Out-of-Pocket Health Care Spending at Older Ages: Do Caregiving Arrangements Matter?

Abstract

Identifying the correlates of out-of-pocket (OOP) health care spending is an important step for ensuring the financial security of older adults. Whether or not someone has a family member providing assistance is one such factor that could be associated with OOP spending. If family caregivers facilitate better health, health care spending could be reduced. On the other hand, costs would be higher if family members facilitate more (or more costly) care for loved ones. This paper explores the relationship between caregiving arrangements and OOP spending using data from 5045 individuals in the 2000–2016 Health and Retirement Study with Medicare coverage and caregiving needs. We do not find a relationship between family caregiving and OOP health care costs, overall. However, among those with Medicare HMO insurance, having a family caregiver is associated with more spending than having no helper. This is mainly due to differences in spending on prescription medications.

Keywords

activities of daily living caregiving health services

What this paper adds

• Out-of-pocket (OOP) health care spending at older ages is common, but there is limited work on the characteristics of the subpopulations at greatest risk of OOP spending.

• Caregiving arrangements are one potential correlate of OOP health care spending that has not yet been examined in this context.

Applications of study findings

• The impact of caregiving on OOP health care spending has implications for patients, their families, and policymakers.

• Quantifying health care costs of both caregivers and care recipients over the caregiving lifecycle will be important as policymakers consider how to best support older adults with care needs.

Introduction

Despite near universal Medicare coverage among adults 65 and over, health care spending at older ages is common. The median older household devotes 12% of its budget to health care and health insurance (Gruber & Levy, 2009). While there is a large range in medical spending among adults ages 65 and over, individuals in the top 5% of the distribution are responsible for half the total OOP costs, spending close to $98,000 per year (De Nardi et al., 2016). In addition, adults with functional limitations—those typically in need of care—face higher than average out-of-pocket (OOP) health care expenditures (Van der Heyden et al., 2015). While Medicare covers hospital stays, doctor visits, and, more recently, prescription drugs, older adults incur OOP health care costs because there can be substantial copays. Copays may be covered through private insurance, Medicare Health Maintenance Organization (HMO) plans, Medicaid, or other public insurance. Nearly 20% of medical spending is paid for OOP (De Nardi et al., 2016).

While there is a fair amount of work quantifying OOP spending in later life, less attention has been paid to the factors that put people at greatest risk of experiencing OOP costs. One factor that has been identified is whether someone has supplemental insurance, which is associated with lower OOP spending (Levy, 2020; Narang & Nicholas, 2017), but with higher insurance premium payments (Levy, 2020). Among patients newly diagnosed with cancer with some type of supplemental insurance, those with private insurance incurred the lowest out-of-pocket costs and those with Medicare HMO supplementary insurance the highest (Narang & Nicholas, 2017). Other factors associated with higher OOP spending include being older, female, more highly educated, having higher relative wealth, and poorer health (Narang & Nicholas, 2017; Yabroff et al., 2019). This paper builds on this body of work by exploring another potential correlate of OOP health care spending: caregiving arrangements. We argue that health care costs, including those paid OOP, could be associated with unpaid care provided by family and friends, whom provide the overwhelming majority of long-term care (Friedman et al., 2015; Harris-Kojetin et al., 2013), although the directionality is not clear. As we show in the conceptual framework depicted in Figure 1, we expect caregiving arrangements to be related to health behaviors, outcomes, and health care utilization which, in turn, would affect out-of-pocket spending. We also hypothesize that, consistent with other work (Narang & Nicholas, 2017), type of supplemental insurance, if any, will be associated with OOP spending. In our models, we explore the relationship between caregiving arrangements and OOP costs overall and by supplemental insurance status.

Figure 1.

Conceptual framework of the relationship between caregiving and out-of-pocket health care spending, Medicare-eligible population with care needs.

One way family caregivers could reduce health care costs is by reducing rates of hospitalization or other health care utilization, which could occur if family caregivers improve health outcomes, as shown in other work (Coe et al., 2019). On the other hand, having family caregivers could lead to higher costs if family members facilitate more (or more costly) health care utilization for loved ones. Paid, professional caregivers may also reduce health care costs, as home care provided by trained staff could improve health relative to care from untrained family and friends. On the other hand, family and friends are closer to the care recipient and may be more diligent about their care or may be more aware of changes in behaviors that could suggest new health care needs. While very little work examines this, one paper (Tobica, et al., 2015) finds that a 10% increase in paid caregiving costs results in a 1.2% reduction in total health costs.

A recent scoping review (Friedman et al., 2019) identified only three studies that explicitly examine assistance from a caregiver (not merely access to one) and its relationship to health care costs, but findings are inconsistent regarding whether receiving family care increases, decreases, or makes no difference for expenditures (Coe et al., 2019; Torbica et al., 2015; Van Houtven & Norton, 2008). For instance, one study uses Medicare claims data matched to a subsample of the HRS, and an instrumental variable approach to examine the effect of assistance from children on the health care costs of parents, and finds that receipt of care from children (compared to not getting care from children) is associated with decreased costs to Medicare (Van Houtven & Norton, 2008). This work does not examine helpers other than children. Another uses data on 532 stroke patients in Italy and shows increased health care costs after a stroke among those with a family caregiver (Torbica et al., 2015), but this work is limited to a particular sample and disease conditions. The third paper uses Medicaid claims data and finds that assistance from family caregivers improves health and decreases health care utilization, but the difference to overall costs for Medicaid is not statistically significant (Coe et al., 2019).

We use data on community-dwelling adults with functional limitations from the 2000 to 2016 waves of the Health and Retirement Study (HRS) to take a first look at the relationship between caregiving arrangements (i.e., receipt of help from different combinations of types of care providers) and OOP health care costs. The goal of this paper is to descriptively explore whether and which caregiving arrangements might be associated with health care spending. Because spending on long-term care is highly correlated with caregiving type, our focus is on health care spending only, and we do not include spending on nursing homes or the costs of hiring paid caregivers. We expand on existing work in several key ways. First, among older adults with disabilities, we examine the relationship between caregiving arrangements and OOP health care costs. Out-of-pocket health costs have not been examined in this context in prior work. Second, we go beyond family care to look at the impact of care from several combinations of caregiving arrangements, including receiving help from: family/friends only, paid caregivers only, both family and paid caregivers, or receiving no help at all. The current literature on this topic compares family help to no help at all, but does not consider how care from family compares to care from direct care workers. Finally, we also explore interactions between caregiving arrangements and supplemental insurance types and identify the types of health care services responsible for differences in OOP costs.

We explore three research questions:

(1) Are caregiving arrangements associated with OOP health care spending among older adults with care needs?

(2) How does this relationship compare for individuals with different types of supplemental insurance?

(3) How does this relationship compare for different types of health care expenditures?

Methods

Data

Data were drawn from the 2000 to 2016 waves of the HRS, a multi-cohort longitudinal biennial survey of a nationally representative sample of community-dwelling adults ages 51 and over and their spouses. The study was first launched in 1992 and is refreshed periodically to remain representative of the U.S. population over age 50. Data have been collected every 2 years since 1998. We pooled data across all available years and limited our analytic sample to waves in which respondents were 65 and older and reported having insurance coverage through Medicare. This allowed us to focus on a more homogeneous group of older adults with health care insurance. We also exclude anyone residing in a nursing home or other facility that provides health services to residents. To ensure that we were capturing individuals with care needs, we further limited the sample to respondents reporting difficulty with at least one Activity of Daily Living (ADL). We lagged our outcomes by one wave (2 years) to allow for better estimates of directionality; that is, we wanted to be sure we were capturing spending at a later point in time than caregiving and supplemental insurance. All covariates are captured at the wave prior to the OOP spending outcomes.

We began with a sample of 83,800 person-wave observations representing 19,515 community-dwelling respondents age 65 and over and excluded 67,639 person-waves in which individuals do not have an ADL and/or are not on Medicare. In addition, because we capture information on spending from the wave following reporting of caregiving information, we also excluded 5938 person-waves for which we did not have two consecutive waves of information (this includes the exclusion of the last observed record for each individual). Finally, we excluded an additional 317 person-waves with missing covariates. This gave us a total sample size of 9906 person-wave observations representing 5045 unique individuals aged 65 and over, with Medicare coverage, who reside in the community at every time point, have at least one ADL, and have observations for two waves of data or more.

Measures

Independent Variable: Caregiving Arrangements

The HRS data include rich information on in-home caregiving for assistance with ADLs (eg eating, toileting, dressing, bathing, and walking across a room) and instrumental ADLs (IADLs; eg preparing meals, grocery shopping, making phone calls, taking medications, managing money). These data sources ask respondents who report limitations in ADLs or IADLs whether anyone helps and, if so, the relationship of the caregiver, the amount of care, and whether care was paid for. We used this information to assess whether the respondent has caregivers of different types. We divided care into non-familial paid care (referred to as paid care) and care from family and unpaid friends (referred to as family care or family/friend care). Paid care is defined as all care from helpers with the following relationship categories: “paid helper,” “professional,” “employee of institution,” and “organization.” We also categorized as paid caregivers those caregivers who are coded as “other individual” and are also reported as paid. We did not count relatives of any type in this category. Family care includes help from any family member (e.g., spouse, child, sibling, etc.) and/or unpaid friends. Our key independent variable is a four category mutually exclusive measure of care coded as: (1) no help; (2) help from family and friends only; (3) help from paid helpers only; and (4) help from both paid and family helpers.

Dependent Variable: Out-of-Pocket Spending

We use the RAND-HRS version of the OOP data, as those data are cleaned and streamlined for consistency across waves and also imputed for missing data. Health-related OOP spending is asked every wave and respondents are asked about spending over the last 2 years/since the last wave. Because respondents are asked about the time period since the last wave, to ensure we have the appropriate ordering of events, we draw this information one wave following reports of caregiving arrangements. We combine five types of health care expenses reported in the HRS: hospital, physician, dentist, outpatient surgery, prescription medications, and spending on health insurance premiums for a total measure of OOP health care costs. Spending on long-term care would be highly correlated with caregiving arrangements; thus, we do not include long-term care spending (eg spending on nursing home, home health, etc.) in our measures of health care spending. All OOP cost measures are in constant 2016 dollars. In line with other work (Levy, 2020; Narang & Nicholas, 2017) to get annual measures of costs, we divide OOP spending in half (because respondents are asked about spending between waves, or every 2 years).

Covariates

Other individual covariates used in our models include: age in years; years of schooling completed; race (White vs. Non-White); an indicator for currently married or partnered, male, total number of ADLs; total # of IADLs; whether respondent had heart attack or stroke past 2 years (since last wave); income tertiles; and wealth tertiles. Respondents were classified into 4 possible insurance categories, which included supplemental insurance arrangements, in any, for those enrolled in fee-for-service Medicare and a category for those in a Medicare HMO. These categories were (1) Public (Medicaid or VA plans); (2) Medicare HMO (Advantage plan), (3) Private (employer-paid or other private insurance), and (4) None (no supplemental coverage). This four category schema is consistent with other work (Narang & Nicholas, 2017) and combines Medicaid and VA supplemental insurance into one category of public insurance, and employer-sponsored and Medigap categories into private supplemental insurance.

Statistical Analysis

We used the Stata twopm command to estimate a two-stage model. The first regression used a probit model to examine whether an individual had any OOP medical expenditures in a given time period, and the second model (generalized gamma with log link) estimated the dollar costs for individuals who had any costs. Models used survey weights and adjusted for clustering and sample design. In all models, we also controlled for sociodemographic, health, and supplemental insurance measures.

After running the models, we obtained predicted means using the Stata Margins command, with covariates held at their mean values (means include zeros) and report mean costs for each caregiving arrangement and for fully interacted models that interact covariates with supplemental insurance type. Corresponding standard errors are computed automatically by Stata and use the Delta method. To examine whether there are significant differences among the caregiving arrangement types for the models interacted by supplemental insurance, we run two sets of posthoc tests after model estimation. First, we use a Wald test to examine whether there are any differences by caregiving arrangements for each category of supplement insurance. For categories where there are statistically significant differences, we also run pairwise tests comparing each type of caregiving arrangement to identify which groups are significantly different from one another.

Results

Sample Characteristics

Table 1 shows the descriptive characteristics of the analytic sample of community-dwelling adults 65+, with Medicare coverage and at least one ADL limitation. This sample of older adults with reported functional limitations is more than 60% female, almost three quarters white, 46% report being married or partnered, and respondents are, on average, almost 77 years old. This sample has a little over 11 years of schooling, on average. The sample respondents report an average of 1.8 ADLS (range: 1–5, note that respondents must have at least one ADL to be included) and about one IADL (range: 0–5). About half the sample has had a heart attack or stroke in the past 2 years. About 22% of the sample are on Medicaid or other public insurance (e.g. VA), 23% are insured through Medicare HMO/Advantage plans, 37% have private insurance and 18% have no supplemental insurance.

Table 1.

Characteristics of Sample, Community-Dwelling Adults ages 65 and Over, with Medicare Coverage, and at least one ADL Limitation, HRS 2000–2016.

	Proportion or Mean (sd)
Sociodemographic characteristics
Male	.371
White	.742
Married or partnered	.462
Age in years (mean, sd)	76.67 (7.811)
Education in years (mean, sd)	11.397 (3.691)
Health and functioning
Total # ADL limitations (mean, sd)	1.821 (1.131)
Total # IADL limitations (mean, sd)	1.046 (1.400)
Heart attack or stroke past 2 years	.520
Income (in approximate tertiles)
First (≤15k)	.321
Second (15k–32k)	.321
Third (>32k)	.358
Assets (in approximate tertiles)
First (≤2.2k)	.303
Second (2.2k–69k)	.332
Third (>69k)	.365
Supplemental insurance
Public	.224
Medicare HMO	.229
Private	.369
None	.178
Caregiving arrangements
No help	.451
Family/Friend help only	.434
Paid help only	.032
Both paid and family/friend help	.082
Annual out-of-pocket spending
Any health care utilization	.992
Any out-of-pocket spending	.878
Mean spending	$3236 ($9918)
Median spending	$1781
Spending for 90th percentile	$7032

Notes: Results weighted using individual weights and adjusted for survey design. ADL = activity of daily living; HRS = health and retirement study. Data on out-of-pocket spending are in constant 2016 dollars. N = 9906 observations; 5045 unique individuals.

There is fair amount of variation in caregiving arrangements. Nearly 45% do not report having any helpers; just over 43% report that a family or friend helps with their functional limitations; just over 3% report having only a paid helper; and almost 9% report paid and family helpers. On average, annual OOP costs for this high need population is about $3236, with a median of $1781, and 90th percentile of $7032. Nearly 90% of the sample had some OOP costs.

Model Results

Table 2 shows the predicted mean OOP health care spending overall, by caregiving arrangements, and for caregiving arrangements by supplemental insurance type, as estimated from two-stage models. The underlying models are fully adjusted, controlling for age, race, gender, years of schooling, number of ADLs, number of IADLs, whether heart condition/stroke, whether married, whether/type of supplemental insurance, and income and wealth categories. This table shows that, overall, predicted annual mean OOP spending with all covariates held at their average values is $2931. When broken down by caregiving arrangement, we see some variation by caregiving type, but these differences are not statistically significant. Respondents who reported having paid helpers are estimated to spend the most OOP, $3232. They are followed by people receiving help from family/friends ($3011), those with both family/friend and paid help ($2952), and those with no help ($2824) spent the least, on average. Appendix Table A1 in the Online Supplement shows the detailed results of both of the two stages of the model. Of note, and as depicted in Table A1, those with paid help or both paid and family help are less likely than those with no help to incur any out-of-pocket costs, but we do not see significant differences among the caregiving groups for total costs.

Table 2.

Predicted Mean Annual Out-of-pocket Spending Overall, by Caregiving Arrangement, and by Caregiving Arrangement and Supplemental Insurance Type, HRS 2000–2016 Among Community-Dwelling Adults Ages 65+ with Medicare Coverage, and At Least One ADL Limitation.

			Obs	Mean OOP Spending ($s)	Standard Error ($s)
Overall			9906	$2931	$96
By caregiving arrangements only		No help	4316	2824	133
		Family/Friend only	4472	3011	146
		Paid only	293	3232	391
		Both	825	2952	257
By caregiving arrangements and supplemental insurance status	Public	No help	800	1392	126
		Family/Friend only	1091	1623	139
		Paid only	143	1265	311
		Both	370	1691	320
	Medicare HMO	No help	1089	2826	176
		Family/Friend only	1006	3848^a	430
		Paid only	31	5016	1490
		Both	108	4459^a	798
	Private	No help	1695	4099	383
		Family/Friend only	1472	3961	273
		Paid only	72	5852	1190
		Both	199	3917	493
	None	No help	732	3341	281
		Family/Friend only	903	3094	184
		Paid only	47	3340	634
		Both	148	2537	390

Notes: Predicted means are based on results of two-stage models where stage 1 uses a probit model for any spending and stage-two uses a GLM model with log link to predict spending amount with covariates held at their mean values (means include zeros). Models control for age, race, gender, years of schooling, number of ADLs, number of IADLs, heart condition/stroke, whether married, type of supplemental insurance, income tertiles, and wealth tertiles. Out-of-pocket spending is in constant 2016 dollars. Results weighted using individual weights and adjusted for survey design. OOP = out-of-pocket health care spending; ADL = activity of daily living. N = 9906 observations; 5045 unique individuals.

Significantly different from no help group at p < 0.05.

Where we see significant differences by caregiving arrangements is once we interact all model covariates with supplemental insurance type in both parts of the model. From these fully interacted models, we see that among respondents with Medicare HMO supplemental insurance, having a family caregiver is associated with mean spending of $3848, which is significantly higher than OOP spending for those with no helpers ($2826). Older adults with both paid and family helpers are also paying significantly more than those with no help ($4459). People with paid helpers are paying even more in OOP costs: $5016, but this does not reach statistical significance, likely due to the small sample sizes for this group.

Among those with HMO supplemental insurance, we also examined which of the six health care and insurance costs are associated with caregiving arrangements. Predicted probabilities drawn from two-stage models that disaggregate the different types of health care costs are depicted in Table 3. As this table shows, prescription drugs are the highest cost across all groups, and are significantly higher for those with family helpers relative to those with no help. Among those with Medicare HMO and no help, average annual OOP spending on prescription drugs is $1193, but OOP spending among those with family help is much larger than those for the total population, $2211. This results in a significant difference between respondents with no help relative to those with family help, of more than $1000 per year. We also see a small but significant difference in outpatient spending between family/friend help and those with no help relative to those with paid help. Appendix Table A2 of the Online Supplement shows the results for all insurance-by-caregiving arrangements categories. While only HMO supplemental insurance showed significant differences by caregiving arrangements for total out-of-pocket costs, when costs are parsed into their component parts, other differences by caregiving group start to emerge. Which groups differ from one another and the directionality of the differences are conditional on supplemental insurance status as well as the type of cost incurred.

Table 3.

Predicted Means and Standard Errors for Annual Out-of-Pocket Spending by Caregiving Arrangement and Type of Expenditure, HRS 2000–2016 Community-Dwelling Adults Ages 65+ with Medicare Coverage, At Least One ADL Limitation, and Enrolled in Medicare HMO Insurance Plan.

	Obs	Drugs	Insurance	Dentist	Hospital	Doctor	Outpatient
No help	1089	1193^a (102.7)	799.6 (62.60)	275.2 (30.6)	249.8 (47.8)	218.6 (23.2)	24.0^b (4.3)
Family/Friend only	1006	2211^c (322.1)	734.0 (67.09)	251.5 (22.8)	232.5 (40.2)	233.4 (21.0)	34.1^b (7.7)
Paid only	31	1851 (549.7)	679.9 (223.1)	471.7 (270.8)	963.6 (612.1)	273.4 (157.9)	3.4^a,c (2.9)
Both	108	1688 (387.5)	1237 (313.7)	234.0 (52.5)	616.6 (479.4)	516.6 (263.1)	32.4 (17.7)

Notes: Standard errors in parentheses. Predicted means are based on results of two-stage models where stage 1 uses a probit model for any spending and stage-two uses a GLM model with log link to predict spending amount with covariates held at their mean values (means include zeros). Models control for age, race, gender, years of schooling, number of ADLs, number of IADLs, heart condition/stroke, whether married, income tertiles, and wealth tertiles. Out-of-pocket spending is in constant 2016 dollars. Results weighted using individual weights and adjusted for survey design. OOP = out-of-pocket health care spending; ADL = activity of daily living.

^aSignificantly different from family and friends only at p < 0.05.

^bSignificantly different from paid only at p < 0.05.

^cSignificantly different from no help group at p < 0.05.

Sensitivity Analysis

Because the number of times we observe individuals vary in the data (ie ranges from 2 to 7 waves of observation), the independence assumption for costs over time implicit in our models may be violated. To examine whether this may bias the results, we reran the analyses with only the first two waves of observation for each individual. Results from this analysis are consistent with those reported here.

Discussion

Out-of-pocket health care costs are associated with cost-related medication non-adherence (Briesacher et al., 2007; Osterberg & Blaschke, 2005) and adverse health outcomes, including premature death (Heisler et al., 2010; Ramsey et al., 2016). While there is already work quantifying total OOP costs among older adults, there has been less focus on the subset of the population at greatest risk of incurring such costs. Whether someone has any supplemental insurance, type of insurance coverage, and demographic factors have been found to be key correlates in other work (Levy, 2020; Narang & Nicholas, 2017; Yabroff et al., 2019). We extend this body of work by examining another potentially relevant and, until now, unexplored factor: caregiving arrangements.

In our sample of community-dwelling adults ages 65 and over with Medicare coverage and at least one ADL, we find about $3000 on average being spent OOP on health care and health insurance per year. We also find a fair amount of variation in caregiving arrangements. Nearly 45% of this population do not report any helper; just over 43% report that a family or friend helps with their functional limitations; about 3% report having only a paid helper; and almost 9% report both paid and family helpers. This variation enables us to compare OOP costs among these groups.

Consistent with other work showing that assistance from family caregivers is not statistically significantly associated with cost savings to Medicaid (Coe et al., 2019), we do not find significant differences by caregiving arrangements in individual OOP health care spending. We do, however, see notable differences by caregiving arrangement once we explore interactions by supplemental insurance type. While we do not see differences among those without any supplemental insurance—the group at risk of highest out-of-pocket costs—we see differences among older adults with Medicare HMO insurance. For this group, having a family caregiver is associated with mean spending of $3848, which is nearly $1000 more per year than what we see among those with no helpers, a statistically significant difference. This difference is mostly due to differences in spending on prescription drugs. Having both family and paid care is associated with even greater out-of-pocket costs and is significantly higher than for those with no help. While only HMO supplemental insurance showed significant differences by caregiving arrangements for total OOP costs, differences emerge for other supplemental insurance categories when costs are parsed into their component parts, suggesting that this relationship may depend on both the type of supplemental insurance and specific health care costs incurred.

Higher OOP spending may be desirable if costs arise because family caregivers help older adult receive necessary care. Lower cost, on the other hand, could be related to lack of access to necessary or preventative healthcare. To fully interpret the impact of costs will ultimately require that OOP costs be examined in the context overall Medicare and Medicaid expenditures for these individuals. For example, if the costs we observe are because caregivers facilitate better care, higher OOP costs for prescription medications may be offset in the longer term by reduced formal care expenditures. For instance, the availability of family caregivers may assure that appropriate medications are acquired and taken as prescribed (Fernandez-Lazaro et al., 2019), thereby reducing the probability of emergency room visits and hospitalizations. This work also has implications for the health and health care costs of individuals who need help with at least one ADL, but who are not receiving assistance from family or paid caregivers. This group spends less OOP, which may suggest that they are saving money in the short term, but could have more significant health events requiring hospitalization and even greater costs in the longer run if those costs were necessary for meeting care needs.

This work also has implications for policymakers as they plan for the future of long-term care in the United States. Only three other papers to our knowledge examine the relationship between receipt of care and health care costs for the care recipient, but they do not explicitly explore OOP costs, and results are mixed (Coe et al., 2019; Torbica et al., 2015; Van Houtven & Norton, 2008). A report by the National Academies of Sciences, Engineering, and Medicine acknowledges that strategies to effectively support and engage family caregivers could be costly. But, suggest that “. . . some portion of new investments will be offset by savings—from reductions in use of nursing home, home health, emergency room, and inpatient hospital care” (Schulz & Eden, 2016, p. 258). This is a first step to thinking about the impact of caregiving on health care costs. More work is needed to get a more comprehensive estimate of health care costs across payers, types of costs, and over the longer term. In addition, because caregiving could come with health care needs for caregivers, future work could assess health care utilization for both care recipient and caregiver. A dyadic approach may be needed for assessing the full costs of care across dyads; other work already suggests that the needs of care recipients are linked to caregiver wellbeing (Beach & Schulz, 2017).

It is important to consider these findings in the context of Medicare policy. Several policy changes occurred during the 16-year time period of the study that could affect the interpretation of findings. For example, there were changes in Medicare Part D cost sharing over the study period, with Medicare prescription drug plans covering more of the cost of prescription drugs starting in 2006. Thus, differences in out-of-pocket prescription payments may be less relevant in recent years. Another notable change is the increase in payments to Medicare HMO plans, which also provide more coverage of various services and prescription drugs and may serve to reduce differences between groups in out-of-pocket spending. Our findings suggest that policies to reduce out-of-pocket costs through greater Medicare spending could potentially reduce differences in spending among individuals with different caregiving arrangements. While supplementing Medicare with other types of insurance can do this as well, they come at a financial cost to older adults.

Limitations

As with every study, this work has several limitations that need to be acknowledged. First, we examine associations between care and OOP costs, but we cannot be certain of the causal directionality of the findings. This work is a descriptive first look at this topic. We capture spending after care arrangements are established which helps with directionality, but this strategy does not fully address issues related to causation. The main novelty of this work is to advance the idea that caregiving arrangements may be another correlate of out-of-pocket spending that has been absent from these models and may have implications for identifying the benefits to different types of care arrangements. Causal inference is an important next step, but is not without its own limitations. For instance, it is generally restricted to a much narrower range of variation in the processes under study than the broader population of interest.

In addition, while care arrangement could cause differences in spending, people in greater need of care may be more likely to have caregivers and have more spending. We attempt to address this by reducing the heterogeneity of the sample used. We limit the analytic sample to individuals aged 65+ with Medicare and with self-reported care needs. We also control for a variety of covariates including the functional status of the respondent and presence of chronic health conditions.

Relatedly, people with adverse health outcomes may be more likely to purchase supplemental insurance. Bias may also occur over time if people who receive care are healthier, as they may be overrepresented in the data. Excluding people without two consecutive waves to capture causal directionality may exacerbate this as sicker people may be more likely to be excluded if they are more likely to exit the sample due to non-response or death. These individuals may also be more likely to receive caregiver support and have higher OOP spending.

Another limitation is that we examine health care OOP costs, but do not include long-term care costs, such as OOP payments for home health care. Most long-term care is paid OOP and an estimate of that would increase costs for those with paid helpers, but this introduces other analytic complexities. Future work could examine a combined measure of spending on long-term care with health care at the household level to capture overall household cost savings from having an unpaid caregiver. In addition, the sample size for people receiving only paid care is relatively small. This may limit our ability to capture significant differences in our interacted models and when we disaggregate costs by type.

We do not have the data to examine OOP costs paid for by caregivers or other family and friends for a respondent's care, yet a recent AARP report suggests that half of caregivers use their own money for a care recipient's household expenses. Medical costs (i.e., health care costs, in-home care, and medical equipment) accounted for 17% of this spending (Skufca & Rainville, 2021). For this paper, we use a relatively simple version of caregiving that focuses on whether care is received from different combinations of types of caregivers (family, formal, both, neither). Of course, caregiving arrangements are much more complex and include people of different relationships to the care recipient, different size caregiving networks, and different intensities of care overall and across caregiver types.

Conclusion

This work is a first step to quantify the extent to which caregiving arrangements mitigate and exacerbate health care costs of older adults with care needs. Additional work is needed for a more comprehensive assessment of health care costs that include OOP health care costs, long-term care spending, as well as those covered by other public and private payers, and the costs to both caregivers and care recipients over the caregiving cycle. Future work is also needed to explore the causal links between caregiving arrangements and health care costs. Only with this information can the full costs of care associated with caregiving be assessed. This will be important as new policies are considered to protect the financial stability of and ensure adequate care for the population of older adults in the US.

Supplemental Material

Supplemental Material - Out-of-Pocket Health Care Spending at Older Ages: Do Caregiving Arrangements Matter?

Supplemental Material for Out-of-Pocket Health Care Spending at Older Ages: Do Caregiving Arrangements Matter? by Esther M. Friedman, Scott R. Beach, and Richard Schulz in Journal of Applied Gerontology

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The contents of this publication were developed under a grant from the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR grant number 90 RTGE0002). NIDILRR is a Center within the Administration for Community Living (ACL), Department of Health and Human Services (HHS). The contents of this paper do not necessarily represent the policy of NIDILRR, ACL, or HHS, and you should not assume endorsement by the Federal Government.

ORCID iD

Esther M. Friedman

Supplemental Material

Supplemental material for this article is available online

References

Beach

S. R.

Schulz

(2017). Family caregiver factors associated with unmet needs for care of older adults. Journal of the American Geriatrics Society, 65(3), 560–566. https://doi.org/10.1111/jgs.14547

Briesacher

B. A.

Gurwitz

J. H.

Soumerai

S. B.

(2007). Patients at-risk for cost-related medication nonadherence: A review of the literature. Journal of General Internal Medicine, 22(6), 864–871. https://doi.org/10.1007/s11606-007-0180-x

Coe

N. B.

Guo

Konetzka

R. T.

Van Houtven

C. H.

(2019). What is the marginal benefit of payment-induced family care? Impact on Medicaid spending and health of care recipients. Health Economics, 28(5), 678–692. https://doi.org/10.1002/hec.3873

De Nardi

French

Jones

J. B.

McCauley

(2016). Medical spending of the US elderly. Fiscal Studies, 37(3–4), 717–747. https://doi.org/10.1111/j.1475-5890.2016.12106

Fernandez-Lazaro

C. I.

Adams

D. P.

Fernandez-Lazaro

Garcia-González

J. M.

Caballero-Garcia

Miron-Canelo

J. A.

(2019). Medication adherence and barriers among low-income, uninsured patients with multiple chronic conditions. Research in Social and Administrative Pharmacy, 15(6), 744–753. https://doi.org/10.1016/j.sapharm.2018.09.006

Friedman

E. M.

Rodakowski

Schulz

Beach

S. R.

Martsolf

G. R.

James

A. E.

III (2019). Do family caregivers offset healthcare costs for older adults? A mapping review on the costs of care for older adults with versus without caregivers. The Gerontologist, 59(5), Article e535–e551. https://doi.org/10.1093/geront/gny182

Friedman

E. M.

Shih

R. A.

Langa

K. M.

Hurd

M. D.

(2015). US prevalence and predictors of informal caregiving for dementia. Health Affairs, 34(10), 1637–1641. https://doi.org/10.1377/hlthaff.2015.0510

Gruber

Levy

(2009). The evolution of medical spending risk. Journal of Economic Perspectives, 23(4), 25–48. https://doi.org/10.1257/jep.23.4.25

Harris-Kojetin

L. D.

Sengupta

Park-Lee

Valverde

(2013). Long-term care services in the United States: 2013 overview.

10.

Heisler

Choi

Rosen

A. B.

Vijan

Kabeto

Langa

K. M.

Piette

J. D.

(2010). Hospitalizations and deaths among adults with cardiovascular disease who underuse medications because of cost: A longitudinal analysis. Medical Care, 48(2), 87–94. https://doi.org/10.1097/mlr.0b013e3181c12e53

11.

Levy

(2020). The risk of high out-of-pocket health spending among older Americans. University of Michigan Retirement and Disability Research Center (MRDRC) Working Paper.

12.

Narang

A. K.

Nicholas

L. H.

(2017). Out-of-pocket spending and financial burden among Medicare beneficiaries with cancer. JAMA Oncology, 3(6), 757–765. https://doi.org/10.1001/jamaoncol.2016.4865

13.

Osterberg

Blaschke

(2005). Adherence to medication. New England Journal of Medicine, 353(5), 487–497. https://doi.org/10.1056/nejmra050100

14.

Ramsey

S. D.

Bansal

Fedorenko

C. R.

Blough

D. K.

Overstreet

K. A.

Shankaran

Newcomb

(2016). Financial insolvency as a risk factor for early mortality among patients with cancer. Journal of Clinical Oncology, 34(9), 980–986. https://doi.org/10.1200/jco.2015.64.6620

15.

Schulz

Eden

(Eds.). (2016). Families Caring for an Aging America (p. 258). Washington, DC: National Academies Press.

16.

Skufca

Rainville

(2021). Caregiving out-of-pocket costs study 2021. AARP Research. https://doi.org/10.26419/res.00473.001

17.

Torbica

Calciolari

Fattore

(2015). Does informal care impact utilization of healthcare services? Evidence from a longitudinal study of stroke patients. Social Science and Medicine, 124, 29–38. https://doi.org/10.1016/j.socscimed.2014.11.005

18.

Van der Heyden

Van Oyen

Berger

De Bacquer

Van Herck

(2015). Activity limitations predict health care expenditures in the general population in Belgium. BMC Public Health, 15(1), 1–12. https://doi.org/10.1186/s12889-015-1607-7

19.

Van Houtven

C. H.

Norton

E. C.

(2008). Informal care and Medicare expenditures: Testing for heterogeneous treatment effects. Journal of Health Economics, 27(1), 134–156. https://doi.org/10.1016/j.jhealeco.2007.03.002

20.

Yabroff

K. R.

Zhao

Han

Zheng

(2019). Prevalence and correlates of medical financial hardship in the USA. Journal of General Internal Medicine, 34(8), 1494–1502. https://doi.org/10.1007/s11606-019-05002-w

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.10 MB