Caring Together: Trajectories of Paid and Family Caregiving Support to Those Living in the Community With Dementia

Abstract

Objectives

Paid caregivers (e.g., home health aides) often work with family caregivers to support persons living with dementia at home. We identify (a) unique trajectories of paid and family caregiving support among persons living with dementia with high care needs and (b) factors associated with these trajectories.

Methods

We used group-based multiple trajectory modeling to identify distinct trajectories of paid and family caregiving hours among National Health and Aging Trends Study respondents with dementia who died or moved to a nursing home (n = 334, mean follow-up 5.5 years). We examined differences between trajectory groups and identified factors associated with group membership using generalized estimating equation modeling.

Results

A 3-group model best fit our data: (a) “low/stable care” (61.3% of respondents) with stable, low/no paid care and moderate family care, (b) “increasing paid care” with increasing, moderate paid and family care, and (c) “high family care” with increasing, high family care and stable, low paid care. While both the “increasing paid care” and “high family care” groups were more functionally impaired than the “low/stable care” group, the “high family care” group was also more likely to be non-White and experience multiple medical comorbidities, depression, and social isolation.

Discussion

Study findings highlight the importance of considering unique arrangements in dementia care. Receipt of paid care was not only determined by patient care needs. Creating equitable access to paid care may be a particularly important way to support both persons living with dementia and their family caregivers as care needs grow.

As dementia advances, the care needs of the nearly 5.8 million persons living with dementia will steadily increase (2020 Alzheimer’s disease facts and figures, 2020; Kasper et al., 2015; Reckrey, Bollens-Lund, Husain, et al., 2021). Due to both individual and family preference as well as growing evidence that community-based long-term services and supports (LTSS) can be cost-effective, the locus of long-term care is shifting from institutions into the community (Kaye et al., 2010; Kitchener et al., 2005). Concerns about safety in congregate settings like nursing homes during the coronavirus disease 2019 pandemic may accelerate this trend. As a result, it is essential to understand how care needs of persons living with dementia in the community are met over time, particularly as care needs grow.

Family members are the primary providers of care in the home and are increasingly recognized as an essential part of the LTSS team (Qualls, 2016). Yet the aging population and shrinking numbers of young people available to provide care, along with factors such as the need for adult children to balance elder care with care of young children, will over time result in increased demands on a smaller group of family caregivers. Such shifts may leave many persons living with dementia with no family caregivers at all (Carney et al., 2016). Home-based models of clinical and long-term care may further exacerbate demands on family caregivers as individuals remain living in the community with a greater illness burden for longer (Dong et al., 2019; Samus et al., 2018).

In order to provide additional needed support for persons living with dementia in the home, many family caregivers turn to paid caregivers (i.e., home health aides, personal care attendants, and other direct care workers in the home; Bressan et al., 2020). Over 2 million paid caregivers provide care in the community and this is one of the fastest growing professions in the United States (Spetz et al., 2019). Paid caregivers are funded by a patchwork of payers including Medicaid, long-term care insurance, and personal out-of-pocket spending (Buhler-Wilkerson, 2007; Janus & Ermisch, 2015). In 2015, one quarter of all persons living with dementia in the community received paid care and nearly half of those with advanced dementia received paid care; for those who received paid care, paid caregivers provided on average about half of total care hours received (Reckrey, Morrison, et al., 2020).

Family caregivers play an important role directing care as cognitive impairment in dementia increases (Kasper et al., 2015). Therefore, we have conceptualized a home-based caregiving team in which paid caregivers work closely with family caregivers to support persons living with dementia in the community (Reckrey, Boerner, et al., 2021). Care hours are shared among paid and family caregivers who divide care tasks in a wide variety of ways (Kemper, 1992; Lyons & Zarit, 1999; Noelker & Bass, 1989; Wiles, 2003), and evidence suggests that paid caregivers develop relationships with both the individuals they care for and their families (Ayalon & Roziner, 2016; Kemp et al., 2009, 2018). These care arrangements depend not only on characteristics of the care recipient and their family, but also the larger health care system and community in which they exist; they are dynamic and responsive to growing care needs that arise at the end of life and changes in care context over time.

Existing research on paid caregivers has focused on key issues related to the paid caregiver workforce such as recruitment and retention, job satisfaction, and training (Kelly et al., 2013; Stone et al., 2017). Paid caregivers are rarely examined in relation to the individuals they care for or the families with whom they share care. In order to provide appropriate support tailored to the unique needs of each person living with dementia, it is important to understand if and when paid caregivers are involved in care and the variety of ways that paid and family work together to provide care, particularly as disease progresses and care needs are high. Because evidence suggests that those with dementia have high care needs in the years prior to death or nursing home placement (Reckrey, Bollens-Lund, Husain, et al., 2021), this period is a critical one in which to understand these interrelated caregiving roles.

Using longitudinal data from the National Health and Aging Trends Study (NHATS), this study seeks to: (a) identify unique trajectories of paid and family caregiving support among persons living in the community with dementia with high care needs over time, and (b) identify factors (e.g., sociodemographic, clinical and functional, geographic and living arrangement) associated with these trajectories.

Method

Study Population

We drew data from the nationally representative NHATS. This longitudinal study of late-life disability and function among Medicare beneficiaries ≥65 years of age started in 2011; during the initial NHATS interview and at each annual interview that follows, a wide variety of functional, cognitive, and caregiving information is obtained (Kasper & Freedman, 2020). Of note, NHATS determines dementia status (i.e., probable dementia, possible dementia, or no dementia) for each participant at each annual interview using criteria that incorporate self-report of dementia, proxy responses to the Alzheimer’s disease-8 screening tool, and a cognitive interview that assesses memory, orientation, and function (Kasper et al., 2013).

In order to identify a cohort of persons living in the community with dementia with high levels of care needs that we could follow over time, we focused on individuals with dementia in the period before death or nursing home placement. We first identified all community-dwelling NHATS respondents from the 2011 cohort (n = 7,609) who died or permanently moved to a nursing home between 2011 and 2020 (n = 2,509). Of these, 1,274 had any probable dementia during follow-up. We then limited our sample to those with probable dementia 4 or more years prior to death or nursing home placement in order to examine the experiences of a population of persons living with dementia in the community over time and to ensure adequate data points for a robust trajectory analysis. Only interviews at or after the time of dementia identification were included in the analysis. This resulted in a final sample of n = 334 persons living in the community with dementia with 1,837 included interviews.

Measures

Outcome variables

For each NHATS respondent, we calculated the (a) paid caregiving hours, (b) family caregiving hours, and (c) total caregiving hours received at each annual survey.

During each annual interview, NHATS respondents or proxies are asked whether or not help was needed to perform a particular self-care or mobility task (eating, getting out of bed, showering, toileting, dressing, getting around inside, getting around outside) or if help was needed for health and functioning reasons to perform a particular household activity (laundry, shopping, meal preparation, medication management, bills and banking, addressing money matters, and going to the doctor) in the previous month. If so, the individual providing care is identified and the following information about the caregiver is subsequently obtained: (a) relationship to study respondent, (b) hours of help provided, (c) whether care was paid or unpaid, (d) if paid, sources of payment (e.g., self or family, a government program).

Paid caregiving hours included all paid caregiving hours regardless of source of payment or relationship to caregiving recipient and missing caregiving hours were imputed using guidance from NHATS (Freedman et al., 2014). Family caregiving hours included all hours provided by unpaid caregivers regardless of relationship. In order to ground hours received in the context of an individual person living with dementia and consistent with previous literature (Kelley et al., 2015), we capped total family caregiving hours at 168 hr per week (24 hr × 7 days). Because evidence suggests that paid caregiving does not simply substitute for family care (Lyons & Zarit, 1999; Noelker & Bass, 1989), we separately capped paid caregiving hours for any individual person living with dementia at 168 hr per week.

Covariates

We included a wide variety of covariates from NHATS corresponding to factors we conceptualize as important to the home-based dementia care team, which were obtained via self-report and drawn from the NHATS interview in which dementia was first identified (i.e., at baseline) unless otherwise noted. These included sociodemographic characteristics (e.g., age, gender, marital status, education, race, income, Medicaid status), clinical and functional characteristics (e.g., interview via proxy, number of chronic conditions, depression, social isolation, self-care activities for which help was received, household activities for which help was received), and geographic and living arrangement characteristics (e.g., living in assisted living, lives alone, lives in metropolitan area, availability of LTSS).

NHATS provided imputed values for missing income by applying multiple imputation techniques and income quartiles are based on all reported income for all NHATS respondents in a given survey year (DeMatteis et al., 2016). Chronic conditions were assessed by self-report and included ever having: a heart attack, stroke, cancer, hip fracture, heart disease, high blood pressure, arthritis, osteoporosis, diabetes, lung disease, or anxiety. Depression was determined based on positive responses to Patient Health Questionaire-2 screening. Social isolation was defined as at least four of: being unmarried, not visiting in person with friends or family, not attending religious services, not participating in clubs, classes, or other organized activities, not talking to friends about important things, and not talking to family about important things (Pohl et al., 2017). We counted the number of self-care tasks (eating, getting out of bed, getting around inside, showering, toileting, dressing) for which help was received in the last month. We also counted the number of household tasks (laundry, shopping, meal preparation, medication management, bills and banking) for which help was received in the last month for health or functioning reasons. Following previous literature, we used respondent self-report and facility staff member interview responses to define those living in assisted living facilities (Freedman & Spillman, 2014). Availability of LTSS was determined using the 2020 American Association of Retired Persons (AARP) Long-term Services and Supports Scorecard, which creates state-level rankings of LTSS system performance using 26 indicators across five dimensions (i.e., Affordability and Access, Choice of Setting and Provider, Quality of Life and Quality for Care, Support for Family Caregivers, and Effective Transitions; AARP); because access to paid care may be less in states with weaker LTSS systems, those respondents living in a state in the lowest LTSS system performance quartile were noted to have low LTSS availability.

Analysis

After describing paid, family, and total caregiving hours over the study period, we then used group-based multiple trajectory modeling techniques to simultaneously model trajectories of paid caregiving support and family caregiving support (Nagin et al., 2018). This approach is an extension of univariate group-based trajectory modeling (GBTM), which is widely used in clinical research to identify latent longitudinal clusters of individuals who have a similar developmental trajectory based on a single indicator of an outcome of interest (Nagin, 2014). While GBTM would allow for identification of separate trajectories of paid caregiving support and family caregiving support over time, group-based multiple trajectory modeling considers the interrelationship between both trajectories including trajectories where no paid care is present.

In order to determine the optimal number of paid and family caregiving trajectories within the population, we compared information criteria obtained from models with various number of trajectories, including Akaike information criterion, Bayesian information criterion (BIC), and log likelihood. The model fit indices for 2, 3, and 4 groups are reported in Table 1. We then used visual examination to determine the order of equations (linear, quadratic, or higher order polynomials), which reflected the shape of each trajectory. Finally, we considered model interpretability such as a meaningful label for each group, adequate sample size of a class, and sufficient probability of the class membership (Nagin et al., 2018). We selected the three-group model with a linear equation, which summarized the distinctive pattern of the data with the least negative number of observation-based BIC and subject-based BIC. This model also was clinically interpretable and distinguished the large group with no or minimal paid care.

Table 1.

Model Fit Indices for Group-Based Trajectory Models With 2, 3, and 4 Groups

Number of groups	BIC (observation)	BIC (unique ID)	AIC	Log likelihood
2	−11,117.22	−11,104.21	−11,083.25	−11,072.25
3	−11,043.43	−11,024.51	−10,994.02	−10,978.02
4	−11,052.69	−11,027.86	−10,987.84	−10,966.84

Notes: AIC = Akaike information criterion; BIC = Bayesian information criterion.

We used chi-square and Student’s t test to examine pairwise differences in baseline characteristics among each of the trajectory groups identified as described above. We then used the generalized estimating equation (GEE) modeling for multinomial responses to quantify the relative impact of the factors associated with trajectory membership, while accounting for the correlation of time-dependent variables (Nguena Nguefack et al., 2020). Based on results of the bivariate analysis as well as factors known to be significantly associated with receipt of paid care in the existing literature, variables included in the GEE models were age, help with self-care activities, female gender, White non-Hispanic race, lives alone, depression, and receipt of Medicaid.

In order to test the robustness of our findings, we conducted two sensitivity analyses that examined care trajectories and factors associated with trajectory membership using GBTM and GEE. The first analysis examined a sample consisting of only those who entered the main sample because they died (“death only” group; n = 289). Because services provided by assisted living facilities can vary widely depending on the type of facility and state and local regulations (Aldridge et al., 2021; Freedman & Spillman, 2014) and may in some instances provide a level of care similar to nursing homes, the second analysis examined a sample where assisted living facilities were considered residential care settings; this approach defined community-dwelling as only those living in private homes and included those who died, moved to a nursing home, or moved to an assisted living facility in the sample (“assisted living as high care needs” group; n = 280).

Results

The 334 persons living with dementia included in our sample had an average of 5.5 (SD 1.5) interviews included in analysis. Of these, 289 had died and 45 moved to a nursing home; those who moved to a nursing home were significantly younger than those who died (mean age 81.3 vs 84.1, p < .01) but no other statistically significant differences between groups existed (Supplementary Table 1). Half the sample (174 individuals) had no paid care at any point during follow-up. At the time of the interview prior to death or moving to a nursing home, respondents received on average 70.8 total hours of care per week (58.5 hr of family care and 12.3 hr of paid care). Three years prior to that, respondents received on average 52.7 total hours of care per week (48.3 hr of family care and 4.4 hr of paid care). Six years prior to that, respondents received on average 45.7 total hours of care per week (43.3 hr of family care and 2.3 hr of paid care; Figure 1).

Figure 1.

Average total, paid, and family caregiving hours per week for community-dwelling persons with dementia in the years prior to death or nursing home placement. Paid and family caregiving hours are separately capped at 168 hr per week (24 hr × 7 days). Total caregiving hours are the sum of capped paid and family caregiving hours.

Final groups depicting trajectories of paid and family caregiving hours over time are shown in Figure 2. Group 1 (61.7% of respondents) had minimal paid care hours (average weekly hours ranged from 0 to 3.2) and moderate family caregiving hours (average weekly hours ranged from 23.2 to 31.5); both paid and family caregiving hours remained relatively stable over the study period (Group 1 = “low/stable care”). Group 2 (15.8% of respondents) had moderate paid care hours (average weekly hours ranged 16.1 to 60.2) and moderate family caregiving hours (average weekly hours ranged 39.9 to 60.3); both paid and family caregiving hours increased over the study period (Group 2 = “increasing paid care”). Group 3 (22.4% of respondents) had low paid care hours (average weekly hours ranged 1.6 to 6.4) and high family caregiving hours (average weekly hours ranged 82.0 to 133.9); while paid caregiving hours remained stable, family caregiving hours increased over the study period (Group 3 = “high family care”).

Figure 2.

Trajectories in paid and family caregiving among community-dwelling persons with dementia in the years prior to death or nursing home placement.^aaPaid and family caregiving hours are separately capped at 168 hr per week (24 hr × 7 days).

Table 2 describes characteristics of the three trajectory groups at baseline and presents pairwise comparisons to evaluate differences between the groups. The “low/stable care” group was the largest group. With a mean age of 83.5 years at baseline, 60.1% were female, 59.6% were White non-Hispanic, and 28.1% had Medicaid. About a third (38.5%) lived alone and 25.0% had depression. As compared to the other groups, those in “low/stable care” received help with fewer self-care tasks (0.8 compared to 2.3 in “increasing paid care” and 2.4 in “high family care,” p < .01 for both comparisons) and household activities (1.5 compared to 3.7 in the “increasing paid care” group and 3.4 in the “high family care” group, p < .01 for both comparisons). In addition, the “increasing paid care” was older (86.3 vs 83.5 years in the “increasing paid care” and “low/stable care” groups, respectively, p = .01) and more likely to be female (77.6% vs 60.1% in the “increasing paid care” and “low/stable care” groups, respectively, p = .02) than the “low/stable care” group.

Table 2.

Baseline Characteristics of Community-Dwelling Persons With Dementia by Group Membership

Variables	Full population	Low/stable care group	Increasing paid care group	High family care group	1 vs 2 p Value	1 vs 3 p Value	2 vs 3 p Value
N respondents	334	208	49	77
Sociodemographic characteristics
Age, mean (SD)	83.7 (6.9)	83.5 (7.0)	86.3 (6.4)	82.8 (6.5)	.01	.47	<.01
Female, %	61.1	60.1	77.6	53.3	.02	.30	.01
Married, %	32.9	30.8	24.5	44.2	.39	.03	.03
≥ High school education, %	59.0	60.3	64.6	52.0	.58	.21	.17
White non-Hispanic	55.1	59.6	57.1	41.6	.75	.01	.09
Income quartile					.69	.86	.66
Bottom quartile	48.7	47.8	54.4	47.8
Second quartile	26.5	28.0	13.9	30.2
Third quartile	15.4	16.6	14.9	12.9
Top quartile	9.4	7.7	16.9	9.2
Medicaid, %	31.3	28.1	37.0	36.1	.24	.21	.93
Clinical and functional characteristics
Proxy interview, %	30.5	21.6	34.7	52.0	.05	<.01	.06
Chronic conditions, mean (SD)	3.3 (2.0)	3.1 (2.0)	2.8 (1.7)	3.9 (2.1)	.28	.01	<.01
Depression, %	27.8	25.0	NRc	NR	.50	.01	.02
Socially isolated, %	50.6	46.6	53.1	59.7	.42	.05	.46
Help with self-care activities,a mean (SD)	1.4 (1.9)	0.8 (1.4)	2.3 (2.1)	2.4 (2.3)	<.01	<.01	.68
Help with household activities,b mean (SD)	2.3 (2.0)	1.5 (1.7)	3.7 (1.8)	3.4 (1.8)	<.01	<.01	.47
Geographic characteristics and living arrangements
Assisted living, %	6.6	8.2	NR	NR	.64	.09	.32
Lives alone, %	32.6	38.5	NR	NR	.97	<.01	<.01
Metropolitan area, %	80.5	80.8	NR	NR	.85	.96	.90
Low LTSSd availability, %	38.6	38.5	40.8	37.7	.76	.90	.72

Notes: SD = standard deviation.

^aSelf-care activities include eating, getting out of bed, showering, toileting, dressing.

^bHousehold activities include laundry, shopping, meal preparation, medication management, getting around outside, bills and banking.

^cNR = nonreportable due to sample size constraints.

^dLTSS = long-term services and supports.

The “high family care” group differed from the other two groups in several ways. Overall, respondents in the “high family care” group were less likely to be White non-Hispanic (41.6% vs 59.6% in the “low/stable care” group and 57.1% in the “increasing paid care” group, p = .01 and p = .09, respectively). Those in the “high family care” group were more likely to be married (44.2% vs 30.8% in the “low/stable care” group and 24.5% in the “increasing paid care” group, p = .03 for both comparisons), less likely to live alone (percentages nonreportable based on cell size reporting restrictions but p < .01 as compared both to the “low/stable care” and the “increasing paid care”), and more likely to have interview responses by proxy (52.0% vs 21.6% in the “low/stable care group and 34.7% in the “increasing paid care” group, p < .01 and p = .06, respectively). They were more likely to have a higher number of comorbidities (3.9 vs 3.1 the “low/stable care” group and 2.8 in the “increasing paid care” group, p = .01 and p < .01, respectively) and more likely to report depressive symptoms (percentages nonreportable based on cell size reporting restrictions but p = .01 as compared to the “low/stable care” and p = .02 as compared to the “increasing paid care” group) and social isolation (59.7% vs 46.6% in the “low/stable care” group and 53.1% in the “increasing paid care” group, p = .05 and p = .46, respectively). While they received help with more self-care activities than those in the “low/stable care” group (2.4 vs 0.8 in the “low/stable care” group, p < .01), means were similar between the “increasing paid care” group and the “high family care” group. Of note, there were no statistically significant differences between groups in several key variables including income, Medicaid, and low LTSS availability.

Table 3 presents the relative impact of various factors associated with membership in the trajectory groups with the “low/stable care” group as the referent. As compared to the “low/stable care” group, help with more self-care activities was associated with greater odds of being in the “increasing paid care” and the “high family care” groups (odds ratio [OR] 1.39, 95% confidence interval [CI] 1.23–1.58, p < .01 and OR 1.49, 95% CI 1.35–1.64, p < .01, respectively). Those who were depressed were 40% less likely to be in the “increasing paid care” group as compared to the “low/stable care” group (OR 0.58, 95% CI 0.35–0.95, p = .03). Those who lived alone were more than 80% less likely to be in the “high family care” group as compared to the “low/stable care” group (OR 0.22, 95% CI 0.12–0.41, p < .01).

Table 3.

Multinomial Logistic Regression Comparing Family and Paid Care Group Membership

	Increasing paid care vs low/stable care (ref.)			High family care vs low/stable care (ref.)
Variables	Odds ratio	95% CI	p Value	Odds ratio	95% CI	p Value
Age	1.04	1–1.09	.08	0.99	0.95–1.03	.54
Help with self-care activitiesa	1.39	1.23–1.58	<.01	1.49	1.35–1.64	<.01
Female	2.01	0.93–4.36	.08	0.61	0.33–1.1	.10
White non-Hispanic	0.85	0.41–1.76	.66	0.58	0.30–1.11	.10
Lives alone	0.89	0.47–1.69	.72	0.22	0.12–0.41	<.01
Depressed	0.58	0.35–0.95	.03	1.15	0.79–1.69	.46
Medicaid	1.41	0.74–2.69	.29	1.18	0.67–2.09	.56

Notes: CI = confidence interval.

^aSelf-care activities include eating, getting out of bed, showering, toileting, dressing.

Sensitivity analyses among the “death only” group (n = 289) and the “assisted living as high care needs” group (n = 280) revealed similar group trajectory patterns and factors associated with group membership as compared to the primary analysis (Supplementary Figures 1 and 2; Supplementary Tables 2 and 3).

Discussion

We identified three trajectories of paid and family caregiving support among persons living in the community with dementia with high care needs: “low/stable care,” “increasing paid care,” and “high family care.” Our study suggests that assistance with a greater number of self-care activities does not completely explain differences in paid and family care trajectories over time. In particular, those in the “high family care group” received help with a similar number of self-care activities as compared to the “increasing paid care” group but had more medical comorbidities and increased reports of depression and social isolation. The identification of factors associated with three distinct caregiving trajectories (“low/stable care,” “increasing paid care,” and “high family care”) provides important insights into the lived experience of persons living with dementia in the community over time.

First, the three distinct trajectories of paid and family care underscore the diversity of care experiences of persons living in the community with dementia even when care needs are high. While examination of average care hours in both our study and others suggest slowly increasing care receipt among individuals with dementia over time (Jutkowitz et al., 2020; Reckrey, Bollens-Lund, Husain, et al., 2020), our findings suggest that for the majority of persons living in the community with dementia, care received remained relatively steady prior to death or nursing home placement. Other work examining trajectories of caregiving support provided by spouses over time has also revealed diverse trajectories of family caregiving support (e.g., slowly increasing care, sharply increasing care, decreasing care; Taylor et al., 2008).

Some of this variability may be due to the fact that the trajectory of functional decline in dementia is heterogeneous; while a smaller proportion of persons living with dementia experience precipitous declines in function before death, many individuals experience slow functional and cognitive decline (Haaksma et al., 2019; Melis et al., 2019). However, it may also reflect the fact that many persons living with dementia report high levels of medical comorbidities and that their trajectories of care received are not determined by dementia status alone. In our study, this was most notable in the “high family care” group, who had both the highest numbers of medical comorbidities and also received the highest total care hours. This supports the need for person-centered dementia care that includes not only pharmacologic and nonpharmacologic treatment of dementia itself, but also careful attention to coexisting medical comorbidities and cultivation of a therapeutic community of caregivers and professionals to providing support (Fazio et al., 2018).

Yet it is important to note that care received by persons living with dementia is not determined solely by care needs; unmet care needs and related adverse consequences are common among older adults with functional impairment in general and those with dementia in particular (Beach et al., 2020; Reckrey, Bollens-Lund, et al., 2020). Trajectories of care like those in the “low/stable care” group (which comprised over 60% of the total sample) may either reflect stable care needs or an inability of current systems of support (i.e., paid and family caregivers) to meet growing care needs. Because our sample consists of individuals who died or moved to a nursing home, future work should explore whether or not needs are being met in this high care needs population and how the needs and experiences of this population differ from those who do not require as much care.

It is likely that receipt of paid caregiving support in particular is limited not only by care need or individual and family preference, but also by access to paid care. In general, receipt of paid care among persons living with dementia was uncommon regardless of total care hours; average weekly paid care hours were between 0 and 3.2 in the “low/steady care” group and between 1.6 and 6.4 hr in the “high family care” group. This is consistent with other work that found that while quantity of family care is associated with dementia severity, quantity of paid care is not (Hajek et al., 2016). The fact that paid care in the home is often intermittent rather than a continuous source of support may contribute to this (Cloutier et al., 2019). While paid care in the home matters for the health and well-being of both persons living with dementia and their family caregivers, low levels of paid care may not be enough to contribute meaningfully to outcomes (Reckrey, Boerner, et al. 2021). Current and anticipated shortages of paid caregivers themselves will only exacerbate challenges in access to paid care (Spetz et al., 2019).

It is important to consider what factors beyond care needs drive receipt of paid care in the home. The lack of a cohesive system of long-term care in the United States contributes to highly variable access to paid care and both local and state policies and family finances contribute to what care options for care look like (Kaye, 2014; Kelly et al., 2013; Spetz et al., 2019). In order to evaluate the impact of policy and finances on receipt of paid care, we examined variables related to income, Medicaid, and state LTSS availability in our analysis, but surprisingly these variables were not significantly different between trajectory groups. This may be due at least in part to our inability to capture the nuances of these factors. For example, having Medicaid may be associated with more paid care for someone living in a state such as New Mexico (where 73.5% of Medicaid and state-funded LTSS spending goes to home and community-based services), but having Medicaid may be associated with less paid care for someone living in a state such as Kentucky (where 13.5% of Medicaid and state-funded LTSS spending goes to home and community-based services; AARP).

It is also important to consider questions of equity in access to paid care and the implications of paid care for families. Those in the “high family care group” were more likely than others to be non-White, married, receive more help with self-care activities, and have more medical comorbidities, depression, and social isolation. Evidence suggests that the family caregivers providing high levels of care to persons living with dementia like those in this group are at high risk for caregiver burden and other negative outcomes related to caregiving (Kasper et al., 2015; Riffin et al., 2019). Paid caregiving has the potential to alleviate at least some of this burden (Reckrey, Boerner, et al. 2021), yet paid care receipt in this group was minimal. Given ongoing dialogue about the costs of dementia care in general and medical treatments for dementia in particular (Crosson et al., 2021), greater attention to the value of providing direct care for persons living with dementia is needed.

This study has several limitations. Due to our sample size, we were unable to examine race and ethnicity in a more nuanced way. Further, because we are limited to variables collected via the NHATS survey, we are unable to explore how individual and family preference contribute to use of community-based long-term care in general and paid care in the home in particular. Finally, while limiting the sample to those with high care needs (i.e., those who died or moved to a nursing home) was necessary to meaningfully examine patterns of paid and family care receipt, it may limit the generalizability of our findings to the dementia population at large, particularly those with less care needs. Yet the results of our analyses are highly relevant as preferences for in-home care result in more persons with dementia remaining in the community with advancing disease and high care needs.

It is also important to note that while we analyzed data from respondents only following dementia identification, we cannot determine the severity of dementia. The fact that those in the “high family care” group were more likely to have proxy-reported NHATS responses raises the possibility that this group had more advanced dementia; the finding that living with others was significantly associated with membership in the “high family care” group (as compared to the “low/stable care” group) in multivariable models raises the possibility that measures of help received with self-care activities may not capture the nuances of impairment in dementia. Future work should include more nuanced measures of cognitive function and dementia-related impairment in order to help us better understand the care needs of the growing population of persons living with dementia, especially as care needs advance.

In conclusion, the significant variability in trajectories of paid and family care among persons living in the community with dementia in the years before death or nursing home placement highlights the importance of considering each family’s individual arrangements for dementia care. Care needs play only one part in determining care arrangements and creating equitable access to paid care may be a particularly important way to support family caregivers. This will help ensure that the growing number of persons living with dementia have the support needed to remain living safely in the community as care needs grow.

Funding

Funding was provided by the National Institute on Aging (U01AG032947, P30AG012846, K23AG066930, and R01AG060967). This paper was published as part of a supplement sponsored by the University of Michigan and the Johns Hopkins Bloomberg School of Public Health with support from the National Institute on Aging (U01AG032947 and P30AG012846).

Conflict of Interest

None declared.

Author Contributions

J. M. Reckrey planned the study, supervised data analysis, and wrote and revised the manuscript. L. Li planned the study, conducted and supervised data analysis, and revised the manuscript. S. Zhan and C. Yee conducted data analysis and revised the manuscript. K. A. Ornstein planned the study, supervised data analysis, and revised the manuscript. J. Wolff revised the manuscript.