Longitudinal impact of transition to caregiving on cognitive functioning: a matched case-control study

Abstract

Objectives

Chronic stress has a strong theoretical link to poorer cognitive aging outcomes. Stress from caregiving, especially dementia caregiving, is associated with worse cognition; however, most prior studies have serious methodological limitations. We examined the longitudinal impact of caregiving transitions (including dementia caregiving) on cognition, compared to carefully matched non-caregiver controls, and the possible mediating role of depressive symptoms and perceived stress.

Methods

Participants in the Caregiving Transition Study who transitioned into caregiving (n = 251) were compared to sociodemographically and health-matched non-caregiving controls (n = 251). Data included 14 years of repeated assessments, including timepoints before and after transitions on global cognition (six-item screener), episodic memory (word-list learning, delayed word recall), and verbal fluency (letter and animal fluency).

Results

Compared to non-caregiving, negative associations between caregiving transitions, episodic memory, and global cognition were significant during transitions, but these associations were small (Cohen’s d approximately 0.2 SD units). Compared to non-caregivers, transitioned caregivers showed less decline in word list learning and global cognition during post-transitions. In adjusted subgroup analyses, dementia caregivers performed worse after transitions than non-dementia caregivers in delayed word recall but better on animal fluency. Increases in depressive symptoms mediated caregiving transitions and episodic memory but not global cognition. Perceived stress did not have a mediating role.

Discussion

Significant associations were found between caregiving transitions and cognition, but effects were domain-specific for memory and global cognition, and short-lived. Dementia caregiving also had differential effects on cognition. Future studies should examine whether these small, temporary declines in caregiver cognition improve with caregiver interventions.

Family caregiving is often described as a very stressful life experience associated with negative health outcomes (Allen et al., 2017) and a naturally occurring stress model for studying stress and health. When compared to non-caregivers, caregivers have higher rates of depression (Pinquart & Sörensen, 2003), depressive symptomology (Haley et al., 2020), and perceived stress (Haley et al., 2020). Family caregivers are also exposed to high levels of stress for extended periods (Roth et al., 2018), suggesting that family caregiving may be a chronic stressor. Research on general stress and caregiving stress (Machado et al., 2014) suggests that chronic stress can lead to worse cognition.

The stress process model (Lazarus, 1984), and variants specific to caregiving (Beach et al., 2000; Haley et al., 1987), can be useful in evaluating the impact of caregiving on cognition, including dementia caregiving. The effects of a stressor, like family caregiving, on cognition can be modified by the caregiver’s appraisal of his/her caregiving role. Studies on dementia caregiving report dementia caregivers having more negative appraisals and more physical and mental health problems compared to other caregivers (Sheehan et al., 2021). The well-known Cache County Study by Norton et al. (2010) reported that spousal caregivers for people with dementia had a sixfold increased risk of incident dementia compared to spousal caregivers for persons with no dementia. However, Vitaliano (2010) identified serious limitations of this study in determining whether caregiving explained this finding, due to factors including assortative mating and lack of measures relevant to caregiving-related stress.

There are findings that caregiving can sometimes bring positive outcomes, including meaning and purpose in life for a caregiver that may reduce the negative effects of stress on caregiver health (Haley et al., 2003). Family caregiving might promote cognitively stimulating experiences (Hertzog et al., 2008) and contribute to late-life cognitive reserve (Stern, 2009) through diverse daily activities. These activities include caregiving, meaningful interactions with the care recipient, learning and receiving caregiving information, and training from health care professionals. However, caregiving may also reduce the amount of time invested in cognitively challenging activities outside caregiving duties, such as participating in cultural events, reading books, or social activities outside the home, which may consequently contribute to cognitive decline in caregivers.

Caregiving and cognition

Several investigators used the stress process or cognitive reserve model as a rationale to study the associations between family caregiving and cognition, but results have been mixed. A systematic review of cross-sectional and longitudinal studies concluded that caregivers show worse cognitive function compared to non-caregivers (Romero-Martínez et al., 2018). The authors of this review, however, did report heterogeneity among studies. Included longitudinal studies confirmed that caregiving was associated with worse general cognitive state, memory, processing speed, and vocabulary richness (de Vugt et al., 2006; MacKenzie et al., 2009; Pertl et al., 2015; Vitaliano et al., 2009, 2017). Contrary to these findings, a longitudinal study from this review (Bertrand et al., 2012) reported better memory and faster processing speed for caregivers compared to non-caregivers. Moreover, many of the included studies are limited by their methodological designs. For instance, many studies had small samples (Corrêa et al., 2019; de Vugt et al., 2006; MacKenzie et al., 2009), improper identification of primary caregivers (Dassel et al., 2017; Pertl et al., 2015), and lacked demographically matched caregivers and non-caregivers that could explain differences in the caregiving-cognition relationship (MacKenzie et al., 2009). To better examine stress that is associated with caregiving and health, researchers should recognize diversity among caregivers and the need to consider covariates such as age, economic status, health, and risk factors associated with poor health when making caregiver and non-caregiver comparisons (Brown & Brown, 2014). Another major concern is convenience sampling of caregiving groups (Corrêa et al., 2019; MacKenzie et al., 2009) and the need to examine cognition through multiple cognitive domains (Dassel et al., 2017).

More recent studies found caregiving associated with better cognition or no differences in cognition. A large cross-sectional study (García-Castro et al., 2022) found better immediate and delayed recall and verbal fluency for caregivers compared to non-caregivers. Another cross-sectional study with demographically matched caregiver and non-caregiver groups (Jütten et al., 2020) found that being a caregiver was associated with better immediate recall. Longitudinal studies (O’Sullivan et al., 2019; Yuan & Grühn, 2021) found that being a caregiver and continuous caregiving were associated with better memory and faster processing speed compared to non-caregivers. Yuan and Grühn (2021) found that current and prior caregiving were associated with better memory and verbal fluency, with beneficial effects on memory 2–4 years after the caregiving experience.

A recent, large, population-based study examined the caregiver-cognition relationship over time across multiple cognitive domains using a propensity-matched caregiver and non-caregiver sample (Elayoubi et al., 2023). The study found that there were no differences between caregivers and non-caregivers in cognition at baseline or over time. This study, however, did not identify transitions into caregiving and subsequent changes over time in caregivers’ cognition. This has been a limitation in many past studies (Dassel et al., 2017; Vitaliano et al., 2009).

One promising approach would be to examine cognitive changes among individuals in a population-based study who transitioned into a family caregiving role compared with a matched sample of individuals who did not become caregivers over the same time. The design of the Caregiving Transitions Study (CTS) permits addressing this research question. Prior work from the CTS examined psychological well-being, inflammation, and telomere length as outcomes (Armstrong et al., 2022; Haley et al., 2020; Roth et al., 2020b). A prior CTS study found dementia caregivers reported higher appraisals of stress and burden and more depressive symptoms than non-dementia caregivers (Sheehan et al., 2021). Examining cognitive changes among individuals who transitioned into a family caregiving role compared with a matched sample of individuals who did not become caregivers at the same time could provide more insight into the long-term cognitive outcomes of caregiving. Moreover, based on the stress process model in caregiving (Haley et al., 1987), which proposes that poorer caregiver well-being is associated with higher levels of stressors, high-strain caregiving, like dementia caregiving, is recognized as particularly impactful for research on the association of stress and cognition.

Psychological mechanisms

Depressive symptoms and perceived stress

Caregiving is associated with a greater risk of depression (Pinquart & Sörensen, 2003), higher depressive symptoms, and perceived stress (Haley et al., 2020; Roth et al., 2018). One study comparing caregivers and non-caregivers found evidence that the relationship between caregiving and poorer processing speed and attention was mediated by greater depressed mood (Vitaliano et al., 2009). Recent work through CTS documented an increase in depressive symptoms and perceived stress for transitioned caregivers compared to non-caregivers (Haley et al., 2020). Depressive symptoms have also been associated with a greater risk of cognitive decline (Deckers et al., 2015). Past research found that objective caregiving stressors (care recipient dementia-related behavioral and psychological symptom severity) were only related to caregivers’ executive functioning through subjective appraisal of the caregiving stressor (Pertl et al., 2017). Therefore, it is worth examining depressive symptoms and perceived stress as possible mediators in the association between caregiving transition and cognition, and especially in the association between dementia caregiving and cognition.

Current study

Guided by previous literature, we addressed four hypotheses. Hypothesis 1: Caregivers who transition into the caregiving role for people with chronic illness or disability would have lower initial cognitive function at the first cognitive assessment after transition relative to matched non-caregivers at similar time points. Hypothesis 2: People who transition into caregiving would show more cognitive decline after transitions than non-caregivers. Hypothesis 3: Dementia caregivers, a high-strain caregiving group often vulnerable to negative effects of caregiving, would show the lowest levels of cognitive function at the first cognitive assessment after the transition and greater cognitive decline over time than non-dementia caregivers. Hypothesis 4: Increases in depressive symptoms and perceived stress would mediate the relationship between caregiving transition and cognitive change over time. Depressive symptoms and perceived stress would mediate the relationship between dementia versus non-dementia caregiving and cognitive change over time.

Method

Procedure and participants

REGARDS and CTS procedures and participants

The present study uses data from the REasons for Geographic And Racial Differences in Stroke (REGARDS) ancillary study, the CTS (Howard et al., 2005; Roth et al., 2020a). In 2003–2007, the REGARDS enrolled 30,239 African American and White adults 45+ years old across the United States, with oversampling of African Americans and those living in the stroke belt (North Carolina, South Carolina, Georgia, Tennessee, Alabama, Mississippi, Arkansas, and Louisiana). Participants with informed consent were administered in a baseline computer-assisted telephone interview (CATI) that assessed demographics, medical history, and other risk factors for stroke. Additional information about the REGARDS methodology has been documented elsewhere (Howard et al., 2017).

During baseline CATI, participants were asked, “Are you currently providing care on an ongoing basis to a family member with a chronic illness or disability?” Participants who replied “yes” were categorized as caregivers, and participants who replied “no” were categorized as non-caregivers. After REGARDS second in-home assessment (approximately 11.8 years after the baseline CATI), an updated caregiving status was collected through a Caregiving Screening CATI module administered as part of the standard REGARDS semi-annual follow-up CATI (Roth et al., 2020a). The study included 1,229 participants who transitioned into the caregiving role, i.e., those who answered “yes” to being a caregiver at the Caregiver Screening CATI and had previously answered “no” to being a caregiver at the baseline CATI. Participants who reported being a caregiver at the second CATI answered additional questions on their caregiving exposure during the CTS Enrollment Interview. Non-caregivers at REGARDS baseline and CTS enrollment were considered potential non-caregiving controls. Caregiver and non-caregiver groups were individually matched on demographic and health history factors, including age, sex, race, education, marital status, self-rated health at baseline, and self-reported history of cardiovascular disease (Roth et al., 2020a). The results of the inclusion/exclusion criteria were 251 transitioned caregivers to 251 demographically matched non-caregiving controls. Additional information on the timeline of the various assessments that were administered in REGARDS and CTS and the CTS methodology is documented elsewhere (Roth et al., 2020a). The protocols of REGARDS and CTS were approved by the Institutional Review Boards (IRBs) of the University of Alabama at Birmingham and each participating institution, and written informed consent was provided by all participants. The present study was exempt from IRB approval from the University of South Florida because this study involved secondary analysis and therefore did not constitute research involving human subjects.

Measurements

Caregiving status

Participants were from the CTS. A binary variable indicated whether participants transitioned into caregiving (yes = 1; no = 0). During CTS enrollment, caregivers were asked if they were caring for a person with Alzheimer’s disease, another form of dementia, or a serious memory disorder. Participants either responded “yes” (=1) for dementia caregivers or “no” (=0) for other caregivers, or if they refused to answer, were treated as missing. Caregivers also completed The Eight-item Informant Interview to Differentiate Aging and Dementia Test (AD8), a self-reported measure of impairment in the care recipient, consistent with a validated screening measure for dementia (Galvin et al., 2005). A prior CTS study found that 100% of caregivers who self-reported caring for a family member with dementia reported impairment in the care recipient over the cut point for dementia on the AD8 (Sheehan et al., 2021).

Cognitive functioning and change

REGARDS measured multiple cognitive domains, including global cognition, episodic memory, and verbal fluency, annually and biannually up to 14 years by telephone (Hachinski et al., 2006). Previous studies indicate that these cognitive tests can be measured reliably and precisely by telephone (Unverzagt et al., 2007) and are validated across different races, including Black/African American and White/European Americans (Callahan et al., 2002). Detailed methods of telephone administration and validation of these fluency tasks within the REGARDS sample have been described elsewhere (Marceaux et al., 2019). Because cognitive tests were administered on separate schedules, we established separate analytic samples to preserve cognitive data for all participants. Details on the cognitive measures can be found in Supplementary Methods and Supplementary Table 1.

Every transitioned caregiver had a date (in month/year) for when he/she started the caregiving role. The date of transition into caregiving served as the baseline for caregiving transitions and was coded as time = 0. Two time metrics were calculated: First, a variable indicating the time to/from the caregiving transition date (in years). For all cognitive assessments before the caregiving transition date (approximately 8 years), the time variable had negative values. For all cognitive assessments after transitions (approximately 6 years), the time variable had positive values. Second, a post-transition variable was coded 0 for all observations prior to the caregiving transition and 1 for all observations after the caregiving transition. Since the control group did not have a caregiving transition date value, a midpoint assessment was used. If a non-caregiver participant had an odd number of measurements, the measurement in the middle was used as the transition point. If the number of observations was even, then the first measurement after the mid-point (identified by the ceiling SAS function [ceil(count (*)/2)] was used as the transition date. On average, participants contributed 6.23 years of longitudinal observations (SD = 1.42, range = 1–9).

Depressive symptoms and perceived stress

Depressive symptoms and perceived stress were assessed during the REGARDS baseline CATI with a four-item version of the Center for Epidemiological Studies-Depression (CES-D) scale (Melchior et al., 1993) and a four-item version of the Cohen Perceived Stress Scale (Cohen et al., 1983), respectively. Higher scores indicated more depressive symptoms or perceived stress. During the CTS Enrollment interview, caregivers and non-caregivers repeated the identical four-item version of the Perceived Stress Scale but completed a 10-item CES-D measure (Andresen et al., 1994). For data analysis, the 4-item version of the CES-D was used to estimate a 10-item score, using regression analyses as reported elsewhere (Haley et al., 2020).

Sociodemographic and health variables

In the caregiver subgroup analyses, examining dementia versus non-dementia caregiving cognition, we included several of the sociodemographic and health variables that were used for matching in the CTS as covariates. Specifically, we included age at the time of transition (years) body mass index and self-rated health (excellent = 5, very good = 4, good = 3, fair = 2, poor = 1) measured continuously, all other variables were treated categorically in analyses including, sex (men = 1, women = 2), race (Black = 1, White = 2), region (Stroke Belt = 1, Stroke Buckle = 2, Rest of U.S.=3), education (less than high school = 1, high school graduate = 2, some college = 3, college graduate and above = 4), income (refused = 1; less than $20,000 = 2; $20,000–$34,999 = 3; $35,000–$74,999 = 4; $75,000 and above = 5), marital status (married = 1, single = 2; divorced/separated = 3, widowed = 4), health insurance (yes = 1, no = 0), hypertension (yes = 1, no = 0), and diabetes (yes = 1, no = 0).

Statistical analyses

Hypotheses 1 and 2

To assess whether transitioned caregivers and matched non-caregivers differed in trajectories of cognitive function around the caregiving transition, we conducted multilevel analyses (Singer & Willett, 2003) using PROC MIXED (Littell et al., 2006) in SAS for the outcomes of episodic memory (word list learning, delayed word recall), verbal fluency (animal fluency and letter fluency), and global cognition (Six-Item Score or SIS). We used multilevel logistic regression analyses to assess whether caregiving transition was associated with an increased risk of having impaired global cognition, based on the dichotomized scores from SIS (5 or 6 = intact; 4 or less = impaired).

Hypothesis 3

To examine the role of dementia caregiving status as a predictor of caregiver cognition, we restricted analyses to caregivers only and repeated analyses for all cognitive outcomes. The binary variable indicating dementia caregiving (1 = caregiver to a person with dementia; 0 = caregiver to a person without dementia) was included in the growth curve models as a predictor of the intercept (indicating individual differences at transition) and the slope (indicating individual differences in change). We adjusted for sociodemographic and health variables mentioned above to account for any differences within the caregiving cohort associated with these variables.

Hypothesis 4

If differences were found between either caregivers and controls, or dementia caregivers and non-dementia caregivers, we conducted additional analyses to assess whether changes in depressive symptoms or perceived stress between the first and second in-home assessments mediated the relationship between caregiving status (caregivers vs. controls, dementia caregivers vs. other caregivers) and cognitive functioning in unadjusted and adjusted models. Mediation analyses were restricted to cognitive outcomes that showed an association with caregiving. We used the PROCESS Macro in SAS (Hayes, 2017) to assess whether depressive symptoms or perceived stress mediated the relationship between caregiving transitions and cognition at the time of transition and over time. We used bootstrapped mediation analyses using 10,000 bootstrap samples to assess the significance of the indirect effect. Unadjusted models included the independent variable (caregiving transition status or dementia vs. non-dementia caregiver), the mediator (changes in depressive symptoms or perceived stress), and the dependent variable (cognition at transitions and over time).

When examining the effects of caregiving on various cognitive domains, we did not apply alpha adjustments for multiple testing. This is because we were not testing a joint null hypothesis that was to be rejected by at least one significant test (disjunctive testing; Rubin, 2021), which would require controlling for a familywise error rate. The type of inference involved in these analyses was individual testing of null hypotheses (see Rubin [2021] for an extended review and discussion of alpha adjustment).

Results

Table 1 shows descriptive statistics for study variables for transitioned caregivers and non-caregivers at transitions. Due to the matching procedures, caregivers and non-caregivers did not differ statistically on sociodemographic variables, including age, sex, race, education, and marital status. Both samples were aged on average 72 years, two-thirds of the sample identified as White, and two-thirds were women. Of the transitioned caregivers, 47% cared for a person with dementia. Like a prior CTS (Haley et al., 2020), transitioned caregivers had higher depressive symptoms (M = 6.96, SD = 6.29) compared to non-caregivers (M = 3.29, SD = 4.43), p < .001 during transitions. Caregivers also had increases in depressive symptoms over time (M = 3.47, SD = 6.48) compared to non-caregivers (M = −0.23, SD = 4.69), p < .001. Similarly, caregivers had higher perceived stress after transitions (M = 4.55, SD = 3.01) compared to non-caregivers (M = 2.53, SD = 2.56), p < .001. Transitioned caregivers showed increases in perceived stress over time (M = 1.69, SD = 3.27) compared to non-caregivers (M = −0.52, SD = 2.81), p < .001.

Table 1.

Descriptive statistics for health and demographically matched transitioned caregivers and non-caregivers.

Variable	Transitioned caregivers (n = 251)		Non-caregivers (n = 251)
	(M, SD)	%	(M, SD)	%
Matched variables
Agea	(71.95, 7.95)		(72.08, 7.52)
Sex
Men		33.84%		34.49%
Women		66.16%		65.51%
Race
White		65.44%		64.97%
Black		34.56%		34.03%
Region
Stroke belt		38.17%		35.37%
Stroke buckle		21.01%		21.22%
Rest of US		40.82%		44.40%
Education
Less than high school		3.37%		4.08%
High school graduate		22.88%		24.08%
Some college		31.79%		27.21%
College graduate and above		41.96%		44..63%
Income
Less than $20,000		8.07%		7.35%
$20,000-$34,999		19.81%		19.52%
$35,000-$74,999		41.06%		38.57%
$75,000 and above		22.82%		26.94%
Refused		8.25%		7.62%
Marital status
Single		6.14%		5.10%
Married		76.22%		76.94%
Divorced/separated		10.17%		9.46%
Widowed		7.47%		8.50%
Health insurance
Yes		99.76%		98.44%
No		0.24%		1.56%
Self-rated health
Excellent		24.14%		22.04%
Very good		32.81%		35.92%
Good		32.75%		34.29%
Fair		9.33%		7.76%
Poor		0.96%		0.00%
Hypertension
Yes		69.60%		58.98%
No		30.40%		41.02%
Diabetes
Yes		25.77%		21.09%
No		74.23%		78.91%
Body mass index	(30.46, 6.37)		(28.37, 5.96)
Main study variables
Δ Depressive symptoms	(3.42, 6.29)		(−0.20, 4.59)
Δ Perceived stress	(1.62, 3.24)		(−0.49, 2.84)
Caregiving type
Dementia		46.84%		-
Non-dementia		53.16%		-

^aMeasured in years at Caregiving Transitions enrollment. Covariates measured from REasons for Geographic and Racial Differences in Stroke (REGARDS) firstin-home assessment. Change (Δ) in perceived stress was measured as the difference between Cohen’s four-item Perceived Stress Scale from REGARDS baseline computer-assistant telephone interview (CATI) and Cohen’s four-item Perceived Stress Scale at Caregiving Transitions enrollment. Change (Δ) in depressive symptoms was measured as the difference between depressive symptoms at the REGARDS baseline CATI and depressive symptoms measured at Caregiving Transitions enrollment. Caregivers had significantly higher depressive symptoms and higher perceived stress than non-caregivers at transitions (ps < .001). Caregivers and non-caregivers did not differ significantly on any other study variables (ps > .06).

Cognition of transitioned caregivers and non-caregivers

Multilevel models were run for the five cognitive outcomes (word list learning, delayed word recall, animal fluency, letter fluency, and global cognition). Analyses began with the entire sample of transitioned caregivers and matched non-caregivers, with results shown in Table 2 and Figure 1. After transitions, transitioned caregivers and non-caregivers differed in their performance on word list learning, delayed word recall, and global cognition, such that transitioned caregivers recalled less on word list learning (Cohen’s d = 0.26), word recall (Cohen’s d = 0.17), and global cognition (Cohen’s d = 0.22) than matched non-caregivers. Linear effects of time showed decreases in non-caregivers’ performances in word list learning (Cohen’s d = 0.13) and global cognition (Cohen’s d = 0.10), while their performance in animal (Cohen’s d = 0.05) and letter fluency (Cohen’s d = 0.01) decreased. Transitioned caregivers showed a greater decline in word list learning performance (Cohen’s d = 0.13) but less decline in animal fluency scores (Cohen’s d = 0.05) over time. Moreover, scores on the animal fluency test were lower in the post compared to the pre-transition period (Cohen’s d = 0.11) for non-caregivers only. The effects of caregiving status and the rate of cognitive change specific to the post-transitional period (or similar time period for the non-caregivers) indicated more decline in word list learning (Cohen’s d = 0.11), word recall (Cohen’s d = 0.12), and global cognition (Cohen’s d = 0.03) for non-caregivers but less decline in word list learning (Cohen’s d = 0.11) and global cognition (Cohen’s d = 0.03) for transitioned caregivers. Non-caregivers had less decline for letter (Cohen’s d = 0.19) and animal fluency (Cohen’s d = 0.05) in this period, whereas transitioned caregivers had more decline in animal fluency (Cohen’s d = 0.05).

Table 2.

Level of and changes in cognition for caregivers and matched non-caregivers surrounding transitions.

Variable	Word list learning			Delayed word recall				Letter fluency			Animal fluency			Global cognitive functioning
	b	SE	p	b	SE	p	b		SE	p	b	SE	p	b	SE	p
Fixed effects
Intercept	20.33	0.29	<.001	7.50	0.12	<.001	11.99		0.27	<.001	17.91	0.32	<.001	5.82	0.02	<.001
Caregiver	−1.39	0.41	<.001	−0.35	0.17	.036	−0.50		0.39	.196	0.35	0.09	.689	−0.13	0.04	.001
Time	0.34	0.09	<.001	0.07	0.04	.058	−0.17		0.08	.048	−0.45	0.09	<.001	0.01	0.01	<.001
Caregiver × time	−0.41	0.15	.005	−0.07	0.04	.131	0.15		0.10	.143	0.32	0.11	.003	−0.01	<.01	.176
Post-transitions	−0.16	0.39	.681	0.21	0.17	.228	−0.45		0.33	.178	−0.75	0.35	.034	0.04	0.03	.261
Post-transitions × time	−0.45	0.19	.018	−0.23	0.08	.005	0.42		0.19	.027	0.50	0.18	.005	−0.06	<.01	<.001
Caregiver × post-transitions	0.36	0.58	.532	−0.07	0.26	.768	0.13		0.47	.786	−0.23	0.53	.666	−0.02	0.05	.688
Caregiver × post-­transitions × time	0.61	0.23	.009	0.09	0.13	.469	−0.37		0.27	.171	−0.71	0.23	.002	0.04	0.01	.001
Random effects
Variance of cognition	14.15	1.43	<.001	2.26	0.23	<.001	14.38		1.23	<.001	18.44	1.58	<.001	0.03	0.00	<.001
Covariance of cognition × time	0.33	0.23	.152	0.06	0.03	.102	0.04		0.21	.830	−0.41	0.26	.122	−0.00	0.00	<.001
Covariance of cognition × post-transitions	0.76	0.99	.444	−0.10	0.16	.055	0.08		0.81	.910	0.89	1.04	.391	0.06	0.01	<.001
Covariance of time × post-transitions	0.45	0.09	<.001	0.08	0.01	<.001	0.21		0.06	.001	0.45	0.09	<.001	0.00	0.00	<.001
Residual	8.71	0.28	<.001	1.71	0.05	<.001	6.24		0.20	<.001	9.44	0.27	<.001	0.27	0.01	<.001

Note. b = unstandardized parameter estimate.

Bolded values indicate significant effects. The caregiver variable was coded as 0 = non-caregiver and 1 = transitioned caregiver. The intercept is centered at the time of the caregiving transitions event. Time indicates the number of years to/from the caregiving event. Post-transition is coded 0 for all observations before the caregiver transitions and 1 for all observations after caregiving onset.

Figure 1.

Trajectory of cognitive function for transitioned caregivers and matched non-caregivers surrounding transitions.

Figure shows trajectories using multilevel models 3 years before and 3 years after transitions for the entire sample of transitioned caregivers and matched non-caregivers on word list learning, delayed word recall, animal fluency, and letter fluency.

We did not find caregiving transitions associated with global cognitive impairment after transitions or clinically significant declines in global cognition over time (results not shown). Within the study sample, 2.79% of transitioned caregivers and 4.78% of non-caregivers scored below the cutoff point for cognitive impairment during their final cognitive assessment after transitions.

Cognition in dementia versus non-dementia caregivers

Table 3 shows results for the unadjusted models for the associations between dementia and non-dementia caregiving and cognition. After caregiving transitions, dementia caregivers performed worse than non-dementia caregivers on tests of delayed word recall (Cohen’s d = 0.22) and word list learning (Cohen’s d = 0.21). However, the effect was attenuated to non-significance for word list learning, but the effect remained significant for delayed word recall (b = −0.44, SE = 0.21, p = .042, Cohen’s d = 0.23) (not shown in table) when models were adjusted for health and sociodemographic covariates. Interestingly, after transitions, dementia caregivers did better on animal fluency (b = 1.41, SE = 0.58, p = .01, Cohen’s d = 0.13) in the adjusted model (not shown in the table). There were no differences between dementia caregivers and non-dementia caregivers on the remaining cognitive measures, and no other differences between dementia and non-dementia caregivers over time or in the post-transitions period in unadjusted and adjusted models.

Table 3.

Unadjusted growth curve model for level of and changes in cognition for dementia caregivers compared to non-dementia caregivers.

Variable	b	SE	p	95% CI
Delayed word recall
Intercept	7.56	0.15	<.001	7.26, 7.85
Dementia caregivers	−0.73	0.22	.001	−1.18, −0.28
Time	−0.01	0.03	.736	−0.16, 0.12
Dementia caregivers × time	0.08	0.05	.901	−0.09, 0.10
Post-transitions × time	0.42	0.39	.287	−0.35, 1.19
Dementia caregivers × post-transitions	0.01	0.28	.999	−0.55, 0.55
Dementia caregivers × post-transitions × time	−0.07	0.16	.678	−0.40, 0.25
Word list learning
Intercept	19.64	0.36	<.001	19.07, 20.57
Dementia caregivers	−1.22	0.55	.002	−2.31, −0.14
Time	−0.01	0.07	.815	−0.16, 0.12
Dementia caregivers × time	0.08	0.11	.440	−0.13, 0.30
Post-transitions × time	0.21	0.20	.283	−0.17, 0.61
Dementia caregivers × post-transitions	0.04	0.62	.946	−1.18, 1.27
Dementia caregivers × post-transitions × time	−0.15	0.37	.673	−0.89, 0.58
Letter fluency
Intercept	11.40	0.37	<.001	10.66, 12.15
Dementia caregivers	0.47	0.54	.382	−0.59, 1.55
Time	−0.04	0.07	.542	−0.18, 0.09
Dementia caregivers × time	0.04	0.11	.692	−0.17, 0.26
Post-transitions × time	−0.07	0.17	.682	−0.42, 0.27
Dementia caregivers × post-transitions	−0.22	0.48	.641	−1.16, 0.71
Dementia caregivers × post-transitions × time	0.07	0.31	.813	−0.54, 0.69
Animal fluency
Intercept	17.67	0.42	<.001	16.83, 18.51
Dementia caregivers	0.65	0.61	.286	−0.54, 1.85
Time	−0.28	0.07	.062	−0.58, −0.15
Dementia caregivers × time	0.11	0.01	0.31	−0.10, 0.33
Post-transitions × time	−0.18	0.19	.357	−0.56, 0.20
Dementia caregiving × post-transitions	−0.48	0.54	.369	−1.55, 0.57
Dementia caregiving × post-transitions × time	−0.15	0.34	.647	−0.83, 0.51
Global cognitive functioning
Intercept	5.72	0.03	<.001	5.75, 5.88
Dementia caregivers	−0.07	0.04	.098	−0.16, 0.01
Time	0.01	0.00	.653	−0.02, 0.01
Dementia caregivers × time	0.01	0.00	.683	−0.01, 0.01
Post-transitions × time	−0.05	0.03	.167	−0.31, 0.02
Dementia caregivers × post-transitions	0.06	0.06	.296	−0.05, 0.19
Dementia caregivers × post-transitions × time	−0.03	0.02	.239	−0.08, 0.02

Note. b = unstandardized parameter estimate.

n = 117 for dementia caregivers and n = 134 for non-dementia caregivers. Bolded values indicate significant results. The intercept is centered at the time of the caregiving transition event. Time indicates the number of years to/from the caregiving event. Post-transition is coded 0 for all observations before the caregiver transitions and 1 for all observations after caregiving onset.

Caregiving transition and cognition mediated by depressive symptoms or perceived stress

We performed bootstrapped analyses to assess whether increases in depressive symptoms or perceived stress mediated the relationship between caregiving transitions and word list learning, delayed word recall, and global cognition over time. We also assessed whether increases in depressive symptoms or perceived stress mediated the relationship between dementia vs. non-dementia caregiving and delayed word recall. Bootstrapped analysis revealed a significant indirect effect for increases in depressive symptoms on caregiving transitions and word list learning and delayed word recall such that increases in depressive symptoms mediated a portion of the relationship between caregiving transitions and changes in word list learning (β = −0.23, 95% CI = −0.35, −0.11) and delayed word recall (β = −0.12, 95% CI = −0.17, −0.07) for transitioned caregivers compared to non-caregivers (shown in Figure 2). The caregiving transition was associated with increases in depressive symptoms for word list learning (β = 3.68; 95% CI = 3.26, 4.09) and delayed word recall (β = 3.71, 95% CI = 3.29, 4.12), and increases in depressive symptoms were associated with lower word list learning (β = −0.06; 95% CI = −0.09, −0.03) and delayed word recall scores (β = −0.03, 95% CI = −0.04, −0.02). Increases in depressive symptoms did not mediate the association between caregiving transitions and global cognition, nor did they mediate the relationship between dementia caregiving and delayed word recall. Also, we did not find increases in perceived stress mediating the relationship between caregiving transitions and any of these cognitive outcomes.

Figure 2.

Mediation analyses for the association between caregiving transition and word list learning and delayed word recall.

Changes in depressive symptoms were calculated as the difference between the REGARDS second in-home assessment (T2) and the REGARDS first in-home assessment (T1) for this variable. Mediation analyses assessed whether changes in depressive symptoms mediated the association between caregiving transitions and changes in word list learning performance (A) and delayed word recall (B). Mediation results revealed a significant indirect effect for changes in depressive symptoms, mediating the relationship between caregiving transition and word list learning and delayed word recall over time. **p < .01; ***p < .001.

Discussion

The results of these analyses add to the growing literature on the relationship between family caregiving and cognitive functioning. Caregiving Transitions Study used novel approaches and a population-based sample of transitioned caregivers and matched non-caregivers to examine the impact of caregiving on cognition from before and after the transition into the caregiving role. Our findings generally revealed that the transition to family caregiving had small yet significant effects on cognitive functioning. In line with our first hypothesis, we found transitioned caregivers performed significantly worse on word list learning, delayed word recall, and global cognition after transitions compared to matched non-caregiver controls measured at a similar time point. This is consistent with prior research that found caregivers did worse (Corrêa et al., 2019; Dassel et al., 2017; MacKenzie et al., 2009) than non-caregivers on cognitive functioning. However, Cohen’s standardized effect sizes for these associations were 0.26, 0.17, and 0.22 SD units, respectively. These small effect sizes can have questionable clinical and practical significance. When examining the time for post versus pre-transition, we did not find a caregiving interaction effect. In other words, cognitive functions were similar for both caregivers and non-caregivers when comparing post versus pre-transitions.

We found one significant effect in support of our second hypothesis that people who transition into caregiving would show more cognitive decline after transitions than non-caregivers. Specifically, transitioned caregivers compared to non-caregivers had declines in word list learning over time, but this effect was small with a Cohen’s effect size of only 0.13 SD units and did not remain significant in the post-transition period. Transitioned caregivers did not have cognitive decline in any of the other four cognitive outcomes during transitions. When examining the effects of caregiving status and the rate of cognitive change in the post-transition period (or a similar time period for the non-caregivers), we found two significant effects in that transitioned caregivers had less decline in word list learning and global cognition. Non-caregivers also showed more decline in word list learning, word recall, and global cognition. We did not find caregiving transitions associated with global cognitive impairment or clinically significant declines in global cognition. Since our study allowed us to examine cognitive trajectory surrounding transitions, the negative effects of caregiving on cognition could have been exacerbated after onset, as caregivers take on the new role of caregiving, but then became short-lived as caregivers adjusted to their roles with time. Moreover, our results also suggest that only specific cognitive domains may evidence a relationship with caregiving, since we did not find negative effects for cognitive tests assessing verbal fluency after transitions. There is abundant research on the negative effects of stress on brain health, particularly in the hippocampus (Kim et al., 2015), a brain region linked with memory, relatively more than other cognitive domains. It could be that caregiving creates a mentally stimulating environment where executive functioning and processing speed, which are important for verbal fluency, are optimized, whereas caregivers’ abilities to concentrate and retain information are temporarily compromised due to the overload of caregiving responsibilities.

We found two significant effects that support Hypothesis 3 when analyses were restricted to transitioned caregivers. Specifically, after transitions, dementia caregivers performed worse on word list learning and delayed word recall in the unadjusted models. In the adjusted model, transitioned caregivers performed worse only in delayed word recall and not in any of the other four cognitive outcomes. The effect size for this interaction was also small (Cohen’s d = 0.23). Dementia caregivers, compared to non-dementia caregivers, did not have worse cognitive functioning after transitions on any of the other cognitive outcomes in adjusted models. On the flipside, at transitions, dementia caregivers did better on animal fluency (b = 1.41, SE = 0.58, p = .01, Cohen’s d = 0.13) (not shown in table) than non-dementia caregivers. Although a prior CTS found general perceived stress to be higher in dementia caregivers than non-dementia caregivers (Sheehan et al., 2021), the lack of negative effects between dementia caregiving and cognition may be explained by caregiving being mentally stimulating and having a beneficial influence on cognitive performance. This is in line with our theoretical framework, whereby family caregiving might promote cognitively stimulating experiences (Hertzog et al., 2008) and increase cognitive reserve (Stern, 2009) through diverse daily activities associated with caregiving, meaningful interactions with the care recipient, and learning and receiving care training from health care professionals. Some studies have found that greater exposure to caregiving (e.g., more hours per week of caregiving) was associated with greater health benefits (Fredman et al., 2019). Our finding contradicts the Cache County Study by Norton et al. (2010) that reported caregivers of spouses who experienced incident dementia had a six times greater risk for incident dementia compared to caregivers of spouses who were dementia free. Although we did not specifically examine spousal caregivers, our results showed that dementia caregivers can have better cognition than non-dementia caregivers, especially in verbal fluency.

We found support for Hypothesis 4 such that increases in depressive symptoms mediated the relationship between caregiving transitions and cognitive change. Specifically, depressive symptoms mediated the relationship between caregiving transition and worse performance in word list learning and delayed word recall. However, increases in depressive symptoms did not mediate the association of caregiving transitions and poor global cognition. Moreover, increases in perceived stress did not mediate any other cognitive outcomes. Of note, caregivers enrolled in the CTS were providing care to people who needed assistance with ADLs or IADLs for an average of over 43 hr per week, and for an average duration of over 5 years (M = 5.78 years, SD = 2.54; range = 1.6–12 years) prior to caregiving transitions enrollment. Moreover, 76% resided with the care recipient (Haley et al., 2020; Roth et al., 2020a). Prior work using the same study sample confirmed that transitioned caregivers had significant increases in depressive symptoms and perceived stress with medium standardized effect sizes (0.50 SD units). Therefore, this group of caregivers did report substantial elevations in subjective distress compared to non-caregivers, and increases in depressive symptoms were associated with lower episodic memory over time. Family caregiving was associated with worse cognition after transitions, but the effect sizes for those negative interactions were less than half the effect size of depressive symptoms, and the effects were mediated by depressive symptoms. Prior research found depressive symptoms associated with a greater risk of cognitive decline (Deckers et al., 2015), and our findings are consistent with prior work by Vitaliano et al (2009) that showed depressive symptoms as a mediator in the relationship between caregiving status and poorer cognitive functioning.

Our finding that depression mediated some of the associations of caregiving with cognition, but that perceived stress did not, may provide insights into underlying mechanisms. Specifically, research has shown that chronic stress negatively affects the brain and cognitive function (Marin et al., 2011), whereas depressive symptoms are often associated with fatigue and reduced motivation, which may lead to lower cognitive performance (Snyder, 2013). Therefore, it may be argued that the effects of caregiving on cognition, which are mediated through depressive symptoms, may reflect reduced performance rather than reduced cognitive function, and could therefore be of a temporary nature. However, at the same time, research has shown that although cognitive deficits with depression are partially state-dependent, some deficits are still observed during remission (Bora et al., 2013; Rock et al., 2014). More research is needed to understand the pathways through which caregiving affects cognitive function and the persistence of effects over time.

Strengths and limitations

There were several major strengths of the current study. We used a nested case–control ancillary study from a national longitudinal cohort study on aging and cognition in the United States. Thus, caregivers and non-caregivers were recruited in the same manner, which does not occur in caregiving studies using convenience samples. Also, the caregivers in the CTS have done a substantial amount of caregiving prior to the caregiving transitions enrollment interview. Our 14-year cognitive data span allowed us to assess cognitive changes among individuals who transitioned into a family caregiving role compared with a matched sample of individuals who did not become caregivers over the same time. Our methodological study design using multilevel modeling allowed us to examine intra-individual differences and between-person differences in long-term cognitive outcomes associated with caregiving. Our study design also allowed for longitudinal mediation to examine psychological mechanisms driving the relationship between caregiving and cognition. Prior studies were mainly cross-sectional, and longitudinal follow-ups did not exceed 6 years. Our study also included multiple dimensions of cognition to assess potential differences in cognitive domains such as global cognition, verbal fluency, and episodic memory. This study is limited by not having specific caregiving characteristics, such as how care intensity or care recipient dementia-related behavioral/psychological symptom severity may affect cognition. Our study also did not include many racial/ethnic groups since the primary aim of REGARDS was to examine racial differences in stroke among Black and White adults. Future studies will benefit from the inclusion of different racial/ethnic samples and from examining the long-term effects of caregiving and cognition.

Conclusion

We found that transitioning into caregiving was associated with small yet significant declines in cognition compared to matched non-caregivers, but the declines were domain-specific. Specifically, we found only three significant effects after transitions that showed transitioned caregivers, compared to matched non-caregiver controls, performed significantly worse on three of the five cognitive outcomes, including word list learning, delayed word recall, and global cognition. When examining linear change over time, we found one significant effect for caregiving status and word list learning. Specifically, transitioned caregivers compared to non-caregivers had a greater decline in word list learning over time. Transitioned caregivers did not show declines in the other four cognitive outcomes after transitions. When examining caregiving status and cognition in the post-transitions period (or a similar time period for the non-caregivers), we found two significant effects by which transitioned caregivers reported less decline in two of the five cognitive outcomes, including word list learning and global cognition. Two significant effects showed that non-caregivers had more decline in word list learning, word recall, and global cognition. When examining dementia versus non-dementia caregiving, we found only one significant effect in the adjusted model, whereby transitioned dementia caregivers performed worse in delayed word recall.

Our results suggest that while caregiving can be challenging, attending to caregivers’ symptomology of depression may improve caregivers’ cognition. Caregivers are often resilient, and the effects of caregiving on cognition may be short-lived. Serious cognitive decline is not a common clinically relevant outcome of dementia caregiving. Future research should examine whether caregiver interventions that reduce depressive symptoms also have effects on these small and short-lived declines in some cognitive functions. Interventions that use cognitive reappraisal to find positive gains from caregiving (Cheng et al., 2017, 2020) may help reduce caregivers’ depressive symptoms and improve caregivers’ memory and global cognitive function. In clinical practice, identifying caregiver overload or burnout is needed to help alleviate caregiver depression and improve both caregiver and care recipient overall well-being.

Supplementary material

Supplementary material is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online.

Data availability

The CTS is an ancillary study to REGARDS, and to abide by its obligations with NIH/NINDS and the IRB of the University of Alabama at Birmingham, REGARDS facilitates data sharing through data use agreements. Any investigator is welcome to access the REGARDS data, including statistical code, through this process. Requests for data access may be sent to regardsadmin@uab.edu

Funding

This work was supported by a cooperative agreement (U01 NS041588) cofounded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Services. The CTS was further supported by an investigator-initiated grant (RF1AG050609) from the NIA. Additional support was provided by the Johns Hopkins University Claude D. Pepper Older Americans Independence Center funded by the NIA (P30 AG021334). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or NIA. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis, or interpretation of the data.

Conflict of interest

None declared.