Partner effects on caregiver and care recipient depressed mood: Heterogeneity across health condition and relationship type

Abstract

The well-being of caregivers and their care recipients is interrelated, although conflicting evidence has emerged across different caregiving populations. Using data from the National Health and Aging Trends Survey and the National Study of Caregiving (2015 and 2017, n=742 dyads), we constructed actor-partner interdependence models assessing how spillover (i.e., interdependence) of depressed mood varied by care recipient health condition (specifically cancer, dementia, stroke, and diabetes) and kinship type (spouse/partner, child, other relative, or non-relative). Across condition types, care recipient-to-caregiver partner effects were significantly larger in dyads with vs. without cancer and significantly smaller in dyads with vs. without diabetes (p_interactions<.05). Substantive differences in partner effects were observed by kinship type, although moderation was not statistically significant. The findings highlight potential heterogeneity in caregiver-care recipient interdependence with implications for future research and delivery of supportive care.

Introduction

Across the U.S., an estimated 53 million adults provide unpaid care for a family member or friend with an illness or disability (National Alliance for Caregiving & AARP Public Policy Institute, 2020). These caregivers play a critical role in supporting the emotional and physical health of their care recipients, as well as engaging in care coordination, personal care, and even home medical tasks (National Academies of Sciences Engineering and Medicine, 2016).

A growing body of research has indicated that a caregiver’s well-being is associated with care recipient outcomes, including depressed mood, psychological distress, quality of life, and even mortality (Kershaw et al., 2015; Lwi et al., 2017; Pristavec, 2019; Segrin et al., 2020). The interdependence of caregiver well-being on subsequent care recipient well-being, and vice versa (hereafter referred to as ‘spillover’ or ‘partner’ effects) is of increasing interest to enhance the impact of interventions and supportive care on the well-being of both care recipients and their caregivers. However, the results are not consistent across studies. For example, longitudinal studies have variously shown caregivers’ distress to be predictive of care recipients’ distress (Litzelman & Yabroff, 2015), recipients’ distress predictive of caregivers’ distress (Segrin et al., 2020), partner effects in both directions (Kershaw et al., 2015; Meyers et al., 2020), or no partner effects for psychological distress (Kim et al., 2015; Pristavec, 2019; Shaffer et al., 2016). Cross-sectional studies have also provided evidence of such interrelationships (Badr et al., 2017; Ivziku et al., 2019; Streck et al., 2021; Thomson et al., 2020). These studies differed considerably on several contextual factors such as patient health conditions, caregiver relationship types, as well as outcomes (e.g. depression, quality of life, well-being). Furthermore, much of the work in this area has focused on cancer caregiving and spousal caregiving. Very little attention has been given to examining well-being spillover effects across multiple care recipient health conditions or by the relationship between caregiver and care recipient.Explicit assessment of how spillover effects differ across contexts and populations, engaging the heterogeneity paradigm presented by Bryan and colleagues (2021), is a critical step in developing a nuanced understanding of interdependence that can more fully inform future interventions, policy and practice.

The caregiving stress process model (Pearlin et al., 1990) provides a theoretical touch-point for conceptualizing this potential heterogeneity. Primary stressors such as duration and type of care differ across health conditions and are related to subsequent caregiving burden and mental health (Kim & Schulz, 2008). For example, cancer caregiving is often characterized as being intense and episodic, with high levels of uncertainty (Hunt et al., 2016). Dementia caregivers often provide high-levels of care over long periods of time and experience high levels of burden (National Alliance for Caregiving & the Alzheimer’s Association, 2017). Furthermore, the care recipients’ functional ability is likely to differ both within and across condition types and may impact whether and how the well-being of the caregiver and care recipient interrelate. Secondary stressors may be particularly salient across relationship types, such as experiences of role and relational strain for adult child caregivers. For example, McLean and colleagues (2017) found that the intensity of caregiving was a stronger predictor of outcomes for adult-offspring caregivers compared to spousal caregivers of individuals who die of cancer.

In order to better understand these complex interrelationships and how heterogeneity may operate across contexts and populations, this analysis sought to elucidate how interdependence in caregiver and care recipient well-being differs in a single sample across key subgroups. By using a single nationally representative sample and assessing subgroups commonly used in research and implementation, we were able to identify how patterns of interdependence differ across subgroups in the context of a single methodological approach to data collection. We focus specifically on health conditions and relationship type, given the primacy of these characteristics in research design as well as intervention and supportive care implementation, which typically focus on a specific health condition and/or relationship type rather than (or sometimes in addition to) other factors such as burden. Based on the previous research findings, we hypothesized that interrelationships in depressed mood would be heterogenous across groups, with the most pronounced interdependence expected to be observed in cancer and dementia and among spouse/partner dyads. We are specifically interested in understanding substantive differences in partner effects (i.e., the observed magnitude and statistical significance of the associations among caregiver and care recipient depressed mood within each subgroup) as well as the statistical significance of such differences (i.e., determining whether partner effects are statistically significantly different across dyads with and without each condition type, and across relationship types). The findings have implications for understanding heterogeneity in the interrelationships between caregiver and care recipient well-being in both research and practice.

Materials and Methods

Data were from the National Health and Aging Trends Study (NHATS) and the National Study of Caregiving (NSOC) (National Health and Aging Trends Study). Briefly, the NHATS is a nationally representative dataset of Medicare beneficiaries ages 65 and older. Respondents participated in annual interviews starting in 2011 (Round 1). In 2015 (Round 5), the sample was refreshed with additional individuals from the Medicare sampling frame to maintain a representative sample. In 2011, 2015, and 2017, the NHATS respondents were invited to identify family and unpaid caregivers who offer them assistance with self-care, mobility, medical or household activities. Caregivers were interviewed in the supplemental NSOC. The 2015 sample of caregivers were invited to complete a follow-up interview in 2017 (Round 7). The NHATS is sponsored by the National Institute on Aging (grant number NIA U01AG32947) and was conducted by the Johns Hopkins University. This secondary analysis was reviewed by the Minimal Risk Research Institutional Review Board of the University of Wisconsin-Madison and certified as exempt (2018–1087).

Sample selection and exclusion criteria

Dyads were included in the present study if the caregiver completed the 2015 caregiver interview and the 2017 longitudinal follow-up. Caregivers were eligible if they were: 1) nominated by an NHATS participant in round 5 as someone who provided assistance with self-care, mobility, medical or household activities; 2) a relative or unpaid non-relative (e.g., friend, neighbor) of the care recipient; and 3) providing ongoing care at round 7. Dyads were excluded if the NHATS sample person was deceased or if the caregiver was no longer providing care at the 2017 follow-up, as these participants did not receive questions related to psychosocial well-being, including the outcome variable of interest. Individuals with aggressive or early onset conditions may therefore be underrepresented in this sample.

A primary caregiver was identified for each NHATS participant with linked NSOC data. In most cases, only one eligible caregiver was available in the NSOC dataset (n=573). If more than one eligible caregiver was linked to the NHATS sample person, the caregiver providing the greatest amount of care was selected (n=219). If multiple caregivers provided equivalent care, one caregiver was selected at random (n=2). Fifty-five primary caregivers and 54 care recipients were excluded from this sample due to missing data on the outcome variable (depressed mood) at either timepoint or missing partner depressed mood at round 5. An additional 21 caregivers and 19 care recipients were excluded from the analysis due to missing covariate data. This resulted in a final analytic sample of 742 dyads. Compared to the full sample, caregivers excluded due to missing data were slightly older (mean 66.3 vs. 63.0 years, p=0.05), more likely to be Black non-Hispanic, Hispanic, or other race/ethnicity (50% of excluded vs. 38% of final, p=.05), provided greater hours of care per week (mean 35.1 vs.21.2, p=.03)., and were less likely to care for someone with lung disease (16% vs. 26%, p=.04). Care recipients excluded due to missing data were more likely to be Black non-Hispanic, Hispanic, or other race/ethnicity (50% of excluded vs. 35% of final, p=.03), have dementia (31% vs. 16%, p=.003), or not to report their income (32% vs. 17%, p=.03), and were less likely to be married or partnered (68% vs. 56%, p=.04). Those excluded due to missing did not differ from the final sample on other characteristics, and those with missing covariate data did not differ from the full sample on level of depressed mood.

Measures

Depressed mood was assessed using the two-item Patient Health Questionnaire-2 (Kroenke et al., 2003), which measured the frequency that respondents “had little interest or pleasure in doing things” and “felt down, depressed or hopeless” over the past month (0= not at all, 1= several days, 2=more than half the days, 3= nearly every day). The items were summed and a score of 3 was considered the cut-point for depressed mood (Kroenke et al., 2003). This categorical variable was used in the descriptive analyses while the continuous sum scores at rounds 5 and 7 were used in the multivariable analyses.

Participants’ age, sex (male, female), race/ethnicity (white non-Hispanic, Black, Hispanic, other), marital status (married/living with a partner versus never married, separated, or divorced), education level (high school or less; some college, vocational schools, or associates degree; bachelor’s degree or higher), employment status (employed, not employed), household size (number of people in the household), poverty status (<100%, 100%–<200%, 200%–<400%, or >400% of the federal poverty level), and insurance coverage (all care recipients were enrolled in Medicare thus insurance type included whether they had private or public insurance as well) were assessed in both caregivers and care recipients.

Health conditions were assessed as the presence of.: arthritis, cancer, cardiovascular disease, diabetes, dementia, lung disease, osteoporosis, or stroke. Conditions were not mutually exclusive. Total number of conditions was counted.

Caregiving characteristics included the caregiver’s relationship to the care recipient (i.e., spouse/partner, child, or other relative or non-relative), coresidence with the care recipient, hours of care per week, duration of care (years), care tasks, care recipient activity limitations, and total number of helpers listed by the care recipient. Hours of care per week was highly skewed and divided into approximate tertiles for the descriptive analyses <7 hours, 7–14 hours, or >14 hours. Duration of care was measured by the number of years caregivers provided care to the care recipient. Caregiving tasks included providing personal care (e.g., bathing, eating), medical/nursing tasks, support for health behaviors including diet or exercise, and medical logistics such as making appointments. Care recipients were coded as having an activity limitation if they answered yes to any of the seven items assessing health or functioning limitations (e.g., work, social participation, other activities). Caregiving characteristics were coded as dyad-level variables and applied to both the caregiver and care recipient.

Statistical Approach

Descriptive statistics were calculated for all covariates (percentages or mean/standard deviation). Linear dyadic multilevel models with random intercepts were constructed following the actor–partner interdependence model (APIM) framework (Kenny et al., 2006). Individuals (the caregiver and the care recipient) were nested within the dyad and a two-intercept model was used. The model examined the association between caregiver and care recipient depressed mood at baseline on both self and partner depressed mood at follow-up. Actor effects (the association between the participant’s depressed mood at baseline on their own mood at follow-up) and partner effects (the association between the participant’s depressed mood at baseline with their partner’s depressed mood at follow-up) were calculated for both caregivers and care recipients (full-sample model). To assess effect modification, we then entered an interaction term for each care recipient health condition of interest (cancer, dementia, stroke, diabetes; not mutually exclusive) and for relationship type. The health conditions of interest were selected to represent a range of conditions that typically involve considerable caregiver engagement (as evidenced by the greater number of care tasks and increased likelihood of engaging in nursing/medical tasks by the caregiver in this sample); moreover, cancer and dementia are the most well-represented conditions in the literature on caregiver-care recipient interdependence. Effect sizes (ES) were calculated as the regression estimate divided by the standard deviation of the outcome variable for the group. All models controlled for sociodemographic characteristics, health factors, caregiving characteristics and depressed mood at baseline; continuous variables were centered at the group mean. SAS 9.4 was used for all analyses.

Results

Table 1 summarizes the characteristics of the caregivers and care recipients in the final sample. Caregivers’ mean age was 63 years (standard deviation [SD] = 13 years) and the care recipients’ mean age was 82 years (SD = 8 years). The majority of caregivers and care recipients were women (70% and 67%, respectively). Approximately 62% of caregivers had an education level beyond high-school and most (72%) were not employed. Care recipients had slightly worse psychosocial profiles at both time points and were more likely to report depressed mood than caregivers. Most caregivers lived with the care recipients (65%) and have been providing care for a mean of 8 years (Table 2). Complete characteristics by health condition and relationship type are available in Supplemental Table 1.

Table 1.

Characteristics of caregivers and care recipients

	Caregivers	Care recipients
	% or Mean (SD)	% or Mean (SD)
Age (years)	63.0 (13.2)	81.7 (8.0)
Gender
Female	70.3	66.7
Male	29.7	33.3
Race and Ethnicity
White (non-Hispanic)	61.6	64.9
Black, Hispanic, or Other	38.4	35.1
Marital Status
Married/partnered	64.4	44.0
Not married/partnered	35.7	56.0
Education
High school or less	38.2	60.6
Some college, vocational or associates degree	33.4	23.6
Bachelors or higher	28.4	15.8
Employment Status
Employed for pay or self-employed	27.6	3.2
Not employed for outside the home	72.4	96.8
Household Size Categorya
Single	7.8	26.9
Two Persons	17.2	46.6
Three or More	75.0	26.5
Federal Poverty Level
<100%	20.1	26.8
100%-<200%	18.7	23.3
200%-<400%	19.1	19.3
400%+	22.6	14.0
Unknown/unreported	19.6	16.6
Number of health conditions	1.6 (1.3)	3.0 (1.3)
Depressed mood
Mean score at baseline	1.1 (1.3)	1.5 (1.5)
Clinically significant depressed mood at baseline	14.4	22.5

SD: Standard deviation

Data are from the National Health and Aging Trends Survey (NHATS) and the National Study of Caregiving (NSOC) (2015; n=742 dyads)

^aItem missing in a subset of caregivers (n=61)

Table 2.

Characteristics of the caregiving experience

	%
Caregiver’s Relationship to Care recipient
Spouse/Partner	33.7
Child	51.4
Other Relative	9.2
Other Non-relative	5.8
Distancea
Mean (minutes)	9.7 (35.2)
Co-resident	65.0
<10 Minutes	12.9
10–20 Minutes	12.4
>20 Minutes	9.8
Hours of Care Per Week
Mean (hours)	21.6 (29.3)
<7 hours	37.6
7–14 hours	23.7
>14 hours	38.7
Duration of Care (years)	8.0 (10.2)
Health Conditionsb
Mean number of conditions	3.0 (1.3)
Cardiovascular disease	84.5
Arthritis	76.8
Osteoporosis	34.2
Diabetes	35.3
Lung disease	25.9
Cancer	14.2
Skin	33.3
Breast	18.1
Prostate	12.4
Bladder	<5
Ovarian	<5
Colon	6.7
Kidney	5.7
Other	7.6
Multiple	10.5
Stroke	10.2
Dementia	17.4
One or more activity limitations	60.2
Care tasks
Household chores	79.7
Shopping	84.8
Personal care	51.4
Getting around the homec	51.8
Transportationc	74.5
Banking	67.4
Medical/nursing care	61.6
Health behavior support	42.1
Medical logistics	75.5
Mean number of above tasks	5.9 (2.4)
Total number of helpers	2.4 (1.3)

SD: Standard deviation

Data are from the National Health and Aging Trends Survey (NHATS) and the National Study of Caregiving (NSOC) (2015; n=742 dyads). Cell sizes less than five are redacted.

^aItem was inadvertently skipped and is therefore missing in a subset of caregivers (n=57)

^bHealth conditions are not mutually exclusive

^cMissing on one observation

Table 3 shows the actor and partner effects of the APIM by care recipient health condition (full results tables, including estimates for all covariates, are available in Supplemental Table 2). In the full sample, the APIM indicated evidence for small partner effects on depression scores. Caregiver’s depressed mood at T1 was associated with care recipient’s depressed mood at T2 (β=0.09, ES=0.05, p=.03). Similarly, care recipient’s depressed mood at T1 predicted caregiver’s depressed mood at T2 (β=0.07, ES=.05, p=.02).

Table 3.

Actor-Partner effects of depressed mood on later depressed mood in caregiver-care recipient dyads across major condition types (n=742 dyads)

	Full Sample			Cancera			Dementia			Stroke			Diabetesa,b
	Beta	ES	SE	Beta	ES	SE	Beta	ES	SE	Beta	ES	SE	Beta	ES	SE
Actor Effects
Care recipient	0.43	0.26	0.04 ***	0.45	0.28	0.09 ***	0.47	0.25	0.09 ***	0.45	0.29	0.11 ***	0.44	0.24	0.06 ***
Caregiver	0.44	0.32	0.04 ***	0.40	0.26	0.08 ***	0.46	0.29	0.09 ***	0.32	0.21	0.11 **	0.47	0.33	0.05 ***
Partner effects
Caregiver-to-care recipient	0.09	0.05	0.04 *	0.20	0.12	0.09 *	0.22	0.12	0.11 *	0.02	0.02	0.11	0.16	0.09	0.07 *
Care recipient-to-caregiver	0.07	0.05	0.03 *	0.23	0.15	0.08 **	0.00	0.00	0.07	0.08	0.05	0.11	−0.02	−0.02	0.05

^*p<0.05;

^**p<0.01;

^***p<0.001

^aPartner effect (care recipient to caregiver) significantly different from dyads without the condition (p<.05);

^bPartner effects (caregiver to care

recipient) approaching statistically significant difference from dyads without the condition (p<.10)

ES: Effect size; SE: Standard error

Longitudinal data from the National Health and Aging Trends Survey and National Study of Caregiving (2015 and 2017).

Models control for: age, gender, race/ethnicity, marital status, education, employment status, poverty status, relationship to and coresidence with the care recipient, care recipient’s household size, hours of care per week, years of care, presences of care recipient activity limitations, care recipient health condition(s), care tasks, and total number of helpers.

Effect modification by condition type

There were substantive differences in partner effects among dyads with the health conditions examined (Table 3; Figure 1). Partner effects on care recipients’ depression scores were observed in dyads with cancer (β=0.20, ES=0.12, p<.05), dementia (β=0.22, ES=0.12, p<.05) and diabetes (β=0.16, ES=0.09, p<.05), but not stroke. Similarly, partner effects were observed on caregivers’ depression scores in dyads with cancer (care recipient-to-caregiver: β =0.23, ES=0.15, p<.01). There were no statistically significant associations between care recipient depressed mood at T1 and caregiver depressed mood at T2 in dyads with dementia, stroke, or diabetes. Figure 1 depicts predicted depression scores at T2 as a function of partner’s depressed mood at T1 in the full sample and stratified by care recipient condition type. The figure illustrates the substantive differences in the partner effects.

Figure 1.

Effect modification of care recipient condition on partner effects for depressed mood. Panel A shows predicted associations of caregiver depressed mood with later care recipient depressed mood for the full sample (solid line) and stratified by care recipient health condition. Panel B shows predicated associations of care recipient depressed mood with later caregiver depressed mood. Data are from the National Health and Aging Trends Survey and National Study of Caregiving (2015 and 2017, n=742 dyads); predicted values are calculated from models adjusting for caregiver and care recipient covariates. * indicates slope is significant within the subgroup (p<.05).

Moderation by cancer and diabetes status were statistically significant. Specifically, the care recipient-to-caregiver partner effect was statistically significantly larger in dyads with cancer than those without cancer (p_interaction=.04), and significantly smaller in dyads with diabetes than those without diabetes (p_interaction=.01). Additionally, the caregiver-to-care recipient partner effect was larger in dyads with diabetes than those without diabetes, approaching statistical significance (p_interaction=.09).

Effect modification by relationship type

Partner effects were substantively but not statistically significantly different by relationship type (Table 4; full results tables, including estimates for all covariates, are available in Supplemental Table 3). Partner effects were observed on care recipients’ depression scores among spouse/partner dyads (caregiver-to-care recipient: β =0.14, ES=0.09, p<.05), but not in adult child or other caregiving dyads. Partner effects on caregivers’ depression scores were observed for adult child caregivers (care recipient-to-caregiver: β=0.10, ES=0.07, p<.05). There was no evidence of statistically significant partner effects on caregivers’ depression scores among spouse/partner, other relative, or friend/non-relative caregivers. Figure 2 depicts predicted depression scores at T2 as a function of partner’s depressed mood at T1, stratified by relationship type. Moderation by relationship type was not statistically significant (p>.05), despite the substantive differences observed.

Table 4.

Actor-Partner effects of depressed mood on later depressed mood in caregiver-care recipient dyads across relationship types (n=743 dyads)

	Full Sample			Spouse/partner			Child			Other relative			Friend or other non-relative (n=113 dyads)
	Beta	ES	SE	Beta	ES	SE	Beta	ES	SE	Beta	ES	SE	Beta	ES	SE
Actor Effects
Care recipient	0.43	0.26	0.04 ***	0.47	0.29	0.06 ***	0.42	0.26	0.05 ***	0.38	0.24	0.12 **	0.35	0.24	0.13 *
Caregiver	0.44	0.32	0.04 ***	0.49	0.34	0.06 ***	0.39	0.28	0.05 ***	0.40	0.32	0.11***	0.63	0.48	0.11***
Partner effects
Caregiver-to-care recipient	0.09	0.05	0.04 *	0.14	0.09	0.07 *	0.03	0.02	0.06	0.11	0.07	0.14	0.22	0.15	0.15
Care recipient-to-caregiver	0.07	0.05	0.03 *	0.01	0.01	0.06	0.10	0.07	0.04 *	0.08	0.07	0.09	0.15	0.11	0.10

^*p<0.05;

^**p<0.01;

^***p<0.001; actor and partner effects were not statistically different across relationship categories

ES: Effect size; SE: Standard error; CG: Caregiver; CR: Care recipient

Longitudinal data from the National Health and Aging Trends Survey and National Study of Caregiving (2015 and 2017)

Models control for: age, gender, race/ethnicity, marital status, education, employment status, poverty status, relationship to and coresidence with the care recipient, care recipient’s household size, hours of care per week, years of care, presences of care recipient activity limitations, care recipient health condition(s), care tasks, and total number of caregivers.

Figure 2.

Effect modification of relationship type on partner effects for depressed mood. Panel A shows predicted associations of caregiver depressed mood with later care recipient depressed mood for the full sample (solid line) and stratified by caregivers’ relationship to the care recipient. Panel B shows predicated associations of care recipient depressed mood with later caregiver depressed mood. Data are from the National Health and Aging Trends Survey and National Study of Caregiving (2015 and 2017, n=742 dyads); predicted values are calculated from models adjusting for caregiver and care recipient covariates. *indicates slope is significant within the subgroup (p<.05).

Discussion

This study assessed how the interrelationships between caregiver and care recipient well-being differ across health conditions or caregiver relationship type. The findings indicate that partner effects of caregiver and care recipient depressed mood vary by major caregiving characteristics. In the full sample, we found that caregiver and care recipient depressed mood were both associated with partners’ depressed mood two years later, albeit with small effect sizes. Substantive differences were observed across health condition and relationship types: caregiver-to-care recipient effects were observed among dyads with cancer, dementia, and diabetes, as well as spouse/partner caregivers, while care recipient-to-caregiver effects were observed among dyads with cancer and adult child caregivers. The moderation analyses provide evidence that differences in partner effects were statistically significant in dyads with cancer and diabetes compared to dyads without those conditions. However, moderation by relationship type was not statistically significant in the full model.

These results emphasize the importance of context in considering how caregiver and care recipient well-being are intertwined, and the potential impact of heterogeneity in population-based analyses. In families experiencing cancer, dementia or diabetes, or patients with spouse/partner caregivers (the subgroups in which substantive caregiver-to-care recipient partner effects were observed), it may be particularly important to monitor the caregiver’s well-being and subsequent ripple effects on care recipient outcomes. While we were not able to assess potential mechanisms in the current study, we posit that these caregivers may have a particularly active role in symptom monitoring, care coordination, or supporting patient self-management (Hunt et al., 2016; The Hormone Foundation & The National Alliance for Caregiving, 2010). Caregivers experiencing depressed mood may have lower capacity to engage in these supportive roles, with potential adverse consequences for patients’ well-being (Litzelman, 2019). It is important to note that dyads at the end of life could not be included in this study, and that the sample is likely to underrepresent individuals with aggressive, advanced stage, or early onset disease. This is another important contextual factor that deserves additional evaluation. It could be that the interdependence observed here is a particular feature of dyads with long-standing, chronic, or non-aggressive conditions, and patterns in dyads with severe or quick moving conditions may be different. In addition, future research to assess the mechanisms underlying heterogeneity in partner effects will further inform approaches to intervening at the individual, dyadic, and systems levels.

By contrast, differences in care recipient-to-caregiver spillover may be explained, at least in part, by considering the intensity of secondary caregiving strains, such as role conflict or poor adjustment to the caregiving role. For many caregivers, the transition into a caregiving role is accompanied by elevated levels of distress that resolve as the families adapts to their new roles and circumstances (e.g., (Graf et al., 2017)). In conditions with high levels of uncertainty or that are episodic in nature – such as cancer, in which the patient’s well-being and functional abilities can fluctuate considerably (Hunt et al., 2016)-- caregivers may experience continued or cyclical role strains and interpsychic strains that challenge their ability to adjust and adapt. We also observed care recipient-to-caregiver spillover in adult child and other types of caregivers, for whom role strain and relational changes may be particularly salient (Dellmann-Jenkins et al., 2001), although moderation by relationship type was not statistically significant overall. In these contexts of uncertainty, caregivers may be unable to fully adjust to their caregiving role and thus remain particularly vulnerable to variations in the care recipient’s well-being.

The findings have implications for both practice and research. Assessing caregiver well-being may support risk stratification and early recognition of psychological distress in patients, particularly those with cancer, dementia, or diabetes, and those with spousal caregivers. Dyadic interventions and dyadic coping are of increasing interest for supporting both patient and caregiver outcomes, although systemic reviews and meta-analyses have highlighted inconsistent efficacy of such approaches (Badr et al., 2019; Buck et al., 2018; Pucciarelli et al., 2021). The current findings suggest that contextual factors such as the health condition and relationship type may contribute to heterogenous effects. Other contextual factors are likely also important in considering how well-being is interrelated in caregivers and care recipients and deserve future study, including caregiver and care recipient gender and gender concordance, relationship quality, the caregiving and disease trajectories (including the end-of-life period), the role of disease characteristics such as aggressiveness or early onset conditions, and the caregiver’s and care recipient’s life course. For example, Badr et al. (2017) have demonstrated that in chronic obstructive pulmonary disease, higher levels of depression in a partner were associated with higher levels of depression for women, regardless of whether women were patients or caregivers. Future research and practice should consider how such contexts holistically influence caregiver and care recipient well-being, both independently and intersectionally.

While this analysis sought to tease out heterogeneity in population-based caregiver research, heterogeneity remains an important consideration in interpreting the findings both within and across subgroup analyses. The cancer subgroup, for example, represents a diverse array of diagnoses, prognoses, treatment protocols, and disease progressions that can be expected to impact the interdependence of caregiver and care recipient well-being. The sample of individuals with cancer, in particular, is likely skewed towards those with non-aggressive or well controlled disease while the sample of individuals with dementia likely underrepresents those with early onset conditions. In addition, most care recipients in our sample reported multiple chronic conditions, and the sample likely underrepresents individuals with advanced stage conditions. While this is representative of the reality of older age in the U.S., it remains challenging to assess the role of multiple chronic conditions on care recipient, caregivers, and their interrelationships. Across subgroups, roles and characteristics can be entangled in ways that are difficult to tease out both conceptually and analytically. In the analysis across relationship types, for example, adult child caregivers tend to be younger than spouse/partner caregivers. They are also more likely to have other roles and responsibilities, including employment or childcare, which may reduce their emotional load bearing capacity and create an environment that permits greater spillover of the care recipient’s mood and well-being. Future research is needed to assess the impact of heterogeneity across other key factors, such amount of care, types of care, or overall burden. Applying a systems-thinking lens may facilitate the field’s ability to avoid overly reductionist approaches and embrace this complexity in both research and practice.

The findings from this study should be interpreted in light of several potential limitations. Given the nature of the design and methodology of data collection, selection bias for individuals with relatively chronic, non-aggressive conditions, and the exclusion of those at the end of life, is likely. The results therefore may not be generalizable to those with aggressive or early onset conditions, or those approaching end of life. Even with the large sample size, there was limited power to detect significant differences in the moderation analyses. The statistical significance of the interaction effects should therefore be interpreted cautiously in light of the potential for type II error. This study assessed older adult care recipients only. Patterns of interrelationships are likely to differ for care recipients at an earlier stage in the life course. We assessed only the most involved caregivers who participated in the study. Understanding how these interrelationships operate among networks of caregivers is an important area of future research. Depressed mood was assessed using a two-item screener, which is a blunt instrument for assessing this construct. In clinical practice, an elevated score would be followed by further assessment to determine major depression. Additionally, there is a risk for self-selection bias, as those providing the greatest amount of care or experiencing the greatest levels of burden may have been disproportionately likely to drop out of the study or provide incomplete data. Furthermore, we could not include caregivers no longer providing care at the 2017 follow-up, as only caregivers who were providing ongoing care were asked about psychosocial well-being. The findings may not be generalizable to caregivers no longer providing care. Nevertheless, the findings draw on longitudinal data from a large, national sample and thus provide important insight into the way partner effects of depressed mood differ across health conditions and relationship types.

Conclusion

In conclusion, this study highlighted the potential for heterogenous caregiver-care recipient interdependence across care contexts. Specifically, statistically significant caregiver-to-care recipient spillover was observed in cancer, dementia, and diabetes, and spousal/partner caregiver dyads. By contrast, care recipient-to-caregiver spillover was most pronounced in cancer and in dyads with adult child caregivers. The findings highlight the importance of considering contextual factors in assessing and intervening on dyadic well-being.

Availability of data and material:

Data are publicly available from nhats.org