Implications of Rebalancing Home-Based Care for Family Caregivers to Older Adults

Abstract

Over the last several decades, Medicaid has been “rebalancing” services from institutions to the community, increasing support of home- and community- based services (HCBS). These services could potentially substitute for care typically provided by family members, replacing or reducing care from kin. Leveraging one of the most recent Medicaid rebalancing programs, the Balancing Incentive Program (BIP), and using data from the 2008–2016 Health and Retirement Study on 953 Medicaid beneficiaries ages 65 and over with at least one functional limitation, we examined the relationship between exposure to BIP and family and professional caregiving. BIP exposure was not associated with receipt of care or total hours of care. It was, however, associated with more hours of professional care and fewer hours of family care from non-spouse kin. These findings suggest that recent Medicaid rebalancing programs, while intended to meet the desires of older adults, could also have implications for their families.

BACKGROUND

Family and friends provide 80% of long-term care for older adults with disabling conditions (Friedman et al, 2015). Lack of access to professional care services in the U.S. may partly explain why so much of this care is shouldered by family (Bookman & Kimbrel, 2011). With regard to formal long-term care, Medicaid, the primary public payer of long-term care, has made significant progress over the last couple of decades moving services from institutional to community settings, when appropriate. This “rebalancing” of long-term care from institutions to the community, has resulted in a faster growth in Medicaid spending on home- and community-based services (HCBS) compared with institutional long-term care. Medicaid HCBS expenditures rose from 43 percent of total long-term services and supports (LTSS) in FY 2008 to 58.6 percent in FY 2019 (Murray et al., 2019). In addition, most U.S. states saw an increase in their home care workforce with less growth in their nursing home workforce size (Friedman et al,., 2021). While policies to shift support from institutional to community-based care are being implemented to meet the desires of older adults to receive care in their homes (Binette, 2021), they might have implications for their families, in particular the family members who would otherwise be called upon to provide care.

Consequences of and reasons for family care.

When older adults need assistance with everyday tasks or personal care, they either turn to family and friends, hereafter, family care, or seek assistance from hired caregivers such as home health aides or other paid helpers, hereafter professional care, or some combination thereof. It is important to understand whether policy changes to increase professional care would reduce family care, as caregiving comes at a high cost to families. For one thing, there are significant economic costs of care, such as health care or other out of pocket costs incurred by families (Schulz, 2016, or through changes in work and work hours of caregivers (Martsolf et al, 2020; Fahle and McGarry, 2022). Financial distress due to caregiving may be lasting (Reinhard, Feinberg, Houser, 2019; Schulz, 2016). There is also evidence that caregiving responsibilities place family members at increased risk of adverse physical and mental health outcomes (Bevan and Sternberg, 2012; DePasquale et al., 2017; Greene et al. 2017; Perkins, et al. 2013; Schulz and Sherwood 2008). Caregivers also commonly engage in unhealthy behaviors and neglect self-care (Schulz, et al. 2020).

Despite these potential consequences, family care remains the dominant form of care in later life. There are many reasons for this. Some are financial or related to availability of services. First, the costs of obtaining professional care are significant, and publicly-financed services through Medicaid are limited due to Medicaid’s eligibility requirements. Second, workforce shortages make it difficult to find and retain professional caregivers (Hartoonian, et al, 2023). Third, older adults prefer to age at home or in the community (Kasper, Wolff, Skeehan, 2019; Wolff, Kasper, Shore, 2008), in particular when they have conditions other than dementia. Thus, even people who are eligible for Medicaid-funded long-term care may not seek these services in states that predominantly support institutional care and have less funding for HCBS.

There are also reasons for providing care to loved ones that are emotional and less sensitive to financing or the long-term care marketplace. For instance, caregivers report getting meaning in life from caregiving tasks (Polenick, Kales, Birditt, 2018), and caregiving might be perceived as a family obligation (Aires et al, 2019). The complex array of reasons for providing care make it difficult to know whether increasing access to professional HCBS would alter care provided by families.

Is professional care a substitute or complement for care from family and friends?

The growth in HCBS might have positive implications for family members if paid care from hired caregivers such as home health aides or other paid helpers substitutes for care from family care. If professional caregivers take on caregiving responsibilities that would otherwise fall on family and friends, this increased access to professional care could alleviate the care burden placed on families. On the other hand, professional care might complement care provided by family caregivers and not necessarily replace it. In this complementary role for family caregivers, we would not expect to see an association between increased access to professional care on care provided by family and friends. The evidence for professional care as a substitute or complement to family care is mixed in existing studies in the U.S. (Blackburn et al., 2018; Bonsang, 2009; Gaugler et al., 2004; Golberstein et al., 2009), as well as those in China, Europe, and Canada (Liu, 2021; Stabile at. al 2006; Viitanen, 2007). Two recent working papers use the Health and Retirement Study (HRS) and leverage variation in HCBS spending across U.S. states using quasi-experimental approaches, with mixed results. One such study (Zai, 2020) finds that an increase in Medicaid aging waiver funding increases the likelihood that a respondent becomes a caregiver, suggesting that paid and family care may be complements. There are also gender differences in these effects: only female caregivers provide fewer hours of personal care. A second study showed that the policy’s main effect was to replace care from family members with professional care, suggest a substitution effect of care (Shen, 2021). The current study likewise leverages HRS data, and uses the most recent Medicaid rebalancing program to explore the relationship between access to HCBS and caregiving assistance.

Variation in who cares and how much they help.

Spouses are often the first family members to assist an older adult in need of care, although adult children and their partners are also frequently involved in providing for older parents, especially when they are widowed, or suffer from certain disease conditions such as dementia (Friedman et al, 2015; Spitze and Logan, 1990). Other family and friends also step up when usual caregivers are not available; for instance, siblings are more likely to provide care for a sibling without children (Moyer, 2019). Intensity of help provided varies by caregiver relationship as well (Friedman et al, 2015). The specific family relationship could, therefore, have implications for whether and how much professional help might alter the support behaviors and family and friends. For instance, there is evidence that the substitution effect of professional for family care may be more pronounced for spouses and daughters than other relationships (Shen, 2021). The current study therefore pays special attention to different types of caregivers.

Shifting care from institutions to HCBS.

We build on prior work by examining one of the most recent Medicaid rebalancing programs, the Balancing Incentive Program (BIP), to capture access to HCBS. We focus specifically on this program, as it is the most recent Medicaid rebalancing program designed to increase access to HCBS. This allows us to explore whether massive efforts to rebalance care from institutions to home and community settings could potentially relieve family caregivers, or if it requires their supplemental help to manage complementary tasks to that of HCBS providers, or to serve as facilitators for in-home or community-based care. BIP provided enhanced federal matching funds to eligible states to improve their provisions of HCBS. Eligibility for the Federal Medical Assistance Percentage (FMAP) required a state to spend less than 50% of its total Medicaid LTSS expenditures on HCBS for federal fiscal year 2009. Participating states agreed to make several structural changes in their LTSS systems through the development of a “no wrong door” system, conflict-free case management services, and core standardized assessment instruments. In addition, a specific balancing benchmark, based on a state’s baseline ratio of HCBS to institutional LTSS expenditures, was required by the end of the program. States that spent between 25% and 49% of total LTSS expenditures on HCBS services in 2009 needed to achieve a balancing benchmark of at least 50% by the end of the 2015 fiscal year. States that met this benchmark received a 2% enhanced FMAP on all HCBS expenditures. States that spent less than 25% of total LTSS expenditures on HCBS in 2009 were eligible for a larger enhanced FMAP of 5%, if they met a balancing benchmark of 25% for HCBS relative to LTSS. Of the 38 states in the U.S. eligible for BIP, 21 states applied, and 18 states ultimately participated in the program (Lester et al., 2015). States received approval between March 2012 and July 2014.

We leverage BIP as a source of state-by-year variation in Medicaid long-term care policy to examine whether exposure to BIP (and its associated HCBS benefits) is related to the receipt of care and hours of care older adults receive from both professional caregivers and family caregivers. If help from HCBS providers such as home health and personal care aides is a substitute for family care, BIP should increase care from HCBS providers while reducing care from family. If, on the other hand, professional and family care are complements, shifting funds to more home- and community-based care may not reduce family care. We also explore variation among different types of family members.

RESEARCH QUESTIONS

In this paper, we use data from the 2008 – 2016 waves of the HRS to capture state-by-year variation in BIP exposure and examine the following research questions:

1. Is BIP exposure associated with the likelihood of receiving professional home-based care and/or assistance from family and friends?

2. How do these associations compare for different family members who provide assistance (i.e. spouse, child, other relative/friend)?

3. Is BIP exposure associated with the number of hours of care received overall, from the professional workforce, and from different family members?

DATA AND METHODS

Data.

Data come from the 2008 – 2016 waves of the HRS, a multi-cohort longitudinal biennial survey of a nationally representative sample of older community-dwelling adults ages 51 and over and their spouses. The study was first launched in 1992 and is refreshed periodically to retain representativeness of the population over age 50. Data have been collected biennially since 1998. Although the initial sample included only community-dwelling individuals, participants who enter a nursing home continue to be tracked and surveyed with separate survey weights available for the institutionalized population. The study was approved by RAND IRB.

Our analyses start in 2008 to include data from a few years prior to initial BIP implementation and end in 2016. Our analytic sample is limited to respondents residing in states eligible to implement BIP. To focus on older adults eligible for BIP, we also limit the sample to: community-dwelling respondents 65 years of age or older who are Medicare eligible (majority of HRS respondents are 65+ and Medicare eligible), Medicaid eligible beneficiaries, and individuals reporting difficulty with at least one ADL, as this is a requirement to be eligible for Medicaid HCBS through BIP.

Among the 38 states that were eligible to apply for BIP due to spetablending less than 50% on HCBS compared to total LTSS, 21 states were approved to participate in the program. Three states (Indiana, Louisiana, and Nebraska) withdrew early; therefore, these three are excluded from analyses. We also excluded Mississippi due to its lower spending benchmark of 25%, resulting in 17 “BIP states.” Another 17 states were eligible but did not participate in BIP, which we refer to as “non-BIP” states. The start date for BIP exposure in each state was the date of BIP approval by CMS. Each wave of data was coded as having BIP exposure if the interview date was after the approval date for the state the respondent lived in at that wave. If the BIP approval date was in the first six months of the fiscal year (FY), we treated that FY as the initial year of BIP exposure. If the BIP approval date was in the second six months of the FY, we treated the following FY as the initial year of BIP exposure. Table 1 lists BIP-eligible states along with approval dates. We began with a sample of 36,947 person-wave observations for community-dwelling respondents age 65 and over and exclude 35,390 person-waves of observations due to eligibility criteria. They either did not reside in a BIP-eligible state, were not dual eligible, or did not have at least one ADL. We excluded another 8 observations due to missing data on key covariates. Our final analytic sample is 1,549 person-waves from 953 unique individuals.

Table 1:

BIP Approval Dates by State

State	Approval Date
Arkansas	3/15/2013
Connecticut	12/7/2012
Georgia	6/13/2012
Illinois	6/12/2013
Iowa	6/13/2012
Kentucky	3/15/2014
Maine	6/12/2013
Maryland	3/20/2012
Massachusetts	3/19/2014
Missouri	6/13/2012
Nevada	3/15/2014
New Hampshire	3/1/2012
New Jersey	3/15/2013
New York	3/15/2013
Ohio	6/12/2013
Pennsylvania	7/01/2014
Texas	9/4/2012

MEASURES

Family Caregiver Information:

The HRS data include rich information on in-home caregiving for assistance with activities of daily living (ADLs; e.g., eating, toileting, dressing, bathing, walking across a room) and instrumental ADLs (IADLs; e.g., preparing meals, grocery shopping, making phone calls, taking medications, managing money). Respondents who report limitations in ADLs or IADLs were asked whether anyone helps and, if so, the relationship of the caregiver, the hours of care, and whether care was paid. We use this information to assess the relationships of caregivers to respondents and hours of care provided by each caregiver.

We divide care into non-familial professional care (professional care) and care from family and friends (family care). Within the category of family care, we further categorize types of family relationships. We define professional as care from unrelated caregivers who provide paid services, which includes care from helpers with the following relationship categories: “paid helper”, “professional”, “employee of institution”, and “organization”. The HRS also has a category for “other individual”. In our definition of professional care, we include care from caregivers who are coded as “other individual” if they are paid. We do not count relatives of any type in this category. Spouse caregivers include both spouse and domestic partners. Care from children includes care from respondent’s biological children, stepchildren, and the spouses and partners of biological and step-children. Other relatives and friends includes all other relations (grandchildren, siblings, and other relatives) as well as those whose reported relationship is “other individual” but are not paid.

The HRS asks respondents about both the relationship of caregivers and the hours of care each caregiver provided during the past month. We aggregate these hours by relationship category to determine amount of care provided by each type of caregiver. Consistent with other work (Friedman et al, 2015), we constrain individual caregivers to a maximum of 16 hours per day to account for time spent sleeping. Total care is the sum of hours of care across all categories. Family care is the sum of all categories except professional care.

Other Measures:

Exposure to BIP. Our main predictor is exposure to BIP. To capture this, we use a time-varying binary indicator for whether BIP was implemented or not in the respondent’s state of residence prior to the respondent’s interview date for that wave. ADL Severity. HRS asks a series of questions regarding difficulties performing six Activities of Daily Living (eating, getting up from bed or a chair, toileting, dressing, bathing or showering, and walking across a room). We use these to create a three-category measure of severity of activity limitation. We categorize respondents who report no difficulty with eating, toileting, dressing, or bathing but do report difficulties with getting into or out of bed/chairs and or walking as having mild limitation. Respondents who report difficulty with dressing, bathing, getting in or out of bed/chairs, or walking but no difficulty with eating or toileting are categorized as having moderate limitation. Those who reported difficulty eating or toileting, or who reported difficulties with all ADLs are categorized as having severe or complete limitation (Stineman et al., 2014). Other covariates. Characteristics used in our models for covariate adjustment include: indicators for year of survey; age (categories in 5 year age bands); sex, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Other); indicator for currently married/cohabiting; education indicators (less than high school, high school graduate, some college, college graduate or more); indicator for urban residence (defined as Rural-Urban Continuum Codes of 1, 2, or 3).

METHODS

This paper examines the relationship between BIP implementation and caregiving, including whether help is received, who helps, and hours of help conditional on receiving any help. Because we are interested both in whether any help is received and hours of help, conditional on receiving help, we begin with several probit models predicting the likelihood of receiving help and from whom help is received and then turn to two-stage regression models for hours of help. The two-stage model addresses the fact that not everyone has help of each type and allows us to first estimate whether care was received and then hours of care, conditional on getting any care.

Analysis of who helps.

We begin by using a series of probit models to examine three dichotomous variables capturing type of help: (1) any type of help; (2) help from family/friends; (3) help from professionals. We also examine three dichotomous measures of who specifically helps, among those receiving help from family and friends; namely, whether help is provided by: (1) spouse; (2) children; (3) other relatives and friends. In addition to our time-varying measure of BIP implementation, these models also include indicators for year of survey, gender, race, whether respondent is currently married, education levels, age in five-year categories, whether respondent lives in a metropolitan area, and three categories of ADL severity. For the second set of models, we further limit the analytic sample in two ways: to a subsample of currently married individuals and a subsample of those with at least one living child. All models include survey weights. Standard errors are adjusted for clustering to account for the repeated nature of the data over time.

Analysis of hours of help.

To examine the relationship between BIP and how much care older adults receive from people with different relationships, we explore: total hours of care, total hours of professional care, and total hours of family care. We also break down care by relationship of caregiver and examine hours of care from spouse, hours of care from children, hours of all non-spouse family care, and hours of non-spouse and non-child family care. For each category of care, we first estimate a probit model predicting receipt of this type of care. We generate predicted values from the first-stage model by setting BIP=0 and then setting BIP=1 for every record (using actual or observed values for all other covariates). In the second stage, we estimated hours of care, conditional on receipt of care using a generalized linear model with log link. Using the second-stage model, we predict hours of care (for the entire sample) similarly by setting BIP=0 and then BIP=1 for all records. For each individual, we multiplied the first and second stage predictions from setting BIP=0 and then from setting BIP=1 to get predicted hours of care for each individual in the absence and in the presence of BIP. The difference between the two predictions is the effect of BIP.

We generate standard errors for these estimates with 1,000 bootstrap replications, which involved sampling (with replacement) individuals from the analytic sample. In each bootstrap replicate, all observations for a sampled individual were included. Both the first and second stage models control for all model covariate described above. Additionally, we repeated the full estimation procedure predicting hours of care by spouse for the subsample of respondents who are currently married, and from children for the subsample with living children.

Imputations.

Data on hours of care, for at least one type of care, were missing for about 20% of respondents in the HRS who reported receiving some type of care. We imputed missing hours for professional caregivers, spouse, children, other relatives and friends using chained multiple imputation models with the predictive mean matching method using the MI IMPUTE command in Stata (van Burren, et al., 2006). For each type of care, we imputed hours only for those respondents who reported receiving care from at least one caregiver of that type. Imputation models use information on age, gender, marital status, education, urban residence, number of children, any co-resident children, any children living within 10 miles of respondent, number of household residents, ADL severity, a count of difficulties reported with Instrumental Activities of Daily Living, an indicator for whether respondent ever smoked, and indicators for whether respondent has ever been diagnosed with diabetes, cancer (other than skin cancer), stroke, heart problems, high blood pressure, arthritis, or psychiatric illness.

RESULTS

Table 2 shows the descriptive characteristics of the analytic sample of dual-eligible beneficiaries with at least one ADL. Almost three quarter of the sample is female, 43% are White, 61% have less than a high school education, and 23% are married. Seventy–nine percent live in an urban area. This is a sample with significant functional limitations: 37% have moderate limitations and over half have extreme or complete functional limitations. About a third of the sample had exposure to BIP in at least one year. Seventy-two percent of respondents received some caregiving assistance.

Table 2.

Characteristics of analytic sample, community-dwelling Medicaid-eligible adults with at least one ADL limitation aged 65+, HRS 2008 – 2016

		% or mean	SD
BIP exposure (%)		33.60
Age Categories (%)	Age 65–69	27.27
	Age 70–74	21.28
	Age 75–79	16.87
	Age 80–84	14.95
	Age 85–90	10.03
	Age 90+	9.61
Gender (%)	Female	73.25
Race (%)	White	42.91
	Black	26.79
	Hispanic	26.59
	Other	3.71
Married (%)		23.44
Education categories (%)	Less than HS	60.96
	High school	26.29
	Some college	8.52
	College or more	4.23
Metro (%)	Rural	21.44
	Urban	78.56
Severity of ADLS (%)	Mild	10.99
	Moderate	37.24
	Extreme/complete	51.77
Receipt of help (%)	Any help	72.19
	Family friend help only	42.47
	Paid help only	6.46
	Both paid and family help	23.25
Family helper relationship (%)	Spouse	52.96
	Child	49.61
	Other family/friend	23.02
	Non-spouse unpaid	55.69
Mean hours of help per month	All help	198.6	(185.3)
	Paid	113.5	(129.6)
	Unpaid	166.8	(169.2)
	Spouse	148.4	(161.2)
	Child	141.4	(171.0)
	Other	118.0	(151.4)

Notes: Weighted using individual survey weights. ADL=Activities of Daily Living; HRS=Health and Retirement Study; SD=standard deviation; BIP=Balancing Incentive Program. Number of observations: 1,549; Number of unique individuals: 953.

There is tremendous variation in the helper arrangements for this sample. Forty-two percent received help only from a family or friend, about 6% received only professional help and 23% received both professional and family help. Family help was typically received from a spouse (53%) and/or a child (50%). Twenty-three percent of the sample were receiving help from other family and friends. These categories are not mutually exclusive. Help from other family and friends could be in addition to help from a spouse and/or child. On average, respondents received almost 200 hours of help per month. That is between 40–50 hours per week, the equivalent of a full-time job. Family and friends provided 167 hours of help per month, with spouses providing close to 150 hours per month, children provided just over 140 hours per month and other family and friends provided an average of 118 hours.

We begin our analyses by exploring the relationship between BIP and helper arrangements and, among those receiving help from family and friends, who helps. Table 3 displays the results of probit models of the association between helper arrangement and BIP exposure. We see no significant association between BIP exposure and care receipt. The second panel of Table 3 shows the association between BIP and several types of family caregiver relationships. Once again, we do not see any statistically significant associations.

Table 3.

Marginal effects from separate probit models predicting the relationship between BIP and help arrangements, community-dwelling Medicaid-eligible adults with at least one ADL aged 65+, HRS 2008 – 2016

		N	BIP	Non- BIP	Difference
Type of Help	Any help	1,549	0.750	0.708	0.042
	Professional	1,549	0.309	0.291	0.018
	Family/friend	1,549	0.678	0.647	0.031
Family Caregiver Relationship	Spouse, conditional on married	384	0.590	0.486	0.104
	Non-spouse	1,549	0.578	0.548	0.030
	Child, conditional on having a living child	1,400	0.520	0.485	0.035
	Other relative/friend	1,549	0.218	0.237	−0.019

Notes: Estimates weighted using individual survey weights and adjusted for individual-level clustering to account for multiple observations per person. Models also control for calendar year, 5-year age groups, gender, race, currently married, education, rural/urban residence, and ADL severity index. ADL= Activities of Daily Living; HRS= Health and Retirement Study. BIP= Balancing Incentive Program.

We next turn to the results of the two-stage predictions of hours of care. Table 4 shows the predicted mean hours of care from these models for BIP and non-BIP, along with p-values for significant group differences. There are three key findings from these analyses. First, overall hours of help, when aggregated across professional and family care, is not significantly associated with BIP exposure. This suggests that BIP exposure is not associated with a change in the total hours of help received for this dual-eligible high need sample. Second, upon disaggregating professional care from family care, we show that exposure to BIP is associated with significantly more hours of professional help (43 hours per month for BIP relative to 25 hours per month for non-BIP). Moreover, we see a substitution effect for family care. BIP exposure is associated with fewer hours of help from family caregivers (86 hours) compared to subjects without BIP exposure (133 hours), although this difference is only marginally significant at p<0.07. Third, when we disaggregate family help into specific helper relationships, the biggest differences in mean hours of help by BIP exposure is for non-spouse family caregivers, with 108 hours of care per month received by older adults without exposure to BIP compared to 63 hours for those exposed to BIP. The hours of care from spouses is almost identical for BIP and Non-BIP groups. Most of the difference in hours of family care comes from non-spouse helpers providing less assistance after BIP exposure.

Table 4:

Predicted hours of help from two-stage models of the association between BIP and hours of help, community-dwelling Medicaid-eligible adults with at least one ADL limitation ages 65+, HRS 2008 – 2016

		N	BIP	Non-BIP	Diff	P-value
Total hours of help		1,549	134.96	158.04	−23.09	0.407
Type of help	Professional	1,549	43.49	25.19	18.31	0.035
	Family/friend	1,549	86.28	132.64	−46.36	0.073
Family caregiver relationship	Spouse, conditional on married	384	83.92	87.29	−3.37	0.921
	Non-Spouse	1,549	62.55	108.60	−46.05	0.035
	Child, conditional on having a living child	1,400	63.78	77.52	−13.74	0.269
	Other relative/friend	1,549	4.90	38.54	−33.63	0.122

Notes: Estimates weighted using individual survey weights and adjusted for individual-level clustering to account for multiple observations per person. Models also control for calendar year, 5-year age groups, gender, race, currently married, education, rural/urban residence, and ADL severity index. Probabilities are derived from separate models and may not sum to the same totals across models. ADL= Activities of Daily Living; HRS= Health and Retirement Study. BIP= Balancing Incentive Program.

DISCUSSION

With the aging of the U.S. population comes an increased need for long-term services and supports. This can place a significant burden on family members, who are the primary caregivers to older adults. Policies to move care to the home instead of institutional settings are being implemented to meet the desires of older adults, but may have positive benefits for family caregivers as well if they substitute for care typically provided by kin. This paper uses a recent rebalancing program as a source of state variation in Medicaid long-term care policy to evaluate the impact of increased access to home- and community-based care for dual-eligible beneficiaries with functional difficulties. Our findings suggest that BIP was not associated with changes in whether older adults receive care. Where we do see significant findings is in the hours of care received from family and professional caregivers. Among dual-eligible adults with care needs, exposure to BIP is associated with 18 more hours of professional care per month (25 hour per month for non-BIP compared to 43 hours per month among those with BIP exposure). On the flipside, family care is lower among those with BIP exposure, 86 hours compared to 133 for those without exposure to BIP. This difference is driven by non-spouse care.

The amount of care from a spouse remains constant with and without BIP exposure, suggesting that spouse care is less elastic than care from other kin. This also suggests that professional and family care are substitutes for non-spouse relationships, but not for spouses, whose care remains inelastic even after BIP implementation. It is worth noting that the total amount of care that dual-eligible beneficiaries receive does not differ significantly for those with versus without BIP exposure. This finding is consistent with other work that shows elasticities in the supply of home care from family (e.g. Bolin et al., 2008; van Houtven and Norton, 2004; Golberstein et al., 2009; Shen 2021). We expand on prior work by showing that different family members will respond differently to increased availability of professional care and by examining this within the context of the newest Medicaid rebalancing program.

Although we estimated associations and did not identify the causal impact of BIP, our findings provide suggestive evidence that the BIP program – intended to increase access to home-based care - could also have positive implications for beneficiaries’ families. As expected, BIP exposure is associated with more hours of professional care. But it is also associated with less care from family members other than spouses.

This work has implications for family caregivers, many of whom provide this assistance at potential expense to their own health and employment (Bauer & Sousa-Poza, 2015; Martsolf et al., 2020; Schulz & Sherwood, 2008). Care substitution by paid home care workers could result in positive outcomes for families, for instance by freeing up more time to participate in the labor market or by reducing the burden of care associated with adverse health outcomes (Schulz et al., 2020). On the other hand, it is not clear whether professional care in place of family care would have beneficial or deleterious effects to the care recipient’s health. These factors need further examination and could be explored together – by estimating the combined impact of programs increasing access to home-based services on caregiver and care recipient outcomes - to assess the total effects of recent Medicaid programs to expand support for home and community-based programs on families. Analyses of the cost-benefits of such programs should consider spillover effects onto other family members, even those who may not live with the care recipient.

Our finding that the association between BIP implementation and hours of care was not significant for spouses raises questions both about why this is the case and whether there are other potential interventions that might alleviate the caregiving burden on spouses. For instance, spousal caregivers might be more likely to cut back on their hours providing care if they feel comfortable around the professional caregiver, perhaps by having a regular or consistent caregiver. Or they might feel more confident in a professional caregivers abilities to provide care solo, if they completed a training or certification program. It is also possible that spousal care is more difficult to affect through such programs because spouses are providing care for emotional reasons. Spouses may find fulfillment and purpose in life through their provisions of care (Polenick, Kales, Birditt, 2018), feel a strong sense of obligation to provide care to loved ones (Aires et al, 2019) or worry that handing off care to a paid caregiver could hurt their relationship with their partner. While these topics are beyond the scope of this study, they could potentially be explored in future work. Finally, even if spouses do not cut back on care, that does not mean that they do not benefit from having a professional caregiver. Having someone with whom to share caregiving responsibilities may reduce the burden on spouses, even if hours of care are unchanged.

As with all studies, this work has several limitations. Most critically, this work examined associations, leaving us unable to identify a causal relationship between implementation of the BIP program and family care. More work is needed to examine the causal impact of BIP and similar rebalancing programs on caregiving. In addition, given that we are only looking at several years of data post-BIP, we can only speak to changes in care that occurred within the first few years after BIP was implemented (between 1–4 years later, depending on the state). Trends in who provides care may shift with time. Spouses for instance may take longer to cut back on care than others. Future work could examine longer- term impacts of BIP on family care. Finally, we focus this work on the 65 and over population, but BIP has implications for Medicaid-eligible individuals with disabilities of all ages. Although the HRS has data on the over 50 population, we limited analyses to the dually eligible (Medicare-Medicaid eligible sample) which gives us a more heterogeneous group for these analyses, and limits us to the 65 and over population (there are very few dual-eligible beneficiaries under age 65). The impact of BIP on younger adults could be examined using other datasets on younger individuals with disabilities.

Despite these limitations, this is an important contribution to the body of work examining the impacts of increased access to home- and community-based services on caregiving. We provide suggestive evidence that the association might depend on the specific family member. We also show that family caregivers may not always stop providing assistance, but may instead adjust their care hours in the presence of professional care. This work is important for understanding the implications of policies supporting home-based care for older adults and their families, which will become even more relevant as Baby Boomers age and new policies become crucial for ensuring the care needs of the growing U.S. population of older adult and their family caregivers are met.

What this paper adds.

• Medicaid has been rebalancing its portfolio of long-term care from institutional to home- and community- based services (HCBS), which could have implications for use of professional care and may replace or reduce caregiving from family and friends.

• Research to date on access to HCBS and family care have produced mixed results, and do not typically explore variation across different types of family members.

• We examine a recent Medicaid rebalancing program, the Balancing Incentive Program (BIP), and find that exposure to BIP was associated with more hours of professional care and fewer hours of family care, in particular from non-spouse kin.

Applications of study findings.

• This study provides suggestive evidence that Medicaid programs, like the Balancing Incentive Program, could affect how families allocate care.

• The study also highlights the importance of considering how programs affect different family members.

• With family caregivers prone to adverse outcomes due to caregiving, rebalancing programs may also have broader effects on the health and wellbeing of caregivers.