Childhood and adulthood trauma associate with cognitive aging among Black and White older adults

Abstract

Sociocontextual factors powerfully shape risk for age-related cognitive impairment, including excess risk burdening medically-underserved populations. Lifecourse adversity associates with cognitive aging, but harms are likely mitigable. Understanding population-salient relationships and sensitive periods for exposure is crucial for targeting clinical interventions.

Objectives:

We examined childhood and adulthood traumatic events in relation to cognition among Black and White older adults in the Health and Retirement Study (HRS).

Participants:

Participants (N=13,952) aged 55+ had complete lifetime trauma and cognitive testing data at the 2006/08, 2010/12, and/or 2014/16 waves.

Measures:

Trauma indices comprised childhood and adulthood event counts. Outcomes included baseline performance and trajectories on the Telephone Interview for Cognitive Status.

Design:

Main and non-linear trauma effects were modeled via linear regression, and overall contributions assessed with omnibus Likelihood Ratio Tests.

Results:

Black participants (N=2,345) reported marginally lower adulthood trauma exposure than White participants (N=11,607) with no other exposure differentials observed. In White participants only, greater childhood trauma exposure predicted worse baseline cognition but slower change over time. Across race, adulthood trauma robustly associated with baseline cognition. Relationships were frequently non-linear: low but non-zero trauma predicted highest cognitive scores, with much poorer cognition observed as trauma exposure increased. Relationships between adulthood trauma and trajectory were limited to the White sample.

Conclusions:

Traumatic experiences, particularly in adulthood, may impact late-life cognitive health if not addressed. Findings highlight foci for clinical researchers and providers: adverse life events as a source of cognitive risk, and identification of community-specific resources that buffer behavioral, physical, and mental health sequelae of previous and incident trauma.

Objectives

As the U.S. population ages, incidence and prevalence of Alzheimer’s disease and related dementias (ADRD) increase; parallel increases in subclinical cognitive impairment correlate with additional adverse outcomes for older adults, including falls, loss of independence, and clinically-presenting ADRD (1). Fortunately, 30-50% of global ADRD burden is modifiable (2,3); the physical, mental, and fiscal costs for patients, families, and communities highlight the need to identify malleable determinants of cognitive health. Growing evidence suggests exposure to stressors and stress-related sequelae represents one such modifiable source of risk (4-7), operating via environmental, behavioral, psychosocial, and biological pathways (8). Traumatic events are distinguished from other stressors by their severity. Like many ADRD determinants (2), trauma-related risk processes may operate within sensitive periods of influence, when targeted risk reduction and access to clinical and community supports will most substantially improve later-life cognitive outcomes. Substantial theoretical framing and empirical evidence establishes childhood as a sensitive period for long-term health detriment related to traumatic experiences, via persisting changes to physiological stress responses, and lasting impacts on social and economic trajectories (8,9). However, relatively few studies examine later-life cognitive outcomes in this context.(10)

An even smaller body of work focuses on cognitive impacts of trauma during adulthood, despite evidence suggesting stress operates on health throughout life. Adulthood objective and appraised stress associates with volumetric reductions in brain regions key for memory and executive function (11); in experimental settings, introduction of a stressor temporarily reduces abilities in these domains (12). Traumatic life events, even common experiences such as illness or death of a close family member, may also precipitate harmful ADRD risk factors for aging adults: poor management of vascular diseases, social isolation, depression, poor sleep, and even posttraumatic stress disorder (2,13-15).

There is particularly critical need to identify and mitigate salient cognitive risk factors in minoritized populations including Black Americans, whose risk for ADRD and cognitive impairment is nearly twice as high as White peers (16) and even more likely to be attributable to modifiable factors (3). Prevalence of modifiable excess risk results from structural, institutional, and interpersonal racism that perpetuates inequalities in neighborhood environments, educational and occupational opportunities, legal system contact, and health care access (17-19). Adverse social conditions expose the people living in those conditions to stressful events, and constrain the resources needed to cope with such events (9,20,21). Recent work suggests that stress exposures partially explain racial disparities in cognitive aging and may be a key determinant of cognitive health in older Black adults (4,5,22).

Recognizing the need to identify population-salient ADRD risk factors, and sensitive periods with distinct translational implications, we examined associations of childhood and adulthood trauma exposure with later-life cognitive test performance within Black and White samples of older adults. We used a within-group approach to account for experiences and exposures created through racialization across the life course.

Methods

Population and Setting

Data are from the Health and Retirement Study (HRS), a nationally representative panel study of the population in the U.S. over age 50. Each full wave is completed across two yearlong periods. Data are collected in home at the enrollment baseline interview and by telephone at each biennial follow-up wave. A self-administered questionnaire, including lifetime trauma assessment, is left with respondents at the end of baseline interviews. Data from waves 2006/2008 through 2014/2016 are included in analyses. Participants were included in a cross-sectional analysis of baseline cognitive test performance if they were aged >= 55 years and had complete life trauma and cognitive data available from at least one wave (see Figure, Supplementary Digital Content 1, for CONSORT diagram). Participants were included in a longitudinal analysis if they had two or more waves of cognitive data available. The HRS (NIA U01AG009740) is conducted by the University of Michigan.

Measures

Childhood traumatic events

Data on childhood traumatic events (CT) were drawn from responses to 11 items regarding lifetime trauma on the leave-behind questionnaire (23). For participants who completed the questionnaire at two or more timepoints, first timepoint with complete trauma data was used. Four items asked specifically about childhood (e.g., Before you were 18 years old, were you ever physically abused by either of your parents?). Participants were also asked whether seven additional events occurred at any age (e.g., Were you the victim of a serious physical attack or assault in your life?) and, if yes, the most recent year that the event occurred. Age at event was calculated as the difference between birth year and year of event. A CT index score was computed based on number of “Yes” responses to childhood-specific items together with the number of “Yes” responses to any-age items reported as occurring before age 18 (potential CT index range, 0-11).

Adulthood traumatic events

Data on adulthood traumatic events (AT) were drawn from the seven any-age items described above. An AT index score was computed as total number of items with “Yes” response with age of event being 18 or older, (potential AT index range, 0-7).

Cognitive outcomes

Cognitive outcomes were assessed with the 27-item adapted Telephone Interview on Cognitive Status (TICS). The TICS provides a global cognition score as well as subscores for verbal learning and memory. For each measure, lower scores indicate poorer performance.

Self-identified race and covariates

HRS participants are asked to specify their race and Hispanic/Latino ethnicity (“Do you consider yourself Hispanic or Latino”) at their baseline interview. The current study included those participants who identified as White/Caucasian or as Black/African American and excluded those who identified as Hispanic/Latino. Covariates included basic demographic characteristics, chosen a priori based on empirical correlation with both cognitive test scores and with self-reported experiences of trauma: age at interview, gender, and educational attainment. Age was calculated for each interview based on date of birth and analyzed as a continuous variable. Gender was self-reported as “male” or “female” and analyzed as a binary variable. Educational attainment was categorized as “under high school” (<12 years), “high school” (equal to 12 years), “some college” (greater than 12 years; less than 16 years), and “college grad” (>= 16 years).

Analyses

In order to assess within-group relationships of childhood and adulthood trauma with cognitive outcomes, the analytic sample was stratified by self-identified race into two samples, non-Hispanic Black/African American (referred to from this point forward as Black) and non-Hispanic White (referred to from this point forward as White). Characteristics of Black and White participants were compared using Chi-squared and Mann-Whitney U tests. In preliminary analyses, a priori modeling of trauma-cognition relationships revealed both linear and parabolic forms to be relevant. Accordingly, performance on each of three key cognitive scores (global cognition, immediate recall, and delayed recall) were examined with respect to two trauma terms: (i) main effects for CT and AT index scores, and (ii) a quadratic of the relevant trauma index score. Continuous trauma index scores were centered prior to squaring and analysis. Two modelling setups were used. First, cross-sectional relationships between main and quadratic trauma effects and baseline cognitive performance were examined using linear regression, adjusting for baseline age, gender, and educational attainment. Second, to examine relationships between traumatic event indices and cognitive trajectory across time, a two-stage process was used. In the first stage, we calculated each participant’s cognitive slope using linear regression with longitudinal cognition as the outcome, and longitudinal age as the only covariate. These calculated individual slopes were then used as the outcome in models otherwise similar to the cross-sectional analyses, though age is not included given it was used / accounted for in the first stage.

“Omnibus” trauma and cognition

Wald tests assessed significance of main and quadratic pieces of the form separately. To assess if trauma indices, overall, associated with each outcome, Likelihood Ratio Tests (LRT) were used to compare models with the linear and quadratic effects of CT and AT (“omnibus” trauma) to models without either.

Statistical significance was assessed at the 5% level throughout. For each inferential test type (main CT and AT effects, quadratic effects, omnibus tests), Benjamini-Hochberg corrections were performed to mitigate false discovery rate (FDR) at 5% across the 12 models. All analyses and graphics were conducted using R version 4.0.0 and utilized the lme4, boot, tableone, and ggplot2 packages. Model diagnostics for residual trends, heteroscedasticity, outliers, residual distribution, random effects distribution, and residuals vs random effects trends revealed no issues.

Results

Sample characteristics

A total of 2,345 Black and 11,607 White participants were included in baseline cross-sectional analyses. 1,510 Black and 8,434 White participants were included in cognitive slope analyses. Participant characteristics for the whole sample and the slope subgroup are presented in Table 1. Black participants were slightly younger than White participants and more likely be female. Relative to White participants, a greater proportion of Black participants reported completing fewer than 12 years of education, and a smaller proportion reported completing at least 16 years. Average numbers of endorsed traumatic life events in childhood and in adulthood were low in both groups. White participants reported marginally more AT events than did Black participants but otherwise surveyed trauma exposure did not vary by race. While CT and AT events were only analyzed as cumulative scores in line with accumulation-of-risk models [Hostinar et al 2015], the frequencies of endorsement for each individual event are provided (see Figure, Supplementary Digital Content 2, for sample proportions endorsing traumatic events in childhood and adulthood).

Table 1:

Participant characteristics of Black and White analytic samples

	Cross-sectional Baseline
	Cognitive Performance analyses				Cognitive Slopes analyses
Characteristic	Black/African American	White	p		Black/African American	White	p
N	2345	11607			1510	8434
Age, years (mean (SD))	64.73 (8.59)	68.07 (9.61)	<0.001		64.48 (7.55)	67.58 (8.64)	<0.001
Male (N (%))	834 (35.6)	4979 (42.9)	<0.001		505 (33.4)	3551 (42.1)	<0.001
Education (N (%))
< High School	679 (29.0)	1603 (13.8)			410 (27.2)	1036 (12.3)
High School	711 (30.3)	4154 (35.8)	<0.001		473 (31.3)	3059 (36.3)	<0.001
Some college	581 (24.8)	2758 (23.8)			380 (25.2)	2000 (23.7)
College grad	374 (15.9)	3092 (26.6)			247 (16.4)	2339 (27.7)
Childhood Trauma Index (mean (SD))	0.50 (0.80)	0.51 (0.78)	0.347		0.49 (0.78)	0.50 (0.77)	0.275
Adulthood Trauma Index (mean (SD))	1.03 (1.15)	1.05 (1.10)	0.046		1.03 (1.15)	1.05 (1.09)	0.139
Global Cognition (mean (SD))	13.29 (4.36)	16.06 (3.97)	<0.001		13.73 (4.06)	16.48 (3.66)	<0.001
Immediate Recall (mean (SD))	5.12 (1.53)	5.65 (1.56)	<0.001		5.24 (1.44)	5.80 (1.46)	<0.001
Delayed Recall (mean (SD))	3.73 (1.86)	4.60 (1.88)	<0.001		3.90 (1.76)	4.78 (1.77)	<0.001

Statistical tests between racial groups’ characteristics was conducted using Mann-Whitney U tests for continuous items, and Chi-squared tests of association for differences in Gender proportion (df = 1) and Education level (df = 3).

Trauma and cognition

Relationships of main and quadratic effects for CT and AT indices with baseline cognitive performance in three domains are presented in Table 2. Omnibus effects, described here, are fully presented in supplementary materials (see Table, Supplemental Digital Content 3, for detailed model characteristics and results). Relationships between number of childhood and adulthood trauma counts and predicted level of performance for global cognition and memory subscores, by race, are shown in Figure 1. In the Black sample there were no statistically significant associations between CT index and baseline TICS performance in any domain. However, AT index associated with baseline level of performance in global cognition, F(2, 2345) = 8.37 p=<0.001 and on the delayed recall subscale, F(2, 2345) = 3.80, p=0.02. A comparison of trauma and age coefficients as they predict global cognition (see Table, Supplementary Digital Content 3, for covariate estimates) are helpful to illustrate clinical significance of effects and occasional non-linearity of trauma-cognition relationships. In the Black sample, the addition of one traumatic event from a starting point of zero exposure is “protective” for global cognition, the cognitive equivalent of 4.02 fewer years of chronological age. However, the addition of one event from a starting point of four events is equivalent to 4.02 additional years of chronological aging. For White participants, CT index associated with level of performance in global cognition, F(2, 10899) = 6.55, p=0.002 in White participants, but not memory subscores. Comparing trauma and age coefficients as they predict global cognition, an addition of one childhood event from zero exposure is equivalent to 1.11 additional years of chronological age while the addition of one event from a starting point of four events is equivalent to 2.04 additional years of chronological age. AT index associated with level of performance in global cognition, F(2, 11607) = 9.01, p<0.001; immediate recall, F(2, 11607) = 12.00, p<0.001; and delayed recall, F(2, 11607) = 6.25, p=0.002. An addition of one adulthood event from zero exposure is equivalent to 1.48 fewer years of chronological age while the addition of one event from a starting point of four events is equivalent to 0.95 additional years of chronological age.

Table 2:

Beta coefficients and confidence intervals for trauma-baseline cognition regression analyses

Race	Trauma life period	Model Outcome	Stress Coefficient component	N	Coefficient Estimate	df	t value	Estimate's 95% CI	p value	adj. p value
Black	Childhood	Global Cognition	Main effect	2008	0.208	2000	1.268	−0.11 — 0.53	0.2050	0.2734
			Quadratic effect	2008	−0.083	2000	−0.958	−0.25 — 0.09	0.3380	0.4506
		Immediate Recall	Main effect	2008	0.043	2000	0.709	−0.08 — 0.16	0.4786	0.5743
			Quadratic effect	2008	−0.026	2000	−0.809	−0.09 — 0.04	0.4188	0.4569
		Delayed Recall	Main effect	2008	0.120	2000	1.602	−0.03 — 0.27	0.1093	0.1639
			Quadratic effect	2008	−0.077	2000	−1.961	−0.15 — −0.01	0.0500	0.0999
	Adult	Global Cognition	Main effect	2345	0.356	2337	4.070	0.18 — 0.53	<0.0001	0.0002
			Quadratic effect	2345	−0.121	2337	−2.948	−0.2 — −0.04	0.0032	0.0302
		Immediate Recall	Main effect	2345	0.056	2337	1.724	−0.01 — 0.12	0.0849	0.1455
			Quadratic effect	2345	−0.032	2337	−2.107	−0.06 — −0.01	0.0352	0.0955
		Delayed Recall	Main effect	2345	0.106	2337	2.668	0.03 — 0.18	0.0077	0.0185
			Quadratic effect	2345	−0.042	2337	−2.254	−0.08 — −0.01	0.0243	0.0955
White	Childhood	Global Cognition	Main effect	10044	−0.140	10036	−2.115	−0.27 — −0.01	0.0344	0.0689
			Quadratic effect	10044	−0.015	10036	−0.390	−0.09 — 0.06	0.6962	0.6962
		Immediate Recall	Main effect	10044	0.007	10036	0.257	−0.05 — 0.06	0.7971	0.8502
			Quadratic effect	10044	−0.027	10036	−1.813	−0.05 — 0.02	0.0698	0.1197
		Delayed Recall	Main effect	10044	−0.006	10036	−0.189	−0.07 — 0.06	0.8502	0.8502
			Quadratic effect	10044	−0.016	10036	−0.886	−0.06 — 0.01	0.3755	0.4506
	Adult	Global Cognition	Main effect	11607	0.156	11599	4.189	0.08 — 0.23	<0.0001	0.0002
			Quadratic effect	11607	−0.041	11599	−2.056	−0.08 — −0.01	0.0398	0.0955
		Immediate Recall	Main effect	11607	0.073	11599	4.89	0.04 — 0.10	<0.0001	<0.0001
			Quadratic effect	11607	−0.022	11599	−2.805	−0.04 — −0.01	0.0050	0.0302
		Delayed Recall	Main effect	11607	0.060	11599	3.333	0.03 — 0.10	0.0009	0.0026
			Quadratic effect	11607	−0.011	11599	−1.143	−0.03 — 0.01	0.2532	0.3798

All models additionally adjust for baseline age, gender, and educational attainment

Adjusted p-values were conducted using the Benjamini-Hochberg procedure across the trauma coefficient main effects and quadratic effects separately (12 items in each adjustment).

Figure 1: Baseline cognitive performance as predicted by trauma index scores.

Shaded regions represent the 95% confidence interval for the estimated relationship between trauma and cognitive test score. Predictions and CI's are calculated for age being set to the median age in the data (66 years old), and using the modes for gender (Female) and educational attainment (High school). Graphic panel columns represent the two different trauma measures (Childhood and Adulthood), and the graphic panel rows represent the three different cognitive test outcomes (Global Cognition, Immediate Recall, and Delayed Recall).

Relationships of main and quadratic effects for CT and AT indices with slope of cognitive test performance across time in three domains are presented in Table 3. Omnibus effects, described here, are also fully presented in supplementary materials (see Table, Supplemental Digital Content 3, for detailed model characteristics and results). There were no significant associations between CT or AT and cognitive slope for Black participants in any cognitive domain. Among White participants, CT index associated with scores on global cognition, F(2, 7840)=5.05, p=0.006 and immediate recall, F(2, 7840)=5.07, p=0.006. In these same White participants, AT index associated with change in those same domains of global cognition, F(2, 8427) = 5.28, p=0.005 and immediate recall, F(2, 8427) = 6.62, p<0.01. Figure 2).

Table 3:

Beta coefficients and confidence intervals for trauma-cognitive slope regression analyses

Race	Trauma life period	Model Outcome	Stress Coefficient component	N	Coefficient Estimate	df	t value	Estimate's 95% CI	p value	adj. p value
Black	Childhood	Global Cognition	Main effect	1372	−0.011	1365	−0.261	−0.09 — 0.07	0.7942	0.7942
			Quadratic effect	1372	0.015	1365	0.592	−0.03 — 0.06	0.5536	0.7583
		Immediate Recall	Main effect	1372	0.022	1365	1.206	−0.01 — 0.06	0.2282	0.4564
			Quadratic effect	1372	−0.009	1365	−0.891	−0.03 — 0.01	0.3729	0.7583
		Delayed Recall	Main effect	1372	−0.037	1365	−1.693	−0.08 — 0.01	0.0907	0.3628
			Quadratic effect	1372	0.022	1365	1.735	−0.01 — 0.05	0.0829	0.7583
	Adult	Global Cognition	Main effect	1510	0.009	1503	0.414	−0.03 — 0.05	0.6788	0.7714
			Quadratic effect	1510	−0.006	1503	−0.570	−0.02 — 0.01	0.5687	0.7583
		Immediate Recall	Main effect	1510	0.007	1503	0.773	−0.01 — 0.03	0.4398	0.5864
			Quadratic effect	1510	−0.001	1503	−0.265	−0.01 — 0.01	0.7909	0.8057
		Delayed Recall	Main effect	1510	0.004	1503	0.376	−0.02 — 0.03	0.7071	0.7714
			Quadratic effect	1510	−0.005	1503	−1.029	−0.01 — 0.01	0.3039	0.7583
White	Childhood	Global Cognition	Main effect	7847	0.019	7840	1.321	−0.01 — 0.05	0.1866	0.4564
			Quadratic effect	7847	0.008	7840	0.948	−0.01 — 0.02	0.3431	0.7583
		Immediate Recall	Main effect	7847	0.006	7840	0.871	−0.01 — 0.02	0.3837	0.5756
			Quadratic effect	7847	0.005	7840	1.397	−0.01 — 0.01	0.1626	0.7583
		Delayed Recall	Main effect	7847	0.008	7840	1.066	−0.01 — 0.02	0.2863	0.4908
			Quadratic effect	7847	0.003	7840	0.676	−0.01 — 0.01	0.4990	0.7583
	Adult	Global Cognition	Main effect	8434	−0.022	8427	−2.703	−0.04 — −0.01	0.0069	0.0413
			Quadratic effect	8434	0.001	8427	0.246	−0.01 — 0.01	0.8057	0.8057
		Immediate Recall	Main effect	8434	−0.011	8427	−3.185	−0.02 — −0.01	0.0015	0.0174
			Quadratic effect	8434	0.001	8427	0.576	−0.01 — 0.01	0.5645	0.7583
		Delayed Recall	Main effect	8434	−0.005	8427	−1.262	−0.01 — 0.01	0.2068	0.4564
			Quadratic effect	8434	−0.001	8427	−0.332	−0.01 — 0.01	0.7402	0.8057

All models additionally adjust for gender and educational attainment

Adjusted p-values were conducted using the Benjamini-Hochberg procedure across the trauma coefficient main effects and quadratic effects separately (12 items in each adjustment).

Figure 2: Change in cognitive performance as predicted by trauma index scores.

Shaded regions represent the 95% confidence interval for the estimated relationship between trauma and cognitive test score. Predictions and CI's are calculated for age being set to the median age in the data (66 years old), and using the modes for gender (Female) and educational attainment (High school). Graphic panel columns represent the two different trauma measures (Childhood and Adulthood), and the graphic panel rows represent the three different cognitive test outcomes (Global Cognition, Immediate Recall, and Delayed Recall).

Conclusions

In this nationally-representative sample of older adults, we examined relationships between experiences of childhood and adulthood traumatic events and cognitive outcomes in later life. We conducted within-group analyses by race in accordance with disparities frameworks (9,18) and evidence that measured and unmeasured differences in the lived experiences of Black and White older adults have implications for exposure distributions and magnitudes of effect (24). To our knowledge, this is the first exploration of population-specific relationships of distinct sensitive periods for trauma exposure with cognitive health.

Our finding that Black and White participants reported similar exposure to traumatic events during childhood, with White adults reporting marginally greater exposure during adulthood, was somewhat surprising within a larger body of work on racialized stress exposures. However, results were consistent with prior findings on traumatic event indices specifically, highlighting the importance of delineating distinct stress constructs. Several studies have reported on racial disparities in chronic stress and discrimination exposures among HRS participants (20,25,26), but in a sample of Black and White men drawn from the first cohort to complete the leave-behind questionnaires, those disparities did not extend to the surveyed childhood and adulthood traumatic events (26). These findings are echoed in a latent cluster analysis of Adverse Childhood Experiences (ACEs) in another large cohort; among non-Hispanic Black adolescents, ACEs representing socioeconomic adversity, not trauma, drove disproportionate overall prevalence of ACEs (27).

Childhood trauma significantly associated with cognition only in White participants, with relationships differing qualitatively for baseline cognition and for cognitive change. A primarily linear negative association with baseline performance on global cognition contrasts with an apparent protective effect on trajectory. Negative associations with baseline cognition echo prior work with other outcomes in the full, predominantly White HRS cohort. The few studies exploring surveyed childhood traumatic events as a predictor of health have reported associations between childhood trauma items and physiological detriment including telomere attrition (28) and all-site cancer prevalence (29). Slowed decline is not consistent with theories of accelerated aging or weathering under adverse conditions and in general, causal protective relationships between high exposure to childhood trauma and preserved late-life cognition are not plausible. Inconsistency could be explained by atypical resilience prevalent in a sample of older adults able to select into a longitudinal study. Stable TICS score over time may also reflect baseline scores that underestimate cognitive function in persons who have experienced social disadvantage, a systematic test bias common to cognitive screeners. Within the Black/African American sample, the type of traumatic experiences assessed (or absent) may again explain our wide confidence intervals and subsequently null findings for childhood trauma. Zhang and colleagues focused on an index of childhood socioeconomic disadvantage, and found that it associated with 12-year dementia risk in the full HRS sample and partially accounted for racial disparities therein (30). Meanwhile, another HRS study using a broad index of childhood stressors to predict functional limitations, inclusive of the trauma items in the current study, found, as we did, that associations observed in White participants were absent in Black participants (31). In other cohorts, absence of relationships or even positive relationships have been observed (32,33). A systematic review attributed inconsistencies in findings on childhood adversity as a determinant of later-life health to issues of internal validity, and potential for effect modification by other demographic characteristics such as age and gender (10). Notably, within the Black sample for this study, our wide confidence intervals and subsequent null findings for childhood trauma could be due to several factors including the relatively smaller sample of Black participants, the few participants reporting high numbers of traumatic events in this group, and the type of traumatic experiences assessed or absent.

That significant cognitive detriment was observed for both Black and White participants at higher exposures to adulthood traumatic events, with substantial quadratic effects driving particularly strong negative relationships in the Black sample, is not surprising. One prior study of lifetime accumulated stressors and cognitive aging in a much smaller and volunteer-based cohort described associations of adversity with cognition across both Black and White middle-aged and older adults – but relationships were both greater in magnitude and expanded to more cognitive domains in Black participants, compared to White peers (5). Similar results have also been recently observed in much larger, population-based cohorts (22,34). The aforementioned HRS-based study on traumatic life events and men’s functional health echo the current results as well; trauma exposure associates with functional health across racial groups but also partially accounts for racial disparities in functional limitations despite equivalent exposure, indicating greater magnitude of effect in older Black men (26). The current study did not make between-race comparisons of the relationships between trauma and health, but prior findings on disparate magnitudes of effect can help guide mechanistic hypotheses and future research questions. Differential vulnerability to trauma in the absence of differential exposure is likely to be explained by the inequitable contexts in which traumatic events occur. Cumulative stress indices do not account for co-occurring race-based stressors such as everyday discrimination (20). Additionally, if coping resources – emotional, instrumental, financial, physiological – are already strained by existing chronic adversity, then the impact of acute stressors is likely to extend and expand (21).

In conjunction with a small but rigorous body of previous work, our findings also demonstrate the importance of testing for non-linear health effects of social adversity. As noted, in both Black/African American and White samples, relationships between traumatic experiences in adulthood and cognitive test performance were modestly positive when traumatic events were present but rare. When these low-exposure groups dominate a sample, regression coefficients for linear terms are driven by that positive association, obscuring detriment occurring at high exposure. Nonlinear modeling of risk factors is uncommon in cognitive aging research. However, studies elsewhere have described these “U-shaped” relationships between cumulative adversity and health outcomes, wherein nonzero but low adversity scores predict better mental and physical function (35,36). A frequent hypothesis is that when rare, adverse experiences require adaptation that fosters personal resilience. Few studies report positive associations between adversity and cognitive outcomes, but protective effects of specific adverse childhood experiences, such as not having enough food to eat, being underweight, and living with someone with mental illness, have been observed for the cognitive health of older ethnoracially diverse and Black cohorts (19,32). We note that the cumulative traumatic event exposures are right-tailed in our data, with most participants experiencing no or few events, and very few experiencing a large number of events. The relatively small (and thus more variable) number of high cumulative totals could ostensibly influence the nonlinear trends we observed. Additionally, if surveyed traumatic events represent a proxy for correlated, potentially unmeasured trauma and adversity, exacerbated associations at high cumulative totals would also be expected. The need for future studies of trauma and adversity, including non-linear considerations, is clear.

Finally, our study is not the first to find that stress-related exposures associate primarily with a baseline level of cognitive health rather than with rate of decline across time. Our findings in this regard are similar to previous studies reporting that measures of adversity primarily associate with a reduction in level of cognitive health at an early point in the measured trajectory (37), likely fomenting social disparities via a later-life “peak” closer to thresholds of impairment.

Limitations

Limitations in the described analysis draw attention to crucial future studies and research directions. First, our analyses did not include HRS participants who identified as Hispanic/Latino. Intersectionality between ethnicity and racialization requires additional within-group analyses that consider heterogeneity across culturally unique subgroups of broader Latino populations, as well as integration of factors such as nativity and immigration status. Additionally, as with many studies utilizing self-reported exposure data and cognition as an outcome, the potential for exposure misclassification is a concern, though we expect that underreporting should attenuate rather than inflate risk estimates in this study. Further, as discussed, selection and operationalization of distinct stress constructs inform results. In line with cumulative stress models, we focused on a count of life traumas within a relatively brief index. Trauma exposures were low; it is noteworthy that associations with cognitive health were detectable given this analytic constraint. Nonetheless there are notable considerations for such models. All experiences are equally weighted, but though the nature of “traumatic events” relative to “stressful events” is in their seriousness and severity, there may still be age, period, cohort, or other axes of variation in both objective impact and in subjective appraisal. In Black populations in particular, event indices are likely to underestimate exposure (20) and to fail to account for racial trauma or the adaptation and coping strategies developed to thrive in a context of racism (38). Complementary existing and future research guided by other life course models is needed. In a sweeping study of ethnoracial disparities in stress exposures and health outcomes, Sternthal, Slopen, and Williams describe unique implications for stress modeling, between and within populations (39). For instance, lifetime cumulative models can help account for bias resulting correlations between stressors, including unmeasured stressors, that individuals are exposed to. In other cases, examining the population-specific importance of distinct events (14), or clusters or chains of events that tend to occur in conjunction with each other or sequentially (27,31) can provide crucial information for screening and mitigation purposes, particularly within the clinical setting. Finally, as with all non-experimental studies, residual confounding of trauma-cognition relationships may be present. In particular, early-life exposures such as childhood socioeconomic status, quality of education, and chronic stress potentially intersect in complex ways with the primary relationships of interest. These intersections are likely to vary with racialization (39). Studies that model this complexity with intentionality can parse not only unique contributions but also mediation and moderation effects.

Implications for clinical research and translation

Despite its limitations, this study adds to a growing body of literature suggesting accumulating life stressors can undermine successful aging in older populations including Black/African American communities. In these samples, at high levels of trauma exposure, one additional adulthood trauma is equivalent to between one (White participants) and four (Black participants) years of chronological aging. Black Americans have been historically underserved by healthcare institutions. Clinical, public health, and health policy teams dedicated to equitably improving outcomes for older adults are uniquely situated to assess and mitigate health detriment associated with traumatic experiences given their close work with patients, families, and healthcare systems.

In terms of research implications, it is noteworthy that magnitude and even direction of effects for risk or protective factors in predominantly White samples cannot be assumed to hold for other populations; equity-focused gerontological science requires inclusive cohorts adequately powered for within-group analyses. Within and beyond race and ethnicity, this must include assessment of intersections between axes of both vulnerability and resilience. For instance, several studies have shown that trauma during early- and middle-adulthood associate with functional limitations in older age (23,31). If accommodations are absent, functional limitations and disability may compound health effects of accumulating stressors by constraining access to activities and resources that facilitate coping and recovery. Establishing population-specific physiological, psychological, behavioral, and environmental mediators and moderators will help to pinpoint salient intervention loci across stages of life and at all levels of function.

The role that clinical practices can play in mitigating health impacts of trauma is also increasingly clear. Traumatic experiences during childhood and adulthood have been robustly tied to disruption of brain-healthy behaviors, depressive symptoms, and symptoms of posttraumatic stress disorder in older adults. While studies of resilience – thriving under adverse conditions –often focus on individual-level traits, a more impactful approach from a clinical and public health perspective includes facilitating and sometimes forging connections between older adults and meaningful resources. Empirically, both clinical supports including medication and psychotherapy, and community assets including interpersonal relationships and access to religious services, have been found to help older adults recover following traumatic psychosocial and physical stressors (15,40,41). While acute events themselves may not always be preventable, clinicians play an irreplaceable role in helping older patients and families to navigate coping with and recovering from those events, even – especially – in a context of more chronic adversity.

Supplementary Material

1

Figure, Proportion of samples endorsing traumatic events in childhood and adulthood

2

Table, Omnibus effects for trauma based on Likelihood Ratio Testing

3

Figure, CONSORT Diagram

4

Table, Beta coefficients and confidence intervals for trauma-baseline cognition regression analyses (covariates only)

Data Statement

These data have not been previously presented orally or by poster at scientific meetings.