Developmental Pathways of the Family Bereavement Program to Promote Growth 15 Years After Parental Death

Abstract

Although parental death increases the risks of negative developmental outcomes, some individuals report personal growth, an outcome that has received little attention. We tested a developmental cascade model of post-loss growth in 244 parentally bereaved youth (ages 8–16 at baseline) from 156 families who participated in a randomized controlled trial of a family-based intervention, the Family Bereavement Program (FBP). Using five waves of data, the present study examined the prospective associations between the quality of parenting immediately following the FBP and post-loss growth six and 15 years later, and whether these associations were mediated by changes in intra- and inter-personal factors (mediators) during the initial 11 months following the FBP. The mediators were selected based on the theoretical and empirical literature on post-loss growth in youth. Results showed that improved quality of parenting immediately following the FBP was associated with increased support-seeking behaviors and higher perceived parental warmth at the 11-month follow-up, both of which were related to post-loss growth at the six-year follow-up and 15-year follow-up. No support was found for the other hypothesized mediators that were tested: internalizing problems, intrusive grief thoughts, and coping efficacy. To promote post-loss growth for parentally bereaved youth, bereavement services should target parent-child relationships that help youth feel a sense of parental warmth and acceptance and encourage youth to seek parental support.

As of 2023, approximately 5.8 million U.S. children had experienced the death of a parent by the age of 18 (Judi’s House, 2023). Estimates from the COVID-19 pandemic suggest that rates of parental bereavement among children under the age of 18 in the U.S. increased by 25% (Burns et al., 2023). The vast majority of research on parental bereavement among children has focused on the negative effects, including depression (Berg et al., 2016; Luecken, 2000), prolonged grief (Melhem et al., 2013), posttraumatic stress disorder (Melhem et al., 2008), and drug and alcohol dependency (Giordano et al., 2014; Hamdan et al., 2013). Nevertheless, a few studies have examined positive psychosocial consequences such as post-loss growth (Şimşek Arslan et al., 2020; Vloet et al., 2017). This study uses a secondary dataset to investigate the pathways through which a preventive intervention, the Family Bereavement Program (FBP) led to post-loss growth over a 15-year time period in a sample of parentally bereaved children.

Intervention studies with bereaved youth and their families have provided evidence on effective ways to promote resilience, but to our knowledge, no study has examined whether or how interventions for parentally bereaved youth might promote growth. While many terms such (e.g., adversarial growth, perceived benefits, and posttraumatic growth (PTG)) have been used to refer to growth in response to adversity of many kinds (e.g., illness, natural disaster, war), in this study we will use “post-loss growth” to refer to the positive changes that can result from parental bereavement. Although resilience and post-loss growth are related constructs, they are different; resilience is conceptualized as adaptation or lack of negative outcomes in the context of adversity (Masten, 2001) while post-loss growth is not only a lack of negative outcomes but positive psychological change as a result of adversity (Tedeschi & Calhoun, 2004). Research has found that growth in bereaved youth was uncorrelated with negative outcomes such as complicated grief or depression (e.g., Salloum et al., 2019), suggesting the importance of understanding growth as a unique developmental outcome. Considering the importance of a strength-based perspective and the need to better support parentally bereaved children (Eppler, 2008), the current study investigates a cascade model of growth (Figure 1) using secondary data from a randomized controlled trial of the FBP, one of the few preventive, evidence-based programs for parentally bereaved youth (Bergman et al., 2017).

Figure 1.

Conceptual model.

The cascade model of FBP’s effect on post-loss growth

We drew on Kilmer et al.’s (2014) conceptual model of PTG in children to select mediators that may explain the FBP’s cascading effects on post-loss growth. Kilmer et al.’s model posits that the family environment and how the parent responds to the child regarding the stressful event (e.g., warm versus cold, expressive versus inhibitive) may influence the child’s functioning and intra-personal resources that support or hinder the development of PTG (Kilmer et al., 2014). Specifically, a positive parent-child relationship can bolster the child’s efficacy beliefs, augment the child’s seeking support behaviors from the parent, and increase the parental warmth perceived by the child, all of which affect PTG. Kilmer et al.’s model also suggests that the parent-child relationship impacts both the child’s post-trauma functioning (e.g., internalizing problems) and intrusive grief thoughts, which are both linked to PTG.

According to these assumptions, we selected five mediators of post-loss growth: coping efficacy, support-seeking from the parent, perceived parental warmth, internalizing problems, and intrusive grief thoughts. Our cascade model (Figure 1) begins with the previously demonstrated effect of the FBP to increase the quality of parenting at post-test (path a; Sandler et al., 2003). Consistent with Kilmer et al.’s assumptions, we hypothesized that FBP-induced improvements in the quality of parenting at post-test would be associated with changes in the five mediators at the 11-month follow-up (b1 path). In addition, the five mediators were expected to relate to post-loss growth at six- and 15-year follow-ups (b2 & c5 paths). Our cascade model also posited that there would be continuity in post-loss growth from the six-year follow-up to the 15-year follow-up (b3 path). Below we discuss the empirical support for the hypothesized mediating pathways.

Linking quality of parenting to the 11-month mediators (b1 path)

Although others have demonstrated that the FBP improved children’s positive coping, a general composite of active coping and coping self-efficacy (Sandler et al., 2003), no study has examined whether the FBP had an indirect effect to improve coping efficacy specifically via its effect on the quality of parenting. One study found that improvements in the quality of parenting that resulted from participation in a parenting-focused preventative intervention for divorced families led to more coping efficacy among children (Vélez et al., 2011). Other cross-sectional examinations have found significant, positive associations between the quality of parenting and children’s coping efficacy (Smith et al., 2006; Wolchik et al., 2008). Given these prior findings and Kilmer et al. (2014)’s assumptions, we hypothesized that FBP-induced improvements in the quality of parenting at post-test would be associated with greater coping efficacy at the 11-month follow-up, controlling for baseline coping efficacy (a path). Moreover, we hypothesized that FBP-induced improvements in the quality of parenting would be associated with greater support-seeking from a parent and perceived parental warmth (Ray et al., 2013; Smith et al., 2006). Previous studies showed that the FBP reduced internalizing problems at 11-month follow-up and this effect was mediated by the quality of parenting at post-test (Tein et al., 2006). Additionally, child-centered parenting (e.g., parental warmth and engagement, open communication, and provision of a stable environment) is associated with better psychological outcomes among bereaved youth (Howell et al., 2015; Lin et al., 2004; Saldinger et al., 2004). As such, we hypothesized that higher quality parenting at post-test would be associated with less internalizing problems at the 11-month follow-up. Finally, although the FBP has been shown to reduce intrusive grief thoughts, the mediating influence of the quality of parenting has not been examined (Sandler et al., 2010). Wolchik et al. (2008) found that quality of parenting assessed at baseline was negatively related to intrusive grief thoughts at 11-month follow-up. We hypothesized that FBP-induced improvements in the quality of parenting at post-test would be associated with less intrusive grief thoughts at the 11-month follow-up.

Linking the mediators to post-loss growth (b2 & c5 paths)

Few studies have explicitly examined the relation between coping efficacy and post-loss growth in children. In one of the studies, Cryder et al. (2006) found that, among children displaced by Hurricane Floyd in North Carolina, those who indicated greater perceived efficacy to adjust in the face of stress also reported greater post-loss growth. Self-reported efficacy beliefs were also positively associated with the provision of a supportive, nonjudgmental environment by parents (Cryder et al., 2006). These results are in line with Kilmer et al. (2014)’s assumptions. Therefore, we hypothesized that greater coping efficacy at 11-month follow-up would be related to greater post-loss growth (b2 and c5 paths).

Studies support the notion that seeking support from a parent can facilitate children’s later loss-related growth. Wolchik et al. (2009) found that in a sample from the control group of the current FBP sample, seeking support from adults, especially from parents, was related to greater levels of post-loss growth in children six years later. Other studies have found similar associations, including family social support’s association with post-loss growth among middle schoolers affected by an earthquake in China (Zhou et al., 2017) and general social support’s association with post-loss growth among adolescents living in a war-afflicted Israeli community (Kimhi et al., 2009). As discussed above, Kilmer et al. (2014) emphasized the role of parental support in facilitating PTG. Thus, we hypothesized that support-seeking from the parent, assessed at 11-month follow-up, would be associated with post-loss growth (b2 and c5 paths).

Minimal research has investigated the link between parental warmth and children’s reports of growth following a traumatic event. A retrospective study of nearly 4,000 U.S. adults found that parental warmth during childhood was associated with positive relations with others and personal growth a decade later (Chen et al., 2019). Another study of flood survivors found a positive relation between strength-based parenting and post-loss growth (Zavala et al., 2022). Informed by these findings and Kilmer et al. (2014)’s assumptions, we hypothesized that parental warmth at 11-month follow-up would be related to post-loss growth (b2 and c5 paths).

We also explored two additional potential mediators at the 11-month follow-up without explicit hypotheses –internalizing problems and intrusive grief thoughts. The literature consistently supports that quality of parenting would be associated with decreases in internalizing problems and intrusive grief thoughts (Tein et al., 2006; Wolchik et al., 2008), i.e., the b1 path in our model. However, Kilmer et al. (2014) posited that more intrusive grief thoughts and internalizing problems are directly related to greater post-loss growth (b2 path). Empirical evidence on this link is scant and mixed. Cross-sectional studies found that higher internalizing problems were associated with lower post-loss growth in children (Milam et al., 2005; Phipps et al., 2007; Vaughn et al., 2009), while a longitudinal analysis of parentally bereaved children (Wolchik et al., 2009) found that more internalizing problems were related to greater post-loss growth six years later. The variable of intrusive grief thoughts has scarcely been examined in relation to post-loss growth in children. A couple of studies found that greater intrusive thoughts among children after exposure to a natural disaster were positively related to post-loss growth one year later (Andrades et al., 2021; Kilmer & Gil-Rivas, 2010). Given the scarcity and inconsistent findings of empirical studies that have examined the longitudinal relations of intrusive grief thoughts and internalizing problems to post-loss growth in children, no a priori hypotheses were generated.

The current study

Our cascade model focused on the quality of parenting as the first mediator immediately after the intervention because parenting has been found to mediate many of the effects of the FBP in prior studies (Sandler et al., 2023; Tein et al., 2006; Zhang et al., 2021) and because the Kilmer et al. (2014) conceptual model proposed that positive parent-child relationships can promote post-loss growth in children. Indeed, others’ work consistently emphasized the key influence of caregivers on children’s outcomes after parental death (Alvis et al., 2022; Jiao et al., 2021). Based on prior work (e.g., Tein et al., 2006), we expected that the “a” path from FBP to post-test parenting would be statistically significant.

H1: We hypothesized that FBP-induced improvements in the quality of parenting would be associated with greater post-loss growth at six-year follow-up through increases in three proximal mediators– coping efficacy, perceived parental warmth, and support-seeking from parent (b1 and b2 paths)– after controlling for the direct effect of quality of parenting on post-loss growth (c4). In addition, we explored two additional mediators– internalizing problems and intrusive grief thoughts.

H2: We hypothesized that the cascading effects (H1) on post-loss growth at six-year follow-up, measured by the Relating to Others subscale of the Posttraumatic Growth Inventory (PTGI; Tedeschi & Calhoun, 1996), would extend to post-loss growth at 15-year follow-up, measured by the Personal Growth subscale of the Hogan Grief Reaction Checklist (HGRC; Hogan, 2001) (b3 path) because both measures focus on positive changes as a result of their parental loss. We hypothesized that this path would be significant after accounting for the direct path from the mediators at 11-month follow-up to post-loss growth at 15-years (c5).

Method

We report how our sample size was determined and describe all data exclusions and manipulations as well as all measures in the study.

Participants

The sample in the FBP randomized controlled trial was comprised of 244 children/adolescents (46.7% female) aged 8–16 years (M = 11.39, SD = 2.43) from 156 families and their surviving caregiver at the study baseline. Two-thirds (67%) of the sample identified as non-Hispanic, 16% Hispanic, 7% African American, 3% Native American, 1% Asian or Pacific Island, and 6% other. Sixty-three percent (63%) of participating caregivers were the mother, 21% were the father, and the remaining 16% were adult family caregivers (e.g., grandparent, aunt, etc.; also referred to as parents in this paper). The median family income was between $30,001– $35,000. The primary causes of parental death were illness (67%), accident (20%), and homicide or suicide (13%). The average time since death at the study baseline was 9.77 months (SE = 5.70, range: 3–26).

Procedures

Details of the study procedure and content of the FBP have been discussed in previously published work (e.g., Sandler et al., 2003, 2013) and so will briefly be summarized here. Families were recruited by mail and media advertisement and via local community agencies in a Southwestern metropolitan area in the United States. Parents were screened by phone to determine whether they met the eligibility criteria: a child aged 8–16 years experienced the death of a biological parent between three and 30 months prior, and neither the child nor the surviving parent was currently receiving mental health or bereavement services. If the child or parent endorsed suicidal ideation or met diagnostic criteria for one or more mental disorders that could interfere with participation in the study, the family was excluded and referred for treatment.

All study procedures were approved by the University of Arizona’s Institutional Review Board. Consent and assent (for youth under age 18) were obtained. The baseline assessment was conducted in the family’s home by trained staff members prior to random assignments of families into either the FBP (n = 90 families with 135 children; 55.3%) or a literature control group (n = 66 families with 109 children) using a computer-generated randomization sequence.

In the FBP group, parents and children met separately once a week for 12 2-hour sessions, four of which included joint parent-child activities. The intervention contained a youth component and a parent component. The parent component addressed parenting skills to improve parent-child relationships, increase the use of effective discipline, decrease parents’ psychological distress, and decrease children’s exposure to stressful events after the death. The youth component addressed children’s coping skills, control beliefs, appraisals of stressful events, emotional expression as well as positive parent-child relationships. relationship. Caregivers and children in the literature control group each received three grief books accompanied by a summary of key components in each book at one-month intervals.

All families participated in four additional at-home interviews at post-test (three months after baseline; 98% retention), 11-month follow-up (87% retention), six-year follow-up (87% retention), and 15-year follow-up (80% retention). Participants were paid for participating in all interviews, but not for their participation in the program. At baseline, post-test, and 11-month follow-up, families with one child were compensated with $40 per interview plus $30 for each additional child that participated. Because the six-year follow-up was more time-intensive for participants, each was paid $175 plus $100 for each additional child that participated from the family. At the 15-year follow-up, young adults who completed an interview were compensated with $100. The current study used data from all five time points. The CONSORT diagrams of participant randomization and retention during the 15 years of data collection have been published (Sandler et al., 2003, 2010, 2018).

Measures

Quality of Parenting.

Quality of parenting was assessed based on a multi-method multi-informant measurement model with multiple, highly correlated measures of the parent-child relationship and parental discipline practices (Kwok et al., 2005). Parents and children completed parallel versions of the following five measures to capture the parent-child relationship: the Acceptance subscale (16 items; e.g., “Your caregiver/you enjoyed doing things with you/your child”) and the Rejection subscale (16 items, e.g., “Your caregiver said you are a big problem” or “You said your child was a big problem”) of the Child Report of Parenting Behavior Inventory (CRPBI; Schaefer, 1965), the Dyadic Routines Scale (Wolchik et al., 2000; 7 items, e.g., “Your caregiver/you had some time each day for just talking to you/your child”), and the Stable Positive Events Scale (Sandler et al., 1991; 5 items; e.g., “Household routines got done smoothly”). Parents completed the Talk with Reassurance subscale of the Parents Expression of Emotion Questionnaire, which asks parents to consider and rate the likelihood of them responding to their child in a particular manner (Jones & Twohey, 1998; 6 items, “Reassure [child] that although you are feeling sad, you are dealing with your sadness”). Children completed the Sharing Emotions with Parents (SEP) Scale (Ayers et al., 1998; 10 items, e.g., “Your [parent/guardian] understands your sad feelings”). The Inconsistent Discipline subscale (8 items, e.g., “Your [parent/guardian] soon forgot a rule he/she made” or “You soon forgot a rule you made”) of the CRPBI (Schaefer, 1965) and an adapted version of the Parent Perception Inventory (Hazzard et al., 1983; 8 items, e.g., “How often does your [parent/guardian] thank you for doing things?” or “How often do you thank your child for doing things?”) were each completed by both parents and children to capture consistent discipline. Additionally, parents completed an adapted version of the Follow-Through subscale of the Oregon Discipline Scale (Oregon Social Learning Center, 1991; 6 items, e.g., “How often was your child able to get around the rules you had set?”). The Cronbach’s alphas (αs) of these scales, as applicable, demonstrated satisfactory levels of internal consistency within the sample (αs > .74), and all scales have previously exhibited adequate validity and reliability.

In addition to questionnaires, parent-child relationships were assessed through behavioral observation of parent-child dyads. Two family issues were selected from the Parent Issues Checklist (Prinz et al., 1979) and were discussed by the dyad during the 12-minute, video-recorded family interaction task. Trained coders rated parents’ behaviors, focusing on positive affective tone and attending (backchanneling and head nodding). Interrater reliability was adequate for all three codes (Cohen’s κ ≥ .77). The measurement model (Kwok et al., 2005) showed a good fit to the data: χ2(113) = 207.45 (p < .001), CFI = .93, RMSEA = .06, SRMR = .06. The factor scores were computed from the model and used in the present analyses.

Coping Efficacy.

Children completed a seven-item measure of general coping efficacy (Sandler et al., 2000). Children responded on a four-point scale from 1 to 4 with varying labels to items (e.g., “Overall, how satisfied are you with the way you handled your problems during the last month?”). Items were averaged and z-scored into a composite measure of coping efficacy with higher values indicative of greater efficacy (baseline α = .72; 11-month follow-up α = .79). A confirmatory factor analysis found that the one-dimensional scale fit the data well and demonstrated adequate validity (Sandler et al., 2000).

Support-seeking from parent.

Children completed four items on support-seeking behaviors from parents in a general coping scale developed for the FBP study (Program for Prevention Research, 2006): “You told your caregiver how you felt,” “You asked your caregiver for help figuring out what to do,” “You told your caregiver how you felt about the problem,” and “You asked your caregiver for help in figuring out what to do.” Children reported the frequency with which they did each item on a scale from 1 (never) to 4 (most of the time). Scores were averaged and z-scores were computed. Higher scores are indicative of greater support-seeking from the parent (baseline α = .83; 11-month follow-up α = .88).

Perceived parental warmth.

To assess children’s perceptions of parental warmth, a latent variable was created for the present study using the child-reported SEP Scale and the Acceptance and Rejection subscales of the CRPBI, as described above. These three measures were selected to reflect children’s felt sense of responsivity and warmth from their parent (α = .83-.94). Confirmatory factor analysis (CFA) tested how well these three indicators represent the latent construct of perceived warmth at baseline and 11-month follow-up. In the CFA, two latent constructs (perceived warmth at baseline and at 11-month follow-up) had three indicators at the scale level, and each indicator measure at baseline was correlated with the same measure at 11-month follow-up. The model fit the data well: χ2 (3) = 5.08 (p = .17), CFI = 1.00, RMSEA = .05 (90% CI: 0, .13), SRMR = .02. Factor loadings ranged from .64 - .96. The latent variables were used in the analyses.

Intrusive grief thoughts.

Children completed the nine-item Intrusive Grief Thoughts Scale developed for the FBP study (IGTS; Program for Prevention Research, 2006) (e.g., “How often did you find yourself thinking about how things might have been different if your parent was still alive, even when you didn’t want to think about it?”). On a scale from 1 (several times a day) to 5 (not at all), children indicated the extent to which they experienced each item in the past month. Item scores were reverse-scored, averaged, and z-scored such that higher scores reflected more intrusive grief thoughts (baseline α = .89; 11-month follow-up α = .93). The IGTS has been found to have strong internal consistency and validity (Sandler et al., 2010).

Internalizing problems.

Children’s internalizing problems were assessed based on two measures of anxiety and depression. Children reported on the 28-item Revised Children’s Manifest Anxiety Scale (RCMAS; Reynolds & Richmond, 1978) which assesses chronic anxiety (e.g., “You were afraid of a lot of things”). Children responded with a “yes” if the statement was true or mostly true or “no” if the statement was not true or mostly not true within the past month. Children also reported on the 27-item Children’s Depression Inventory (CDI; Kovacs, 1981) by selecting which statement best described them within the past two weeks. An example of the statements is: “(1) I am sad once in a while,” “(2) I am sad many times,” and “(3) I am sad all the time.” Both scales have demonstrated strong construct validity (Carey et al., 1987; Reynolds, 1980; Reynolds & Richmond, 1979). RCMAS (baseline α = .90; 11-month follow-up α = .93) and CDI items (baseline α = .87; 11-month follow-up α = .85) were averaged and z-scored with higher scores indicating more severe anxiety or depression. Because the RCMAS and CDI composite scores were highly correlated (rs = .65 and .74 at baseline and 11-month follow-up, respectively), they were averaged to generate a score on internalizing problems.

Post-loss growth at six-year follow-up.

Youth completed the 7-item Relating to Others subscale of the PTGI (Tedeschi & Calhoun, 1996) which asks how much positive change they experienced in interpersonal relationships since the death of their parent (from 0 [did not experience this change] to 5 [experienced this change to a very great degree]) (e.g., “You learned a great deal about how wonderful people are”). Items were summed to generate a composite score such that higher scores correspond with greater levels of post-loss growth (α = .91). The subscale has good construct validity (Taku et al., 2008).

Post-loss growth at 15-year follow-up.

The Personal Growth subscale (12 items) of the HGRC (Hogan, 2001) was used to assess post-loss growth at 15-year follow-up (e.g., “I care more deeply for others”). Respondents rated how well each statement describes them on a scale from 1 (does not describe me at all) to 5 (describes me very well) in the past two weeks. Higher scores reflected greater levels of post-loss growth (α = .94). The HGRC has demonstrated strong construct, convergent, and divergent validity (Hogan, 2001).

Baseline Covariates.

We controlled for (1) time since death, (2) parent’s cause of death (two dummy coded variables: accident vs. illness and violent vs. illness), (3) child age (in years), and (4) gender (male = 0, female = 1) on all mediators and dependent variables in the models. The corresponding baseline measure of the post-test and 11-month follow-up measures was also controlled for in the models. Because post-loss growth was not assessed at baseline, we do not have a baseline covariate for the six-year and 15-year follow-up post-loss growth measures. To address this issue, a baseline psychosocial risk index, which is a composite of standardized parent and child reports of children’s externalizing and internalizing problems, was used as a covariate for post-loss growth measures at six- and 15-year follow-ups. Specifically, externalizing problems were measured with the 33 items from the Child Behavior Checklist (CBCL; Achenbach, 2001) completed by the caregiver (e.g., “Gets in many fights”, α = .90) and 30 items from the Youth Self-Report scale (Achenbach, 1997; e.g., “Threatens people”; α =. 86). Internalizing problems were measured with 31 items from the CBCL (e.g., “Cries a lot”; α = .87) and the child-reported 27-item CDI and 28-item RCMAS (as described above).

Statistical Analyses

This study’s hypotheses and plan of analysis were preregistered; see https://doi.org/10.17605/OSF.IO/EH3QJ. In our pre-registration, we stated that we would use a latent variable to assess children’s efficacy beliefs, but because the proposed factors did not adequately load onto a latent factor, we ultimately used just the coping efficacy scale. We also decided to exclude a factor included in the pre-registered parental warmth variable as it was completed by caregivers and not children. The data and analytic code can also be accessed online: https://doi.org/10.17605/OSF.IO/3UNYV.

For hypotheses testing, five path models were estimated in Mplus 8 (Muthén & Muthén, 2017), one for each of the mediators measured at 11-month follow-up (Figure 1). To test the hypothesized cascade effects (i.e., a→b1→b2 and a→b1→b2→b3), we used the joint significance test (Taylor et al., 2008), which has the best balance of Type I error and statistical power compared to other approaches (MacKinnon et al., 2002; Taylor et al., 2008). It suggests that the three-path mediation or the four-path mediation exists when each of the regression coefficients along the mediation pathways is significant (p < .05) or marginally significant (p < .10). In our models, in addition to the direct effect of quality of parenting on post-loss growth (c4) and the direct path from the mediators at 11-month follow-up to post-loss growth at 15-years (c5), we also added the direct path from the FBP to each of the measures at 11-month (c2), six-year (c3), and 15-year (c3) follow-up assessments, although we did not expect that there would be direct effects of FBP on post-loss growth outcomes due to the focus of the FBP on mental health problems.

A sandwich estimator was used to account for the family clustering effect due to some families including multiple children. Model fit indices (CFI > .9, RMSEA < .1, and SRMR < .08) were used to evaluate whether the model fits the data well. Rates of data missingness among study variables were between 0% and 15.6% for baseline, post-test, 11-month, and six-year follow-up data and 29.9% for 15-year follow-up data on Personal Growth. The assumption of missing completely at random on study variables was not rejected based on Little’s Missing Completely at Random (MCAR) tests (ps > .05). Missing data was handled using Full Information Maximum Likelihood estimation in Mplus (Muthén & Muthén, 2017).

Results

Descriptive statistics and bivariate correlations among variables are presented in the Online Supplementary Materials. Bivariate correlations showed that six-year Relating to Others PTG was correlated with 11-month coping efficacy and baseline support-seeking (r = .20 and .18, respectively). Personal Growth at the 15-year follow-up was correlated in the expected directions with post-test quality of parenting (r = .17) and baseline and 11-month coping efficacy (r = .23 and .24, respectively). More 15-year Personal Growth was related to fewer baseline intrusive grief thoughts (r = −.23) and baseline and 11-month internalizing problems (r = −.16 and −.17, respectively). Six-year Relating to Others PTG and 15-year Personal Growth were positively correlated with each other (r = .29). Child age was correlated with Relating to Others PTG at the six-year follow-up (r = .19). Gender was also correlated with Relating to Others PTG at six-year follow-up (r = .20) such that girls reported greater Relating to Others PTG scores than boys, which is consistent with the adult literature on PTG as women tend to report more PTG than men. There were no associations between growth at either time point and the cause of death.

Hypothesis Testing

Model results are summarized in Table 1. Model fit indices of all models met the criteria (CFI > .9, RMSEA < .1, and SRMR < .08). Consistent with prior work, in all models, we found a significant “a” path from FBP to post-test quality of parenting (ps < .01), indicating that parents who participated in the FBP demonstrated higher quality of parenting at post-test than those in the control condition. As expected, six-year Relating to Others PTG was positively related to 15-year Personal Growth (ps < .05) in all models after controlling for the covariates and effects from the FBP intervention, quality of parenting at post-test, and mediators at 11-month follow-up.

Table 1.

Estimated regression coefficients of the five mediation models

Models	Estimate	FBP → Parenting (post-test) [a]	Parenting (post-test) → Mediator (11 months) [b1]	Mediator (11 months) → Growth (6 years) [b2]	Growth (6 years) → Growth (15 years) [b3]
Support-seeking	b	0.127**	0.338**	1.119†	0.025**
	SE	0.046	0.130	0.610	0.009
	β	0.117	0.184	0.134	0.236
Perceived parental warmth	b	0.127**	0.470**	1.661*	0.027**
	SE	0.046	0.147	0.728	0.009
	β	0.116	0.295	0.172	0.257
Coping efficacy	b	0.127**	0.176	1.696*	0.023*
	SE	0.046	0.107	0.670	0.009
	β	0.116	0.100	0.195	0.225
Internalizing problems	b	0.127**	−0.238*	0.479	0.027**
	SE	0.046	0.106	0.645	0.009
	β	0.117	−0.145	0.051	0.258
Intrusive grief thoughts	b	0.127**	−0.170	0.746	0.027**
	SE	0.046	0.120	0.635	0.009
	β	0.117	−0.082	0.101	0.261

Note.

^†p < .10,

^*p < .05,

^**p < .01.

FBP = Family Bereavement Program. b = unstandardized coefficients. SE = standard error of the unstandardized coefficient. β = standardized coefficients.

Two out of the five models had statistically significant developmental cascade pathways. First, after controlling for baseline support-seeking, improved quality of parenting as a result of random assignment into the FBP (vs. control) group was associated with greater support-seeking from the parent at 11-month follow-up (b = .34, SE = .13, β = .18, p = .009) and support-seeking at 11-month follow-up was in turn related to greater Relating to Others PTG at the six-year follow-up (b = 1.19, SE = .61, β = .13, p = .067). Second, after controlling for baseline perceived parental warmth, improved quality of parenting at post-test as a result of random assignment into the FBP (vs. control) group was also associated with greater perceived parental warmth at the 11-month follow-up (b = .47, SE = .15, β = .30, p = .001), which was in turn associated with greater Relating to Others PTG at six-year follow-up (b = 1.66, SE = .73, β = .17, p = .022). As mentioned earlier, six-year Relating to Others PTG was positively related to 15-year Personal Growth after controlling for the effects of the 11-month mediators.

As such, the findings support that the FBP’s effects on quality of parenting had cascading effects on the relational aspect of post-loss growth during adolescence and early adulthood via increasing children’s support-seeking behaviors and perceived parental warmth. Moreover, the relational aspect of post-loss growth was associated with general post-loss growth nine years later during adulthood. As a post-hoc analysis, we ran a cascade model that included both support-seeking and parental warmth at 11-month follow-up. Results revealed that neither the mediation effect of support-seeking nor the mediation effect of perceived parental warmth remained significant. Given that support-seeking and the indicators of perceived parental warmth at 11-month follow-up were moderately related (rs = .31-.44), this suggests that the shared variance between support-seeking from parent and perceived parental warmth, instead of the unique variance of each measure, accounted for the cascade effects.

While the other 11-month mediators (coping efficacy, internalizing problems, and intrusive grief thoughts) were not found to mediate the effects of program-improved quality of parenting on post-loss growth, a few additional findings are noteworthy. First, although post-test quality of parenting was not related to 11-month coping efficacy per se, coping efficacy was associated with six-year Relating to Others PTG (b = 1.70, SE = .67, β = .20, p = .011). Second, post-test quality of parenting was related to fewer internalizing problems at 11-month follow-up (b = −0.24, SE = .11, β = −.15, p = .025) but this did not extend to greater six-year Relating to Others PTG. Third, coping efficacy had a significant direct effect on 15-year Personal Growth (b = .14, SE = .07, β = .15, p = .046) after controlling for the b3 path from coping efficacy to Relating to Others PTG. Finally, we did not detect longitudinal associations from intrusive grief thoughts to either six-year Relating to Others PTG or 15-year Personal Growth (ps > .05).

Discussion

This is the first study to examine the cascading pathways through which improvements in quality of parenting after participation in a family-based intervention for parentally bereaved youth led to greater post-loss growth six and 15 years later. We found that improved quality of parenting immediately following the program, a general composite measure of overall parenting practices, was related to more frequent youth-reported support-seeking from their parent and greater levels of perceived parental warmth at 11-month follow-up. Each of these two factors was in turn related to post-loss growth in the relationship domain (Relating to Others PTG) in adolescence and early adulthood (i.e., six-year follow-up), which was then related to a broader sense of post-loss growth (Personal Growth) in adulthood (i.e., 15-year follow-up). In addition, youth-reported coping efficacy was related to post-loss growth many years later. Finally, intrusive grief thoughts and internalizing problems were not significantly related to post-loss growth longitudinally. These findings have theoretical implications for the understanding of the development of post-loss growth in parentally bereaved children, as well as clinical implications to inform service providers who work with this population. They support the use of programs, such as the FBP, that strengthen the quality of parenting as an approach to promoting long-term post-loss growth for children who experience the death of a parent.

Support-seeking from parents and child-perceived parental warmth each mediated the impact of higher-quality parenting due to the FBP on later reports of post-loss growth. These findings support Kilmer et al.’s (2014) assumptions of PTG. It is likely that a positive parent-child relationship fosters a sense of security that breeds a supportive atmosphere conducive to support-seeking (Saldinger et al., 2003). In providing support, parents may foster their children’s growth by modeling their own resilience and growth or facilitating “family meaning-making” centered around personal strength and nurturing relationships (Nadeau, 1998). These findings align with the work of Wolchik et al. (2009) which found that support-seeking from a parent at baseline assessment was significantly related to more Relating to Others PTG at six-year follow-up in a subsample of the control group in the current sample. Of note, the three indicators of the youth-reported parental warmth variable were part of the measures of post-test quality of parenting. Thus, the significant pathway in part represents stability in these three aspects but also suggests that other aspects of high-quality parenting, such as discipline, listening skills, and dyadic activities, may contribute to more parental warmth over time. Youth-perceived parental warmth was directly linked to greater post-loss relationship growth more than five years later, which is consistent with prior research (Chen et al., 2019), and the effects on personal growth nine years later further extend understanding of the long-term effects. The post-hoc analysis, indicated that the shared variance between support-seeking behaviors and perceived warmth likely was responsible for the effects on post-loss growth but neither support-seeking nor perceived warmth has a unique predictive effect when controlling for the other. Both support-seeking and parental warmth reflect secure parent-child attachment, which might be the “active ingredient” that explains how FBP’s effects on the quality of parenting might have promoted post-traumatic growth. Indeed, a meta-analysis found that secure attachment was positively associated with PTG in adult samples affected by various kinds of trauma (Gleeson et al., 2021). Additional research is needed to further probe the association between secure attachment and post-loss growth among parentally bereaved children specifically.

Coping efficacy was related to post-loss growth in the relationship domain five years later and general post-loss growth 14 years later, although we did not find a significant path between post-test parenting and coping efficacy at 11-month follow-up. However, it is worth noting that Sandler et al. (2003) found that the FBP improved positive coping at the 11-month follow-up, in which positive coping was a composite measure of coping efficacy and active coping (taking cognitive and behavioral actions to address the stressor). While we chose to test coping efficacy instead of general coping behaviors or skills because Kilmer et al. (2014) emphasized that it is children’s efficacy beliefs that promote one’s PTG, future research may explore different coping constructs in relation to post-loss growth.

Finally, higher levels of internalizing problems did not correlate with more post-loss growth. While Kilmer et al.’s model posits that more internalizing problems precede the development of PTG in children, empirical evidence on the role of internalizing problems in PTG is inconsistent (Meyerson et al., 2011) and similar to the current analyses some cross-sectional studies have failed to find a significant association between psychological distress and PTG (Milam et al., 2004; Salloum et al., 2019). Moreover, contrary to theories suggesting that intrusive grief thoughts are a critical precursor to the development of PTG following a highly stressful event (Kilmer et al., 2014; Tedeschi & Calhoun, 2004), we did not find evidence for the prospective relation between intrusive grief thoughts and post-loss growth outcomes over the five- and 14-year time lag. Previously, among a subset of the control group in the current sample, Wolchik et al. (2009) also did not find significant associations between intrusive grief thoughts at baseline and any of the five subscales of the PTGI at six-year follow-up. Given that most of the existing studies on PTG have been conducted in adults or older adolescents, it is possible that internalizing problems or intrusive grief thoughts in children might operate differently across development, so future studies might investigate whether age moderates the relation between intrusive grief and post-loss growth. This non-significant relation might also have occurred because on average the children in the current sample had experienced the death nearly 21 months prior to the 11-month follow-up, and it may be that reports of high internalizing problems and intrusive grief thoughts at this stage may be more indicative of problems or risks rather than meaning-making processes that occur shortly after the adverse event.

The current study has several limitations. First, our sample was predominantly non-Hispanic White, which limits its generalizability to diverse populations. Future research should examine these pathways in a more racially diverse sample of parentally bereaved children. Second, while most of the literature uses the full PTG scale, our post-loss growth measure at six-year follow-up used one subscale that assessed interpersonal relationships, and our post-loss growth measure at 15-year follow-up assessed more general grief-related personal growth. The difference in measures makes it difficult to compare the findings of our study to those of other studies. Next, growth was not assessed at baseline, and thus there was no baseline covariate for post-loss outcomes, so we were unable to study the effects of the FBP or the cascading model on changes in post-loss growth over time. Finally, our measures of post-loss growth were self-reported perceptions and do not reflect objective positive changes, which can be assessed via other informants. Future research should incorporate assessments of growth from other informants and assess growth at baseline assessment in order to examine changes over time. It would also be valuable to include a comparison group of non-bereaved youth to differentiate growth that is specifically loss-related and growth that is a result of typical development.

Regardless of the limitations, the present study provides an important contribution to the literature on the cascading effects of a preventive family-based intervention for parentally bereaved youth on post-loss growth outcomes 15 years later. The findings underscore the efficacy of a preventive, parenting-focused intervention in generating stronger parent-child bonds and the mediating pathways in which the intervention can have “cross-over” promotive effects on long-term post-loss growth outcomes as well as previously demonstrated effects to reduce problem outcomes (Sandler et al., 2010; 2023). While the understandable negative aspects of parental loss in childhood have been well documented and continue to be examined in the literature, our findings align with a strength-based approach and suggest that program effects to improve quality of parenting not only reduces problem outcomes but also boosts post-loss growth in parentally bereaved youth.