), substance abuse disorders (if any), majorlife events, teachers and/or best friend at the time, landmark or transitionevents, and place of residence. By this means, the 33 items used to assesslifetime exposure to adversity were set in time relative to the onset of disorder.This procedure and the fact that the recall period for this young adult populationwas relatively short argues for the reliability of the data used in theseanalyses.
Ethnicity was measured by the respondents' self-reported ethnic groupidentification. Because our subjects were in the transition to adulthood,socioeconomic status (SES) was estimated in terms of parental education, income,and occupational prestige level.53 These datawere primarily obtained from parent interviews. However, it was necessaryto rely on information provided by the participants for 33% of the sample.Scores for these 3 status dimensions were standardized, summed, and dividedby the number of status dimensions for which the parent or participant waswilling and able to provide information.
Our analyses were multivariate and based on discrete-time event historyregression.54,55 This method usedonly data covering the time each individual was at risk for the event of interest.That included the entire period of observation for those for whom the eventhad not yet occurred, which were referred to as right-censored observations.Time at risk was divided into discrete periods (eg, years). Data for the earliest5 years were collapsed into a single period because there was inadequate variationfor analysis within the earliest individual years. The remaining informationwas grouped into 17 one-year intervals representing ages 6 to 22 years. Survivaltime to the first onset of a depressive or anxiety disorder among the 354respondents who met criteria for 1 or more of these diagnoses, as well asthe entire time at risk among the 1426 right-censored subjects, was dividedinto 26 241 person-periods.
We refer to the period for which the conditional hazard of onset isestimated as the index period. In some of the analyses, the models distinguishbetween distal and proximal adversities. Distal adversities are counts ofevents occurring during any period earlier than the year prior to the indexperiod. Proximal adversities are counts of events that occurred in the yearpreceding the index year. The effect of time is modeled as a cubic function.The coefficients for distal and proximal adversities reflect their independentassociations with the conditional hazard of onset in that period. This analyticapproach parallels that previously used to assess the linkage between cumulativeadversity and DSM-IV drug dependence.56
Table 1 presents the 33adversity items with rates of lifetime occurrence and the adjusted odds ratiosfor a depressive or anxiety disorder given the previous experience of eachevent. The odds ratios were derived from event history analyses of individualadversities, controlling for time, sex, ethnicity, and SES. This method ensuresthat the association does not reflect adversities occurring after the onsetof disorder and that any causal connection in reported associations goes fromthe event to the depressive or anxiety disorder.
Of the 9 items listed as major events, 6 were associated with a significantlyelevated risk of subsequent disorder onset. Rates of exposure to all 9 itemsvaried by ethnicity, with 7 of the 9 reported more frequently by African Americans.Six of the 9 occurred significantly more often among women. Ten of the 13items listed as life traumas significantly predicted the subsequent onsetof a depressive or anxiety disorder, 5 of which differed in prevalence acrossethnicity. White non-Hispanic individuals reported the highest rates of sexualmolestation and physical abuse by someone other than a partner or parent.The rates of having been shot or shot at (or otherwise attacked with a deadlyweapon) were highest among African Americans and higher among men in general.Five events were more frequently reported by women: sexual molestation, rape,physical abuse by a parent, emotional abuse by a parent, and physical abuseby a spouse or partner.
Five of the 6 items listed in the "witnessed violence" category wereassociated with a significantly higher risk of depressive or anxiety disorders,and African Americans experienced all 6 more frequently. Men experienced 5of the 6 more often than women. With respect to items listed as traumaticnews, hearing that someone the respondent knows was attacked with a deadlyweapon and hearing that such a person was killed were dramatically more commonamong African American respondents, whereas hearing of the suicide or rapeof someone they know was significantly more prevalent among white non-Hispanicsubjects. There were no significant sex differences in reports of traumaticnews. All 5 experiences were indicative of increased risk for the subsequentoccurrence of a depressive or anxiety disorder.
Mean differences in cumulative lifetime adversity across sex, ethnicity,and SES are presented in Table 2 alongwith the corresponding distributions of the lifetime occurrence of depressivedisorders, anxiety disorders, and disorders combined. Cumulative lifetimeadversity is represented by a simple count of the number of different events(of 33) reported. Multiple occurrences of the same event are not includedin the count. This procedure presumably yields conservative estimates of thesignificance of prior stress exposure for the onset of depressive and anxietydisorders. Significant differences in the mean number of reported adversitiesare observed across categories of all 3 status variables. Men reported moreaccumulated adversities than women, whereas African Americans generally experiencedmuch higher levels of adversity than all other subgroups. The relationshipbetween SES and level of adversity is not monotonic, but relative to theirmore advantaged counterparts, those in the lower half of the distributiontend to have experienced significantly higher levels of adversity.
The substantially higher lifetime prevalence of depressive and anxietydisorders among women and the significantly lower prevalence of depressivedisorders among African Americans correspond with findings from the NCS.7 Despite the clear link between SES of origin and lifetimerisk of exposure to major stressors, these results offer no indication thatthose who grew up in poorer economic circumstances are any more or less likelyto have experienced a depressive or anxiety disorder.
The results of hierarchical event history analyses that distinguishbetween distal and proximal occurrences are presented in Table 3. As previously noted, events are categorized as distal ifthey occurred earlier than the year prior to the index year (the year forwhich hazard is estimated). Proximal events are those occurring in the yearprior to the index year.
The upper section of Table 3 showsthe associations previously presented between sex, ethnicity, and psychiatricdisorder and demonstrates their independent significance. We trimmed SES fromthese models on the basis of preliminary analyses that revealed no independentrelationship with disorder. Model 2 indicates that distal adversities aresignificantly associated with disorder. When proximal adversities are added(model 3), distal events remain significant, and the magnitude of the distalevents association is little diminished. The nonutility of stress exposuredifferences for explaining ethnic variations in risk is clearly demonstratedin these results. When both distal and proximal adversities are held constant,the coefficients for each ethnic group contrasted with African Americans increaserather than decrease. Computing the change in these coefficients from model1 to model 3 reveals that if other groups were as exposed to adversity asare African Americans, the occurrence of depressive and anxiety disorderswould increase by 32% to 35% rather than decrease.
Clearly, there is a compelling relationship between cumulative adversityand risk of a depressive or anxiety disorder. The magnitude of this associatedrisk can be illustrated in terms of the odds ratio for disorder onset at highcompared with low levels of exposure to adversity. At age 18 years, the mean± SD number of preonset adversities experienced was 4.21 ± 3.57for distal events and 0.88 ± 1.19 for proximal events. Based on theeffects in Table 3, the odds ratiofor disorder onset among subjects with distal adversity scores 1 SD higherthan the mean is double that for those with scores 1 SD lower than the mean(odds ratio = 2.0) with proximal adversities controlled. The comparable oddsratio for the effect of proximal adversity with distal adversities controlledis a lower but still substantial 1.53. Additional analyses of interactioneffects (data not shown) revealed that the rate at which increases in lifetimestress are translated into increased risk for disorder is generally equivalentacross ethnicity and sex.
The lower section of Table 3 presentsan effort to assess the plausible alternative hypothesis that problematicindividuals tend to place themselves in circumstances in which stress exposureis more likely, on the one hand, and are at elevated risk for psychiatricdisorder on the other. This analysis controls for childhood conduct disorder,which reflects serious conduct problems up to age 15 years, and attention-deficit/hyperactivitydisorder as well as the prior occurrence of posttraumatic stress disorderand alcohol or drug dependence. Although attention-deficit/hyperactivity disorderrepresents a significant independent predictor of psychiatric disorder, theinclusion of this set of previous conditions produces no substantial changesin the observed effects of cumulative adversity.
A second competing hypothesis is that these results arise from statedependence bias. State dependence bias refers tothe tendency for persons with a current disorder to be more likely to rememberand/or report having experienced stressful events than those same individualswhen they are relatively free of disorder symptoms. We evaluated the plausibilityof this hypothesis in 2 ways. First, we compared the total number of adversitiesreported by participants with a current (last 6 months) psychiatric (excludingposttraumatic stress disorder) or substance abuse disorder with the reportsof those who met criteria for 1 or more of these disorders but for an earlierperiod in their lives. The mean scores for these 2 groups were highly similar(9.99 among those with a current disorder and 10.23 among those with a previousdisorder; P = .82). We also extended the analysisshown in model 4 of Table 3 bycontrolling for level of depressive symptoms at the time of the interview,as estimated by scores on the Center for Epidemiologic Studies DepressionScale. Although the coefficients for distal and proximal adversities werediminished, they remained statistically significant. These twin results makeit unlikely that the observed links between cumulative adversity and psychiatricdisorder can be attributed to state dependence.
As previously reported,56 these resultsindicate that exposure to major and potentially traumatic events is commonplaceamong young people, at least in South Florida. The typical African Americanin the sample had experienced more than 9 such events, and the remaining 3groups averaged more than 6. A total of 26 of the 33 events examined wereassociated with significantly increased risk for a depressive or anxiety disorder.In some cases the experience itself may be implicated in the observed elevationin risk, whereas in others the event may represent simply a marker for theoccurrence of other stressors and/or the presence of other significant riskfactors.
For both sexes and for all 4 ethnic groups, increases in lifetime exposureto adversity were associated with increased risk for psychiatric disorder.Moreover, the rates at which increases in exposure were translated into increasesin risk were generally equivalent for all subgroups. These latter findingsmake it unlikely that the relatively low prevalence of affective and anxietydisorders among African Americans is attributable to social or cultural differencesin resiliency or other stress-moderating factors. More plausible are the possibilitiesthat response differences to diagnostic measures systematically underestimatethe presence of disorder among African Americans or that stress exposure ismore often translated into physical disorders that are dramatically elevatedin that population.
These results also demonstrate that accumulating adversities, both distaland proximal, contribute significantly and independently to the predictionof disorder with demographic factors controlled. Moreover, the role of prioradversity remains clearly observable when child conduct disorder, attention-deficit/hyperactivitydisorder, prior substance dependence, and prior posttraumatic stress disorderare held constant.
This set of findings along with evidence contrary to the state dependencehypothesis suggests that high levels of lifetime exposure to adversity maybe causally implicated in the onset of depressive and anxiety disorders. Althoughthis statement appears equally applicable across ethnic groups, it is clearthat exposure differences cannot account for the relatively low prevalenceof depressive and anxiety disorders among African Americans observed in thisas well as previous studies. If exposure to stressors were equalized acrossethnic groups, the more favorable outcome found among African Americans wouldbe substantially more rather than less pronounced.
The finding of no SES variations in risk of depressive and anxiety disordersis of special interest in the context of the clear links between SES and stressexposure and between stress exposure and the occurrence of such disorders.Because the SES measure is based on parental education, occupation, and incomeand must therefore roughly index participants' developmental social environment,the lack of any association with the disorders examined would appear to challengesocial causation perspectives.
Given the well-established relationship between adult SES and risk ofpsychiatric disorder,6,57,58 thefact that the young adults in this study with a lifetime occurrence of depressiveor anxiety disorder appear to be randomly recruited with respect to SES oforigin would traditionally be held as indirect but compelling evidence forthe primacy of biogenetic etiology. In contrast, the clear relationship betweensocial stress and the occurrence of these disorders and the fact that stressexposure is significantly elevated among those with parents of low SES providedirect and persuasive support for the hypothesis that variations in the socialenvironment represent crucial causal contingencies in psychiatric disorder.The combination of these seemingly contradictory results suggests that levelof exposure to social stress during the developmental years may influencean individual's eventual socioeconomic achievement, at least partially throughaffecting the risk for disorders that tend to impede such achievement. Thus,with respect to the widely documented social class–mental health association,these findings suggest that social selection phenomena may arise substantiallyfrom prior social causation processes rather than simply reflecting the consequencesof biogenetically determined disorders.
The limitations of this study, as previously noted,32 includethose characterizing prior studies that have derived diagnoses from a singlestructured interview that does not involve clinical judgment. The data arecross-sectional, so estimates of the lifetime occurrence of depressive andanxiety disorders rely entirely on retrospective recall. Whereas the youngage of this cohort presumably minimizes recall problems, they remain of someconcern. The fact that study participants represent a narrow age range (approximately93% were aged 19 to 21 years) advises caution in generalizing these findingsto other age groups, particularly given the remaining years of high risk forthe onset of the disorders considered. Finally, although we believe our sampleto be representative of young adults in Miami-Dade County, Florida, the distinctnature of the resident Hispanic population suggests caution in generalizingto other areas of the country.
Corresponding author: R. Jay Turner, PhD, Center for Demography andPopulation Health, Bellamy Bldg 617, Florida State University, Tallahassee,FL 32306-2240 (e-mail: email@example.com).
Submitted for publication September 22, 2003; final revision receivedNovember 21, 2003; accepted December 15, 2003.
This study was supported by grant RO1 DA10772 from the National Instituteon Drug Abuse, Bethesda, Md.