Six-Year Follow-up of a Preventive Intervention for Parentally Bereaved Youths: A Randomized Controlled Trial

Objective  To evaluate the efficacy of the Family Bereavement Program (FBP) to prevent mental health problems in parentally bereaved youths and their parents 6 years later.

Design  Randomized controlled trial.

Setting  Arizona State University Prevention Research Center from November 2002 to July 2005.

Participants  Two hundred eighteen bereaved youths (89.34% of 244 enrolled in the trial 6 years earlier) and 113 spousally bereaved parents.

Interventions  The FBP includes 12 group sessions for caregivers and youths; the literature control (LC) condition includes bereavement books for youths and caregivers.

Main Outcome Measures  Comparisons of youths in the FBP and LC on a measure of mental disorder diagnosis, 5 measures of mental health problems, and 4 measures of competent functioning; and comparisons of spousally bereaved parents on 2 measures of mental health problems.

Results  Youths in the FBP as compared with those in the LC had significantly lower externalizing problems as reported by caregivers and youths (adjusted mean, −0.06 vs 0.13, respectively; P = .02) and on teacher reports of externalizing problems (adjusted mean, 52.69 vs 56.27, respectively; P = .001) and internalizing problems (adjusted mean, 47.29 vs 56.27, respectively; P = .002), and they had higher self-esteem (adjusted mean, 33.93 vs 31.91, respectively; P = .005). Parents in the FBP had lower depression scores than those in the LC (adjusted mean, 5.48 vs 7.83, respectively; P = .04). A significant moderated program effect indicated that for youths with lower baseline problems, the rate of diagnosed mental disorder was lower for those in the FBP than in the LC.

Conclusion  This study demonstrates efficacy of the FBP to reduce mental health problems of bereaved youths and their parents 6 years later.

Trial Registration  clinicaltrials.gov Identifier: NCT01008189

Approximately 3.4% of youths in the United States experience parental death.¹ Research has found that parental death increases children's short-term and long-term risk for mental health problems² and their parents' risk for depression.³ Given these negative outcomes, the development of effective preventive interventions has high public health significance. However, there is no empirical support from randomized trials for the efficacy of programs for parentally bereaved children. A recent meta-analysis of 13 controlled trials found a nonsignificant overall effect size.⁴ Further, no studies assessed program effects longer than 1 year after program completion and few studies have addressed subgroup differences in benefits, issues of critical importance for prevention research.⁵ Although several reviews have proposed that only survivors of traumatic deaths and those with elevated levels of symptoms would benefit,^4,6 there is limited evidence on this issue for bereaved youths.

This study presents findings from a 6-year follow-up of a randomized experimental trial of the Family Bereavement Program (FBP), a brief dual-component intervention designed to improve outcomes for parentally bereaved youths and their parents. The FBP was designed to prevent negative outcomes by changing multiple, empirically supported risk and protective factors.⁷ The FBP demonstrated significant benefits at 11-month follow-up to reduce mental health problems for girls and for youths with greater mental health problems at program entry⁸ and showed significantly reduced caregiver mental health problems at the posttest.

It was hypothesized that at the 6-year follow-up, the FBP as compared with a literature control (LC) would lead to lower prevalence of mental disorders, lower levels of mental health problems, and higher levels of positive developmental outcomes of youths as well as reduce mental health problems of their spousally bereaved parents. Although there is no evidence that bereaved children are at elevated risk for high-risk sexual behavior and substance use, we examined these as secondary outcomes because of their developmental significance. Based on the findings from the FBP⁸ and a meta-analysis of interventions for bereaved individuals,⁹ we hypothesized that the benefits would be greater for those with higher initial levels of mental health problems, females, and those whose parents experienced a violent death. We also explored whether the FBP had differential effects across youths' age.

Participants were recruited by mail solicitation, presentations to agencies, and media presentations. Referred individuals who met eligibility criteria assessed by telephone were invited to participate in in-home recruitment visits. Eligibility criteria were the following: (1) death of a biological parent or parent figure; (2) death occurred between 4 and 30 months prior to the program; (3) at least 1 youth was aged between 8 and 16 years; (4) at least 1 youth and 1 caregiver (the term caregiver is used to refer to surviving parents and others in the parental role) were willing to be randomly assigned to the group or self-study program and participate in assessments; (5) the caregiver and youth could complete the assessments in English; (6) neither the caregiver nor the youth was currently receiving mental health or bereavement services; (7) the youth was not in a special class for mentally challenged persons; and (8) the family planned to stay in the area for the next 6 months.

Families were excluded and referred for treatment if the youth or caregiver endorsed suicidal ideation that included an intent or plan or met diagnostic criteria for mental disorders that might interfere with participation in the program (ie, major depressive disorder for caregivers; conduct disorder, oppositional defiant disorder, or attention-deficit/hyperactivity disorder not being treated with medication for youths assessed by the Diagnostic Interview Schedule for Children–Child Informant [DISC-C] and the Diagnostic Interview Schedule for Children–Parent Informant [DISC-P]^10-13).

Following the pretest, eligible families were randomly assigned to the FBP or LC condition by a research assistant using a computer-generated algorithm, with a greater proportion of families being assigned to the FBP to ensure that the groups would be clinically viable. Enrollment occurred across 6 cohorts between 1996 and 1998. Groups ranged from 5 to 9 individuals (caregivers: mean [SD], 8.1 [1.11]; children: mean [SD], 6.9 [1.04]; and adolescents: mean [SD], 7.0 [1.50]).

As shown in the Figure, of the 617 families referred, 156 (41.16% of those invited) were eligible and randomly assigned to the FBP (n = 90 families, 135 youths) or LC (n = 66 families, 109 youths). Using data from state death certificates, sample representativeness was assessed by comparing ethnicity, sex, and cause of death in the study sample with those of deaths of individuals aged between 28 and 58 years (range that included 90.00% of the study families) in the county where the program was conducted.¹⁴ No significant differences were found in ethnicity (χ²₅ = 1.47, n = 148; P = .91), sex (χ²₁ = 0.17, n = 153; P = .69), or cause of death (χ²₂ = 1.28, n = 153; P = .53).

At the 6-year follow-up, data were collected for 218 adolescents or young adults in 140 families, 89.34% of youths and 89.74% of families randomly assigned to a condition. Interviews were conducted with 209 youths and 143 caregivers; caregivers of 9 youths who did not participate were interviewed, so data were available for 218 youths. Of the caregivers, 113 were spousally bereaved. The rate of follow-up did not differ significantly across the FBP (86.67%) and LC (93.94%) (χ²₁ = 0.12, n = 156; P = .73).

The study was approved by the institutional review board at Arizona State University. Prior to the interview, participants aged 18 years and older signed informed consent forms; those younger than 18 years signed assent forms.

There were 4 assessments: pretest, posttest, and 11 months and 6 years after the posttest. Data for this study were collected 6 years after the posttest (November 2002 to July 2005). Family members were interviewed (usually in their homes) by separate trained interviewers. Interviewers were kept blind to experimental group assignment. Responses to questions after the interviews indicated that 96.50% of the interviewers were blind to assignment.

The FBP caregiver component focused on strengthening caregiver-youth relationship quality and effective discipline and on decreasing caregivers' mental health problems and youths' exposure to negative events.⁷ The child and adolescent components taught developmentally appropriate skills to strengthen the youths' relationship(s) with their caregiver(s), increase adaptive beliefs about why stressors occur and positive coping, and decrease negative thoughts about stressors and the inhibition of emotional expression.

Led by 2 master’s-level clinicians, groups met weekly for 12 two-hour sessions. Four sessions included conjoint activities for youths and caregivers. Two individual meetings were held with each family to review their use of program skills. Process evaluation data indicated high fidelity of program implementation. More detailed information about implementation is available in previous publications.^7,8

Caregivers, children, and adolescents in the LC each received 3 books about grief and a syllabus to guide their reading at 1-month intervals. Of the caregivers, 42.37% reported that they read at least half the books; 70.91% and 38.46% of the children and adolescents, respectively, reported that they read at least half the books.

Mental disorder in the past year was assessed using the computer-assisted caregiver, adolescent, and young adult versions of the DISC (child and adolescent scoring algorithm version N of the Parent or Youth DISC-IV [Diagnostic Interview Schedule for Children] and young adult version C of Parent or Young Adult YADISC-IV [Young Adult Diagnostic Interview Schedule for Children]).¹⁵ Diagnosis was based on meeting symptom criteria according to caregiver or youth/young adult report and meeting criterion D for intermediate or severe impairment. An overall dichotomous score for mental disorder or substance abuse disorder was calculated.

The broadband dimensions of internalizing problems and externalizing problems during the past 6 months were assessed using the Child Behavior Checklist (CBCL)¹⁶ and Youth Self-report (YSR)¹⁶ for youths younger than 18 years. The Young Adult Behavior Checklist (YABCL)¹⁷ and Young Adult Self-report (YASR)¹⁷ were used for youths aged 18 years or older. Because the measures for adolescents and young adults are not identical, we applied item response theory to conduct an equating transformation that selected conceptually equivalent items and put the scale scores on a common metric¹⁸ using a large data set obtained from Achenbach (Thomas M. Achenbach, PhD, unpublished raw data from the CBCL, YABCL, YSR, and YASR, 2003) that contained self- and parent-report scores (n = 800) on the CBCL/YABCL and YSR/YASR. For internalizing problems, the resulting 23-item CBCL, 19-item YABCL, 22-item YSR, and 22-item YASR subscales had Cronbach α (indicating internal consistency) of .86, .90, .90, and .88, respectively. For externalizing problems, the resulting 35-item CBCL, 34-item YABCL, 32-item YSR, and 27-item YASR subscales had α of .92, .93, .88, and .87, respectively.

Using the DISC, dichotomous scores were calculated to indicate whether the criterion was met for any broadband externalizing disorder and any internalizing disorder. The modules for internalizing and externalizing problems were identified using consensus classification of 3 clinicians (I.S., T.S.A., and S.W.). Also, a continuous score of total internalizing symptoms and externalizing symptoms was derived as the number of symptoms reported by either the caregiver or adolescent/young adult on the modules classified as internalizing or externalizing problems.

To reduce the number of measures, a 2-dimensional multimethod, multirater measurement model of internalizing problems and externalizing problems as latent factors was tested with confirmatory factor analysis using scores from the DISC symptom scores, with CBCL/YABCL and YSR/YASR scores as manifest variables. We added correlations between the residuals of caregiver reports of CBCL/YABCL internalizing problems and externalizing problems and between the residuals of adolescent/young adult reports of YSR/YASR internalizing problems and externalizing problems to account for shared method variance due to common reporters. The fit of the 2-factor model was adequate (χ²₆ = 18.28, n = 218; P = .006; comparative fit index = 0.98; root mean square error of approximation = 0.09; standardized root mean square residual = 0.06) and all loadings were 0.55 or greater. Composite scores were constructed by summing the standardized scores of internalizing and externalizing problems, weighted by the standardized factor loadings.

For youths in junior high or high school, teachers reported internalizing problems (α = .90) and externalizing problems (α = .90) (2001 Teacher Report Form¹⁶). Of the 122 youths enrolled in school, 117 Teacher Report Forms were collected (95.90%). The cutoff score¹⁶ of T ≥ 60 was used to identify participants who were above the marginal clinical level.

Adolescents/young adults completed the Rosenberg Self-esteem scale¹⁹ (α = .89).

Adolescent/young adult report of competence during the past month was assessed with the 6-item academic (α = . 87) and 7-item peer relationship (α = .62) subscales of the Coatsworth Competence Scale.²⁰ Data on academic competence were obtained for 154 of 159 adolescents/young adults (96.86%) who were in school when interviewed.

Grade point average was calculated based on the mean grade over the last 2 semesters in English, math, science, and social studies using transcripts of 114 of the 123 youths (92.68%) who had been in high school at least 1 year before the interview.

Items from the Monitoring the Future Scale²¹ were used to assess alcohol and drug use in the past year (7-point scale of times used [a score of 1 indicates that alcohol and drugs were used 0 times; a score of 7, they were used 40 times]). Polydrug use was assessed by counting the number of different drugs used. Number of sexual partners in the past year was assessed by self-report. To maximize validity of responses, items on drug use and sexual partners were self-administered.

Nonspecific psychiatric distress was assessed using the Psychiatric Epidemiology Research Interview²² (27 items; α = .93). Depression symptoms were assessed with the Beck Depression Inventory²³ (21 items; α = .90). Beck Depression Inventory scores of 10 or higher indicate moderate or greater levels of clinical depression.²⁴

The following variables assessed at baseline were used as covariates for the analyses of the same measures at 6-year follow-up (α at baseline is provided): academic competence (α = .90), peer relationship competence (α = .77),¹⁹ parent mental health problems (Beck Depression Inventory scores,²³ α = .95; Psychiatric Epidemiology Research Interview scores,²² α = .93), and teacher report of externalizing and internalizing problems (1991 version of the Teacher Report Form,¹¹ α = .92 and .87, respectively). The baseline measure of self-esteem was the global self-worth subscale of the Self-perception Profile for Children²⁵ (α = .79). For measures not assessed at baseline (eg, diagnosis of mental disorder), composite baseline scores of internalizing and externalizing problems were used as covariates using the mean of the standardized scores on caregiver and youth report measures. For this measure, baseline caregiver and youth reports of externalizing problems were assessed with the 1991 CBCL²⁶ (α = .87) and YSR²⁷ (α = .87), and internalizing problems were assessed with the 1991 CBCL²⁶ (α = .87), Children's Depression Inventory²⁸ (α = .87), and Revised Children's Manifest Anxiety Scale²⁹ (α = .90).

Attrition effects and group × attrition effects in relation to the baseline measures were evaluated either using 2 (assessed or not assessed at 6-year follow-up) × 2 (FBP or LC) analyses of variance for continuous measures or using χ² test or logistic regression for discrete measures.³⁰

Intervention main effects were examined with analysis of covariance for continuous measures, with logistic regression for binary variables, and with Poisson analysis for count scores, controlling for the corresponding baseline covariate. In addition, we examined the interactions between group and 5 potential moderators of program effects: (1) baseline composite scores of internalizing and externalizing problems; (2) sex; (3) youth's age; (4) time since death; and (5) cause of death (dichotomized as violent death [accident, suicide, homicide] vs death from illness⁶). All analyses used an intent-to-treat approach³¹ based on full-information maximum likelihood estimation for handling missing data,²² making use of all available data for all individuals. To adjust for multiple testing, the false discovery rate, which controls for the expected proportion of false-positives among all significant hypotheses, was applied³² for tests of effects on primary (adolescent/young adult mental health and competence; parent mental health) and secondary (high-risk behaviors) outcomes separately.

All analyses on adolescent/young adult outcomes were done recognizing the clustering due to multiple offspring within families.³³ We used Mplus software³⁴ for the analyses with continuous, categorical, and count outcome variables. Hypothesis tests were conducted using 2-tailed α = .05. Optimal design³⁵ was applied for estimating power in clustered data. Assuming a correlation of 0.30 between baseline and follow-up measures, the power of the test for group differences for the current sample was greater than 0.80 for detecting small to medium effects depending on the size of the intraclass correlation. Power analyses for dichotomous measures (ie, diagnoses) assumed a base rate of 30% in the control group and that the intervention would reduce this to 10% (a reduction that is a medium effect size according to Cohen³⁶). The power of the test for this reduction was 0.82 to 0.95 depending on the intraclass correlation. The intraclass correlations within families for baseline measures ranged from 0.00 to 0.62, with a mean of 0.22.

Data were collected from 218 adolescents/young adults from 140 families (102 adolescents/young adults from 62 families in the LC; 116 adolescents/young adults from 78 families in the FBP). Of the adolescents/young adults, 53.67% were male. The mean (SD) age was 17.64 (2.42) years (range, 14-23 years). Ethnicity was 67.89% white, non-Hispanic; 14.22% Hispanic; 6.42% African American; 3.67% Native American; 1.38% Asian/Pacific Islander; and 6.42% other. Of the 140 caregivers who were interviewed, 77.86% were female. The mean (SD) age of the spousally bereaved caregivers was 47.79 (6.97) years (range, 34-64 years). The mean (SD) annual income at follow-up was $30 466 ($19 199) (range, $1050-$124 800). Parental death occurred within a mean (SD) of 10.81 (6.35) months (range, 3-29 months) prior to the baseline assessment. Cause of death was natural cause (ie, illness) in 73.38% and violent death in 26.62% (accident, 16.03%; homicide, 2.68%; suicide, 7.91%). Table 1 compares the FBP and LC groups on the demographic variables and baseline covariates. Of the 25 comparisons, 1 significant difference occurred: the percentage of non-Hispanic white participants was lower in the FBP than in the LC. No significant attrition or attrition × group effects were found on the 25 baseline continuous or demographic measures.³⁰

As shown in Table 2, the main effect on the rate of mental disorder was not significant. However, a significant program × baseline mental health problem interaction was found. Post hoc analyses indicated that the program benefit was stronger for those with lower baseline problems.

The FBP participants had significantly lower levels of externalizing problems on the composite of caregiver and youth report measures (adjusted mean, −0.06 vs 0.13, respectively; P = .02; Cohen d = 0.31) and on the teacher report measures (adjusted mean, 52.69 vs 56.27, respectively; P = .001; Cohen d = 0.59) of externalizing problems. Those in the FBP had a marginally significantly lower rate of externalizing disorder diagnosis compared with those in the LC after adjusting for multiple tests (rate, 15.45% [95% confidence interval (CI), 8.70%-22.20%] vs 27.37% [95% CI, 18.40%-36.34%], respectively; adjusted odds ratio = 1.57 [95% CI, 1.01-2.45]; number needed to treat³⁷ = 8.97; P = .09). No significant difference was found for the internalizing disorder diagnosis on the composite of caregiver and adolescent/young adult report measures of internalizing problems. Teacher reports of internalizing problems were significantly lower for those in the FBP than those in the LC (adjusted mean, 47.29 vs 56.27, respectively; P = .002; Cohen d = 0.57). Fewer adolescents in the FBP than in the LC were above the marginal clinical cut point for teacher report of internalizing problems (rate, 5.00% [95% CI, 0.00%-10.51%] vs 19.30% [95% CI, 9.50%-29.55%], respectively; adjusted odds ratio = 2.22 [95% CI, 1.56-3.13]; number needed to treat = 6.99; P = .02). Adolescents/young adults in the FBP had significantly higher self-esteem than those in the LC (adjusted mean, 33.93 vs 31.91, respectively; P = .005; Cohen d = 0.40). There were no significant interactions with any of the predicted moderators after adjusting for multiple tests. No significant program effects were found for the secondary outcomes of risky behaviors.

Parents in the FBP had lower scores than those in the LC on the Psychiatric Epidemiology Research Interview demoralization scale²² (adjusted mean, 1.84 vs 2.04, respectively; P = .03, Cohen d = 0.42) and Beck Depression Inventory (adjusted mean, 5.48 vs 7.83, respectively; P = .04, Cohen d = 0.40). Also, parents in the FBP were less likely than those in the LC to be above the cut point for moderate or higher levels of depression³⁸ (rate, 21.31% [95% CI, 11.03%-31.59%] vs 34.62% [95% CI, 21.69%-47.55%], respectively; adjusted odds ratio = 1.69 [95% CI, 1.00-2.22]; number needed to treat = 7.51; P = .05).

To our knowledge, this is the first randomized trial to report significant long-term benefits of a preventive intervention for parentally bereaved youths. Program benefits included lower levels of mental health problems (particularly externalizing problems), improved self-esteem, and reduced rates of diagnosed mental disorder for adolescents/young adults who had lower levels of mental health problems at program entry. Also, the FBP significantly reduced mental health problems of spousally bereaved parents.

The convergence of findings across teachers (who were blinded to the program conditions), caregivers, and adolescents/young adults increases confidence in the program benefit to reduce externalizing problems. Because externalizing problems in adolescence and young adulthood are associated with a wide array of social adaptation and mental health problems later in development,³⁹ the current findings have significant public health implications. The program effect on teacher report of internalizing problems is notable because it included reductions in the level of problems and likelihood of exceeding the clinical cut point (odds ratio = 4.48). The program effect on self-esteem is meaningful given that poor self-esteem is a risk factor for depressive disorders in young adulthood.⁴⁰ These findings complement previously reported effects of the FBP to reduce problematic levels of grief and cortisol dysregulation at 6-year follow-up.^41,42

Nearly all of the significant program effects were main effects, which does not support the hypothesis^4,6 that bereaved youths with greater symptoms and those who experienced traumatic deaths would benefit most from the FBP. The one significant interaction indicated a reduced rate of diagnosed mental disorder for youths with lower but not higher levels of mental health problems at program entry. This interactive effect raises caution about using the FBP with youths with clinical levels of mental health problems. Such youths likely require more intensive interventions. In light of evidence that young spousally bereaved individuals are at risk for mental health problems many years following the death,³ the effect of the FBP to reduce such problems indicates that the program had a double-prevention effect on parents and children.

Three limitations should be noted. First, the findings should not be generalized to families who were excluded from the trial such as those receiving mental health services. Second, a placebo control or one that provides supportive assistance was not used. Thus, we cannot rule out that the effects are due to nonspecific aspects of the program such as group support. Findings that the program effects to reduce youths' mental health problems at earlier waves were mediated by improvements in positive parenting and child coping⁴³ provide evidence that program components designed to strengthen these factors were at least partially responsible for program effects. Nevertheless, future evaluations should test the FBP against supportive interventions that do not include the skills taught in the FBP. Third, the sample size precluded testing program effects across other subgroups such as ethnic minorities or those experiencing parental suicide.

Correspondence: Irwin Sandler, PhD, Prevention Research Center, Department of Psychology, Arizona State University, PO Box 876005, Tempe, AZ 85287-6005 (irwin.sandler@asu.edu).

Accepted for Publication: April 22, 2010.

Author Contributions: Drs Sandler and Tein had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Sandler, Ayers, Wolchik, Kaplan, and Luecken. Acquisition of data: Ayers, Kaplan, Luecken, and Schoenfelder. Analysis and interpretation of data: Sandler, Ayers, Tein, Wolchik, Millsap, Khoo, Ma, and Coxe. Drafting of the manuscript: Sandler, Ayers, Tein, Wolchik, Millsap, Khoo, Ma, and Schoenfelder. Critical revision of the manuscript for important intellectual content: Sandler, Ayers, Tein, Wolchik, Millsap, Kaplan, Luecken, and Coxe. Statistical analysis: Tein, Millsap, Khoo, Ma, and Coxe. Obtained funding: Sandler, Ayers, Tein, Wolchik, and Luecken. Administrative, technical, and material support: Sandler, Ayers, Kaplan, and Schoenfelder. Study supervision: Ayers and Tein.

Financial Disclosure: None reported.

Funding/Support: This work was supported by grants R01 MH49156 and P30 MH068685 from the National Institute of Mental Health.

Role of the Sponsor: The National Institute of Mental Health had no direct role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.

Additional Contributions: We thank all of the caregivers, children, and adolescents for their participation in this trial. Michelle McConnaughay, BA, Toni Genalo, BS, and Monique Lopez, BA, contributed to data collection and data management, and Janna LeRoy, BS, provided technical assistance.