Abstract
Purpose
Examine 1) the distribution of experiencing the death of a parent or sibling (family death) by race/ethnicity, and 2) how family death affects attaining a college degree.
Methods
Participants (N=8,984) were from National Longitudinal Survey of Youth 1997 aged 13–17 at baseline in 1997, and 29–32 in 2013. We examined the prevalence of family deaths by age group and race/ethnicity, and used covariate-adjusted logistic regression to assess the relationship between a family death and college degree attainment.
Results
4.2% of white youth experienced a family death, as did 5.0% of Hispanics, 8.3% of Blacks, 9.1% of Asians, and 13.8% of American Indians (group test p<0.001). A family death from ages 13–22 was associated with lower odds of obtaining a Bachelor’s degree by ages 29–32 (OR=0.65, 95%CI=0.50, 0.84), compared to no family death. The effect of a death was largest during college years (age 19–22) (OR=0.57, 95%CI=0.39, 0.82).
Conclusions
Young people of color are more likely to have a sibling or parent die; and family death during college years is associated with reduced odds of obtaining a college degree. Racial disparities in mortality might affect social determinants of health of surviving relatives, and college policies are a potential intervention point.
Keywords: social determinants of health, health disparities, bereavement, education
INTRODUCTION
Racial and ethnic disparities across many health outcomes are well-documented [1], and rates of death in the U.S. are higher for American Indian and Black populations than for white populations up to age 54 [2]. Given this early mortality, it is probable that young Black and American Indian people are more likely to experience the death of a close family member than white people. Indeed, results from the National Longitudinal Survey of Youth 1997 (NLSY97) showed that the risk of a sibling’s, father’s, and mother’s death were all higher for Black people from childhood through midlife, compared to white people [3]. However, neither that study, nor the broader literature, has examined the distribution of family deaths among other racial groups, or the prospective effects of family deaths on college education.
Associations between race and family deaths, and between family deaths and college degree completion, suggest institutional racism, i.e., policies and institutions that “work better for white people than for people of color, often unintentionally (p. 15)” [4]. Institutional racism, in turn, patterns “social determinants of health” – non-medical predictors of health, often unequally distributed by race, like working conditions or educational opportunities [5]. For example, the death of a parent or sibling (i.e., immediate family death) early in life is devastating, and it may affect subsequent social determinants of health and perpetuate racial disparities.
Education is an important social determinant of health that operates through many pathways including work, social standing, knowledge [5], health insurance, income, and social networks [6]. For example, those with less education tend to obtain lower-paying jobs, which have health-harming physical and mental conditions and fewer health benefits [5]. Cross-sectional surveys have linked the death of a friend or family member during college to poorer academic performance [7] and insomnia [8]. Moreover, an immediate family death during childhood has been linked with lower educational attainment [9]. However, the effect of a family death during the college years on college degree attainment has not been examined in prospective or population-based studies.
Racial education disparities are strong; Black and American Indian students who enroll in college are half as likely to graduate within four years compared to white and Asian students [10], even though disadvantaged groups experience greater gains in physical and mental health from education [11]. If Black and American Indian students are more likely to experience family death, and family death disrupts education, then this may exacerbate racial health disparities and create disadvantage over the life course. This framing broadens the focus of racial disparities in life expectancy to include the impacts on surviving family [12].
Education disparities may exacerbate health disparities, so an important next step is to explore the associations among race, family deaths, and educational outcomes in order to inform higher education policies. To illuminate these understudied associations, we take advantage of longitudinal data from a population-based, nationally representative sample of respondents initiated during adolescence. In this manuscript, we ask: 1) what is the racial distribution of adolescents (ages 13–18) and young adults (ages 19–22) who have experienced the death of a close family member? And 2) what are the impacts of being exposed to a family death at different ages (e.g., adolescence versus early adulthood) on attaining a college degree by age 29–32?
METHODS
DATA
We used 4 rounds of longitudinal data over 17 years (1997 to 2013) from the ongoing National Longitudinal Survey of Youth 1997 (NLSY97), designed to be representative of people in the United States at Round 1 in 1997. Of those who were eligible to be surveyed in 1997 90.7% agreed to participate. Participants in NLSY include 8,984 Millennials born between 1980 and 1984 who were aged 13–17 in 1997 (Round 1), and aged 29–32 in 2013 (Round 16). Participants were interviewed every year or every other year with 80% retention of the baseline sample in 2013. Interviews were done using in-person computer-assisted personal interviewing (84% of interviews in 2013) or over the phone (16%). The NLSY97 is made up of two subsamples: a sample representative of the United States, and an oversample of Hispanic and Black participants, resulting in a total sample that was 52% White, 26% Non-Hispanic Black, 21% Hispanic, and 1% Non-Hispanic Other. The full sample of 8,984 was retained in these analyses by multiply imputing missing data on covariates, exposure, and outcome.
VARIABLES
Exposure.
Family death was defined as death of a parent/step-parent or sibling when participants were 13 to 22 years old. Grandparent deaths were not classified as a family death because they are more common and expected, and not usually part of the immediate family. Deaths of spouses and children were not included in the measure because they were rare (both <0.5% of the sample). Deaths of ‘other’ relatives were not included because we could not further define those relationships. Participants were asked three times (2002, 2007 and 2013), “In the last five years, that is since you were [respondent’s age 5 years ago] years old, has a close relative of yours died?” and then identified each as mother/step-mother, father/step-father, brother or sister, spouse or partner, child, or other. The question does not allow separation of deaths of step-parents from parents. Respondents reported their own age at the time of each death. For the 2007 round, those who missed the survey were asked again in 2008 or 2009 with reference to the same timeframe as those who responded in 2007.
We classified family member death in three categories: none, family death during adolescence (13–18 years old), or family death during early adulthood (19–22 years old). We chose to measure family deaths at these ages to examine the hypothesis that a family death would affect college education outcomes during the years someone is typically in college, but not if the death happened before college-age years. Three participants appeared to report the same parent death at different ages, so we classified them at the earlier age. Ten additional participants experienced a family death during both adolescence and early adulthood. In sensitivity analyses, results were similar after changing the family death age category of the 10 participants, so rather than choose one age category over the other, we categorized them as missing, and multiply imputed them (see below).
Outcome.
Educational attainment was captured for everyone in 2013 (Round 16) when participants were 29–32 years old with a variable indicating the highest degree achieved at their 2013 interview. From this question, Bachelor’s degree attainment was coded as a binary variable indicating the respondent received a Bachelor’s degree, compared to no Bachelor’s degree. We chose Bachelor’s degree because of its importance to earnings [13], and because there is a more predictable age range for pursuing Bachelor’s degrees compared to other degrees.
Race.
Race was categorized differently in the two aims. In aim 1, we produced descriptive estimates of 5 self-reported race/ethnicity groups: Black, Hispanic, American Indian, Asian/Pacific Islander, and white. The categories of ‘other’ and ‘mixed’ were excluded from Figure 1 because prevalence estimates were unstable and imprecise due to empty cells. In aim 2, we collapsed race/ethnicity into 4 categories (white, non-Hispanic Black, Hispanic, non-Hispanic other) for two reasons. First, sample sizes were small for American Indians and Asian/Pacific Islanders (e.g., only 2 American Indians reported a family death during ages 19–22), so collapsing was necessary to avoid positivity violations. Second, the 4-category race/ethnicity measure was provided by NLSY97 to account for the sampling design. This 4-category measure combined American Indian, Asian/Pacific Islander, and other into “non-Hispanic other”, and redistributed the respondents who self-reported mixed race into one of these four categories, although that algorithm is not publicly available.
Figure 1:
Directed Acyclic Graph (DAG): Family Death and College Degree Attainment
This DAG informs our selection of covariates. Baseline academic performance would be a collider if adjusted for in this model, so we did not include it as a covariate in regression models, but its pathways are controlled for by including SES in models. Adjusting for baseline family health and baseline respondent health controls for the confounding pathways with genes, although genes here are loosely defined and unmeasured. There is likely a relationship between SES and baseline family health, although the directionality is unclear, so we did not put the arrow in this DAG.
Covariates.
We selected covariates for the regression model based on a Directed Acyclic Graph (DAG), a causal inference method used to reduce confounding [14]. Based on the DAG (Figure 1) we included the following baseline variables as regression covariates, measured prospectively in adolescence in 1997 (Round 1): household income quartile, mother’s education and father’s education (no high school diploma, high school diploma, 2 years of college, 4 years of college), parent self-rated health and participant self-rated health (excellent, very good, or good compared to fair or poor), urban/rural, census region (Northeast, North Central, South, West), and binary sex. Covariates were modeled as indicator variables.
ANALYTIC PLAN
For aim 1, describing the racial distribution of adolescents and young adults who experienced a family death, we estimated weighted univariate and bivariate analyses, including estimates of the prevalence, by age group and race, of exposure to the death of a close family member. We used a set of custom weights for 1997–2013 data that we used (Rounds 1–16), as recommended by NLSY97.[15] For aim 2, estimating the impacts of exposure to a close family member death during adolescence or early adulthood on attaining a college degree, we used logistic regression adjusted for covariates. The regression analyses accounted for the survey design, and adjusted for race/ethnicity rather than incorporating weights, as recommended by NLSY97 [16], We output odds ratios (OR), with 95% confidence intervals, to model the odds of completing a Bachelor’s degree (vs. not) by aged 29–32, comparing those who experienced a family death in early adulthood, or in adolescence, to those who did not experience a family death between 13 and 22 years old. We tested for an interaction between family death and race, to see if the effect of family death on college attainment differed by racial/ethnic group.
We included all respondents (n=8,984) enrolled in the NLSY97 cohort. On average, 4.7% of data were missing (range: 0% to 17% depending on the variable), so we imputed missing data using Stata version 15.0’s multiple imputation ice command [17] (50 imputations). This method used all variables in the model to impute missing data. Results with and without imputed data were similar, so we present imputed results. Table 1 includes proportion of data missing overall for each variable, as well as broken down by exposure category.
Table 1.
Sample Descriptives Overall and by Immediate Family Death, National Longitudinal Survey of Youth 1997 (N=8984)
| Overall 100% (N=8984) | No family death 88% (N=7953) | Family death age 13–18 2% (N= 209) | Family death age 19–22 3% (N= 294) | |
|---|---|---|---|---|
| Bachelor’s degree or higher in 2013 | ||||
| No Bachelor’s degree | 69 (6533) | 68 (5643) | 77 (168) | 82 (243) |
| Bachelor’s degree | 30 (2355) | 32 (2237) | 21 (36) | 17 (43) |
| missing | 0.9 (96) | 0.8 (73) | 2.4 (5) | 1.3 (4) |
| Baseline Characteristics 1997 | ||||
| Respondent self-rated health | ||||
| 1)Excellent, Very Good, Good | 96 (8522) | 96 (7547) | 96 (195) | 93 (274) |
| 2)Fair or Poor | 4.4 (457) | 4.4 (401) | 4.8 (14) | 6.6 (20) |
| missing | 0.1 (5) | 0.1 (5) | 0 (0) | 0 (0) |
| Race/Ethnicity (Figure 1: aim 1 prevalences) | ||||
| 1)White | 67 (4413) | 67 (3910) | 52 (71) | 58 (116) |
| 2)Hispanic | 13 (1901) | 13 (1719) | 11 (35) | 14 (60) |
| 3)Black | 15 (2335) | 15 (91) | 28 (108) | 23 (103) |
| 4)Asian/Pacific Islander | 2.4 (156) | 2.3 (131) | 5.5 (8) | 3.4 (6) |
| 5)American Indian | 1.0 (43) | 0.1 (36) | 3.1 (4) | 0.1 (2) |
| 6)Mixed (excluded in Fig 1) | 1.2 (83) | 1.3 (76) | 0 (0) | 1.1 (2) |
| 7)Other (excluded in Fig 1) | 1.0 (53) | 0.1 (48) | 0 (0) | 0 (0) |
| Race/Ethnicity (Figure 2: aim 2 models) | ||||
| 1)White | 67 (4413) | 67 (3910) | 52 (71) | 58 (116) |
| 2)Hispanic | 13 (1901) | 13 (1719) | 11 (35) | 14 (60) |
| 3)Non-Hispanic Other | 5.1 (335) | 5.0 (291) | 8.6 (12) | 5.5 (10) |
| 4)Non-Hispanic Black | 15 (2335) | 15 (2033) | 28 (91) | 23 (108) |
| Sex | ||||
| Male | 51 (4599) | 51 (4059) | 52 (107) | 48 (138) |
| Female | 49 (4385) | 49 (3894) | 48 (102) | 52 (156) |
| Urban or Rural | ||||
| Rural | 26 (2030) | 27 (1819) | 23 (39) | 26 (69) |
| Urban | 69 (6754) | 69 (5810) | 72 (159) | 70 (210) |
| missing | 4.5 (380) | 4.4 (324) | 4.8 (11) | 4.1 (15) |
| Census Region | ||||
| Northeast | 18 (1585) | 18 (1396) | 20 (43) | 14 (40) |
| North Central | 26 (2050) | 26 (1822) | 17 (29) | 25 (67) |
| South | 34 (3359) | 34 (2965) | 40 (94) | 40 (128) |
| West | 21 (1990) | 21 (1770) | 23 (43) | 20 (59) |
| Father’s highest education | ||||
| 1) Less than 12th grade | 16 (1637) | 16 (1453) | 18 (44) | 20 (62) |
| 2)12th grade | 31 (2728) | 31 (2400) | 35 (66) | 34 (98) |
| 3)Two years of college | 16 (1293) | 16 (1156) | 13 (23) | 13 (32) |
| 4)Four years of college | 20 (1470) | 21 (1336) | 8.9 (17) | 12 (25) |
| missing | 17 (1856) | 16 (1608) | 25 (59) | 21 (77) |
| Mother’s highest education | ||||
| 1) Less than 12th grade | 17 (1963) | 17 (1718) | 27 (64) | 23 (78) |
| 2)12th grade | 34 (3038) | 34 (2672) | 33 (71) | 36 (107) |
| 3)Two years of college | 23 (1892) | 23 (1678) | 22 (39) | 23 (59) |
| 4)Four years of college | 19 (1403) | 19 (1283) | 12 (20) | 8.7 (24) |
| missing | 6.9 (688) | 6.8 (602) | 6.2 (15) | 9.6 (26) |
| Parent income quartile | ||||
| 1)$0 – $15,000 | 18 (2124) | 17 (1848) | 30 (82) | 27 (100) |
| 2)$15,001 – 35,000 | 22 (2102) | 22 (1888) | 25 (49) | 27 (79) |
| 3)$35,001 – 59,999 | 26 (2073) | 26 (1847) | 24 (39) | 26 (67) |
| 4)$60,00 – maximum | 29 (2117) | 30 (1933) | 17 (29) | 16 (36) |
| missing | 6.2 (568) | 5.2 (437) | 4.2 (10) | 3.5 (12) |
| Parent’s self-rated health | ||||
| 1)Excellent, Very Good, Good | 78 (6757) | 79 (6056) | 65 (131) | 70 (195) |
| 2) Fair or Poor | 11 (1179) | 10 (1007) | 21 (48) | 19 (66) |
| Missing | 11 (1048) | 11 (890) | 14 (30) | 11 (33) |
Survey weighted proportions of original data before imputations, and original N’s.
Covariate distributions not broken out for those with missing family death data (N=528, 5.9%).
Variables without a row for missing data were complete for all participants.
RESULTS
Table 1 presents descriptive statistics for the exposure, outcome, and baseline covariates. Overall, 5.0% of participants experienced at least one family death between the ages of 13 and 22. The proportion of participants who obtained a Bachelor’s degree by ages 29–32 was 29.7%. See Table 1 for other demographic descriptives.
Figure 2 presents the family death exposure by race and age groups (adolescence, aged 13–18, versus early adulthood, aged 19–22). Overall, 3.0% of NLSY97 respondents in early adulthood, and 2.0% in adolescence reported an immediate family death. This varied by race with 4.6% of white respondents, 5.2% of Hispanics, 8.7% of Blacks, 9.8% of Asians, and 14.0% of American Indians (group test p<0.001). When combined into a single group, participants of color were significantly more likely to experience a family death than white participants (p<0.001).
Figure 2:
Percent of Respondents who Experienced Exposure to a Sibling or Parent Death, by Age and Race/Ethnicity, National Longitudinal Survey of Youth 1997
these are survey-weighted means and their corresponding confidence intervals. Compared to whites, the overall prevalence of family death was statistically higher for Blacks (p <.001). The prevalence of family death was also statistically higher for Blacks (p =.002) compared to Hispanics. In addition, the prevalence of family deaths among participants of color (7.2%, CI 6.1% – 8.3%) as a group is statistically higher than whites (4.6%, CI 3.9% – 5.4%; p<.001).
Results from the adjusted logistic regression model (Figure 3) show that those who experienced a family death during the typical college years of ages 19–22 were 43% less likely to attain a Bachelor’s degree by ages 29–32 (OR=0.57, 95% CI =0.39, 0.82) compared to those who had not experienced a family death between ages 13 to 22 years old. Adolescents who experienced a family death during adolescence (aged 13–18) were not significantly affected in attaining a Bachelor’s degree compared to those without a family death, although the point estimate was in the harmful direction (OR=0.77, 95% CI = 0.53, 1.13). The effect of a family death on degree attainment was not significantly different between the two age groups (p=0.26). When collapsed on age, a family death during any age (13–22) was significantly associated with a lower adjusted odds of attaining a Bachelor’s, versus no death (OR=0.65, CI: 0.50, 0.84), although this was weaker than the early adult age-specific effect. There also was not a significant overall interaction between the 4-category race variable and 3-category exposure to family death (p=0.96), indicating the effect of family death on education is similar across race/ethnicity (no effect modification by race; see Appendix Figure 1).
Figure 3:
Odds of attaining a Bachelor’s degree (and 95% CI) or not by ages 29–32, comparing those who were exposed to a sibling or parent death, in either adolescence (age 13–18) or early adulthood (age 19–22), and overall (age 13–22) versus those who did not experience a sibling or parent death (N=8984)
Adjusted for baseline race/ethnicity, household income quartile, mother’s education and father’s education, parent self-rated health, participant self-rated health, urban/rural, census region, and sex. Based on 50 multiple imputations for missing data.
Reference group: No sibling or parent death.
Sibling or parent death at age 13–18: OR= 0.77 (0.53, 1.13)
Sibling or parent death at age 19–22: OR = 0.57 (0.39, 0.82)
Sibling or parent death at age 13–22: OR = 0.65 (0.50, 0.84)
DISCUSSION
The prevalence of young people of color who experienced a family death was higher than their white peers. Notably, American Indian participants were most likely to have a family member die. Although this estimate was imprecise, it aligns with evidence that American Indians experience elevated age-adjusted mortality compared to other race/ethnicities [19]. Early deaths are something that one third of American Indians think about daily, and consistent with historical trauma [20].
In addition to being unequally distributed by race, experiencing a family death during college years was associated with 43% lower odds of obtaining a college degree. This suggests that a family death may influence educational attainment if a student is enrolled, or planning to enroll in college. A difference in exposure prevalence by race, in conjunction with an exposure-outcome association, can contribute to disparities (in this case, in college degree attainment) even in the absence of a race interaction [18], as demonstrated visually in Figure 4, quadrant ‘B’.
Figure 4:
Four example interaction analysis scenarios
Scenario B is the one proposed in this paper, where there is no interaction present, but the exposure contributes to a disparity in the outcome.
Reprinted from Annals of Epidemiology, Vol 29, Julia B. Ward,Danielle R. Gartner,Katherine M. Keyes,Mike D. Fliss,Elizabeth S. McClure,Whitney R. Robinson, How do we assess a racial disparity in health? Distribution, interaction, and interpretation in epidemiological studies, Pages 1–7, Copyright (2019), with permission from Elsevier.
One possible mechanism for this finding is via students’ financial stability. A review of the impact of family deaths on health showed that when a death caused a drop in economic resources, the surviving family (usually a spouse) experienced worse outcomes [21]. Given that the size of financial aid packages have a larger impact on drop-out rates in students of color [22], financial aid policies may be important. Other possible mechanisms to explore include impaired academic performance [7], problems sleeping [8], or mental and physical health, and the ways that colleges policies modify the association between family death and college degree attainment.
Adolescents who experienced a family death had 23% lower odds of obtaining a college degree than those with no family death, but it was not statistically significant. Although this estimate did not differ statistically from that of a family death during early adulthood, we tested these age exposure windows because we posited a priori that the etiologic period of the exposure may be important. Substantively, the association for a death during early adulthood was nearly twice as large compared to adolescence. Perhaps in regards to obtaining a college degree, losing a family member is a stronger short-term risk. Conceptualizing the death of a family member as a stronger short-term than long-term exposure has support from the literature. For example, youth and young adults ages 7–25 whose parents died by suicide, injury, or sudden natural death had a higher incidence of PTSD than the non-bereaved comparison group, but not in the second year after death [23].
STRENGTHS AND LIMITATIONS
As with any observational study, our conclusions could be biased by unmeasured confounding. We mitigated this in regression models by controlling for comprehensive measures of SES and health at baseline (at ages 13–17). The unequal distribution between race and family deaths may be partially explained by family size and parental age; although Umberson et al. found that the association between Black or white race and family deaths was strong after accounting for those factors [3]. The operationalization of the exposure is limited, as it does not allow us to estimate family deaths before age 13, and defines family deaths only as parents, step-parents, and siblings.
This study’s strengths include that it uses a population-based, nationally representative prospective cohort study from adolescence through adulthood. Therefore, we have measures of family death and educational attainment collected periodically since childhood. Moreover, NLSY97 oversampled Black and Hispanic children, allowing for more precise estimation of health disadvantages among these groups. We conducted multiple imputations for missing data, which is appropriate for maintaining the sample and for population-based inference.
PUBLIC HEALTH IMPLICATIONS
Mortality is conceptualized in the health literature as the “ultimate” health outcome, but someone’s death might extend the harm to their social networks. An unequal racial distribution of the death of a family member, combined with the connection to obtaining a college degree and the evidence that education is a social determinant of health [24], frames health disparities as a collective disadvantage among families and communities. The public health literature focuses on the ways social determinants of health can produce health disparities [25, 26], and it is also important to consider that health disparities can influence social determinants of health.
Public health scientists have been called to examine “structures, policies, practices, and norms to identify the mechanisms of institutionalized racism [27].” Policies can be “color-blind” and still produce structural discrimination [28]. Although structural discrimination is recognized in public health as a cause of health inequities, a review of U.S. public health literature found only 25 articles in high-impact journals that contained ‘institutionalized racism’ in their titles or abstracts [29]. Acknowledging that certain hardships are concentrated among disadvantaged groups can help us focus on solutions. Although this analysis focused on sibling and parent deaths, future research can look into other health-related hardships, including deaths of other family members or friends, and caregiving responsibilities.
Whether the association between family death and college degree attainment is related to finances, mental health, physical health, or other pathways, it is important to consider from a policy perspective how to weaken the association. For example, attention to institutional racism would address the fact that, historically, universities were designed by, and for, groups who experienced fewer disruptive life events. Are there assumptions, solidified in policies, that students have resources to recover from bumps in the road? Differences in college graduation rates by race, combined with social science research about racism in higher education [30], suggest that institutional racism may be operating. Low SES is an obstacle for completing a Bachelor’s degree but not a sub-Bachelor’s degree [31], so community college policies might inform how to serve a wide variety of students. For example, perhaps inflexible policies regarding course attendance, dropped classes, and tuition forfeiture can create barriers in traditional academic settings.
Little prior literature has explored how family deaths reverberate through a social network to affect social determinants of health among young adults, such as educational attainment. It will be important to probe these relationships to understand how to mitigate such risks.
Acknowledgements/Funding Source:
This work was supported by National Institutes of Health (NIH) grant R01 HD090014 (Dr. Osypuk, PI). The authors gratefully acknowledge support from the Minnesota Population Center (P2C HD041023) funded through a grant from the Eunice Kennedy Shriver National Institute for Child Health and Human Development (NICHD). Funders did not have any role in design or conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Acronym list
- CI
confidence interval
- NLSY97
National Longitudinal Survey of Youth 1997
- OR
odds ratio
- SES
socioeconomic status
Appendix Figure 1.
Odds of the effect of family member death at ages 13–18 and 19–22, compared to no family death, by race/ethnicity
Adjusted for baseline race/ethnicity, household income quartile, mother’s education and father’s education, parent self-rated health, participant self-rated health, urban/rural, census region, and sex. Based on 50 multiple imputations for missing data.
Reference group: No sibling or parent death.
Sibling or parent death at age 13–18: White OR= 0.88 (0.50, 1.56); Hispanic OR= 0.49 (0.14, 1.68); NH Other OR= 0.75 (0.18, 3.06); NH Black OR= 0.68 (0.33, 1.41)
Sibling or parent death at age 19–22: White OR= 0.62 (0.38, 1.68); Hispanic OR= 0.44 (0.16, 1.18); NH Other OR= 0.38 (0.08, 1.69); NH Black OR= 0.57 (0.29, 1.09)
Footnotes
Financial Disclosure Statement: The authors have no financial relationships relevant to this article to disclose.
Conflict of Interest Statement: The authors have no conflicts of interest relevant to this article to disclose.
Competing interests statement: The authors have no competing interests to report.
REFERENCES
- [1].National Center for Health Statistics. Health, United States, 2015: with special feature on racial and ethnic health disparities. 2016. [PubMed]
- [2].Centers for Disease Control and Prevention National Center for Health Statistics. Underlying Cause of Death 1999–2016 on CDC WONDER Online Database. Data are from the Multiple Cause of Death Files, 1999–2016, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. December 2017.
- [3].Umberson D, Olson JS, Crosnoe R, Liu H, Pudrovska T, Donnelly R. Death of family members as an overlooked source of racial disadvantage in the United States. Proceedings of the National Academy of Sciences. 2017;114(5):915–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Nelson J Racial Equity Toolkit: An Opportunity to Operationalize Equity. In: Equity GAoR, editor. September 2015. [Google Scholar]
- [5].Braveman P, Egerter S, Williams DR. The social determinants of health: coming of age. Annual review of public health. 2011;32:381–98. [DOI] [PubMed] [Google Scholar]
- [6].Cutler DM, Lleras-Muney A. Understanding differences in health behaviors by education. Journal of health economics. 2010;29(1):1–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Association ACH. American College Health Association-National College Health Assessment II: Undergraduate Student Reference Group Executive Summary Spring 2018. Silver Spring, MD: American College Health Association; 2018. [Google Scholar]
- [8].Hardison HG, Neimeyer RA, Lichstein KL. Insomnia and complicated grief symptoms in bereaved college students. Behavioral sleep medicine. 2005;3(2):99–111. [DOI] [PubMed] [Google Scholar]
- [9].Høeg B, Johansen C, Christensen J, Frederiksen K, Dalton SO, Bøge P, et al. Does losing a parent early influence the education you obtain? A nationwide cohort study in Denmark. Journal of Public Health. 2018. [DOI] [PubMed] [Google Scholar]
- [10].U.S. Department of Education NCfES, Integrated Postsecondary Education Data System (IPEDS),. Spring 2002 through Spring 2013 and Winter 2013–14 through Winter 2016–17, Graduation Rates. 2017.
- [11].Vable AM, Cohen AK, Leonard SA, Glymour MM, dP Duarte C, Yen IH. Do the health benefits of education vary by sociodemographic subgroup? Differential returns to education and implications for health inequities. Annals of epidemiology. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Umberson D Black Deaths Matter: Race, Relationship Loss, and Effects on Survivors. Journal of health and social behavior. 2017;58(4):405–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Autor DH. Skills, education, and the rise of earnings inequality among the” other 99 percent”. 2014. [DOI] [PubMed]
- [14].Glymour MM. Using causal diagrams to understand common problems in social epidemiology In: Oakes JM, Kaufman JS, editors. Methods in Social Epidemiology. San Francisco: Jossey-Bass; 2006. [Google Scholar]
- [15].U.S. Bureau of Labor Statistics. NLS Custom Weights [Available from: https://www.nlsinfo.org/weights/nlsy97.
- [16].National Longitudinal Survey of Youth. National Longitudinal Survey of Youth 97 (NLSY97) Washington, DC: Bureau of Labor Statistics; 2017. [Available from: https://www.nlsinfo.org/content/cohorts/nlsy97/intro-to-the-sample/nlsy97-sample-introduction-0. [Google Scholar]
- [17].Royston P, White IR. Multiple imputation by chained equations (MICE): implementation in Stata. J Stat Softw. 2011;45(4):1–20. [Google Scholar]
- [18].Ward JB, Gartner DR, Keyes KM, Fliss MD, McClure ES, Robinson WR. How do we assess a racial disparity in health? Distribution, interaction, and interpretation in epidemiological studies. Annals of Epidemiology. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Woolf SH, Chapman DA, Buchanich JM, Bobby KJ, Zimmerman EB, Blackburn SM. Changes in midlife death rates across racial and ethnic groups in the United States: systematic analysis of vital statistics. Bmj. 2018;362:k3096. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Whitbeck LB, Adams GW, Hoyt DR, Chen X. Conceptualizing and measuring historical trauma among American Indian people. American journal of community psychology. 2004;33(3–4):119–30. [DOI] [PubMed] [Google Scholar]
- [21].Stroebe M, Schut H, Stroebe W. Health outcomes of bereavement. The Lancet. 2007;370(9603):1960–73. [DOI] [PubMed] [Google Scholar]
- [22].Chen R, DesJardins SL. Investigating the impact of financial aid on student dropout risks: Racial and ethnic differences. The Journal of Higher Education. 2010;81(2):179–208. [Google Scholar]
- [23].Brent D, Melhem N, Donohoe MB, Walker M. The incidence and course of depression in bereaved youth 21 months after the loss of a parent to suicide, accident, or sudden natural death. American Journal of Psychiatry. 2009;166(7):786–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Hummer RA, Lariscy JT. Educational attainment and adult mortality. International handbook of adult mortality: Springer; 2011. p. 241–61. [Google Scholar]
- [25].Marmot M, Friel S, Bell R, Houweling TA, Taylor S, Health CoSDo. Closing the gap in a generation: health equity through action on the social determinants of health. The lancet. 2008;372(9650):1661–9. [DOI] [PubMed] [Google Scholar]
- [26].Braveman P, Gottlieb L. The Social Determinants of Health: It’s Time to Consider the Causes of the Causes. Public Health Reports. 2014;129(Suppl 2):19–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Jones CP. Confronting institutionalized racism. Phylon (1960-). 2002:7–22. [Google Scholar]
- [28].Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. The Lancet. 2017;389(10077):1453–63. [DOI] [PubMed] [Google Scholar]
- [29].Hardeman RR, Murphy KA, Karbeah JM, Kozhimannil KB. Naming Institutionalized Racism in the Public Health Literature: A Systematic Literature Review. Public Health Reports. 2018;133(3):240–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Harper SR. Race without racism: How higher education researchers minimize racist institutional norms. The Review of Higher Education. 2012;36(1):9–29. [Google Scholar]
- [31].Rosenbaum JE, Ahearn CE, Rosenbaum JE, Becker KI. Beyond Earnings and Social Reproduction: Can College Lead to Good Jobs Without Reproducing Social Inequalities? RSF: The Russell Sage Foundation Journal of the Social Sciences. 2016;2(1):90–110. [Google Scholar]