Parental Death and Cognitive Impairment: An Examination by Gender and Race–Ethnicity

Abstract

Objectives

We provide the first nationally representative longitudinal study of cognitive impairment in relation to parental death from childhood through early adulthood, midlife, and later adulthood, with attention to heterogeneity in the experience of parental death.

Methods

We analyzed data from the Health and Retirement Study (2000–2016). The sample included 13,392 respondents, contributing 72,860 person-periods. Cognitive impairment was assessed using the modified version of the Telephone Interview for Cognitive Status. Discrete-time hazard regression models were estimated to predict the odds of cognitive impairment.

Results

Both exposure and timing of parental death were related to the risk of cognitive impairment in late life and associations vary by gender. The detrimental effect of a father’s death was comparable for daughters and sons although exposure to a mother’s death had stronger effects on daughter’s than son’s risk of cognitive impairment. Father’s death at younger ages had the strongest effect on sons’ late-life risk of cognitive impairment, whereas mother’s death in middle adulthood had the strongest effect on daughters’ risk. We found no significant racial–ethnic variation in the association between parental death and cognitive impairment.

Discussion

It is important to explore the gender-specific pathways through which parental death leads to increased risk of cognitive impairment so that effective interventions can be implemented to reduce risk.

The death of a parent is the most common cause of bereavement among Americans. About 90% of Americans experience the loss of one or both parents by age 64, with a father’s death often preceding a mother’s death (Scherer & Kreider, 2019). Scholars argue that the death of a parent has profound and wide-ranging consequences for adults (Umberson, 2003). Yet, we know little about how the death of a parent may shape the risk of cognitive impairment, a serious and emerging public health concern in aging populations. Today, more than 16 million older Americans are living with cognitive impairment (Centers for Disease Control and Prevention, 2011), and a significant proportion has progressed or will progress to dementia that is associated with disability, increased need for medical and personal care, and premature death (Alzheimer’s Association, 2020). Recent research has identified risk factors—predominately behavioral and biological factors—in order to develop prevention strategies for cognitive impairment and dementia. An important next step is to identify specific life course events that contribute to cognitive impairment.

In this study, we examine the death of a parent as an underexplored risk factor for cognitive impairment. Using data from the Health and Retirement Study (HRS) 2000–2016, we provide the first nationally representative longitudinal assessment of the effects of parental death throughout the life course on the risk of cognitive impairment. Guided by the stress and life course perspective (Elder et al., 2003), we assess both exposure and timing of parental death from childhood through later adulthood in relation to the risk of cognitive impairment. Although most Americans will lose one or both parents by mid- to later life, the degree of risk for cognitive impairment may depend on when in the life course the loss occurred. Loss at younger ages may activate lifelong patterns of psychological distress that increase late-life risk of cognitive impairment, whereas loss in later life is more normative and may have less impact. Because gender shapes relationships, we distinguish death of the mother and death of the father in affecting both daughters’ and sons’ cognition. We further consider whether the relationship between parental death and cognitive impairment varies by race–ethnicity, given that racial–ethnic minorities, in particular Black Americans, are more likely than White Americans to experience cognitive impairment (Zhang et al., 2016) and early-life parental loss (Umberson et al., 2017). Findings will facilitate group-specific policy and intervention strategies to address bereavement experiences that may reduce the risk of cognitive impairment.

Parental Death and Health

Research on the effects of parental death has primarily focused on parental death in childhood or adolescence, finding substantial negative impacts including increased risk of depression, health-damaging behaviors, morbidity, and mortality from childhood to later adulthood (Smith et al., 2014). A few studies have also examined parental loss in later life, focusing exclusively on effects on physical and mental health. These studies suggest that bereaved adult children experience increased psychological distress and diminished physical health (Leopold & Lechner, 2015; Marks et al., 2007).

Only a few studies have explored the association between early parental death and late-life cognitive health, based primarily on regional samples or data outside the United States. These studies provide suggestive evidence that parental death, especially that experienced in early life, may be a risk factor for cognitive impairment (Conde-Sala & Garre-Olmo, 2020). For example, results from the Survey of Health, Ageing and Retirement in Europe suggested that experiencing parent death before age 16 was associated with an increased risk of dementia among older Europeans (Conde-Sala & Garre-Olmo, 2020). A study of Israeli men also found that parental death during childhood increased dementia risk in late life, but parental death after age 18 was associated with decreased risk (Ravona-Springer et al., 2012). Results based on a Chinese sample showed that older men who experienced the death of a mother before age 16 had higher odds of cognitive impairment than those who did not experience loss (Fu, 2019). The only study on parental death and cognitive health conducted in the United States was the Cache County Memory Study that analyzed older residents in Cache County of Utah and found a significant association between early-life death of a father (but not of a mother) and increased risk for dementia in late life (Norton et al., 2009, 2016). To the best of our knowledge, there are no nationally representative studies on the effects of parental death on cognitive impairment in the United States.

Parental Death and Cognitive Impairment: A Stress and Life Course Perspective

A life course approach suggests how major life events can influence lifelong trajectories of health and well-being (Elder et al., 2003); and stressful life events can trigger turning points that disrupt individual life trajectories to affect health (Pearlin & Skaff, 1996). The death of a parent is a life course turning point that increases stress levels in childhood as well as adulthood in ways that undermine mental and physical health (Marks et al., 2007; Umberson, 2003). Glymour and Manly (2008) emphasize that life course stressors also shape cognitive trajectories, arguing that “cognitive function is a developmental trajectory, and harmful exposures may influence the likelihood of impairments in old age by derailing the maturation trajectory, promoting pathological processes, or restricting compensation or resilience after pathological events” (p. 2240). Stress is pathogenic and can raise individuals’ vulnerability to cognitive impairment (Rothman & Mattson, 2010).

The death of a parent leads to many changes in the life of a child (Umberson, 2003). In particular, the death may activate psychosocial and health-related mechanisms (such as mental and physical health, health behaviors) that may increase the risk for cognitive impairment throughout life. For example, elevated stress following the loss of a parent may cause emotional and behavioral problems, such as feeling depressed, reduced social activities, smoking, and drinking (Umberson, 2003). Smoking increases white blood cells in the central nervous system, which can lead to neurological damage and impaired cognitive function (Swan & Lessov-Schlaggar, 2007). Heavy drinking may damage the brain’s white matter and increase the risk of both adverse brain outcomes and steeper declines in cognitive abilities (Hayes et al., 2016). Both reduced social engagement (Lee & Kim, 2016) and depression (Byers & Yaffe, 2011) are identified as risk factors for cognitive impairment and dementia. Stress from the death of a parent is also detrimental to physical health and increases the risk of chronic conditions, such as cardiovascular diseases and diabetes, which have been linked to cognitive impairment (Feinkohl et al., 2015). Thus, we hypothesize that:

Hypothesis 1 (Exposure to Death): Exposure to parental death at any point in the life course is associated with higher risk of cognitive impairment in late life (Hypothesis 1a); and this relationship arises in part through psychosocial and health-related pathways (Hypothesis 1b).

Timing of Parental Death

The timing of parental loss may also influence risk for cognitive decline: “Life course theory is temporal and contextual in locating people in history through birth years and in the life course through the social meanings of age-graded events and activities” (Elder & O’Rand, 1995, p. 454). Loss of a parent during middle and late adulthood is often expected, while loss of a parent during earlier life (i.e., childhood through early adulthood) is typically unexpected and more likely to result in deprivation of emotional and material resources. Moreover, an early loss may put the bereaved child at risk earlier in the life course, increasing their cumulative risk for cognitive impairment over a longer period of time. Both theoretical and empirical work suggests that early-life adversity has cumulative effects on health (e.g., self-rated health, functional limitations, mortality) throughout the life span (Haas, 2008). Advantages or disadvantages experienced in earlier life launch individuals onto distinct developmental pathways that persist throughout life, playing a more salient role in defining life course trajectories than those experienced in later life when individuals have more resources to cope with stress (Dannefer, 2003). Therefore, we hypothesize that:

Hypothesis 2 (Timing of Death): Loss of a parent at younger ages has stronger effects than loss at older ages on the risk of cognitive impairment.

Gender

Gendered relationship experiences begin early and continue throughout the life course (Stockard, 2006). For example, adult daughters are generally closer to their aging parents and are more likely than adult sons to serve as a caregiver to a sick parent (Fingerman et al., 2020). The emotional and social closeness of daughters to their parents suggests that the death of a parent may lead to more emotional distress for daughters than for sons. The parent’s gender may also matter. Children of all ages tend to feel closer and be more involved with mothers than fathers, and adult children also tend to provide more support and care to older mothers than fathers (Fingerman et al., 2020). Therefore, the loss of a mother may have a stronger impact on adult children’s (perhaps more so for daughters than for sons) cognitive health than the loss of a father.

Hypothesis 3 (Gender of Child): Loss of a parent (both exposure and timing) has stronger effects on daughters’ than sons’ risk of cognitive impairment.

Hypothesis 4 (Gender of Parent): Loss of a mother (both exposure and timing) has stronger effects than loss of a father on children’s risk of cognitive impairment, and more so for daughters than for sons.

Race–Ethnicity

Race–ethnicity is another fundamental structural factor to define individual life context and shape the process of cognitive aging (Zhang et al., 2016). Compared to White Americans, Black Americans have a much higher risk and earlier onset of cognitive impairment, with this disparity increasing with age (Zhang et al., 2016). Although research on Hispanic populations is more limited, some evidence suggests that Hispanic adults are at greater risk for cognitive impairment than other racial–ethnic groups (Novak & Riggs, 2004). Because Black Americans have a lower life expectancy than White Americans, Black people are more likely to experience parental death at younger ages (Umberson et al., 2017), which may create greater cumulative stress and lead to more cognitive impairment over time. Although Hispanics have an overall lower mortality rate than non-Hispanic Whites despite their lower socioeconomic status (SES)—often referred to as the Hispanic Paradox—this paradox has great diversity due to heterogeneity of the Hispanic population, for example, by nativity and countries of origin (Borrell & Lancet, 2012). Moreover, a recent analysis of data from the 2014 Survey of Income and Program Participation found that Hispanics experience parental death at earlier ages than non-Hispanic Whites but later than non-Hispanic Blacks (Scherer & Kreider, 2019). Moreover, compared to their White counterparts, both Black and Hispanic Americans may be more vulnerable to parental death due to more stress burden associated with racism, discrimination, poverty, and illness. Thus, parental loss may have a stronger impact on racial–ethnic minority Americans than White Americans. We expect the negative effects of parental loss may be stronger for Black than Hispanic populations, because Hispanic cultural values of “familism” that emphasize strong commitment to extended family ties and support may buffer some negative effects of parental loss (Landale et al., 2006).

Hypothesis 5 (Racial–Ethnic Variation): Loss of a parent (both exposure and timing) has stronger effects on risk of cognitive impairment for Black and, to a lesser extent, Hispanic Americans than White Americans.

Other Covariates

Both parental death and cognitive impairment are related to sociodemographic factors. The likelihood of experiencing parental death and cognitive impairment increases with age. People with higher SES (e.g., higher education and income, being married, higher childhood SES) have a lower risk of cognitive impairment (Garcia et al., 2020; Liu et al., 2020) and also tend to experience parental death at older ages than people with lower SES (Scherer & Kreider, 2019). People born within the United States or in the South tend to have poorer cognitive health and higher mortality risk than people born outside of the United States or in non-Southern regions (Garcia et al., 2020; Glymour et al., 2011). Finally, other confounders such as shared genetics may predict the likelihood of both parental death and cognitive impairment.

Method

Data and Sample

We used data from the 2000 wave of the HRS with follow-up until 2016. The 2000 wave of the HRS surveyed a national sample of 19,579 noninstitutionalized adults born before 1948 (aged 52 and older in 2000) and their spouses (HRS, 2017). Since 2000, HRS has collected consistent measures on cognition (Crimmins et al., 2016). The survey oversamples Black and Hispanic people and collects (by telephone or in person) detailed information on cognitive, physical, economic, work, and family conditions approximately every 2 years. The HRS has high response rates (81%–89%) in each wave and provides a unique opportunity to address parental death and cognitive health because of its large sample size, long-term follow-up, and high-quality measures of cognition and other key variables.

About 5% of the interviews in the analytic sample were conducted through proxies (spouses or children) for those who could not participate in the survey due to health issues or death. In the analysis, we included both self- and proxy reports (and control for proxy interviews) to avoid underestimating cognitively impaired cases. We restricted the sample to 19,172 non-Hispanic White (hereafter “White”), non-Hispanic Black (hereafter “Black”), and Hispanic respondents due to concerns of small sample size in other subgroups. After excluding respondents who had cognitive impairment at the baseline survey, the final analytic sample included 13,392 community-dwelling respondents (contributing 72,860 person-periods) aged 52–100 in 2000 and had no signs of cognitive impairment at the baseline survey.

Measures

Cognitive impairment

For self-respondents, HRS assessed cognitive function via the modified version of the Telephone Interview for Cognitive Status. We followed previous studies in summing the following cognitive items: immediate and delayed recall of a list of 10 words (1 point for each), five trials of serial 7s (i.e., subtract 7 from 100, and continue subtracting 7 from each subsequent number for a total of five trials, 1 point for each trial), and backward counting (2 points). The final summary score ranges from 0 (severely impaired) to 27 (high functioning; Crimmins et al., 2016). We used the summed scores to identify two categories: cognitive impairment (0–11) and no cognitive impairment (12–27; Crimmins et al., 2016).

For individuals who were unable to participate in the cognitive tests due to health issues, cognitive status was measured with an 11-point scale using the proxy’s assessments of (a) the respondent’s memory (0 = excellent, 4 = poor); (b) the respondent’s limitations in five instrumental activities of daily living: managing money, taking medication, preparing hot meals, using the phone, and shopping for groceries (0–5); and (c) interviewer’s assessment of the respondent’s difficulty completing the interview because of cognitive limitations (0 = none, 1 = some, and 2 = prevented completion). Proxy respondents with a summary score of 3–11 were classified as having cognitive impairment, and those with a score of 0–2 were classified as having no cognitive impairment (Crimmins et al., 2016).

Parental death

Four questions were asked about parental death: (a) “Is your mother still living?”; (b) “If not living, in what year did she die/pass away?”; (c) “Is your father still living?”; and (d) “If not living, in what year did he die/pass away?” Based on responses to these questions, we included a series of time-varying measures for parental death to reflect both the exposure and timing of the death. First, we started with a basic measure of experiencing any parental death at any life stage (1 = yes; 0 = no). Second, we considered the cumulative exposure of parental death: one parent died, both parents died, and no parent died (i.e., both parents alive, reference). Third, we further distinguished mother death (1 = yes; 0 = no) and father death (1 = yes; 0 = no). Finally, we included three variables to measure the timing of parental death: age at first parent death, age at mother’s death, and age at father’s death, each of which was categorized into death at ages 0–29 (early life), 30–59 (middle adulthood), 60 and older (late adulthood), and parent alive (reference). Additional analysis (results available upon request) using alternative categorizations of age at death (e.g., 0–17, 18–49, 50–64, 65+) suggested similar findings.

Psychosocial and health-related factors

We included three sets of psychosocial and health-related covariates (all time-varying): physical health, mental health, and health behaviors. Physical health included self-rated health (1 = poor/fair, 0 = good/very good/excellent) and chronic conditions (a summary score of the presence of four major chronic conditions, ranged 0–4: diabetes, stroke, heart disease, and high blood pressure). Mental health was measured by depressive symptoms of the Center for Epidemiologic Studies Depression scale, including the following feelings (with response options of yes or no): (a) depressed, (b) everything was an effort, (c) sleep was restless, (d) happy, (e) lonely, (f) enjoyed life, (g) sad, and (h) could not “get going.” Items 4 and 6 were reverse-coded. After summing up all eight items, we created a dichotomous indicator of clinically significant depressive symptoms, using a cutoff score of 3 (Steffick, 2000). Health behaviors included smoking, drinking, and social activity. Smoking was a dichotomous measure: 0 = current nonsmoker and 1 = current smoker. Following previous studies (Liu et al., 2020), drinking included three categories: current nondrinker (reference), current light drinker (<seven alcoholic beverages per week), and current heavy drinker (≥seven alcoholic beverages per week). Social activity measured how often the respondents socialized with friends, neighbors, or family (1 = daily to monthly, 0 = yearly to never).

Sociodemographic covariates

We controlled for age (in years), gender (1 = women, 0 = men), race–ethnicity (non-Hispanic White [reference], non-Hispanic Black, Hispanic), education (less than high school [reference], high school graduate, some college, and college graduate and above), born in the South (1 = yes, 0 = no), nativity (1 = foreign-born, 0 = U.S.-born), marital status (1 = unmarried, 0 = married), household income, and childhood adversity. Total household income included the respondent’s and the spouse’s income from all sources. We used the RAND version of household income which included consistently imputed missing values across waves (RAND HRS Data, 2016). We took the natural log of the income to adjust the skewed distribution. Childhood adversity was measured as a summary scale (ranged 0–3) of three dichotomous items: childhood family was poor, mother’s education was less than 8 years, and father’s education was less than 8 years (Zhang et al., 2016). We also included an indicator of whether a proxy respondent was used to assess the cognitive status (1 = yes; 0 = no). Although we did not have direct measures for shared genetics, we controlled for parents’ cognitive status and other family member deaths which accounted for, to some extent, the shared influence of family. Parents’ cognitive status indicated whether either parent had cognitive impairment (1 = yes, 0 = no). Death of other family members was assessed using three dichotomous measures indicating the death of spouse, sibling, and child separately (1 = yes, 0 = no). All covariates were measured as time-varying except for gender, race–ethnicity, education, born in the South, nativity, and childhood adversity, which were measured at the baseline as time-invariant.

Statistical Analyses

We estimated discrete-time hazard models to predict the risk of cognitive impairment. Specifically, we created person-period record files and used a logit model. A respondent contributes an observation for each wave up to the onset of cognitive impairment or censoring (i.e., loss to follow-up or death). Because the sample is restricted to those who had no cognitive impairment at the baseline, the estimates reflect the effects of the independent variables on new onsets of cognitive impairment. The discrete-time hazard model is specified as:

$$\begin{array}{l}log\frac{h(t_{ij})}{1-h(t_{ij})}=\underset{j=1}{\sum ^8}{\mathrm{\alpha }}_j{D}_{ij}+{{\mathrm{X}}_i}^{\prime }{B}_1+{Z_{ij}}^{\prime }{B}_2\end{array}$$ (1)

where $h(t_{ij})$ indicates the discrete-hazard (i.e., conditional probability) of the onset of cognitive impairment for individual i at wave j; $\underset{j=1}{\sum ^8}{\mathrm{\alpha }}_j{D}_{ij}$ represents the set of multiple intercepts for HRS 2000–2016, one per period; ${\mathrm{X}}_i$ indicates the vector of time-invariant covariates; and Z_ij indicates the vector of time-varying covariates. B₁ and B₂ are corresponding coefficient vectors.

We estimated a series of models. Model 1 included exposure to any parental death controlling for sociodemographic covariates. Model 2 added psychosocial and health-related factors as additional covariates to see if the relationship between exposure to parental death and cognitive impairment was partly explained by these mechanisms. Models 3–7 considered specific characteristics of parental death related to cognitive impairment with all covariates controlled. Specifically, Model 3 tested cumulative exposure to parental death by comparing “one parent died” and “two parents died” with “no parent died.” Model 4 included mother death and father death. Models 5–7 included age at first parental death, age at mother’s death, and age at father’s death separately to address the timing of parental death. We included the parental death variables in separate models because of the collinearity of these variables. Missing data, with the most missings for the age at mother’s (14.5%) and father’s (16.6%) death, were handled using the Multiple Imputation via Chained Equations with 10 generated multiply imputed data sets (White et al., 2011).

Results

Table 1 presents weighted descriptive statistics for all variables for the total sample as well as by gender and race–ethnicity in the baseline survey. The results show that Black (12.51%) and Hispanic (11.77%) respondents had much higher proportions of cognitive impairment during the subsequent waves than White respondents (6.90%), while no significant differences in the prevalence of cognitive impairment were detected by gender. More than 90% of respondents had experienced one or both parents’ death by the baseline year, with most of them experiencing first parent death during their middle adulthood (ages 30–59), regardless of gender and race–ethnicity. Women, in general, tended to be more likely than men to lose their parents. Although there were no significant racial–ethnic differences in the overall prevalence of parental loss at baseline, Black and Hispanic respondents were more likely to experience early-life parental loss than White respondents: 35.38% of Black and 32.45% of Hispanic respondents in contrast to 28.97% of White respondents experienced their first parental loss at ages 0–29. Black and Hispanic respondents were almost twice as likely as White respondents to lose their mothers by age 29.

Table 1.

Weighted Descriptive Statistics of All Analytic Variables by Gender and Race–Ethnicity, Health and Retirement Study 2000 (N of respondents = 13,392)

Variables	Total	Women	Men	Non-Hispanic White	Non-Hispanic Black	Hispanic
Cognitive impairmenta	7.46	7.46	7.45	6.90	12.51*	11.77*
Exposure to parental death
Any parent died	94.65	95.07	94.12*	94.64	93.65	96.17
Cumulative exposure
No parent died	5.35	4.93	5.88*	5.36	6.35	3.83
One parent died	22.80	22.14	23.63	22.22	25.39	29.04*
Both parents died	71.85	72.93	70.49*	72.42	68.26*	67.13*
Mother died	76.32	77.16	75.24*	76.56	74.00	75.22
Father died	90.19	90.83	89.36*	90.49	87.90*	88.07
Timing of parental death (ref. = no death)
Age at first parental death
0–29	29.59	30.48	28.46*	28.97	35.38*	32.45*
30–59	62.62	62.33	63.00	63.17	56.38*	61.45
60+	2.44	2.26	2.66	2.49	1.89	2.27
Age at mother’s death
0–29	9.60	10.08	8.99*	8.85	17.01*	12.51*
30–59	53.46	53.67	53.18	53.81	48.43*	54.18
60+	13.26	13.41	13.07	13.90	8.56*	8.53*
Age at father’s death
0–29	22.72	23.25	22.02	22.57	23.91	23.60
30–59	62.11	62.35	61.81*	62.45	58.88*	60.57
60+	5.36	5.23	5.53*	5.47	5.11	3.90
Psychosocial and health-related factors
Poor self-rated health	19.74	20.70	18.50*	18.26	27.01*	35.30*
Chronic conditions	0.80 (0.87)	0.77 (0.86)	0.85 (0.89)*	0.79 (0.87)	1.01 (0.91)*	0.75 (0.86)
Depression	19.88	23.09	15.77*	18.88	25.07*	29.95*
Current smoker	15.76	15.63	15.92	15.53	20.65*	13.14
Current drinker
Nondrinker	67.42	75.83	56.65*	66.07	79.87*	73.73*
Light drinker	23.10	19.35	27.89*	23.95	14.46*	20.05
Heavy drinker	9.48	4.82	15.46*	9.98	5.67*	6.22*
Social activity
Yearly–never	22.45	24.49	19.84*	21.57	27.25*	30.95*
Daily–monthly	77.55	75.51	80.16*	78.43	72.75*	69.05*
Sociodemographic covariates
Female	56.12			55.81	59.74*	56.76
Race–ethnicity
Non-Hispanic White	87.96	87.47	88.60
Non-Hispanic Black	6.84	7.28	6.28*
Hispanic	5.20	5.25	5.12
Age	64.69 (9.34)	65.16 (9.53)	64.08 (9.04)*	65.04 (9.44)	62.27 (8.22)*	61.94 (7.96)*
Education
<High school	16.36	17.35	15.08*	13.51	27.02*	50.45*
High school graduate	37.52	40.73	33.43*	38.61	33.63*	24.32*
Some college	22.80	23.77	21.56*	23.15	22.85	16.85*
College and above	23.32	18.15	29.93*	24.73	16.50*	8.38*
Household income (ln)	10.23 (1.20)	10.10 (1.22)	10.39 (1.64)*	10.32 (1.09)	9.72 (1.45)*	9.31 (1.92)*
Born in the South	28.57	29.29	27.66	25.67	72.71*	19.74
Foreign-born	6.67	6.31	6.95	4.11	6.77*	49.83*
Unmarried	33.93	42.99	22.33*	32.04	52.64*	41.25*
Proxy report	5.34	2.03	9.57*	5.03	7.27*	7.95*
Childhood adversity	0.91 (1.01)	0.92 (1.01)	0.89 (1.01)	0.81 (0.96)	1.43 (1.05)*	1.89 (0.99)*
Parent cognitive impairment	11.84	11.41	12.40	11.69	14.46*	11.03
Spousal death	34.45	46.65	18.84*	34.92	33.90	27.19*
Child death	7.61	8.85	6.03*	6.84	14.68*	11.42*
Sibling death	43.11	43.15	43.05	40.39	62.73*	63.24*
Unweighted N	13,392	7,778	5,614	11,192	1,369	831

Notes: Values for categorical variables are in percent. The mean values, followed by standard deviations in parentheses, are presented for all other variables.

^aAll cognitive impairment cases in the baseline survey were excluded. The reported percentages of cognitive impairment were calculated based on person-period files (N = 72,860), reflecting cognitive impairment onset across subsequent survey waves.

*Indicates statistically significant difference compared to women or non-Hispanic White respondents at the .05 level.

Other variables largely show the expected distribution. Men tended to report better health, less depression, and more social activities but more chronic conditions and drinking than women. Blacks generally reported worse health than Whites except for drinking: Blacks were less likely to be either light or heavy drinkers than Whites. Hispanics were more likely to report poor health, depression, and daily to monthly activities but less likely to be heavy drinkers than Whites. Women and racial–ethnic minority respondents tended to have less education and income and were more likely to be unmarried compared to men and White respondents. Blacks and Hispanics were more likely to experience childhood adversity and child and sibling deaths than Whites. A larger proportion of Black respondents reported cognitive impairment of their parents than White respondents.

Table 2 presents results from discrete-time hazard models estimating the relationship between exposure to any parental death and cognitive impairment. Results of Model 1 suggest that older adults who experienced any parental loss had higher odds of cognitive impairment (odds ratio [OR] = 1.950, p < .001) than those who had both parents alive, after controlling for sociodemographic covariates. The magnitude of this difference decreased by about 10% but remained statistically significant after we added psychosocial and health-related factors in Model 2 (OR = 1.760, p < .01). Poor self-rated health, chronic conditions, and depression were all related to higher odds of cognitive impairment, and more participation in social activities was related to lower odds of cognitive impairment (Model 2). Results from Karlson–Holm–Breen analysis (Supplementary Table S1) suggested that the mediating effect of these psychosocial and health-related factors, especially physical health, was statistically significant (p < .001). Results of sociodemographic covariates given in Table 2 were mostly consistent with prior research: older ages, lower education and income, being racial–ethnic minority, being unmarried, born in the South, proxy report, and parental cognitive impairment were all associated with a higher risk of cognitive impairment.

Table 2.

Estimated Odds Ratios of Cognitive Impairment Incidence by Overall Exposure to Any Parental Death, Health and Retirement Study 2000–2016 (N of person-periods = 72,860, N of respondents = 13,392)

Variables	Model 1		Model 2
	OR	SE	OR	SE
Any parent died	1.950***	0.383	1.760**	0.347
Sociodemographic covariates
Female	0.972	0.033	0.974	0.034
Race–ethnicity (ref: Non-Hispanic White)
Non-Hispanic Black	1.668***	0.082	1.616***	0.082
Hispanic	1.657***	0.098	1.630***	0.095
Age	1.086***	0.002	1.084***	0.002
Education (ref: <high school)
High school graduate	0.911*	0.040	0.943	0.040
Some college	0.818***	0.040	0.862**	0.042
College and above	0.637***	0.040	0.681***	0.043
Household income (ln)	0.933***	0.013	0.957**	0.015
Unmarried	1.156*	0.039	1.129**	0.040
Born-south	1.247***	0.042	1.212***	0.040
Foreign-born	1.100	0.055	1.100	0.057
Proxy	2.300***	0.148	2.305***	0.143
Childhood adversity	0.972	0.015	0.956**	0.015
Parent cognitive impairment	1.641***	0.093	1.646***	0.095
Spousal death	1.329***	0.047	1.319***	0.049
Child death	1.208***	0.057	1.201***	0.057
Sibling death	1.125***	0.025	1.106**	0.025
Psychosocial and health-related factors
Poor self-rated health			1.320***	0.057
Chronic conditions			1.190***	0.020
Depression			1.118*	0.049
Current smoker			1.076	0.064
Current drinker (ref: nondrinker)
Light drinker			0.984	0.034
Heavy drinker			1.051	0.053
Daily–monthly social activity			0.717***	0.026
Wald chi-square	8,472***		8,617***

***p < .001, **p < .01, *p < .05.

Table 3 presents estimated ORs of cognitive impairment predicted by specific characteristics of parental death exposure controlling for all psychosocial and health-related factors and sociodemographic covariates. Model 3 tested the effect of cumulative exposure to parental death, suggesting that respondents who lost either both parents or only one parent had a significantly higher risk of cognitive impairment than those who had both parents alive, and the risk was higher for those who lost two parents (OR = 1.951, p < .01) than one parent (OR = 1.482, p < .05). Model 4 distinguished mother’s death and father’s death, suggesting significant effects of both father’s death (OR = 1.469, p < .05) and mother’s death (OR = 1.296, p < .001) on the risk of cognitive impairment.

Table 3.

Estimated Odds Ratios of Cognitive Impairment Incidence by Cumulative Exposure to Parental Death, Death of Mother and Father and Timing of Parental Death, Health and Retirement Study 2000–2016 (N of person-periods = 72,860, N of respondents = 13,392)

Variables	Model 3	Model 4	Model 5 Age at first parental death	Model 6 Age at mother’s death	Model 7 Age at father’s death
Cumulative exposure (ref. = no parent died)
One parent died	1.482*
Two parents died	1.951**a
Mother died		1.296***
Father died		1.469*
Age at parental death (ref. = no death)
0–29			1.564*	1.123	1.431*
30–59			1.550*	1.308***	1.334
60+			1.427	1.152*	1.281
Wald chi-square	8,584***	8,587***	7,747***	7,705***	7,725***

Note: Models controlled for all sociodemographic covariates and psychosocial and health-related factors.

^aDifference between “Both parents died” and “One parent died” is statistically significant at the .05 level.

***p < .001, **p < .01, *p < .05. Standard errors in parentheses.

Models 5–7 consider how the timing of parental death is associated with the risk of cognitive impairment (Table 3). In general, experiencing the first parental loss during early life (i.e., ages 0–29; OR = 1.564, p < .05) and middle adulthood (i.e., ages 30–59; OR = 1.550, p < .05) was both associated with higher odds of cognitive impairment (Model 5), although experiencing the first parental loss during late adulthood (i.e., 60 and older) was not significantly associated with the risk of cognitive impairment. Such associations, however, differ for mother’s and father’s death. Specifically, mother’s death during middle adulthood (Model 6, OR = 1.308, p < .001) and father’s death during early life (Model 7, OR = 1.431, p < .05) had the strongest effect on children’s risk of cognitive impairment.

To further examine group variation, we stratified the analysis by gender (Table 4) and race–ethnicity (Table 5). Results in Table 4 suggest that for both women and men, experiencing any parent’s death was associated with higher risk of cognitive impairment (Model 2, OR = 2.002, p < .05 for women; OR = 1.677, p < .05 for men), and those who had both parents die were especially vulnerable to late-life cognitive impairment (Model 3, OR = 2.322, p < .01 for women; OR = 1.761, p < .05 for men). Results in Model 4 of Table 4 further suggest that mother’s death was associated with daughters’ risk of cognitive impairment (OR = 1.530, p < .001), but not sons’ risk (OR = 1.116, p > .05); while father’s death was associated with both daughters’ (OR = 1.525, p < .05) and sons’ (OR = 1.433, p < .05) risk of cognitive impairment.

Table 4.

Estimated Odds Ratios of Cognitive Impairment Incidence by Death of Parent for Women and Men

Women (N of person-periods = 43,495, N of respondents = 7,778)
Variables	Model 2	Model 3	Model 4	Model 5 First parent	Model 6 Mother	Model 7 Father
Exposure to parental death
Any parent died	2.002*
Cumulative exposure (ref = no parent died)
One parent died		1.517
Both parents died		2.322**
Mother died			1.530***
Father died			1.525*
Age at parental death (ref. = no death)
0–29				1.744	1.508**	1.398
30–59				1.770	1.584***	1.382
60+				1.569	1.292*	1.331
Wald chi-square	5,114***	5,098***	5,099***	4,582***	4,545***	4,582***
Men (N of person-periods = 29,365, N of respondents = 5,614)
	Model 2	Model 3	Model 4	Model 5 First parent	Model 6 Mother	Model 7 Father
Exposure to parental death
Any parent died	1.677*
Cumulative exposure (ref = no parent died)
One parent died		1.555
Both parents died		1.761*
Mother died			1.116
Father died			1.433*
Age at parental death (ref. = no death)
0–29				1.583	0.833	1.512*
30–59				1.505	1.107	1.298
60+				1.462	1.070	1.277
Wald chi-square	3,456***	3,450***	3,452***	3,129***	3,114***	3,117***

Note: Models controlled for all sociodemographic covariates and psychosocial and health-related factors.

***p < .001, **p < .01, *p < .05.

Table 5.

Estimated Odds Ratios of Cognitive Impairment Incidence by Death of Parent for Non-Hispanic Whites, Non-Hispanic Blacks, and Hispanics

Non-Hispanic Whites (N of person-periods = 62,452, N of respondents = 11,192)
Variables	Model 2	Model 3	Model 4	Model 5 First parent	Model 6 Mother	Model 7 Father
Exposure to parental death
Any parent died	1.780*
Cumulative exposure (ref. = no parent died)
One parent died		1.462
Both parents died		1.967*
Mother died			1.316**
Father died			1.507*
Age at parental death (ref. = no death)
0–29				1.467	1.106	1.406*
30–59				1.435	1.309**	1.292
60+				1.378	1.174	1.258
Wald chi-square	7,135***	7,102***	7,106***	6,349***	6,333***	6,328***
Non-Hispanic Blacks (N of person-periods = 6,501, N of respondents = 1,369)
	Model 2	Model 3	Model 4	Model 5 First parent	Model 6 Mother	Model 7 Father
Exposure to parental death
Any parent died	1.451
Cumulative exposure (ref = no parent died)
One parent died		1.290
Both parents died		1.667
Mother died			1.208
Father died			1.500
Age at parental death (ref. = no death)
0–29				1.495	1.137	1.606
30–59				1.489	1.298	1.508
60+				1.408	1.129	1.463
Wald chi-square	811***	811***	814***	749***	737***	755***
Hispanics (N of person-periods = 3,907, N of respondents = 831)
	Model 2	Model 3	Model 4	Model 5 First parent	Model 6 Mother	Model 7 Father
Exposure to parental death
Any parent died	2.128
Cumulative exposure (ref = no parent died)
One parent died		1.850
Both parents died		2.330
Mother died			1.264
Father died			1.433
Age at parental death (ref. = no death)
0–29				2.044	1.171	1.399
30–59				2.285	1.332	1.513
60+				0.977	0.860	1.127
Wald chi-square	571***	571***	570***	534***	528***	541***

Note: Models controlled for all sociodemographic covariates and psychosocial and health-related factors.

***p < .001, **p < .01, *p < .05.

The effect of parental death timing also depended on the gender of both parent and child (Models 5–7 in Table 4). For daughters, mothers’ death during middle adulthood had the strongest effect on the risk of cognitive impairment (Model 6, OR = 1.584, p < .001) followed by mothers’ death during early life (Model 6, OR = 1.508, p < .01) and late adulthood (Model 6, OR = 1.292, p < .05), while the effect of timing of father’s death was not significant for daughters (Model 7). For sons, losing the father during early life had the strongest effect on their risks of cognitive impairment (Model 7, OR = 1.512, p < .05), while the effect of timing of mother’s death was not significant for sons (Model 7).

We present results stratified by race–ethnicity in Table 5. In general, parental loss was associated with higher odds of cognitive impairment in all racial–ethnic groups, while it was only statistically significant among White respondents, perhaps due to relatively small sample sizes of the minority groups. Additional analysis using both Wald tests and adding interaction terms (results not shown but available upon request) showed no significant racial–ethnic differences in any of the associations between parental loss and cognitive impairment.

Sensitivity Analysis

We applied multiple imputation for missing data on the timing of parental death which accounted for about 20% of the sample. We conducted robustness checks by using different approaches to handle missing values on age at parental death. First, we included those missing cases as a separate category without imputations (Supplementary Table S2). Second, we excluded those missing cases on age at parental death (Supplementary Table S3). Third, our additional analysis (results not shown) suggested that compared to the sample without missing data on age at parental death, respondents who had missing values on age at parental death were more likely to be women, older, less educated, with lower income, and more likely to be proxy respondents than those who had complete data on age at parental death, suggesting a possible selection bias due to missing reports. Therefore, we applied a Heckman-type correction to adjust for this sample selection bias (Heckman, 1979; Supplementary Table S4). This approach consists of modeling the probability that a respondent had no missing values on age at parental death using logistic regression models, conditional on a set of predictors. Then, in the final analysis only including individuals without missing data of age at parental death, cognitive impairment was modeled as a function of a set of independent variables plus the predicted probability of not missing on age at parental death. Following this Heckman-type correction, estimates of cognitive impairment should be interpreted as being adjusted for control covariates, as well as for the selection bias related to missing reports of age at parental death. Finally, we tested age at parental death as a continuous variable (Supplementary Table S5). Results from all these different approaches are robust and consistent with our main results.

Discussion

Bereavement is a stressful life course experience that has lasting effects on the health of surviving family members, with increasing importance in the coronavirus disease 2019 (COVID-19) pandemic. It is estimated that over 40,000 children in the United States have lost a parent to COVID-19 (Kidman et al., 2021). This study extends research on bereavement and health to consider how the death of a parent at various points in the life course is associated with risk for cognitive impairment in late life. We find a significant impact of parental loss on cognitive impairment, a finding that is consistent with our expectation (Hypothesis 1a) as well as the broader bereavement literature suggesting far-reaching detrimental impacts of bereavement on individual health and well-being (Shear et al., 2013).

We go beyond previous literature to show that experiencing a parent’s death was linked to a higher risk of cognitive impairment partially through psychosocial and health-related pathways. A stress and life course perspective suggests that experiencing parental death is stressful and may lead to emotional and behavioral changes, such as feeling depressed, reducing social activities, smoking, and drinking—and in turn have negative health consequences. Such emotional and behavioral problems may also damage cognitive function (Byers & Yaffe, 2011). Our results support this hypothesis (Hypothesis 1b). Yet, we found that the effects of parental death on cognitive impairment remained significant after these psychosocial and health-related mechanisms were taken into account, suggesting that future research should explore other factors (e.g., social support, shared genetic and environmental influences) that help explain the link between parental death and cognitive impairment.

Importantly, our results highlight gender of the parent and the child as key in moderating the impact of parental loss on cognitive impairment. Generally, adult children’s cognition is more affected by the death of their same-gender parent. For example, we found that exposure to a mother’s death had stronger effects on daughters’ than on sons’ risk of cognitive impairment (partly supporting Hypothesis 3), although the detrimental effect of exposure to a father’s death was comparable for daughters and sons. The mother–daughter bond is characterized by greater closeness than other gendered parent–child bonds, and these bonds tend to strengthen with age as mothers and daughters work together to maintain family ties (Fingerman et al., 2020). Therefore, the loss of a mother may result in greater stress and stronger effects on the cognitive functioning of daughters than of sons.

Because mothers generally play a more pivotal role than fathers in providing a range of support to children throughout the life course (Fingerman et al., 2020), we had hypothesized that a mother’s death would have stronger effects than a father’s death on the risk of cognitive impairment (Hypothesis 4), but our results do not support this hypothesis. We found that a father’s death had stronger effects, especially on sons’ risk of cognitive impairment (inconsistent with Hypothesis 4). Although longstanding literature has emphasized the greater closeness of children with mothers than with fathers (Fingerman et al., 2020), recent studies have pointed to the symbolic meanings of the father (e.g., a sense of family, father as a friend), as more important for sons’ than daughters’ well-being (Office on Child Abuse and Neglect, 2006). Loss of the father as a symbolic role model may have far-reaching effects on bereaved sons’ socioeconomic achievement and deviant behavior (Lamb, 2010), triggering turning points in life trajectories and increasing stress that, in turn, potentially damage cognitive functioning.

The timing of parental death also matters for cognitive functioning in ways that vary by gender. Consistent with the stronger same-gender child–parent effects described above, we found that the timing of a father’s death mattered more for sons’ risk of cognitive impairment while the timing of mother’s death mattered more for daughters’ risk. Specifically, a father’s death during early life had the strongest effect on sons’ risk of cognitive impairment (partly consistent with Hypothesis 2), whereas experiencing a mother’s death in midlife had the strongest effect on a daughter’s risk of cognitive impairment (partly consistent with Hypothesis 4). Previous studies suggest that involvement with fathers in childhood to young adulthood is especially important for sons’ development and outcomes (Lamb, 2010). Therefore, the loss of a father earlier in life may be especially detrimental to sons’ cognitive functioning in later life. In contrast, a mother’s death may have more impact on health and cognition for daughters in midlife as they become more focused on relationships with others (Fingerman et al., 2020). Mothers may be especially important to women in midlife, a period marked by stress due to multiple role demands and life changes, including menopause (Freund & Ritter, 2009).

Finally, although racial–ethnic minority groups were more likely to suffer from early-life parental death, we found no evidence for significant racial–ethnic differences in the effects of parental death on cognitive impairment—inconsistent with Hypothesis 5. We note that our sample sizes for racial–ethnic minority groups are relatively small, which may limit our statistical power to detect differences. Moreover, Black older adults are more selective of healthy individuals due to higher rates of mortality than White older adults, leaving a more robust Black population in our sample, which may bias estimates. Notably, even though the general patterns are the same, the specific pathways through which parental death shapes cognitive function may vary across racial–ethnic groups, a topic warranting future research attention. Importantly, Black and Hispanic populations are at much greater risk than Whites of losing a parent during early life, and this disadvantage in bereavement risk should be noted.

Several limitations should be noted. First, the cognitive impairment classification is limited because the battery of neuropsychological tests in the HRS may be less accurate than a standard neurological diagnosis examination. Second, although we worked from a causal framework to develop our hypotheses, we could not fully determine causality because of possible unmeasured spuriousness. In particular, the association between parental death and cognitive impairment might reflect shared genetic and environmental influences that could not be fully addressed using our data. Notably, we were able to control for the cognitive impairment of parents; a significant association between parents’ and children’s cognitive status provided some indirect evidence of shared genetics of cognition. Yet, our findings were robust after controlling for parental cognitive impairment, indicating that the relationship between parental death and children’s cognitive impairment likely goes beyond shared genetics. These findings are also consistent with recent research on bereavement and mortality risk, net of familial genetic confounding factors (Song et al., 2019). Nevertheless, we acknowledge our limited ability to test shared genetics in explaining our findings. Third, most respondents (about 95%) in our sample had experienced the death of one or both parents. We note that parental death is a highly variable exposure, occurring across all life course stages. Isolating its overall effect on the risk of cognitive impairment hides the heterogeneity of the experience. Future studies should continue to understand diverse experiences of parental death as well as their effects on individuals’ health. Moreover, the small size of racial–ethnic minority groups in our sample limits our ability to examine the intersection of gender and race–ethnicity in shaping the link between parental death and cognitive impairment, which warrants future research attention. Finally, cognitive impairment may precipitate death, thus our sample may be selective of those with better cognitive health. This selection may be more pronounced for the Black sample due to higher rates of cognitive impairment and mortality. This may partially explain the lack of racial–ethnic differences in the effects of parental death.

Conclusions

Guided by the stress and life course perspective (Elder et al., 2003), the current study extends our understanding of the long-term effects of parental death on adult children’s cognitive function in late life. Prior research has focused on parental loss from childhood through early adulthood, but the present results underscore the importance of considering parental loss throughout the life course, and in relation to cognitive health—a growing public health concern as the population ages. The results, which are based on longitudinal data drawn from a nationally representative sample of U.S. older adults with 16 years of follow-up, suggest that both exposure and timing of parental death are related to the risk of cognitive impairment, and these associations vary by gender but not by race–ethnicity. These results highlight the importance of designing gender-specific interventions to reduce stress following the death of a parent, with the potential to reduce the long-term risk of cognitive impairment.

Funding

This research was supported by the National Institute on Aging, grants P30AG066614, R01AG054624, and R01AG061118, and by the U.S. Eunice Kennedy Shriver National Institute of Child Health and Human Development, grant P2CHD042849.

Conflict of Interest

None declared.

Author Contributions

H. Liu developed the conceptual framework, drafted the paper, and guided the analysis. Z. Lin conducted the analysis and contributed to writing. D. Umberson contributed to the conceptual framework and writing.