Childhood and Adolescent Predictors of Major Depression in the Transition to Adulthood

The 354 participants in these analyses included 170 men and 184 women, almost all of whom (98.0%) were Caucasian. At age 26 they were predominately working or lower-middle class (22). Most (98.0%) had completed at least a high school education, and 35.9% had 4-year college degrees. Nearly all (90.7%) were employed, with a median income of $25,000. Nearly one-fourth (24.3%) were married, and 20.6% had children.

The original 1977 sample included all youth (50.5% boys) entering kindergarten within one public school system participating in state-mandated preschool testing of developmental, academic, and behavioral factors (N=763). The school district was located in a predominately Caucasian, working-class community in the northeastern United States, and the original sample reflected the composition of the community at that time. Two-thirds of the households were working class or lower-middle class, and nearly all youth were Caucasian (99.3%). Between 1977 and 1998, attrition occurred primarily at the end of kindergarten, when students transferred from the public school system to parochial and private schools. Because data collection occurred within the public schools, those who transferred out of this school system after kindergarten were purposely excluded from follow-up (23). Overall attrition for the 21-year period was 27.6% of the youth remaining in public schools through grade 3.

The participants at age 26 did not differ significantly from those lost from the initial 1977 group on factors known to be associated with the development of psychiatric disorders. These factors include key demographic, academic, health, and behavioral-emotional characteristics, such as gender (χ²=0.45, df=1, p=0.50), parental socioeconomic status (χ²=2.01, df=1, p=0.73), nurse-rated health (t=0.58, df=700, p=0.56), and mother-rated anxious-depressed behavior (t=1.07, df=725, p=0.29) and attention problems (t=1.24, df=756, p=0.22). These results strongly suggest that the sample was not compromised by selective attrition. After a thorough description of the study at each time point, written informed consent was obtained from the mothers and, starting at age 15, from the participants.

The participants were interviewed at ages 18 and 21 by using the National Institute of Mental Health Diagnostic Interview Schedule (DIS), version III, revised (24), to provide diagnoses of DSM-III-R major depression. They were subsequently interviewed at age 26 with version IV of the DIS (25) to derive DSM-IV diagnoses. Participants were considered to have depression in the transition period if they met the criteria for major depression between ages 18 and 26 at any of these three assessment points. Administration of the interviewer-rated Global Assessment of Functioning Scale (GAF) (DSM-IV, p. 32) at age 26 provided information about functional impairment.

The DIS has been used in prior community-based studies of psychiatric disorders, and validity studies have shown significant agreement between DIS diagnoses and other measures (26). This structured interview was administered by graduate-level professionals with clinical and/or research experience. DIS-certified staff conducted a rigorous 2-week training and ongoing monitoring of interviewers. High rates of agreement (98%–99%) were found between the interviewers and staff on practice and in-field interviews over the multiple assessments.

The measures of family factors included demographic information, psychiatric histories of parents and siblings, and indexes of family functioning.

Demographic information, determined from interviews with the mother at age 5, included a measure of family socioeconomic status based on the Hollingshead two-factor index of social position (five categories; 1=highest, 5=lowest) (22). A family composition index was also constructed and included four interrelated factors (r=0.44 to r=0.68, all p<0.05) present at the time of birth: older mother (30 years or older), older father (35 years or older), larger family size (≥4 children), and being third or later in the birth order.

DSM-III-R major depression and substance use disorders in first-degree relatives, evidenced before the participants were age 15, were determined from combined reports of the mother and the participant at age 26 on the Family History Assessment Module (27, 28). Combined reports also provided information about whether a parent or sibling had ever been hospitalized for a psychiatric problem or attempted suicide.

Dichotomous measures of family functioning included self-reports of violence (directed toward the participant or between other family members) by age 15 and the occurrence of abuse by parents or siblings. Reports of abuse were based on retrospective accounts during the interview at age 21 of physical or emotional abuse occurring by age 15. The participants’ views of their role in the family at ages 9 (alpha=0.54) and 15 (alpha=0.47) were measured by seven family-related items from the Piers-Harris Children’s Self-Concept Scale (29), such as “I am a disappointment to my family” (7). Family cohesion at age 15 (alpha=0.82) was measured by the cohesion subscale of the Family Adaptability and Cohesion Evaluation Scales (30).

Reports of childhood and adolescent behaviors were prospectively obtained at ages 9 and 15 and relied on the perspectives of the participants and their mothers.

Participant reports at ages 9 (alpha=0.84) and 15 (alpha=0.83) on the Piers-Harris Children’s Self-Concept Scale provided an index of anxious-depressed behavior. At age 15, depressive symptoms were measured by the Children’s Depression Inventory (alpha=0.85) (31). The Youth Self-Report was also used at age 15 to gather information about behavior problems (32). The internalizing (alpha=0.88) and externalizing (alpha=0.86) scales were used.

At ages 9 (alpha=0.73) and 15 (alpha=0.80), mother-rated anxious-depressed behavior was assessed with the Simmons Behavior Checklist (33). At age 15, the mothers also completed the Child Behavior Checklist (34), the parent version of the Youth Self-Report, to provide information about internalizing (alpha=0.89) and externalizing (alpha=0.89) difficulties their adolescents had experienced in the past 6 months.

Since previous studies have indicated that pathways to depression may differ for males and females (3, 35), an initial series of bivariate analyses tested for gender-by-depression interactions in predictors. For factors measured on a continuous scale, two-way analyses of variance with gender and depression as grouping factors were performed, and the F value for the interaction term was evaluated. For dichotomous variables, the Breslow-Day test of homogeneity of odds ratios, in conjunction with the Cochran-Mantel-Haenszel chi-square test, assessed whether associations between predictors and depression differed by gender. No gender interactions were found in these initial analyses. In subsequent bivariate tests conducted to identify family and behavioral-emotional factors associated with depression, gender was included as a control factor in analyses of covariance (for continuous measures) and Cochran-Mantel-Haenszel chi-square tests (for dichotomous indicators). Adjusted odds ratios (controlling for gender) with 95% confidence intervals (CIs) were also computed. Finally, a multivariable logistic regression analysis determined the most salient subset of predictors. Gender interactions were also tested in the regression model.

Two groups were identified on the basis of DSM diagnosis of major depression: 1) participants experiencing major depression between ages 18 and 26 and 2) those who did not have the disorder during this developmental period. Twenty-two participants with bipolar disorder were excluded. We did not, however, exclude participants who had other psychopathology from ages 18 to 26. We were generally interested in identifying markers distinguishing those who experience any major depression during the transition to adulthood from those who do not.

In this sample, 23.2% (N=82) met the diagnostic criteria for depression in the transition to adulthood, while the majority of participants (76.8%, N=272) did not. At age 26, the group with major depression had significantly lower GAF scores than the group without major depression (t=–3.92, df=351, p<0.0001). And in keeping with previous findings (3, 18), there was no significant difference in the prevalence rates for males and females (21.2% versus 25.0%).

Potential familial and behavioral-emotional predictors of depression were selected from the literature and prior analyses. These candidate predictors and their bivariate associations with depression are summarized in Table 1 and Table 2. Twelve significant markers were identified, and all associations were in the theoretically predicted direction.

In the family domain (Table 1), having a depressed parent or sibling was significantly related to experiencing depression in the transition period. Having a sibling with a substance use disorder by age 15 and having a sibling who attempted suicide were also significantly associated with depression. Furthermore, being born into a large family and/or to older parents was a significant risk factor. Self-reports of family violence and low family cohesion were also significant markers.

Significant behavioral-emotional characteristics included both participant- and mother-rated internalizing problems at age 15 (Table 2). Self-reports of anxious-depressed behavior at ages 9 and 15 proved to be significant indicators, as were self-reports of depressive symptoms in midadolescence.

Because of concerns that anxiety and depressive symptoms in midadolescence might be early manifestations of, rather than potential risk factors for, later depression, we reexamined the bivariate associations between depression and the remaining predictors while controlling for the effects of both self-rated anxiety and depressive symptoms at age 15. The adjusted estimates were not attenuated and were of similar magnitude to those found in the unadjusted bivariate analyses. This suggests that self-reported anxiety and depressive symptoms at age 15 were not early manifestations of later major depression, and these variables were therefore included in the subsequent multivariable analyses.

A multivariable analysis was conducted to determine the most salient subset of predictors for experiencing depression between ages 18 and 26. All 12 variables associated with the outcome in the bivariate analyses were considered in a stepwise selection procedure. The final model was obtained by examining individual Wald test statistics for each factor and including only those with p<0.10. Because predictors of depression might be gender specific, interactions of gender with selected factors were also examined. For variables with quantitative scales, the assumption of linearity on the logistic scale was assessed.

Noteworthy predictors from the multivariable analyses included family violence, family composition, self- and mother-rated internalizing problems at age 15, and lower family cohesion. More specifically, participants exposed to family violence by age 15 were approximately four times as likely (odds ratio=3.88, 95% CI=1.56–9.64) to have major depression in the transition period. Also, those with a score of 4 on the family composition index (i.e., older mother, older father, larger family size, and later in birth order) were almost three times as likely (odds ratio=2.73, 95% CI=1.29–5.81) to meet the criteria for major depression as were adolescents with a score of 0. Both the self- and mother-based ratings of internalizing problems (on the Youth Self-Report and Child Behavior Checklist, respectively) were significant predictors. Participants who were approximately 1 SD (or 10 units) above the mean for internalizing problems were twice as likely (Youth Self-Report: odds ratio=2.00, 95% CI=1.08–3.64; Child Behavior Checklist: odds ratio=2.00, 95% CI=1.29–3.96) to meet the criteria for the disorder as were participants whose scores were 1 SD below the mean. Last, the adolescents who scored approximately 1 SD above the mean (or 7.5 units) on the measure of lack of family cohesion were nearly twice as likely (odds ratio=1.70, 95% CI=0.97–3.07) to meet the criteria for depression as were those scoring 1 SD below the mean.