Introduction
Bereavement during childhood is associated with emotional and behavioral problems after loss (Appel et al., Reference Appel, Johansen, Deltour, Frederiksen, Hjalgrim, Dalton, Dencker, Dige, Bøge, Rix, Dyregrov, Engelbrekt, Helweg, Mikkelsen, Høybye and Bidstrup2013; Berg et al., Reference Berg, Rostila and Hjern2016). Prior work suggests that 5−10% of youth report symptoms of depression, prolonged grief disorder, or posttraumatic stress syndromes following bereavement (Melhem et al., Reference Melhem, Porta, Shamseddeen, Walker Payne and Brent2011, Reference Melhem, Porta, Walker Payne and Brent2013). Individual-level factors such as developmental competence (Brent et al., Reference Brent, Melhem, Masten, Porta and Payne2012), substance use (Kaplow et al., Reference Kaplow, Saunders, Angold and Costello2010), and pre-loss behavioral symptoms (Dowdney, Reference Dowdney2000) are known to influence risk and resilience to adverse mental health outcomes among bereaved youth. In addition, social determinants such as race/ethnicity (Umberson et al., Reference Umberson, Olson, Crosnoe, Liu, Pudrovska and Donnelly2017) and parental educational attainment (Denckla et al., Reference Denckla, Averkamp, Slopen, Dice, Williams, Katherine Shear and Koenen2022) shape the risk of experiencing childhood bereavement. Taken together, findings suggest complex, interlocking influences of individual characteristics and social determinants on the health and well-being of bereaved youth (Cicchetti, Reference Cicchetti2010; Masten, Reference Masten2001). However, key questions remain regarding the influence of sociodemographic factors on the association between bereavement and mental health.
Influential theoretical frameworks that explain social influences on health include Bronfenbrenner’s Ecological Systems Theory (Bronfenbrenner, Reference Bronfenbrenner1994) and Ecobiodevelopmental Theory (Shonkoff et al., Reference Shonkoff and Garner2012). These highly influential models have shaped our understanding of the pathways by which societal level factors influence health outcomes among youth. Ecosocial theory (Krieger, Reference Krieger2001) similarly proposes that health outcomes unfold in nested hierarchical systems from the societal level to the individual, but focuses attention on upstream factors that influence the distribution of health across the population. The four core tenets of ecosocial theory of disease distribution include: 1) embodiment, which refers to the literal biological integration of exposures at every level of socially ascribed categories and identities; 2) an explicit focus on the pathways and processes of literal biological embodiment in the context of intersecting socially structured identities; 3) the interplay of exposure, susceptibility, and resistance across the life course, which emphasizes the timing and societal context of exposures; and 4) accountability and agency across institutions and public health researchers to perpetuate, generate, or address health inequities (Krieger, Reference Krieger2001, Reference Krieger2020). Social determinants of childhood bereavement are increasingly recognized in the literature (Umberson, Reference Umberson2017; Wilson & O’Connor, Reference Wilson and O’Connor2022), yet less work has extended this perspective to understand sociodemographic factors that influence mental health among bereaved youth. Extant literature on the association between childhood bereavement and mental health has largely drawn from clinical perspectives that focus on individual-level variables, thus yielding a robust understanding of the individual-level pathways associated with bereavement and mental health. However, questions remain unaddressed regarding the social determinants of bereavement and the role of these underlying sources of structural inequity in shaping health outcomes (Williams & Collins, Reference Williams and Collins1995).
The aims of the present study are guided by the core tenets of ecosocial theory. Consistent with the first tenet of accountability and agency, selected sociodemographic precursors hypothesized to be associated with childhood bereavement are drawn from a wider literature documenting disparities in adversity exposure and health to include neighborhood-level deprivation (Evans, Reference Evans2006), family-level measures of socioeconomic position (e.g., financial hardship (Slopen et al., Reference Slopen, Koenen and Kubzansky2014), social class (Iveson et al., Reference Iveson, Altschul and Deary2020), parental educational attainment (Miech et al., Reference Miech, Pampel, Kim and Rogers2011), and maternal health behavior (i.e., smoking (Taylor et al., Reference Taylor, Carslake, de Mola, Rydell, Nilsen, Bjørngaard, Horta, Pearson, Rai, Galanti, Barros, Romundstad, Davey Smith and Munafò2017)). The multidisciplinary science of human development reinforces that early childhood experiences, alongside environmental influences, profoundly shape the emergence of disease across the lifespan into adulthood and that early origins in health disparities are critical targets to reverse the effects of diseases that have origins in early childhood (Dunn et al., Reference Dunn, Soare, Raffeld, Busso, Crawford, Davis, Fisher, Slopen, Smith, Tiemeier and Susser2018; Shonkoff et al., Reference Shonkoff, Boyce and McEwen2009; Suglia et al., Reference Suglia, Campo, Brown, Stoney, Boyce, Appleton, Bleil, Boynton-Jarrett, Dube, Dunn, Ellis, Fagundes, Heard-Garris, Jaffee, Johnson, Mujahid, Slopen, Su and Watamura2020). The objective of this aim was to consider sociodemographic precursors to childhood bereavement that had not yet been examined in prior work.
Second, ecosocial theory directs us to consider the cumulative interplay of exposure, susceptibility, and resistance. As applied to the research at hand, individual-level health assets such as cognitive ability may influence the relationship between bereavement and mental health. There is value in examining this relationship with careful consideration of the influence of structural determinants that may confound these pathways of interest. With regard to cognitive ability, the cognitive reserve hypothesis proposes that cognitive ability protects against psychopathology (Barnett et al., Reference Barnett, Salmond, Jones and Sahakian2006; Deary & Batty, Reference Deary and Batty2007). Prior work among a general population suggests that lower cognitive ability (measured as full scale intelligence quotient (FSIQ)) is an antecedent of several common psychiatric disorders (Batty et al., Reference Batty, Wennerstad, Smith, Gunnell, Deary, Tynelius and Rasmussen2009), and is linked with an elevated lifetime risk of some psychiatric disorders (Alnæs et al., Reference Alnæs, Kaufmann, Doan, Córdova-Palomera, Wang, Bettella, Moberget, Andreassen and Westlye2018; Koenen et al., Reference Koenen, Moffitt, Roberts, Martin, Kubzansky, Harrington, Poulton and Caspi2009). Whether cognitive ability exerts a unique protective effect against downstream psychopathology after bereavement remains an open question and should be tested within the ecosocial framework, given the aforementioned importance of social determinants. To address this question, we examine whether pre-loss cognitive ability is uniquely protective against post-loss psychopathology, even after adjusting for key sociodemographic precursors. We draw on longitudinal data to capture proper temporal ordering of pre-loss cognitive ability, bereavement exposure, and post-loss psychopathology.
Finally, ecosocial theory maintains a focus on pathways to embodiment, including investigations into the pathways and processes of literal biological embodiment in the context of intersecting socially structured identities across the lifespan. Here we consider various spatiotemporal scales across the lifespan to contextualize the association between bereavement and mental health outcomes shown previously to be associated with childhood bereavement, including depression (Harrison & Harrington, Reference Harrison and Harrington2001), prolonged grief disorder (Bryant et al., Reference Bryant, Edwards, Creamer, O’Donnell, Forbes, Felmingham, Silove, Steel, McFarlane, Van Hooff, Nickerson and Hadzi-Pavlovic2021), and externalizing disorders. This focus is aligned with developmental perspectives that have shown that the loss of a close friend is an especially salient stressor (Layne et al., Reference Layne, Kaplow, Oosterhoff, Hill and S. Pynoos2017) at a time when youth carry a heightened risk of mood disorders (Cicchetti & Rogosch, Reference Cicchetti and Rogosch2002; Kessler et al., Reference Kessler, Amminger, Aguilar-Gaxiola, Alonso, Lee and Ustün2007) and relationships with peers emerge central features of identify development (Thompson et al., Reference Thompson, Flood and Goodvin2015). Life course theory (Leopold & Lechner, Reference Leopold and Lechner2015; Neugarten, Reference Neugarten1969) suggests that life-stage timing of loss is a key moderator for health outcomes, such that “off-time” loss (e.g., the loss of a parent in childhood) is associated with greater difficulty for the bereaved child compared with “on-time” loss (e.g., the death of a grandparent). The extant literature supports this hypothesis (Kamis et al., Reference Kamis, Stolte and Copeland2022), noting that parental bereavement, in particular, is associated with several adverse health outcomes, including elevated psychopathology (Cerel et al., Reference Cerel, Fristad, Verducci, Weller and Weller2006), increased psychotropic medication use (Høeg et al., Reference Høeg, Christensen, Banko, Frederiksen, Appel, Dalton, Dyregrov, Guldin, Jørgensen, Lytje, Bøge and Bidstrup2021), and elevated suicide incidence (Guldin et al., Reference Guldin, Li, Pedersen, Obel, Agerbo, Gissler, Cnattingius, Olsen and Vestergaard2015). Additional types of “off-time” bereavement, including the loss of siblings or close friends, have also been shown to have adverse impacts on mental health (Brent et al., Reference Brent, Perper, Moritz, Allman, Liotus, Schweers, Roth, Balach and Canobbio1993; Layne et al., Reference Layne, Kaplow, Oosterhoff, Hill and S. Pynoos2017; Rostila et al., Reference Rostila, Saarela and Kawachi2013). While the literature on “on-time” losses is more limited, some studies have noted an association with increased levels of depression following the loss of a grandparent during childhood (Harrison & Harrington, Reference Harrison and Harrington2001). Others note transgenerational consequences of grandparent loss not just on maternal depression but also directly on adolescent boys, independent of maternal effects (Livings et al., Reference Livings, Smith-Greenaway, Margolis and Verdery2022). Overall, however, strong evaluation of the differing effects of different types of loss on youth mental health is lacking. As a result, we examine variation in post-loss emotional and behavioral symptoms based on the type of loss, delineated as “on-time” (grandparent) versus “off-time” (parent, sibling, or friend) as hypothesized pathways to embodiment, consistent with ecosocial theory.
In summary, the present investigation has three interrelated aims, each informed by tenets of ecosocial theory (see Figure 1 for a conceptual diagram): 1) to identify sociodemographic precursors associated with childhood bereavement consistent with the ecosocial theory tenet of agency and accountability, 2) to investigate hypothesized pathways between the cumulative interplay of exposure, susceptibility, and resistance by investigating the moderating effect of individual-level cognitive ability on the association between bereavement and emotional/behavioral symptoms, and 3) to identify pathways to the embodiment of bereavement exposures, taking into account developmental perspectives on life course timing by examining variation in emotional/behavioral problems based on the type of loss. To achieve these aims, we leverage the longitudinal design of the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort, controlling for pre-bereavement mental health and baseline sociodemographic variables, measured in utero or infancy, and ensuring temporal ordering of measured cognitive ability prior to bereavement exposure and mental health outcomes. Together, we aim to improve our understanding of the individual, societal, and bereavement-specific influences that impact the relationship between bereavement and mental health among youth.
Conceptual model applying ecosocial theory (Krieger, Reference Krieger2001) to illustrate associations across the community to the individual level, including variables considered in the present analyses represented in concentric associations.
Methods
Sample
The sample consisted of participants enrolled in ALSPAC, an epidemiological birth cohort study of parents and their children. All pregnant women resident in Avon, UK with expected delivery dates between April 1, 1991 and December 31, 1992 were invited to participate (Boyd et al., Reference Boyd, Golding, Macleod, Lawlor, Fraser, Henderson, Molloy, Ness, Ring and Davey Smith2013; Fraser et al., Reference Fraser, Macdonald-Wallis, Tilling, Boyd, Golding, Davey Smith, Henderson, Macleod, Molloy, Ness, Ring, Nelson and Lawlor2013; Golding et al., Reference Golding, Pembrey and Jones2001). Of the 15,454 pregnancies (15,589 fetuses), there were 14,901 were alive at one year of age. Detailed health and sociodemographic data were collected via self-, maternal-, and paternal-report questionnaires as well as in-person assessment clinics. Details of all the data is available through a fully searchable data dictionary and variable search tool: https://www.bris.ac.uk/alspa c/researchers/our-data/. Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. Informed consent for the use of data collected via questionnaires and clinics was obtained from participants following the recommendations of the ALSPAC Ethics and Law Committee at the time.
At 198 months (16.5 years), the 9516 participants retained in the cohort (68%) were sent questionnaires that inquired about, among other information, the loss of a loved one since the age of 12. Among questionnaires sent, 5131 (54%) were completed. Supplementary Table 1 highlights child sex and parent economic differences between respondents and non-respondents. Of those respondents, 5050 children (98%) answered questions on the loss of a parent, brother, sister, grandparent, or close friend and thus were included in this analysis. Among those with bereavement status ascertained, mental health data were available for 4,093 (81%) children using the Strengths and Difficulties Questionnaire (SDQ). Supplementary Table 2 presents child sex and parent economic differences between children with and without SDQ data. See Figure 2 for a flow chart of participant retention across analytic samples. Of note, the mental health outcome ascertainment rates were relatively similar across bereavement status categories (e.g., SDQ was observed for 74% of those bereaved of a parent, sibling, or close friend, 82% of those bereaved of a grandparent, and 82% of those not bereaved). Given the availability of strong auxiliary variables in the ALSPAC data set, we used multiple imputation with chained equations (m = 20 imputations) to impute all sociodemographic, IQ, and pre-loss emotional/behavioral symptoms covariates among those with bereavement status ascertained, consistent with procedures used previously in the ALSPAC data set (Wiles et al., Reference Wiles, Northstone, Emmett and Lewis2007). We did not use imputed values of our outcome, SDQ at age 16.5. Additional information on variable-level missingness and the imputation model is included in Supplemental Table 3.
Participant retention flow chart for the present investigation.
Measures
Bereavement. At 16.5 years, adolescents reported on the death of a “parent, brother, or sister,” “grandparent,” and “close friend” since the age of 12. To test study hypotheses, bereavement was categorized as “no loss,” “grandparent loss,” or “parent, sibling, or close friend loss.” If a participant reported the death of both a grandparent and a parent, sibling, or close friend (n = 172 youth), we categorized that participant in the “parent, sibling, or close friend loss” group, given the larger impact expected of off-time loss, relative to on-time loss.
Mental Health Outcomes. The SDQ is a 25-item behavioral screening tool (Goodman, Reference Goodman1997). The SDQ was administered to the mother/caregiver about the study adolescent when the study adolescent was 16.5 years old. Each item is rated on a three-point scale ranging from not true to certainly true. Item responses were reverse coded, and total scores were derived by summing. Prorated scores were used throughout to account for missingness on any item. Specifically, if more than eight items were missing, the score was set to missing. In all other cases, a sum score was derived, and the score was weighted according to the non-missing items, resulting in a score that is comparable to one with complete item data (under the assumption that non-missing items follow similar patterns as the missing items). The SDQ contains four subscales assessed by five items, each ranging on an integer scale from 0 to 10 whereby higher scores represent greater problems on the scales of: hyperactivity (e.g., “restless, overactive, cannot stay still for long”), emotional symptoms (e.g., “many worries, often seems worried”), peer problems (e.g., “rather solitary, tends to be alone”), and conduct problems (e.g., “had temper tantrums or hot tempers”). A fifth subscale capturing prosocial behavior (e.g., “considerate of other people’s feelings”) is similarly assessed, except that a higher score reflects more positive prosocial behavior. The SDQ total difficulties score ranges on an integer scale from 0 to 40 and is a weighted sum of the hyperactivity, emotional symptoms, conduct problems, and peer problem subscales. The prosocial subscale is not incorporated into the total difficulties score. We examined the SDQ total difficulties score as our primary outcome, followed by sensitivity analyses among the five SDQ subscales to assist with interpreting findings. In analyses, each score was transformed due to high skew by taking the square root of the total score.
Cognitive ability. Cognitive ability (FSIQ) was assessed at 8 years old using the Wechsler Intelligence Scale for Children (WISC; 3rd UK edition) (Wechsler et al., Reference Wechsler, Golombok and Rust1992). A shortened version of the WISC was used wherein only alternate items were administered by trained psychologists; the coding subtest, however, was administered in full form. The WISC consisted of five verbal sub-tests, including information, similarities, arithmetic, vocabulary, and comprehension and five performance sub-tests, measuring picture completion, coding, picture arrangement, block design, and object assembly. Raw scores, manual age scaled scores, and total scores with prorating were calculated by ALSPAC team in accordance with WISC instructions. This measure of cognitive ability was selected because it temporally preceded bereavement.
Covariates. Sociodemographic covariates measured during pregnancy corresponding to the period in which the index study child was either in utero, or at birth included: Child sex (male or female) was determined from the birth notification. Child race/ethnic group was dichotomized as white or non-white due to the underrepresentation of non-white ethnic groups. At 32 weeks’ gestation, mothers reported their own and their partner’s race/ethnic group. When these data were missing, we drew on reports from the mother when the child was 8 years old whenever possible. If the mother reported a race/ethnic group for herself or her partner that was not white, the child was considered non-white. We also used sociodemographic variables commonly used in prior studies investigating socioeconomic status in the ALSPAC cohort, (Morris et al., Reference Morris, Manley, Northstone and Sabel2016) including: Mother’s educational attainment was self-reported at 32 weeks’ gestation and dichotomized as educational attainment corresponding to more than Certificate of Secondary Education (CSEs) (including vocational, O level, A level, and university degrees) versus no CSE or lower. Father’s educational attainment was reported by the mother at this same questionnaire and coded in the same way. A mother was determined to smoke during pregnancy if she reported smoking one or more cigarettes at present on a questionnaire sent between 24- and 41-weeks’ gestation. Mother’s prenatal financial difficulties score (range 0-15) was assessed at 32 weeks’ gestation based on self-reported difficulties in affording the following items: food, clothing, heating, rent/mortgage, and baby items. Each item was ranked on a scale of 0 (not difficult to afford) to 3 (very difficult to afford). Mother’s and father’s social class were derived according to the Office of Population Censuses and Surveys’ 1990 Standard Occupational Classification system, which classifies based on occupational data alone. In ALSPAC, these classifications were derived using employment information reported by the mother at 32 weeks’ gestation. In this cohort, social classes ranged from 1 (highest social class, inclusive of managers and administrators) to 6 (lowest social class, inclusive of personal and protective service occupations). We grouped classes into three overall categories due to low cell count: 1−2, 3−4, and 5−6 + Armed Forces (AF). Mother’s neighborhood quality was assessed during pregnancy via self-report questionnaire querying the following residential neighborhood characteristics: liveliness, friendliness, noise, cleanliness, attractiveness, and pollution/dirtiness. Each characteristic was ranked as usually, sometimes, or not at all true. Scores could range from 0 (poor quality neighborhood) to 12 (high quality neighborhood). Finally, pre-bereavement emotional/behavioral symptoms were assessed at 9 years of age, temporally prior to bereavement exposure. The SDQ as described above was administered to the mother/caregiver about the study child at 9 years of age.
Analytic approach
We first compared sociodemographic characteristics across bereaved and non-bereaved participants. Then, to address Aim 1 objective of investigating the sociodemographic precursors of bereavement, we regressed bereavement on sociodemographic precursor variables including maternal and paternal educational attainment, maternal smoking, maternal financial difficulties, maternal and paternal social class, and neighborhood quality using univariate logistic regression. Before testing our interaction hypotheses in Aim 2, we explored statistical power using the InteractionPowerR package in R (Baranger et al., Reference Baranger, Finsaas, Goldstein, Vize, Lynam and Olino2022). Assuming a sample size of N = 4500 with exposure prevalence of 40% and alpha = 0.05, we drew on 1000 iterations to explore statistical power. Based on exploratory analyses, we explored scenarios with small correlations between bereavement (X1), IQ (X2), and SDQ score (Y). We identified a few plausible scenarios in which we were sufficiently powered to identify the interaction term of interest. For example, assuming a correlation between X1 and X2 of −0.05, a correlation between X1 and Y between 0.01 and 0.1, and a correlation between X2 and Y between −0.1 and −0.3, we had greater than 90% power to identify a correlation between X1*X2 and Y of magnitude 0.07. We were likely underpowered to identify correlations between X1*X2 and Y much smaller than 0.07. Therefore, interaction analyses were restricted to any bereavement vs. no bereavement comparisons. To test our interaction hypothesis, that cognitive ability moderates the effect of bereavement on emotional/behavioral symptoms, we estimated a model wherein cognitive ability, assessed at age 8, was interacted with “any bereavement” vs. “no bereavement” status between the ages of 12 and 16.5 to predict emotional/behavioral symptoms assessed at age 16.5. All models were adjusted for by child sex and pre-bereavement emotional/behavioral symptoms assessed at age 9 (Model 1), and then further adjusted for sociodemographic precursor variables examined in Aim 1 above (Model 2). Before entering sociodemographic variables into multivariable models, we examined correlations among sociodemographic variables to assess for collinearity. See Supplemental Figure 1 for the correlation matrix or Supplemental Table 4 for a table of correlation coefficients. Pre-bereavement mental health adjustment was matched at the instrument sub-scale level (SDQ). For example, in the model regressing the SDQ prosocial behavior score on bereavement, adjustment was made for SDQ prosocial behavior score at 9 years of age. Measures of mental health outcomes in all analyses were square root-transformed due to skewness and kurtosis; all results are reported in these transformed units.
Finally, to address Aim 3, we used linear regression to examine variation in emotional/behavioral problems among bereaved youth, taking into consideration sociodemographic precursors, to gain unique insights based on the relationship with the deceased. We report the results of models regressing bereavement as a three-level predictor (“no bereavement” vs. “grandparent bereavement” vs. “parent/sibling/close friend bereavement”) on emotional/behavioral symptoms at age 16.5, controlling for pre-bereavement mental health assessed at age 9 and child sex (Model 1). We then further adjusted for eight sociodemographic precursor variables including child race, maternal and paternal qualifications and social class, maternal smoking during pregnancy, maternal prenatal financial difficulties, and neighborhood quality (Model 2).
Results
Comparisons of sociodemographic characteristics across bereaved and non-bereaved participants are reported in Table 1. Between the ages of 12 and 16.5, 1,528 (30.3%) youth were bereaved of a grandparent, whereas 499 (9.9%) were bereaved of a parent, sibling, or close friend. Of those who lost a parent, sibling, or close friend, 425 (85%) lost a close friend, 61 (12%) lost a parent or sibling, and 13 (3%) lost both a close friend and a parent or sibling. Univariate logistic regression results from models regressing bereavement on sociodemographic precursor variables suggested that some socioeconomic precursor variables are significantly associated with an increased likelihood of experiencing the death of a parent, sibling, or close friend (see Table 2). Adolescents whose mother smoked around birth exhibited higher odds of bereavement of a parent, sibling, or close friend compared to counterparts whose mother did not smoke around birth (OR = 1.88, CI:1.44, 2.45). This association was not significant for bereavement of a grandparent (OR = 1.02, CI: 0.83, 1.26). Similar significant associations with bereavement of a parent, sibling, or close friend were noted with lower parental educational attainment (maternal: OR = 1.49, CI: 1.12, 1.99; paternal: OR = 1.41, CI: 1.10, 1.80) and greater financial difficulties during pregnancy (OR = 1.06, CI: 1.03, 1.09). In addition, female participants were at higher odds of bereavement of a parent, sibling, or close friend (versus no loss). Finally, every one-point increase in IQ was associated with 0.99 (CI: 0.98, 0.99) lower odds of bereavement of a parent, sibling, or close friend.
Sample characteristics reported separately for non-bereaved, bereaved of a grandparent, bereaved of a friend, parent, or sibling, and any bereavement
Note. CSE: Certificate of Secondary Education. Social Class: 1 (highest social class) to 6 (lowest); AF = Armed Forces. Neighborhood Quality Index reflects self-rated residential neighborhood quality including liveliness, friendliness, noise, cleanliness, attractiveness, and pollution/dirtiness.
Unadjusted associations between early childhood socioeconomic status and bereavement exposure among youth between the ages of 12-16.5
Note. Bolded values indicate statistical significance at α = 0.05. Correlations measured among n = 5,050 whose bereavement status was ascertained. CSE: Certificate of Secondary Education. Social Class: 1 (highest social class) to 6 (lowest); AF = Armed Forces. Neighborhood Quality Index reflects self-rated residential neighborhood quality including liveliness, friendliness, noise, cleanliness, attractiveness, and pollution/dirtiness.
Next, we tested for a hypothesized protective effect of cognitive ability on post-bereavement emotional/behavioral symptoms by specifying an interaction term to detect effect modification by any bereavement status (see Table 3). While models did identify a significant main effect of increased cognitive ability, assessed at age 8, on reduced emotional/behavioral symptoms at age 16.5 (β = −0.01, SE = 0.01, p < 0.001), we did not find evidence of a significant interaction between bereavement status and IQ on emotional/behavioral problems; results were similar across minimally and fully adjusted models.
Linear regression coefficients for models estimating the association between IQ, any bereavement (grandparent, sibling, parent, close friend), and IQ*bereavement interaction and emotional/behavioral symptoms outcomes
Note. Bolded values indicate statistical significance at α = 0.05. Model 1 adjusts for pre-loss mental health and child sex. Model 2 introduces additional adjustment for peripartum sociodemographic covariates including race/ethnicity, maternal and paternal education levels, maternal smoking, financial difficulties, maternal and paternal social class, and neighborhood quality. Sample size depends on outcome missingness: n = 4093 (SDQ – Total); n = 4117 (SDQ – Emotional); n = 4120 (SDQ – Conduct); n = 4121 (SDQ – Hyperactivity); n = 4114 (SDQ – Peer); n = 4117 (SDQ – Prosocial).
Finally, we found that the association between bereavement and emotional/behavioral symptoms varied by the type of loss. In baseline linear models with bereavement modeled as a three-category predictor (no bereavement vs. grandparent bereavement vs. parent/sibling/close friend bereavement), adjusting for pre-bereavement mental health at age 9 and child sex, we found that bereavement of a parent, sibling, or close friend was strongly and significantly associated with increased SDQ (β = 0.20; 95% CI: 0.10, 0.29) log scores (see Model 1, Table 4). Conversely, bereavement by a grandparent was not significantly associated with emotional/behavioral symptoms. Next, we additionally adjusted all prior models for sociodemographic precursor variables. We found that introduced sociodemographic variables did not fully attenuate the relationship between bereavement of a parent, sibling, or close friend and SDQ-emotional/behavioral symptoms (see Figure 3). For example, bereavement of a parent, sibling, or close friend was still associated with a 0.19-point increase in SDQ total log score (95% CI: 0.10, 0.28), even after adjustment for sociodemographic variables. Finally, in sensitivity analyses we estimated effects of bereavement on SDQ subscales, noting that bereavement of a parent, sibling, or close friend was strongly associated with elevated conduct problems, emotional symptoms, and hyperactivity, weakly associated with lower prosocial behavior, and not associated with peer problems.
Linear regression model coefficients for the association between bereavement (modelled as a three-level predictor) and emotional/behavioral symptoms
Note. Bolded values indicate statistical significance at α = 0.05. Beta estimates reflect the expected change in log mental health score by bereavement status. Model 1 adjusts only for pre-loss mental health and child sex. Model 2 further adjusts for peripartum sociodemographic covariates including race/ethnicity, maternal and paternal education levels, maternal smoking, financial difficulties, maternal and paternal social class, and neighborhood quality. Sample size depends on outcome missingness: n = 4093 (SDQ – Total); n = 4117 (SDQ – Emotional); n = 4120 (SDQ – Conduct); n = 4121 (SDQ – Hyperactivity); n = 4114 (SDQ – Peer); n = 4117 (SDQ – Prosocial).
Estimated change in emotional/behavioral symptoms scores associated with bereavement: results from minimally (“Baseline”) and fully adjusted (“SES”) models examining effect attenuation by sociodemographic covariate adjustment.
Note: SDQ = Strengths and Difficulties Questionnaire; SES = socioeconomic status; Beta estimates reflect the expected change in log mental health score by bereavement status; baseline model (black) adjusts only for pre-loss mental health and child sex. The fully adjusted model (orange) additionally adjusts for sociodemographic precursors. Sample size depends on outcome missingness: n = 4093 (SDQ – Total); n = 4117 (SDQ – Emotional); n = 4120 (SDQ – Conduct); n = 4121 (SDQ – Hyperactivity); n = 4114 (SDQ – Peer); n = 4117 (SDQ – Prosocial).
Discussion
Leveraging data from a longitudinal birth cohort with prospective data on sociodemographic precursors to bereavement, pre-loss cognitive ability, and post-loss emotional and behavioral symptoms, the present investigation applied ecosocial theory (Krieger, Reference Krieger2001) to investigate interrelated influences of individual, familial, and social characteristics relevant to childhood bereavement and mental health outcomes. The ecosocial perspective has been applied in work investigating the social determinants of incarceration in the context of childhood adversity and mental health (Henry, Reference Henry2020), and here we extend this perspective to investigate social determinants of childhood bereavement and mental health outcomes. First, we sought to characterize sociodemographic precursors associated with risk for childhood bereavement. We extend work (see Bindley et al., Reference Bindley, Lewis, Travaglia and DiGiacomo2019 for a review), noting differences in exposure to bereavement by race/ethnicity (Umberson et al., Reference Umberson, Olson, Crosnoe, Liu, Pudrovska and Donnelly2017), and parental education levels (Denckla et al., Reference Denckla, Averkamp, Slopen, Dice, Williams, Katherine Shear and Koenen2022), to note that additional socioeconomic factors such as financial difficulties are also important predictors of childhood bereavement of a close friend, sibling, or parent. Economic resources are well-established predictors of early morbidity and mortality (Amick et al., Reference Amick, McDonough, Chang, Rogers, Pieper and Duncan2002; Jokela et al., Reference Jokela, Elovainio, Singh-Manoux and Kivimäki2009; McDonough et al., Reference McDonough, Duncan, Williams and House1997), and our findings place childhood bereavement within this context, raising the possibility of conceptualizing childhood loss as a downstream indicator of structural inequity. Finally, we also found that maternal health behaviors, indicated by maternal smoking, are associated with an elevated risk of childhood bereavement exposure. Interpreting this finding is limited in the context of a purely associative link given the many sources of residual confounding between maternal smoking and offspring health that are likely more broadly related to the characteristics of parents that smoke (Taylor et al., Reference Taylor, Carslake, de Mola, Rydell, Nilsen, Bjørngaard, Horta, Pearson, Rai, Galanti, Barros, Romundstad, Davey Smith and Munafò2017). Further research is needed to investigate this association further. In summary, our findings are consistent with the tenet of agency and accountability in ecosocial theory (Krieger, Reference Krieger2020), highlighting the early childhood origins of health disparities. Future work focused on the social determinants of childhood bereavement will inform our understanding of pathways by which health inequity adversely impacts youth development, potentially informing interventions to reverse compounded adverse health outcomes for bereaved youth (Slopen & Williams, Reference Slopen and Williams2021).
The second aim, seeking to identify whether cognitive ability moderated the association between childhood bereavement and adverse mental health outcomes, resulted in largely null findings. While greater cognitive ability was significantly yet minorly associated with lower emotional/behavioral symptoms, there was no evidence that this relationship differed by bereavement status. Our results are consistent with the cognitive reserve hypothesis documenting a protective effect of childhood cognitive ability on psychopathology (Iveson et al., Reference Iveson, Altschul and Deary2020; Koenen et al., Reference Koenen, Moffitt, Roberts, Martin, Kubzansky, Harrington, Poulton and Caspi2009), but failed to demonstrate additional protective effects in the context of bereavement. Given that we did not find evidence that cognitive ability was uniquely protective among bereaved youth relative to non-bereaved youth, our results in defense of the cognitive reserve hypothesis may be interpreted relative to a general adolescent population. Importantly, there is a significant debate in the cognitive epidemiology literature on whether childhood cognitive ability or sociodemographic variables are more important for health outcomes. While most evidence in cognitive epidemiology suggests that cognitive ability acts independently on health outcomes regardless of sociodemographic variables of origin (Batty et al., Reference Batty, Wennerstad, Smith, Gunnell, Deary, Tynelius and Rasmussen2009; Iveson et al., Reference Iveson, Altschul and Deary2020; Lubinski, Reference Lubinski2009), other work has found some evidence for attenuation in the association of cognitive ability and health outcomes by sociodemographic variables (Calvin et al., Reference Calvin, Batty, Der, Brett, Taylor, Pattie, Čukić and Deary2017; Hackman et al., Reference Hackman, Farah and Meaney2010). In the present study, the protective effect of cognitive ability on emotional/behavioral symptoms remained even after controlling for sociodemographic variables. This evidence supports the former position that cognitive ability acts independently of sociodemographic variables on health outcomes, though caution is warranted given our null findings in the context of bereavement. Further caution is warranted in interpreting these null findings in the context of the ecosocial theory tenet of cumulative pathways of exposure, susceptibility, and resistance, highlighting the possibility of variation in response to stressor exposure not observable by reliance on between-group mean differences among bereaved vs. non-bereaved youth (Denckla et al., Reference Denckla, Lee, Kim, Spies, Vasterling, Subramanian and Seedat2021). For example, it could be that cognitive ability is protective for some bereaved youth in specific settings, but that, on average, it is not protective for all bereaved youth. A third consideration is that even though cognitive ability, as measured by the WISC-3, did not evidence unique protective effects in the context of bereavement, individual differences in related social-emotional capacities not directly assessed by the WISC-3 may have buffering effects for the impact of bereavement specifically. For example, cognitive reappraisal (Shahane et al., Reference Shahane, Brown, Denny and Fagundes2022), effortful self-control (Miller et al., Reference Miller, Yu, Chen and Brody2015), and help-seeking behaviors (Andriessen et al., Reference Andriessen, Lobb, Mowll, Dudley, Draper and Mitchell2019) are important targets for future research. A final consideration in interpreting the present results in a developmental context warrants consideration of the association between bereavement and academic performance. Prior work has found that bereavement in adolescence is associated with reduced lifetime educational attainment, suggesting that pathways linking bereavement in young adulthood and to decrements in academic performance warrants further research (Breslau et al., Reference Breslau, Lane, Sampson and Kessler2008; Burrell et al., Reference Burrell, Mehlum and Qin2022; Espinosa Dice et al., Reference Espinosa Dice, Ye, Kim, McLaughlin, Amstadter, Tiemeier and Denckla2023; Prix & Erola, Reference Prix and Erola2017).
Finally, we found that the association between bereavement and emotional/behavioral symptoms depended on the type of loss, such that only bereavement of a parent, sibling, or close friend between the ages of 12 to 16.5 was associated with elevated emotional/behavioral problems at age 16.5. Importantly, this main effect held after controlling for emotional/behavioral symptoms assessed at age 9, suggesting that elevated post-bereavement symptoms are not explained by pre-bereavement symptoms. Then, these loss-specific associations remained significant when the same sociodemographic variables shown to be associated with bereavement exposure were introduced into models. Results suggest that while sociodemographic precursors are associated with elevated risk for bereavement, adjusting for these variables does not alter the main effect of bereavement on subsequent elevations in emotional and behavioral symptoms. Consistent with the ecosocial tenant of pathways to embodiment, the developmental perspective also underscores the need for more research on the effects of stressful interpersonal events, including bereavement, during adolescence, a time of rapid cognitive and emotional development wherein the risk of the first onset of psychiatric disorders is elevated (McLaughlin & Hatzenbuehler, Reference McLaughlin and Hatzenbuehler2009). While most research to date has focused on the important effects of parental loss among youth, further research on the loss of close friends and its potential association with elevated mental distress is warranted, given results from this study and prior work noting similar associations (Balk & Corr, Reference Balk, Corr and Stroebe2001; Balk et al., Reference Balk, Zaengle and Corr2011).
By using the SDQ as an outcome measure, we were able to acquire additional exploratory information on mental health domains impacted by childhood bereavement, while considering the type of loss. For example, while total SDQ was elevated among children experiencing the death of a parent, sibling, or close friend, this elevation was driven by greater emotional/behavioral symptoms across the subdomains of hyperactivity/inattentiveness, conduct problems, and emotional symptoms. While most prior work on the mental health effects of childhood bereavement has found a strong association with mood psychopathology, including depression (Pham et al., Reference Pham, Porta, Biernesser, Walker Payne, Iyengar, Melhem and Brent2018), our findings augment the evidence base by identifying elevated conduct and hyperactivity symptoms among bereaved youth. Future research may examine the effect of externalizing symptoms among bereaved youth on academic performance. Such association may help to explain the previously observed relationship between lower educational attainment among bereaved youth, relative to non-bereaved youth (Elsner et al., Reference Elsner, Krysinska and Andriessen2021; Liu et al., Reference Liu, Grotta, Hiyoshi, Berg and Rostila2022), and represent a potential target for intervention.
Several limitations should be considered when interpreting findings reported here. Most importantly, the structural relationship with the deceased loved one was defined in rather broad categories (e.g., parent, sibling, or close friend as one exposure group) at data collection, and further details of the relationship with the deceased was not available. Then, we were underpowered due to small sample sizes of exposed individuals to consider parental, sibling, and close friend bereavement separately, or to consider multiple losses as a category. The sample of children bereaved in the parent, sibling, and close friend cohort was predominantly composed of close friend bereavement (85%), thus our results could be under representative of the impact of loss of a sibling or parent. Future studies in cohorts with more bereaved children are necessary, though the pragmatic barriers to collecting such data given the relatively low prevalence of early parental loss in childhood is an ongoing challenge in psychiatric epidemiology. Fourth, while we could delineate a four-year period during which bereavement was experienced, we did not have further details on the exact date of the bereavement event within this four-year window. Therefore, we were not able to distinguish youth who might be experiencing acute grief immediately after a loss from those potentially experiencing a more chronic course. Fifth, emotional and behavioral symptoms were derived from parent report, not direct youth report. Relatedly, relying on observed mother/caretaker report of child SDQ outcomes may exclude individuals who experienced maternal loss, though the expected frequency of maternal loss is low. Sixth, missing data and differential loss to follow-up pose a significant limitation to the present study. While we did not find evidence for differences in observation of our primary outcome variable by bereavement status, sociodemographic comparisons of respondents and non-respondents suggest that our study conclusions are not representative of the full ALSPAC cohort. Rather, our analytic sample over-represented individuals whose parents exhibited higher economic privilege and may have resulted in biased estimates. While we were able to employ a robust multiple imputation model for missing IQ and covariates given the many proxy variables available in the ALSPAC data set, we recognize that attrition among bereaved youth is a significant concern in longitudinal cohort studies. Finally, lack of representation of diverse race/ethnicities and socioeconomic backgrounds in the current data limit generalizability to diverse populations (Boyd et al., Reference Boyd, Golding, Macleod, Lawlor, Fraser, Henderson, Molloy, Ness, Ring and Davey Smith2013).
In conclusion, while some socioeconomic factors were associated with elevated risk for experiencing bereavement, these factors did not account for the association between bereavement and elevated emotional/behavioral symptoms. The ecosocial approach is a promising framework to focus attention on upstream factors that influence the distribution of health across the population of bereaved youth. While cognitive ability was associated with fewer emotional/behavioral symptoms, its effect was small and did not modify the association between bereavement and mental health. Results suggest that policies designed to address social inequity could support preventable health disparities by reducing the likelihood of bereavement among youth.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0954579423000512.
Acknowledgments
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses.
Funding statement
Support for the analyses in the present report was funded by the National Institute of Mental Health (grant ref: 1K23MH117278) to CAD.
UK Medical Research Council and Wellcome (Grant ref: 217065/Z/19/Z) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors, and Christy Denckla will serve as guarantor for the contents of this paper. A comprehensive list of grants funding is available on the ALSPAC website (http://www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf). The parent study was specifically funded by Wellcome Trust and MRC (Grant ref: 092731).
The sponsor was not involved in study design, collection, analysis, and interpretation of data, writing of the report, and decision to submit the paper for publication.
Competing interest
All authors have no conflicts of interest or financial disclosures related to this work to report.
