Psychosocial Disability Before, During, and After a Major DepressiveEpisode: Title and subTitle BreakA 3-Wave Population-Based Study of State, Scar, and Trait Effects

Depression is among the top 5 leading causes of disability-adjustedlife-years and premature death worldwide.¹ Unipolarmajor depressive episodes (MDEs) are associated with significant psychosocialdisability, similar to or exceeding that noted in common medical illnesses.^{2 - 9} Thelevel of psychosocial disability varies as a function of depressive symptomseverity,^{10 - 14} andeffective treatment of depression improves depressive symptoms and disabilityoutcomes.^{15 - 18} However,it is unclear how much psychosocial disability during an MDE was already presentbefore the MDE, how much remains after the MDE, and why.

Most studies comparing psychosocial functioning in people with remissionfrom an MDE with healthy control subjects found mild postmorbid disability.^{4 ,14 ,19 - 21} Intheory, postmorbid disability could have the following 3 origins: (1) thecontinuation of premorbid disability (trait effect); (2) disability due topostmorbid residual depressive symptoms (residual symptom state effect); and(3) disability that was not present before the MDE but developed during theepisode and lasted beyond recovery of the episode (scar effect). Scarringcould result from enduring changes, induced by the MDE, in appraisal and copingstyles and in exposure to stress. Scarring may not occur routinely. Scarringmight occur only in people who suffer a severe or long-lasting first episode,making the experience of psychosocial disability sufficiently severe and prolongedto solidify and endure; or only during a recurrent MDE because of the demoralizingeffects of the recurrence of psychosocial disability.

To date, knowledge of postmorbid disability comes from cross-sectionaland short-term treatment studies and a few long-term cohort studies of depressedpatients. To decompose postmorbid disability into state, trait, and scar effects,a prospective multiwave population study is required, in which a large sampleis followed up to identify first and recurrent episodes, with assessmentsof psychosocial disability before, during, and after the episode. It is essentialto have premorbid disability data of first-MDE subjects, since the disabilityin depression-free people with a history of depression confounds trait andearlier scar effects. In addition, the timing of the premorbid and postmorbiddisability assessments is important. If the premorbid assessment takes placelong before MDE onset but the postmorbid assessment takes place shortly afterMDE remission, the within-subject before-after MDE comparison may be biasedby residual symptoms, wrongly suggesting scarring. Likewise, if the postmorbidassessment occurs long after MDE remission but the premorbid assessment shortlybefore MDE onset, the before-after comparison may be biased by prodromal symptoms,wrongly suggesting improved functioning, ie, the opposite of scarring.

To our knowledge, the 3-wave Netherlands Mental Health Survey and IncidenceStudy (NEMESIS) in the Dutch general population^{6 ,22 - 25} isthe first study that separated trait, scar, and residual-symptom state effectsand evaluated the consequences of different premorbid and postmorbid periodson the before-after MDE comparison. As indicators of psychosocial disability,NEMESIS assessed functioning in the following 4 major domains of functioning:spouse/partner, work/employment, housekeeping, and leisure time activities.Dysfunction was defined as 1 or more deficiencies in the ability to performactivities and behaviors compared with what is considered normal for thatdomain by significant others.^{3 ,11 ,26 - 27}

The NEMESIS provided the data presented in this article. Design, sample,and instruments of NEMESIS have been extensively described elsewhere.^{22 - 24} Briefly, NEMESISis a prospective psychiatric epidemiological survey in the Dutch adult generalpopulation (aged 18-64 years) with waves in 1996 (T1), 1997 (T2), and 1999(T3). It used a multistage, stratified, random-sampling procedure. From eachhousehold we selected randomly 1 respondent. Interviewers made up to 10 telephonecalls or visits to an address at different times of the day and days of theweek to make contact. To optimize response and offset any seasonal influences,the initial fieldwork extended from February to December 1996.

In the first wave, we collected data by interview from 7076 persons,a response rate of 69.7%. At T2, 1458 respondents (20.6%) dropped out, anda further 822 (14.6%) dropped out at T3. We obtained General Health Questionnaire12 data from 44% of the T1 nonresponders.²² TheT1 nonresponders had slightly lower average General Health Questionnaire 12scores (1.16 vs 1.22, suggesting better mental health), a lower average age(40.2 vs 41.2 years), and a higher proportion of women (54.4% vs 53.3%). Theirpsychiatric morbidity (estimated by a logistic regression model) did not differfrom that of the responders.²² Attrition atT2 and T3 was weakly associated with psychopathology at an earlier wave.²⁵ Agoraphobia (odds ratio [OR], 1.96; 95% confidenceinterval [CI], 1.31-2.85) and social phobia (OR, 1.37; 95% CI, 1.07-1.76)at T1 predicted attrition at T2, and major depression (OR, 1.37; 95% CI, 1.05-1.54),dysthymia (OR, 1.80; 95% CI, 1.34-2.18), and alcohol dependence (OR, 1.83;95% CI, 1.18-2.67) at T2 predicted attrition at T3 (ORs were adjusted fordemographic factors). The present report uses the 4796 respondents who participatedin all 3 waves.

At T1, 1- and 12-month and lifetime prevalence rates of major depressionfor women were 3.4%, 7.5%, and 20.1%, respectively; for men, 1.9%, 4.1%, and10.9%, respectively.²³

We diagnosed psychiatric disorders according to DSM-III-R with the Composite International Diagnostic Interview (CIDI), version1.1.²⁸ The CIDI is a structured interview developedby the World Health Organization²⁹ that hasacceptable interrater and test-retest reliability for most nonpsychotic diagnoses,including major depression.^{30 - 32} AtT2 and T3, the CIDI lifetime framework was adapted to the 1-year T1-T2 intervaland the 2-year T2-T3 interval. We assigned DSM-III-R diagnoseswithout the imposition of hierarchical exclusion rules and without the applicationof the functional impairment criterion. Persons with bipolar disorders werenot included in the MDE group.

The MDEs were qualified in terms of severity as mild, moderate, or severeaccording to the DSM-III-R criteria. Duration wasassessed with the Life Chart Interview^{33 - 34} anddichotomized at 6 months.

We constructed several groups from the CIDI data obtained at T1, T2,and T3 for between- and within-group comparisons (Figure 1). The never-MDE group (A in Figure 1) refers to respondents who never had an MDE and in whomnone developed within the T1-T3 interval. The MDE-history-only group (B in Figure 1) refers to respondentswho had at least 1 MDE before T1 and in whom no recurrence developed withinthe T1-T3 interval. Two other groups are the remitted first-MDE subjects (Cin Figure 1) (first episodeafter T1 and no MDE at T3 [n = 216]) and the remitted recurrent-MDE subjects(D in Figure 1) (recurrent episodeafter T3 and no MDE at T3 [n = 118]). Furthermore, we identified first-MDEsubjects who still met MDE criteria at T3 (E1 [n = 46]) and remitted first-MDEsubjects who met MDE criteria at T2 (E2 [n = 39]). For recurrent-MDE subjects,the corresponding groups were F1 (n = 29) and F2 (n = 25). In addition, wetook the time from T1 to MDE onset (the premorbid period) and from MDE remissionto T3 (the postmorbid period) into account, using data from the CIDI and theLife Chart Interview.³³ We split the premorbidand the postmorbid period into 1 year or less vs more than 1 year. This yielded4 first (C1-C4) and 4 recurrent MDE subgroups (D1-D4).

The key dependent measures examined in this article are psychosocialfunctioning in the following 4 major domains of life (or social roles): spouse/partner,work/employment, housekeeping, and leisure time. At each wave, we administered4 scales of the self-report version of the Groningen Social Disability Schedule,which has good reliability and validity properties in the Dutch population.^{3 ,11 ,26 - 27 ,35} Eachscale measures the extent to which respondents experience limitations in theirfunctioning in the respective domain. An important aspect of the GroningenSocial Disability Schedule is that many questions ask the respondent to indicatewhether his or her functioning deviates from the norm as told by significantothers. The scales included the 9-item Employment scale (Cronbach α,0.61-0.64; eg, "My boss was not satisfied with my work" and "I have had difficultykeeping up with my work"); the 10-item Partner scale (Cronbach α, 0.82-0.84;eg "Recently, my behavior has irritated my partner" and "I have avoided mypartner lately"); the 6-item Housekeeping scale (Cronbach α, 0.68-0.69;eg, "I can't get my housework done on time" and "Others have complained thatI am not doing the housework properly"); and finally, the 6-item Leisure Timescale (Cronbach α, 0.77-0.80; eg, "I cannot relax in my spare time"and "I do not get along with the people I spend leisure time with"). A totaldisability score was constructed by adding the 4 scale scores and dividingby 4. Higher scores indicate more psychosocial disability.

We tested between-subject and within-subject differences in disabilityusing t tests. We also performed repeated-measuresanalysis of variance (ANOVA) to adjust for possible retest and time effectsacross T1, T2, and T3 as observed in the never-MDE group.

To facilitate interpretation and comparison of between- and within-subjectdifferences, we express these differences as effect sizes (ESs). Effect sizeis defined as the mean difference in disability between the contrasted groupsor times, divided by the standard deviation of disability in the 2 groups.An ES of less than 0.20 is generally considered negligible; from 0.20 to 0.40,small; from 0.41 to 0.70, moderate; and greater than 0.70, large.

Of the 4796 individuals who participated in all 3 waves, an MDE developedafter T1 in 409, and of these, 334 (81.7%) had recovered at T3. Of the 216remitted first-MDE subjects, MDEs were mild in 36%, moderate in 31%, and severein 32%. Mean episode duration was 6.1 months (SD, 5.7 months). Of the 118remitted recurrent-MDE subjects, MDEs were mild in 25%, moderate in 35%, andsevere in 41%; they lasted on average 4.9 months (SD, 3.7 months). Of thosewith a first MDE, 64% were women, and the mean age was 40 years. For recurrentMDEs, these rates were 75% and 37 years, respectively.

We evaluated state effects by comparing the premorbid T1 disabilityscores of first-MDE subjects who still met MDE criteria at T2 or T3 with theirscores during the MDE (groups E1 + E2). We did the same for recurrent MDEsubjects (groups F1 + F2). Table 1 showsthat disability was higher during the episode than before the episode in all4 domains of functioning, for first and recurrent MDEs. All differences werestatistically significant, except for housekeeping. State ES ranged from 0.33to 0.79. Repeated-measures ANOVA showed that the state effects were not dueto possible retest or time effects. The results suggest moderate state effects.

We evaluated trait effects by contrasting the remitted first-MDE groupwith the never-MDE group with regard to T1 disability. Table 2 presents the results. The premorbid T1 disability scoresof the remitted first-MDE group had an ES of 0.42 to 0.57 U higher than theT1 scores of the never-MDE group. We repeated this comparison for only thoseindividuals who had a premorbid period of more than 1 year after T1 to reducepossible prodromal symptom bias. The differences remained significant, althoughthey dropped with an ES of about 0.10 U. These results suggest moderate traiteffects.

We evaluated scar effects, for remitted first- and recurrent-MDE subjectsseparately, by comparing premorbid T1 disability scores with their postmorbidT3 disability scores. Scar effects imply higher postmorbid than premorbiddisability scores. The postmorbid T3 total disability score of remitted first-MDEsubjects did not differ from their premorbid T1 score (Table 3). Before-after differences were not found for the individualdomains of functioning (data not presented). The same held for recurrent-MDEsubjects. Adjusting for time and retest effects did not yield evidence ofscarring either.

Although these results suggest that scarring does not occur routinely,it might occur in subcategories of MDEs. Therefore, we performed the samebefore-after disability comparison for severe, long-lasting, and severe long-lastingMDEs. We found more disability after rather than before the episode in thesubcategory of severe recurrent MDEs (n = 45) (total disability score, 11.41vs 12.38; t = 2.7 [P<.01];ES, 0.37 U [95% CI, 0.09-0.65]).

We next examined the effect of prodromal and residual symptoms on thepremorbid-postmorbid difference in disability by comparing different premorbidand postmorbid periods (Table 3).Premorbid and postmorbid periods are equal in subgroups C1 and C2 (in C1,both >1 year; in C2, both ≤1 year), and we did not find a premorbid-postmorbiddifference in disability. However, in subgroups C3 and C4, where the premorbidand postmorbid periods differ, significant differences in disability arise.In subgroup C3, where the premorbid period is longer than the postmorbid period,we found higher disability scores after than before the episode, probablybecause of fewer prodromal symptoms at T1 than residual symptoms at T3. Thereverse occurred in subgroup C4, where the premorbid period is shorter thanthe postmorbid period. The 4 subgroups of subjects with recurrent MDE (D1-D4)yielded similar patterns (data not shown).

Although psychosocial functioning was mildly impaired more than 1 yearbefore the MDE compared with the level of psychosocial functioning in thenever-MDE group (trait effect), and, in addition, further deteriorated duringthe MDE (state effect), it returned to its premorbid level after remissionof the MDE (no scar effect). Hence, we found evidence of trait and state effects,but not a scar effect. Although scarring in psychosocial functioning did notoccur routinely, the finding of more disability after the MDE than beforethe MDE in the subcategory of severe recurrent episodes suggests, if replicated,that scarring might occur in severe recurrent depression. Our findings pointat the following 2 independent processes: (1) the ongoing expression of traitvulnerability to depression in the form of mild impairments in psychosocialfunctioning; and (2) synchrony of change between severity of depressive symptomsand severity of psychosocial disability. The observed synchrony of changeis consistent with longitudinal studies of clinical samples showing that psychosocialdisability varies directly and largely as a function of the severity of depressivesymptoms.^{10 - 12 ,15 - 18} Whatour study adds to this literature is that this also holds for MDEs in thegeneral population and for first and recurrent episodes separately.

Before-after episode comparisons of disability may suggest scarringif the premorbid period is longer than the postmorbid period and the oppositeof scarring if the premorbid period is shorter. These misleading results areprobably due to the state effects of prodromal and residual symptoms on psychosocialfunctioning, suggesting that synchrony of change is not limited to the MDEbut extends to its prodromal and residual symptom phases. State effects developand disappear probably gradually, in synchrony with the rise of prodromaland the fall of residual symptoms, respectively. We hypothesized that thefew studies reporting scar effects^{2 ,19} hadon average longer premorbid than postmorbid periods, favoring an excess ofresidual symptoms at the postmorbid assessment compared with the prodromalsymptom level at the premorbid assessment. Little is known about possibletime lags in the synchrony of change between disability level and symptomseverity.

The reader should evaluate our findings and interpretations in the contextof 2 serious limitations. First, we did not measure current symptoms at thetimes we assessed psychosocial disability (T1, T2, and T3). Therefore, wecould not examine whether the disability scores of respondents were increasedin the absence of any symptoms. Hence, we cannot rule out that premorbid andpostmorbid disability might be entirely due to (chronic) subthreshold symptoms.However, for 3 reasons we think that this is not the case and that our datasupport the notion of premorbid impairments in psychosocial functioning asan expression of trait vulnerability to depression. First, mild impairmentswere already present more than 1 year before the first lifetime MDE. Second,the existence of a trait effect is consistent with reports of psychosocialdisability in patients during their asymptomatic phases.^{12 ,19 - 21} Third,even if chronic subthreshold symptoms are present long before and after theMDE and account for the premorbid and postmorbid disability, it is more appropriateto consider chronic subthreshold symptoms and the associated disability asexpressions of underlying trait vulnerability.

The second serious limitation is that we measured psychosocial disabilitywith self-report questionnaires. The NEMESIS did not collect data from independentsources (family members and employers) or performance-based data (absenteeismand work productivity). Hence, we cannot discriminate between the competingexplanations of increased self-reported disability, ie, real change vs reportingbias owing to the effects of depression symptoms on self-perception and recallprocesses.^{36 - 38} Webelieve that real change is most likely. The few validity studies of self-reportedpsychosocial disability report substantial correlations between self-reportmeasures and interview-,^{27 ,35} performance-,³⁹ and employer-based⁴⁰ data.An advantage of our measures is that they directly assess functioning andare not weighted toward symptoms (eg, "to what extent has your health interferedwith" measures).⁴¹

Other less serious study limitations include the retrospective assessmentat T3 of the duration of the MDE,³⁴ therebypossibly introducing recall bias. Because treatment data are lacking, we cannotexclude that, at the group level, effective treatment may have improved functioningof some subjects so much that this has neutralized scar effects of nontreatmentin others. Finally, we could not examine long-term adverse social outcomeslike divorce and job failure.

Major strengths of our study include the sample size, the random populationsample, the structured diagnostic assessment procedures, the largely nonselectiveattrition, and the prospective and longitudinal design made possible by assessmentin 3 waves during a 3-year period. This way we could identify 216 individualswith a first lifetime MDE that started after the first assessment wave andremitted before the last one, avoid referral filter and lead time bias, andassess disability before, after, and during the MDE.

The occurrence of first MDEs in the population is common, with a 1-yearincidence in the Netherlands of 1.72 for men and 3.90 for women per 100 atrisk.⁴² About 50% of these episodes do notremit within a few months and 20% last more than 2 years.³⁴ Hence,the finding that scarring does not occur routinely in major domains of functioningis good news, given the delay in treatment seeking⁴³ andnoncompliance by patients and providers.^{44 - 45} Thebad news is that functioning was impaired long before and after the MDE. Iftreatments could ensure remission of the episode and achieve normal functioning,people experiencing depression might be better off substantially.

Corresponding author: Johan Ormel, MD, Department of Psychiatry,University of Groningen, PO Box 30.001, 9700 RB Groningen, the Netherlands(e-mail: j.ormel@med.rug.nl).

Submitted for publication August 2, 2002; final revision received November10, 2003; accepted November 19, 2003.

This work was supported by the Netherlands Ministry of Health, Welfareand Sport, The Hague; the Netherlands Organization for Scientific Research,The Hague; and the National Institute for Public Health and Environment, Bilthoven,the Netherlands.

The analyses and writing of this article took place partly while DrOrmel was a temporary fellow-in-residence at the Netherlands Institute forAdvanced Study in the Humanities and Social Sciences. We thank the anonymousreviewers of the manuscript for their very helpful comments..