The Incidence and Course of Depression in Bereaved Youth 21 Months After the Loss of a Parent to Suicide, Accident, or Sudden Natural Death

The participants were from 120 parentally bereaved families and 91 nonbereaved families, which comprised 83.1% of the 239 families described previously and an additional 15 families recruited subsequently (4) . The deceased probands were between the ages of 30 and 60 years, had biological offspring between the ages of 7 and 25, and had died within 24 hours of the precipitating event from suicide, an accident, or sudden natural cause. Families in which there were multiple deaths or injuries were excluded. The accidental deaths consisted of 12 motor vehicle accidents, 11 drug overdoses, three falls, and five other causes (e.g., drowning, exposure to cold). At intake the drug overdoses were carefully reviewed to rule out subjects with possible suicidal intent, and those probands had a much lower rate of previous suicide attempt (none of 11, or 0.0%) than did suicide victims (11 of 37, or 29.7%) (p=0.03, Fisher’s exact test). The sudden natural deaths were due to myocardial infarction (N=37), infections (N=5), and 10 less frequent causes (e.g., diabetes, stroke, aneurysm, gastric bypass surgery). In the bereaved families, the caretaking parent was almost always female (112 of 120, or 93.3%) and offspring in bereaved families were almost always the biological offspring of the caretaking parent (151 of 176, or 85.8%). The nonbereaved offspring had two living biological parents, lived in the home of at least one of them, and had no first-degree relatives who had died within the previous 2 years. Four offspring from two comparison families who experienced the loss of a parent during the follow-up period were excluded from these analyses.

Bereaved families were recruited through coroner’s reports and newspaper advertisements, and 71% of the potential subjects participated (4) . The demographic characteristics of the enrolled probands were similar to those for all people who had died by suicide or accident. The accident and suicide probands in the families recruited by advertisement and those recruited from the coroner’s office were similar with respect to method of death and demographic variables. Because people who die from sudden natural causes are not under the coroner’s jurisdiction, the representativeness of this proband group cannot be assessed. The comparison probands were frequency matched to the deceased parents by neighborhood, age, and gender, and they were recruited either by random-digit dialing or by newspaper advertisement, with an overall acceptance rate of 55% (4) . This study was approved by the University of Pittsburgh institutional review board, and all participants gave written consent or assent.

A total of 401 bereaved and nonbereaved youth were available for follow-up, of whom 344 (85.8%) were interviewed ( Table 1 ). Those retained in the study were similar to those who were lost, except that the caregivers who were retained had higher levels of functioning at baseline according to the Global Assessment Scale (GAS) (6) (mean=78.4, SD=11.0, versus mean=71.1, SD=14.4; t=3.8, df=35.9, p=0.008) and were more likely to be Caucasian (85.1% versus 70.3%; χ ² =5.34, df=1, p=0.02). Group comparisons showed similarities among probands, caregivers, and offspring, except that the caregivers of children whose parent committed suicide were more likely to be female than were the caregivers of the comparison children and that the length of follow-up was about 3 weeks longer in the group with sudden natural parental death than in the comparison group.

The participants were interviewed at two points in time, first around 9 months after the death, as previously reported, and then around 1 year later, i.e., 21 months after the death (4) . The latter assessments were timed to avoid conducting an interview on the second-year anniversary of the parent’s death. Socioeconomic status and household income were rated by using Hollingshead’s scale (5) . Psychiatric disorders in offspring younger than 18 were assessed by using the Schedule for Affective Disorders and Schizophrenia for School-Age Children—Present and Lifetime Version (7), and for adult offspring and caregivers, the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) was used (8) . Psychiatric assessment of the proband was conducted by using a psychological autopsy procedure (9) and the SCID-I (8) . The course of disorders in offspring was documented by using the Longitudinal Interview for Follow-Up Evaluations (10) or the adolescent version. Functional status was determined by using the Children’s Global Assessment Scale (11) or the GAS for adults (6) . High interrater reliability was maintained on psychiatric diagnoses and global impairment, as indicated by kappa values of 0.74–0.85 and an intraclass correlation coefficient of 0.88.

The Circumstances of Exposure to Death, a semistructured interview, was used to assess the offspring’s experience surrounding and following the death of the parent (12) . A 28-item version of the Inventory of Complicated Grief modified for children and adolescents was used to assess complicated grief (13), and the adult version was used for adult offspring (14) .

Severity of depression, anxiety, PTSD, and suicidal ideation was assessed in the offspring by using the Mood and Feelings Questionnaire (15), the Screen for Child Anxiety-Related Emotional Disorders (16), the Child PTSD Symptom Scale Interview (17), and the Suicidal Ideation Questionnaire-JR (18) . Parallel measures in adults were obtained by using the Beck Depression Inventory (19), the Beck Anxiety Inventory (20), the PTSD Symptom Scale Interview (21), and the Suicidal Ideation Questionnaire (22) . Intercurrent life events were assessed by using the Life Events Checklist (23) and a shortened version of the Holmes-Rahe Social Readjustment Rating Scale (24), in offspring under and over 18, respectively.

Family cohesion, social support, coping, and self-esteem in youth were assessed by means of the Family Adaptability and Cohesion Evaluation Scales-II (25), the Survey of Children’s Social Support (26), the Kidcope (27), and the self-esteem subscale of the Weinberger Adjustment Inventory (unpublished 1987 work of Weinberger et al.), respectively. Social support and coping were assessed in adult offspring by using the Multidimensional Scale of Perceived Social Support (28) and the Ways of Coping questionnaire (29), respectively.

We compared the bereaved and comparison groups on rates of current disorders by using either the chi-square statistic or Fisher’s exact test. Difference in the severity of a disorder was examined with analysis of variance. When different measures for the same domains were used for adults and children, the data were z-transformed. Unless stated otherwise, df=1. Bonferroni’s correction for multiple contrasts was used as follows: for current psychiatric disorders, with seven categories, α=0.05/7=0.007; for current symptoms, with six measures each, α=0.05/6=0.008. Post hoc pairwise contrasts among the four groups overall and among the three bereaved groups were also protected for the number of contrasts (α=0.05/6=0.008 and α=0.05/3=0.017, respectively).

The cumulative incidence of depression was compared in the four groups by using Kaplan-Meier survival analysis. The correlates of depression occurring between 9 and 21 months were examined by means of standard parametric univariate statistics. Logistic regression, after application of appropriate data imputations, was used to identify the most parsimonious set of variables that predicted depression occurring during the 9–21-month follow-up period. We included a clustering term since multiple offspring from the same family often participated. In this study group, only 10% of the participants had missing data for more than four variables that were significantly associated with follow-up depression; thus, the missing-at-random assumption is reasonable. For statistical analyses with missing data, we used the general approach of Little and Rubin (30), applying multiple imputation by chained equations (MICE), implemented in STATA version 10.0 (Stata Corp, College Station, Tex.), which makes use of auxiliary variables that are correlates of being missing and/or the variables of interest (31) . Finally, path analyses were conducted by using MPlus version 4.1 (Muthén & Muthén, Los Angeles), beginning with the variables selected by logistic regression to identify possible causal pathways between bereavement and subsequent depression in offspring (32) .

In relation to the nonbereaved comparison group, the total group of bereaved offspring had higher rates of major depression (10.2% versus 2.4%; p<0.007, Fisher’s exact test) and alcohol or substance abuse (4.5% versus 0.0%; p=0.008, Fisher’s exact test). As shown in Table 2, post hoc contrasts showed that the offspring of parents who died by suicide and offspring of those who died in accidents showed higher rates of major depression than did the comparison subjects, and the children of suicide victims showed a higher rate of alcohol or substance abuse disorders than did the comparison subjects. Overall differences were found for severity of anxiety symptoms and level of functioning ( Table 2 ). Post hoc comparisons showed that anxiety was higher in youth whose parents died through accidents than in the comparison subjects and functional impairment was greater in the offspring of parents who died by accidents or by sudden natural death than in the comparison subjects. There were also group differences with respect to scores for complicated grief; the post hoc contrasts indicated that the level of complicated grief was higher in youth whose parents died through accidents than in those whose parents died by sudden natural death.

The incidence rate ratio for depression was greater in the bereaved than in the comparison offspring; the incidence rate ratio was 2.7, with a 95% confidence interval (CI) of 1.6 to 4.8. There were significant differences between each of the three bereaved groups and the comparison subjects (incidence rate ratios, 2.1–3.4) but with no significant differences among the three bereaved groups ( Figure 1 ). Bereavement had a strong effect on the incidence of depression in the 9 months following the parent’s death (23.9% versus 5.4%; χ ² =23.3, df=1, p<0.0001) but not during the subsequent year (8.5% versus 7.1%; p=0.69, Fisher’s exact test). The post hoc contrasts showed that the rate of incident depression between the 9- and 21-month assessments was higher in the offspring who lost a parent to suicide (nine of 53, 17.0%) than in those whose parents died by sudden natural death (two of 79, 2.5%) (p=0.007, Fisher’s exact test). The rates of depression remission observed in those who became depressed during the first 9 months after the death were similar in the bereaved (30 of 42, 71.4%) and comparison (six of nine, 66.7%) groups (p=1.00, Fisher’s exact test).

The incidence of PTSD was higher in the bereaved group than in the comparison group during the first 9 months after the death (8.5% versus 0.0%; p<0.0001, Fisher’s exact test) but not during the second year of follow-up (2.8% versus 0.0%; p=0.06, Fisher’s exact test). Of the 15 individuals with incident PTSD within 9 months of the death, all but one experienced a remission. During the second year, two of the five who experienced incident PTSD had remissions.

At some point during the second year of observation (from 9 to 21 months after the death), 35 offspring had a depressive episode, 22 bereaved youth and 13 comparison subjects. These depressive episodes (“follow-up depression”) included both those that were ongoing from the time of the 9-month assessment (N=8) as well as those that were incident during this time interval (N=27). Correlates of follow-up depression were female gender, older age, a previous history of anxiety or depression, and a history of sexual abuse ( Table 3 ). Characteristics assessed at the 9-month assessment that were associated with follow-up depression were greater functional impairment; higher levels of self-reported depression, anxiety, PTSD, suicidal ideation, and impulsive aggression; negative coping; greater number of stressful life events; lower family cohesion; and lower perceived social support. Many of these same variables assessed at the 21-month follow-up were also associated with a depressive episode during the follow-up period. Having had an episode of depression during the first 9 months was predictive of depression in the following year. Depression at 21 months was more likely to be a persistence or recurrence from the first observation period in the bereaved group (12 of 176, 6.8%) than in the nonbereaved offspring (three of 168, 1.8%) (p=0.03, Fisher’s exact test).

The best-fitting, most parsimonious model that predicted the occurrence of depression between 9 and 21 months included previous history of a depressive episode (odds ratio=8.6, 95% CI=1.7 to 43.8), self-reported anxiety at 9 months (odds ratio=1.6, 95% CI=1.1 to 2.4), and negative life events at 9 months (odds ratio=1.9, 95% CI=1.2 to 2.9) and 21 months (odds ratio=1.7, 95% CI=1.2 to 2.5). Compared to the other bereaved and nonbereaved groups, the offspring of parents who died sudden natural deaths had a lower risk of depression (odds ratio=0.01, 95% CI=6×10 ^–5 to 0.99). There was an interaction between bereavement and previous history of depression, meaning that a previous history of depression conveyed an even greater risk for subsequent depression in the nonbereaved group (odds ratio=23.5, 95% CI=1.6 to 350.7). In this regression, the diagnosis of bipolar disorder in the proband, which was an important correlate of incident depression within the first 9 months after parental loss, was associated with a threefold (odds ration=3.15, 95% CI=0.98 to 10.2) increased risk of depression, although it was not statistically significant (p=0.06).

Within the bereaved group, depression between 9 and 21 months was associated with the loss of a mother rather than a father (41.9% versus 13.7%; p=0.004, Fisher’s exact test) and a previous history of any psychiatric disorder (57.1% versus 32.9%; χ ² =4.7, df=1, p=0.03). The risk of follow-up depression tended to be higher in those whose parent died by suicide than in those with parental bereavement due to natural death (18.9% versus 1.3%) (p=0.05, Fisher’s exact test). Higher scores for complicated grief at 9 months and lower self-esteem scores (mean=13.0, SD=5.1, versus mean=18.6, SD=7.0; t=–3.5, df=21.8, p=0.002) were associated with subsequent depression, as were a similar set of measures of self-reported symptoms and psychosocial characteristics at both 9 and 21 months, as noted already. Those who developed depression in the first 9 months after the death were more likely to have an incident depressive episode between 9 and 21 months (54.5%) than were those without depression in the first 9 months (19.6%) (χ ² =12.9, df=1, p<0.001). Those with PTSD during the first 9 months were also more likely to develop depression during the next assessment period (18.2% versus 5.2%; p=0.05, Fisher’s exact test). Risk for depression during the second year of follow-up was also higher among bereaved offspring whose last conversation with the deceased was discordant (26.3% versus 6.1%; p=0.02, Fisher’s exact test) or who felt that others were accountable for the death (47.6% versus 21.6%; χ ² =6.6, df=1, p=0.01).

Logistic regression identified the following as the most parsimonious, best-fitting set of predictors of depression among bereaved offspring between 9 and 21 months: baseline low self-esteem (odds ratio=1.16, 95% CI=1.05 to 1.28) and high negative coping (odds ratio=1.34, 95% CI=1.03 to 1.75) at 9 months. With the two variables in the model, parental bereavement through suicide was, as predicted, associated with a higher risk for depression between the two assessments, relative to accidents (odds ratio=2.9, 95% CI=0.83 to 10.0) and sudden natural death (odds ratio=3.5, 95% CI=0.95 to 12.8), but these findings escaped statistical significance.

A good-fitting path model predicting depression between 9 and 21 months was developed by using variables selected by logistic regression (chi-square test of model fit=10.4, df=8, p=0.24; comparative fit index=0.98; Tucker-Lewis index=0.96; root mean square residual=0.03; weighted root mean square residual=0.58). In this model, bereavement had an indirect effect on depression at 21 months, mediated by the relationship between bereavement and incident depression within the first 9 months of depression ( Figure 2 ). A previous history of depression, antedating the period of observation, was related to depression at 21 months through two pathways: 1) history of depression was related to negative life events, which in turn increased the risk of depression between 9 and 21 months, and 2) history of depression was related to a higher score for self-reported anxiety symptoms at 9 months, which was associated with incident depression at 9 months, which then increased the risk for depression during the subsequent year.