Informal caregivers and the health of older adult care-recipients

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations Fall 2016 Informal caregivers and the health of older adult care-recipients Andrew Joseph Potter University of Iowa Copyright 2016 Andrew Joseph Potter This dissertation is available at Iowa Research Online: Recommended Citation Potter, Andrew Joseph. "Informal caregivers and the health of older adult care-recipients." PhD (Doctor of Philosophy) thesis, University of Iowa, Follow this and additional works at: Part of the Health Services Administration Commons

2 INFORMAL CAREGIVERS AND THE HEALTH OF OLDER ADULT CARE- RECIPIENTS by Andrew Joseph Potter A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Health Services and Policy in the Graduate College of The University of Iowa December 2016 Thesis Supervisor: Professor Fredric D. Wolinsky

3 Copyright by Andrew Joseph Potter 2016 All Rights Reserved

4 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL PH.D. THESIS This is to certify that the Ph.D. thesis of Andrew Joseph Potter has been approved by the Examining Committee for the thesis requirement for the Doctor of Philosophy degree in Health Services and Policy at the December 2016 graduation. Thesis Committee: Fredric D. Wolinsky, Thesis Supervisor Kanika Arora Padmaja Ayyagari Keith J. Mueller Kathleen C. Buckwalter

5 To Elizabeth Potter, my partner in this endeavor and many others. ii

6 Acknowledgements This dissertation would not have been possible without the diligent involvement of my advisor, Fred Wolinsky. Dr. Wolinsky s attention to detail has made this a better work, and has made me a better researcher. During my years in this doctoral program, I have had the opportunity to learn from working with professors across the Department. My academic advisor, Dr. Padmaja Ayyagari has supported me in a multitude of ways over the past several years. I have also benefited from the leadership of Drs. Marcia Ward and George Wehby, who have gone to great lengths to ensure that doctoral students have opportunities to develop. I have been privileged to collaborate with Drs. Keith Mueller, Brad Wright, and Brian Kaskie on additional publications. Through their participation on my dissertation committee, Drs. Kathleen (Kitty) Buckwalter and Kanika Arora have broadened the way I think about the issues explored in this dissertation. During my time in the program, I have worked closely with staff of the Department of Health Management and Policy, including Torrie Malichky, Jean Sheeley, Betsy Winter, Diane Schaeffer, Kristi Yeggy, Megan Watson, Kim Merchant, and Fred Ullrich. Their ability to juggle so many different responsibilities while remaining responsive to students and faculty alike is truly amazing. I am indebted to recent graduates of the PhD program for their guidance, from the early years through the dissertation phase, especially Paula Weigel, Suzanne Bentler, and Michelle Lampmann. For their camaraderie during these last several years, my fellow PhD candidates Jure Baloh, Nabil Natafgi, Matt Nattinger, and Patience Ugwi also deserve special mention. iii

7 Finally, the support of my family has been critical to my success in the program. My wife, Elizabeth Potter, has shared with me the journeys of parenting, career, and Iowa, all at the same time. My parents, Phoebe and David Potter, have encouraged me to value truth and human dignity in every aspect of my life, and I credit these values with my circuitous route to my current career path. My in-laws, Barbara and Ronald Dorr, have enthusiastically provided help and unconditional support throughout the duration of our time in Iowa. I thank my children Cecilia, Ezra, and Malcolm for the full embrace of life that they bring to every moment. This work was supported by a Post Comprehensive Research Award from the University of Iowa Graduate College, and by funds from the John W. Colloton Chair. iv

8 Abstract Family and friends provide substantial daily assistance to older adults with disabilities, but little is known about how that caregiving impacts the health of older adults. Using survey data on a nationally-representative sample of older adults and their caregivers, matched with Medicare claims, I explored several aspects of the relationship between caregiving and the health of older adults receiving care. I relied largely on Andersen s behavioral model, which describes health services use as a product of predisposing, enabling, and need characteristics. I found that 33%-37% of older adults with unmet care needs at one point in time still have them one year later. I also found frequent change in the composition of older adults networks of caregivers. Having a formal caregiver was associated with lower odds of unmet care needs for women; conversely, men receiving care from a spouse had three times lower odds of unmet care needs than those receiving care from a non-spousal caregiver. Some caregivers use services such as respite care, training, and support groups, but some caregivers lack access to desired services. I found that caregivers reporting unmet service need were more likely to be Black and Hispanic, while service use was higher among caregivers providing help on a regular schedule. Care-recipient health and function was associated with both service use and unmet service need, but available measures of local caregiver service supply were not. Older adults frequently seek care in emergency departments (EDs), but this care may not always be necessary or desirable. I hypothesized, but did not find, that caregiver service use was associated with care-recipient ED use. Instead, chronic health conditions v

9 and other need factors were the strongest predictors of ED use. I found that carerecipients of male caregivers had lower rates of ED utilization for reasons that were urgent and not preventable. I also found that care-recipients of adult children had lower rates of non-urgent ED utilization than care-recipients of more distantly-related or unrelated caregivers. These findings suggest that unmet care needs might be reduced by improving access to formal care for older women with functional limitations and monitoring unmarried older men. In addition, improving service access for Black and Hispanic caregivers, and for caregivers who provide unscheduled care, could reduce caregivers unmet service need. Finally, targeting training and support services to male spousal caregivers could improve access both to needed daily care and to emergency care. Future research should focus on replicating these analyses after more data become available and on developing alternate measures of caregiver service supply. vi

10 Public Abstract Many adults age 65 and older require daily assistance from caregivers, including family and friends. Little is known about the relationship between the nature of this assistance, characteristics of caregivers, and older adults health. Older adults needs for assistance are not always met. I found substantial persistence in older adults unmet care needs over one year, and frequent year-to-year change in the caregivers assisting each older adult. Older women reported fewer unmet needs when receiving care from a professional caregiver, while men reported fewer unmet needs when receiving care from their spouse. Caregivers have uneven access to potentially useful services, such as training. I found that Black and Hispanic caregivers were more likely to seek services without using them, as were caregivers without a regular caregiving schedule. The supply of local services was not generally related to caregivers service use. Finally, I found that caregivers use of services was not related to the emergency department use of their older adults receiving care. Older adults receiving care from their adult children or male caregivers had fewer emergency department visits, but these were for different types of visits. These findings suggest that older women would benefit from greater access to professional caregiving, while older unmarried men should be frequently monitored. Caregiver services should be scheduled, designed, and marketed to meet the needs of diverse caregiver populations. These services should not be expected to reduce health care spending. Future research should develop new measures of the supply of caregiver services. vii

11 Table of Contents List of Tables... ix List of Figures... xi Chapter 1 : Care of Older Adults, Caregiving Outcomes, and the Effects of Caregiver Support Services... 1 Chapter 2 : Care Configurations and Unmet Care Needs in Older Men and Women Chapter 3 : Comparing and Contrasting Caregivers Service Use and Unmet Caregiver Service Need Chapter 4 : Caregivers Use of Support Services and Emergency Department Utilization by Their Older Adult Care-Recipients Chapter 5 : Discussion and Conclusion References viii

12 List of Tables Table 1-1: Relating Selected Caregiving Theories to Andersen s Behavioral Model Table 2-1: Derivation of the Analytic Sample Table 2-2: Characteristics of the Analytic Sample Table 2-3: Transition Probabilities for Care Configurations among Women Table 2-4: Transition Probabilities for Care Configurations among Men Table 2-5: Transition Probabilities for Unmet Care Needs among Women Table 2-6: Transition Probabilities for Unmet Care Needs among Men Table 2-7: Adjusted Odds Ratios from the Logistic Regression Models of Unmet Care Needs on Care Configuration and the Percent of Care Provided by Formal Caregivers, by Sex Table 2-8: Adjusted Odd Ratios from the Logistic Regression Models of Unmet Care Needs on Spousal and Non-spousal Care Configurations Table 3-1: Description of the NSOC Caregivers by Service Use and Unmet Service Need Table 3-2: Multinomial Logistic Regression Models of Service Use/Unmet Need on Various Categories of Predictors Table 3-3: Multinomial Logistic Regression Model of Service Use/Unmet Need Categories (Full Model) Table 3-4: Logistic Regression Models of Respite and Training/Support Group Use Table 3-5: Linear Probability Models for Service Use and Unmet Service Need, with State Fixed Effects Table 3-6: Unweighted Multinomial Logistic Regression Models of Service Use/Unmet Need on Categories of Predictors Table 3-7: Unweighted Multinomial Logistic Regression Model of Service Use/Unmet Need (Full Model) Table 4-1: Sample Derivation Table 4-2: Characteristics of the Analytic Sample by Caregiver Service Use Status ix

13 Table 4-3: Results of Negative Binomial Regression Models of Emergency Department Visits in the Year Following the NSOC Survey on Caregiver Service Use and Other Variables Table 4-4: Results of Negative Binomial Regression Models of Emergency Department Visits in the Year Prior to the NSOC Survey on Caregiver Service Use and Other Variables Table 4-5: Results of Negative Binomial Regression Models of Selected Categories of Emergency Department Visits in the Year Following the NSOC Survey on Caregiver Service Use and Other Variables x

14 List of Figures Figure 1-1: Participation in National Health and Aging Trends Study (NHATS) and the National Survey of Caregiving (NSOC) Figure 4-1: Timing of the Measured Outcomes and Independent Variables xi

15 Chapter 1 : Care of Older Adults, Caregiving Outcomes, and the Effects of Caregiver Support Services Informal care constitutes the single largest source of long-term services and supports (LTSS) for older adults in the United States (Gibson & Houser, 2007; Hagen, 2013). Caregiving, however, has become so medically complex (Reinhard, Levine, & Samis, 2012) that ensuring its high quality for community-residing older adults is a challenge (Miller, Allen, & Mor, 2008). In recognition of that challenge, the Health and Medicine Division (HMD) of the National Academy of Sciences, Engineering, and Medicine convened a committee to make recommendations to support the capacity of family caregivers to perform critical caregiving tasks, to minimize the barriers that family caregivers encounter in trying to meet the needs of older adults, and to improve the health care and long term services and supports provided to care recipients (National Academies of Science, Engineering, and Medicine, 2016). This charge reflects three broad areas whose importance has been shown in the literature, but for which key policy-relevant questions remain unanswered. First, the importance of sufficient caregiving capacity is well-known from studies linking unmet need for caregiving with adverse health outcomes (Allen & Mor, 1997; Gaugler, Kane, Kane, & Newcomer, 2005; Hass, et al., 2015). Less is known, however, about prevention of unmet need. Second, designing caregiver support services (e.g., respite care, support groups) to address barriers faced by caregivers is known to reduce caregiver burden (Knight, Lutzky, & Macofsky-Urban, 1993; Sörensen, Pinquart, & Duberstein, 2002; Thompson et al., 2007; Young, 2015; Zarit, Gaugler, & Jarrott, 1999), but caregivers ability to identify and access these services is uncertain (Friedemann, Newman, 1

16 Buckwalter, & Montgomery 2014; Herrera, George, Angel, Markides, & Torres-Gil, 2013; Hong, Hasche, & Lee, 2011; Scharlach et al. 2008; Whittier, Scharlach, & Dal Santo, 2005). Third, many studies have evaluated the effect of caregiver services on caregivers themselves, but fewer have measured care recipient outcomes, limiting the ability to reach conclusions on the overall impact of these services (Van Houtven, Voils, & Weinberger, 2011). Healthcare providers and policymakers have a new incentive to focus on the nonmedical determinants of health. National efforts to simultaneously improve health, reduce health care costs, and improve health care quality (the triple aim) have been underway for several years (Berwick, Nolan, & Whittington, 2008). They have intensified with the emergence of Accountable Care Organizations (ACOs) and other initiatives related to the Patient Protection and Affordable Care Act of 2010 (ACA). ACOs seek to make healthcare providers accountable for achieving the triple aim by allowing provider organizations to share in the savings from reducing the cost of providing care to a defined population, provided that these organizations also score well on quality metrics (Fisher & Shortell, 2010). Other payment innovations (e.g., bundled payments, capitation) similarly seek to incentivize simultaneous improvements in cost and quality. Evidence is growing that health care organizations are most able to meet these seemingly contradictory goals in certain segments of their patient populations who have especially high rates of health care utilization (Colla et al., 2012; Peikes et al., 2013). One predictor of high utilization is the presence of functional limitations; patients with functional limitations generally must obtain assistance with daily activities, and the assistance provided by older adults caregivers makes these caregivers part of the health 2

17 care team. Hospitals may benefit from incorporating caregiver assessments and the provision of caregiver services into clinical practice, and in other actions that reflect a recognition of caregivers as partners (Gaugler, Potter, & Pruinelli, 2014). Given this context, I provide empirical research in several important areas of caregiving. In this chapter, I begin with an overview of theoretical models of elder caregiving. I then summarize what is known about each of three aspects of caregiving - the relationship between care configurations and older adults unmet care needs, factors that are associated with caregivers likelihood of using caregiver services, and the effect of those services on care recipients health and healthcare utilization. This chapter concludes with a brief discussion of the advantages provided by the data source used for the empirical analyses in Chapters 2, 3, and 4. Theories of Caregiving Although a wide variety of models and theories have been developed to explain various aspects of the caregiving experience, here I provide an overview of only a few theories that inform the empirical analyses in Chapters 2, 3, and 4. The Behavioral Model of access to health care developed by Andersen (1968) and extended by Aday and Andersen (1974) provides a means of organizing several theories that focus on predictors of informal care. In this model, predisposing characteristics are those that predispose an individual to seek care, such as a strong commitment to residing in the community; enabling characteristics are resources that allow an individual to obtain care that s/he seeks, such as the availability of nearby relatives who could potentially devote time to providing care; and, need reflects the underlying conditions for which an individual may wish to obtain services, such as difficulties performing self-care or essential household 3

18 tasks. These three sets of characteristics jointly determine an individual s utilization of health care. Bass and Noelker (1987) have noted that, to properly apply Andersen s behavioral model to informal care utilization, it is necessary to consider the predisposing, enabling, and need characteristics of the caregiver as well as the care recipient. Predisposing characteristics are sometimes equated with demographic factors. However, with respect to LTSS, the most important predisposing characteristics have been reframed as psychosocial factors (Bradley et al., 2002). This concept encompasses attitudes and social norms related to LTSS, and is closely tied to cultural beliefs. While these concepts take some effort to measure, race and ethnicity have been used as a rough proxy for culture in some research. Empirical work that compares caregiving across racial and ethnic groups highlights the role of expectations in determining who will provide care (Young, 2015). On the other hand, race and ethnicity are also correlated with socioeconomic characteristics that affect families abilities to access paid home- and community-based services (HCBS, Scharlach, et al., 2008), and therefore the greater reliance on informal care among non-white caregivers may not be purely the result of cultural beliefs. Rosenthal s theory of kinkeeping (1985) focuses on the division of labor within a family. Rosenthal observes that one family member, usually a woman, is often tasked with maintaining contact between family members and across generations, and that this role is inherited from older family members. This theory has been influential in family caregiving research because of the likelihood that a family s kinkeeper would also be tasked with providing elder care. The kinkeeping concept is related to the idea of reciprocal social supports, and is listed below under enabling characteristics. However, 4

19 due to her position, when the kinkeeper herself needs care, she might have closer relationships with the younger family members who might be able to provide this care. Viewed in this way, kinkeepers might also be predisposed to seek care. Cantor s hierarchical-compensatory model (1979) and Becker s time-use allocation theory (1965) focus on enabling characteristics, specifically, the informal care resources available to an individual. Cantor describes potential caregivers within a hierarchy, with close relatives (spouses, adult children) being most highly preferred, while more distant relatives (in-laws, grandchildren) and non-kin (friends and neighbors) provide care if the closer relatives are unavailable, unable, or unwilling. Becker focuses on the opportunity cost of time spent on household tasks, postulating that an individuals receipt of care should depend on the other potential uses of their family members time. While Becker and Cantor both focus on the availability of family, Becker s focus is on the decision-making of each potential caregiver, while Cantor takes availability as fixed, focusing on who provides care when a preferred family member is not available. Finally, Litwak s theory of task specificity (1985) emphasizes the need characteristics of the older adult. Litwak posits that the mixture of formal and informal care provided to older adults is determined by their specific needs and the appropriateness of different caregivers for different tasks. On the one hand, needs that are predictable and/or highly technical are well-suited to formal caregiving. This might include assistance with meal preparation, wound care, or other activities that may be predictably scheduled or routinized. On the other hand, the family unit may provide care that is neither routine nor scheduled, such as assistance with going to the bathroom 5

20 overnight. Similarly, the family often provides care that is the result of emerging needs, such as increased supervision of a family member whose cognition has recently declined. Outcomes of Caregiving: Unmet Care Needs For older adults with functional limitations, unmet care need is the most intuitive measure of inadequate care; individuals who cannot complete all their daily tasks independently should ideally be able to complete these tasks with appropriate help. Some research suggests that unmet needs can be quite prevalent in certain subpopulations; indeed, 60% of community-residing disabled older adults with a paid caregiver had unmet care needs (Freedman & Spillman, 2014), while Medicare-Medicaid dual-eligibles were twice as likely to experience consequences from unmet need as Medicare enrollees who were not eligible for Medicaid (Allen, Piette, & Mor, 2014). Moreover, long-term health consequences have been associated with the lack of needed assistance, including falls, nursing home placement, emergency department (ED) visits, hospitalizations, and death (Allen & Mor, 1997; Gaugler et al., 2005; Hass, et al., 2015). Previous research has identified several predictors of unmet care needs, including severity of need (Allen, 1994), access to formal and informal care services (Lima & Allen, 2001), and use of assistive technology (Agree & Freedman, 2003). The potential relationship of unmet care needs and demographic factors (e.g., gender, race, education, and age) has been hypothesized, but results have been inconsistent. For example, with respect to gender, most studies have found that women report greater unmet care needs, but a handful do not (Kennedy, 2001). In addition, it has been hypothesized that factors predicting unmet care needs would differ between men and women, but studies evaluating this hypothesis have so far found little support (Allen, 1994; Lima & Allen, 6

21 2001). Nevertheless, studies of caregiving continue to find that women are more likely to be socialized into a caregiving role (Allen et al., 2012), and men may therefore be less prepared to care for their wives than women are for their husbands. Thus, in spite of the lack of direct empirical support, it also seems premature to conclude that the same factors predict unmet care needs for men and women. In addition, while one previous study suggests that unmet care needs do not persist over time (Tennstedt et al., 1994), this study relied on a relatively long follow-up period to reach that conclusion. Recent studies have concluded that older adults care configurations change frequently over time (Allen et al., 2012), further suggesting a need to reexamine the relationship between unmet care needs and care configurations longitudinally. Therefore, in Chapter 2 I examine the hypothesis that the care configurations characteristics of the caregiver networks that provide care to older adults associated with unmet care needs are different for men than for women. In addition, I examine changes in care configurations and unmet care needs over time, to assess the viability of intervening to alleviate unmet care needs. The findings from this examination have important implications for interventions to reduce unmet care needs. Caregiver Service Utilization A variety of useful services have been promoted as means of generally enhancing caregivers performance in the caregiving role, including respite care, individual and group psychoeducational programs, and social support groups (Van Houtven et al., 2011; Zarit, Gaugler, & Jarrott, 1999). These services have been provided 7

22 most often by community organizations rather than by healthcare providers. Since 2002, the National Family Caregiver Support Program (NFCSP) has provided federal grants to states for caregiver supports in the following categories: information about support services; assistance accessing support services; counseling, support groups, and training; respite; and supplemental services (e.g., transportation). The NFCSP is funded at about $150 million per year, which is less than $10 per informal caregiver (Administration for Community Living, 2015). This is not enough to provide services on demand to all caregivers in the country; however, funding may not be the limiting factor with respect to caregiver services. Many caregivers do not seek out caregiver support services, either because they do not feel they need them, or because they are unaware that such services exist. The capacity to deliver services is also constrained, and may be particularly limited in certain geographic areas (Whittier et al., 2005). The caregivers who are most likely to receive services are those who provide a high level of care and those who have provided care for many years (Friedemann et al. 2014; Hong, Hasche, & Lee, 2011). The evidence on whether men or women use more caregiver services is contradictory, with some studies indicating greater service use among male caregivers (Allen, 1994; Hong et al., 2011), and others indicating no difference (Scharlach et al., 2008). Studies considering racial disparities in service use similarly fail to reach a consensus (Herrera et al., 2013; Scharlach et al., 2008). Other potential factors that might affect the use of caregiver services include the care recipient s primary disease (e.g., Alzheimer s, cancer, stroke, Parkinson s), which is known to be an important factor in the caregiving experience (Pinquart & Sorensen, 2003), and the caregiver s source of information about services (Herrera et al., 2013). 8

23 In addition to the lack of consensus on many factors that could predict caregiver service use, it is not clear how to interpret some findings. For example, one prevalent barrier to service utilization is awareness of the service itself (Hong et al., 2011). However, because awareness logically precedes other potential factors, such as service availability, this finding leaves unanswered the critical question of whether services are either sufficient or well-targeted to caregivers who have shown some interest in services. Indeed, some studies have found no relationship between caregivers reported need for services and their service use, suggesting that access barriers remain important (Friedemann et al. 2014). Therefore, in Chapter 3 I extend previous analyses of caregiver service utilization and potential barriers to access. In addition to the benefits derived from simultaneously considering self-reported caregiver and care recipient characteristics, my ability to compare caregivers who used caregiver services with those caregivers who sought information but did not use services is another contribution. This permits an evaluation of barriers to service use among only those caregivers who were aware of these services, including potential differences in the sources of information used by caregivers when seeking services. Also, paralleling my approach in Chapter 2, I hypothesized that predictors of caregiver service use would differ based on the caregiver s gender. Effects of Caregiver Support Services on Care Recipients While the services described in the previous section may be desired by some caregivers, the effects that these services have on care recipients has not been fully explored. Meta-analyses and systematic reviews conclude that caregiver services, especially multicomponent interventions and individual psychoeducational interventions, 9

24 are moderately efficacious in reducing caregiver burden (Knight et al., 1993; Sörensen et al., 2002; Young, 2015). Aside from caregiver burden, it is likely that these services also impact other aspects of the caregiving relationship, but that effect is less welldocumented. In caregiver intervention studies, the use of a wide variety of additional outcome measures (including caregiver s psychological well-being, caregiver s physical health, care recipient health, and care recipient healthcare utilization) makes it difficult to draw additional conclusions about the effects of caregiver services beyond caregiver burden (Van Houtven et al., 2011; Young, 2015). The lack of conclusive evidence on the outcome of caregiver services is particularly evident with respect to care recipient outcomes that might be intervention targets for the healthcare system reforms described above. For example, reducing rates of avoidable ED use is desirable, because older adults quality of life/care during an ED visit can be very low (Boltz et al., 2013). This reduced utilization can also cut the cost of providing health care. Unfortunately, the effects of caregiver interventions on these outcomes have rarely been studied (Van Houtven et al., 2011), and I have not been able to find any studies that examine these outcomes at the population level. Therefore, in Chapter 4 I examine the relationship between caregiver support services (respite, training, support groups, financial help) and care recipients patterns of ED use. Using existing claims-based taxonomies, I classify ED visits based on their intensity and the potential that they could have been avoided or treated in an outpatient setting. The findings in Chapter 4 provide a national perspective that complements the numerous evaluations that focus on the outcomes of individual service models. 10

25 Data The analyses that constitute Chapters 2, 3, and 4 rely on data from the National Health and Aging Trends Survey (NHATS), a nationally-representative annual survey of a closed cohort of Medicare beneficiaries, and the National Survey of Caregiving (NSOC), a single cross-sectional survey of caregivers identified by NHATS participants at baseline. During Round 1 (conducted in 2011), NHATS participants were asked to provide contact information for up to five caregivers previously identified in the survey. The NSOC attempted to contact all individuals identified through this mechanism. Figure 1-1 shows the relationship between NHATS and NSOC participants. NHATS has several advantages for answering the research questions posed in Chapters 2, 3, and 4. First, the ability to link caregivers and care recipients between NHATS and NSOC allows for simultaneous consideration of caregiver and care recipient characteristics. Also, while NSOC was a single cross-sectional survey of caregivers, NHATS participants are re-interviewed annually, with four waves of NHATS survey interview data available so far. This allows for care recipient outcomes and key covariates (e.g., functional limitations) to be measured longitudinally. In addition, the ability to merge crude geographic indicators (city and state) with other variables permits consideration of local factors that might be associated with caregiver service use. Finally, the availability of Medicare claims facilitates analyses that combine self-reported measures (e.g., caregiver support use) with claims-based measures of healthcare utilization. For those NHATS participants who were enrolled in Medicare prior to 2011, this includes available claims during a lookback period before the initiation of NHATS. I describe the approach to each analysis more fully in Chapters 2, 3, and 4. 11

26 Dissertation Structure Chapters 2, 3, and 4 present, as independent manuscripts, the three analyses described above. Each chapter will be submitted to a high-quality health services research or gerontology journal to be considered for publication. Chapter 5 concludes the dissertation by synthesizing the findings of these analyses in the context of the existing literature. Because this dissertation follows a three-paper format, it is relatively concise when compared to a traditional dissertation. This is because the publication requirements imposed by particular journals limit the length of each of the empirical chapters. In the Department of Health Management and Policy, the three papers included in all dissertations following a three-paper model must be distinctive from each other, while still forming a coherent body of scholarly research. 12

27 Table 1-1: Relating Selected Caregiving Theories to Andersen s Behavioral Model Predisposing Enabling Need Concepts Psychosocial factors (Bradley, et al.) Time-use allocation (Becker) Hierarchical compensation (Cantor) Kinkeeping (Rosenthal) Task specificity (Litwak) Race/ethnicity Marital status Health conditions Culture Expectations Number of children Children s gender Functional limitations Indicators (or Markers) Children s proximity Spouse/children s income Spouse/children s workplace policies Caregiving policies and programs Care recipient s wealth Potential caregivers health and well-being 13

28 Figure 1-1: Participation in National Health and Aging Trends Study (NHATS) and the National Survey of Caregiving (NSOC) 14

29 Chapter 2 : Care Configurations and Unmet Care Needs in Older Men and Women Abstract Adults over age 65 frequently experience adverse consequences as the result of unmet care needs. Previous studies have identified factors that predict unmet need, including demographics, family composition, and access to formal caregiving services. Despite the role of gender in predicting unmet care need, these studies have rarely examined whether the association between care configurations and unmet care need is different for men and women. Using data from the National Health and Aging Trends Study, I conducted a sex-stratified analysis to model the risk of unmet care needs, focusing especially on the relationship between care configuration and unmet need. I found that 33% of men, and 37% of women with unmet care needs continue to report those unmet care needs one year later. I also found substantial change in communityresiding individuals care configurations from year to year, and high mortality among individuals with multiple caregivers and facility residents. In multiple regression analyses, formal care was negatively associated with unmet care need among women, but not among men. Conversely, men receiving care from one non-spousal caregiver had three times the odds of unmet care than men receiving care from their spouse, but there was no association between spousal care and unmet need in women. These results suggest that different interventions are needed to prevent or resolve unmet care needs for men and women, including enhanced access to formal care for women and monitoring of men who do not receive spousal care. 15

30 Introduction During the aging process, the accumulation of cognitive and physical limitations can make independently completing tasks such as preparing meals, getting dressed, or getting around inside and outside the home difficult. To carry out these important daily functions, older adults in the United States rely primarily on unpaid assistance from family and friends (Gibson & Houser, 2007; Hagen, 2013). A number of trends, however, threaten older adults continued reliance on this source of care including labor force participation, fertility rates, divorce, and geographic mobility. These trends suggest that there will be fewer potential informal caregivers relative to care recipients, and also that these potential caregivers will have less opportunity than individuals in previous generations to provide care (Wolff & Kasper, 2006). In addition, the care needs of community-residing individuals, particularly those who might have been candidates for nursing home placement in the past, have increased. Not only do older adults require more assistance in order to remain in their homes, that assistance is also increasingly complex due to advances in medical care. For example, today s caregivers manage a large number of medications and provide a great deal of wound care to their care recipients (Reinhard, Levine, & Samis, 2012). In sum, these trends suggest that family and friend caregivers will need additional support to continue meeting the needs of an aging population (Miller, Allen, & Mor, 2008). Given the potential strain on older adults caregiver networks, the concern that older adults might not obtain necessary daily assistance, a phenomenon known as unmet care need, has received substantial attention. One study found that 32% of older adults who have difficulty with or receive help performing daily activities have recently 16

31 experienced negative consequences (e.g. not getting dressed, going without food when hungry) directly related to unmet care needs (Freedman, Spillman, & Kasper, 2014). The prevalence of such consequences is even greater among certain high-risk groups, such as those who are eligible for both Medicare and Medicaid (Allen, Piette, & Mor, 2014). In addition to these direct consequences, unmet care needs have also been linked to other adverse outcomes, including a heightened risk of falls (Allen & Mor, 1997; LaPlante et al., 2004), emergency department visits (Hass et al., 2015), hospital readmission (DePalma et al. 2013), nursing home placement, and death (Gaugler et al., 2005). Characteristics of an older adult s network of caregivers, including its size, relationship of its members, and other characteristics of a person s caregivers, may be relevant to his/her unmet care needs. Previous studies that have investigated the relationship between these care configurations, as I will call them here, and unmet care needs have not reached consensus. For example, indicators of inadequate informal care access, such as living alone, have been associated with a greater likelihood of unmet care need (Desai, Lentner, & Weeks, 2001; LaPlante et al., 2004). Receiving formal care after a hospital discharge, however, has been associated with better recovery of functional capacity (Arbaje et al., 2008; Penrod et al., 1998), and improved capacity should logically reduce the need for care. Receiving inadequate care has been associated with a high level of functional limitation, while access to formal and informal care, including unmarried status, fewer family members, and lack of insurance, are associated with the receipt of no care (Lima & Allen, 2001). Gender has also been considered as a predictor of unmet care need. Although findings have not been consistent (Kennedy, 2001), many studies have shown that 17

32 women were more likely than men to report unmet need (LaPlante et al. 2004; Li et al. 2012). It has been further hypothesized that the predictors of unmet care need differ for men and women, because women are more likely to have been socialized into a caregiving role, and may be more able to perform that role for their husbands than their husbands can for them (Allen, 1994; Lima & Allen, 2001). Women also provide more hours of care than men, and are less likely than men to transition out of a caregiving role (Allen et al. 2012). For these reasons, spousal care might be more beneficial for men than for women. Despite its theoretical and intuitive appeal, empirical studies of this hypothesis have been limited. Two decades ago, Allen (1994) found no difference in the unmet personal care needs of married men and married women in a relatively small sample of patients undergoing treatment for cancer in three Northeastern states. In a larger study of a nationally-representative sample from , Lima and Allen (2001) found no significant interactions involving gender when modeling unmet care needs. These older studies remain influential, and this particular hypothesis has not received much attention since then. Newer studies have continued to affirm the importance of gender in caregiving-related research (Allen et al. 2012; Szinovacz & Davey, 2007). Therefore, using newer data and a nationally-representative sample of older adults, I aimed to determine whether care configurations were associated with unmet care needs, and whether that association was moderated by sex. I hypothesized that some care configurations would be associated with higher probabilities of unmet care needs than others. However, the direction of the association was theoretically ambiguous, given that informal care might be more effective at reducing unmet need in older adults with relatively straightforward care needs, while formal care might be helpful in reducing 18

33 unmet care needs for medically-complex care recipients. I further hypothesized that men would benefit more from informal care provided by spouses, but that women might benefit more from non-spousal informal care. Finally, I hypothesized that the apparent stability of care configurations and unmet need would be sensitive to the length of time between repeat observations, with greater apparent stability at shorter intervals. I sought to test whether transition probabilities would differ when examining one-year, two-year, and three-year follow-up periods. Methods Data I used data from the first four waves of the National Health and Aging Trends Study (NHATS) public use data. NHATS is sponsored by the National Institute on Aging through a cooperative agreement with the Johns Hopkins Bloomberg School of Public Health (NIA U01AG032947). NHATS is an annual survey of a nationally-representative sample of Medicare beneficiaries over age 65, beginning with 8,245 participants in The NHATS sample was drawn using a stratified, multi-stage cluster sampling design. Among the advantages of NHATS for this study are its detailed questions on participants functioning, the in-depth information it provides on potential informal caregivers, and the number of annual follow-ups. Primary Outcome NHATS asks participants about their level of difficulty performing four self-care tasks (bathing, dressing, eating, and toileting), as well as three mobility tasks (going outside, getting around inside, and getting out of bed). NHATS deviates slightly from the 19

34 traditional approach of measuring activities of daily living (ADLs), which include the four self-care tasks plus transferring and grooming/hygiene (Katz et al. 1963). Respondents who reported any difficulty with a task were then asked if they experienced a consequence related to their inability to perform that task when help was unavailable. For example, respondents who reported difficulty getting out of bed were asked if there were times in the past month they did not get out of bed because nobody was available to help them. For respondents who reported difficulty getting out of bed, but who did not report receiving any help, the question is modified to ask if there were times in the past month they did not get out of bed because it was too difficult. Parallel questions were asked for individuals who reported difficulty with any of the other self-care or mobility items. While NHATS participants were asked similar questions about their difficulty conducting instrumental activities of daily living (IADLs) such as shopping and driving, the literature on unmet need does not imply the same long-term consequences related to unmet needs for assistance with IADLs, and therefore I restricted my focus to the identification of individuals who have unmet need with self-care or mobility items listed above. In developing a measure of unmet care needs, I excluded the question on getting around outside because it is the only one of the seven questions to focus on tasks that require leaving the home, and is also far more subject to environmental and climactic variations that could complicate the relationship between the availability of care and older adults ability to perform this task. Also, early results including this item were highly suggestive of reverse causality (i.e., more hours of care was associated with higher risk of unmet need). Participants were considered to have unmet need for assistance if 20

35 they reported that they did not conduct one or more of the remaining six activities at any time in the past month, either due to lack of assistance or, for those who did not receive any assistance, because it was too difficult. Specifically, unmet need was coded 1 for any individual who answered yes to any of the items on failure to conduct activities, and 0 otherwise. By grouping participants with multiple unmet care needs together with those reporting only one unmet care need, this approach captures the presence, rather than the intensity, of unmet care needs. I also conducted analyses in which I considered a yes answer to at least one of the four self-care activities as an indicator of unmet self-care need, and a yes answer to either of the mobility activities as an indicator of unmet mobility care need. Those results were generally consistent with my overall findings and are not presented here. It is worth noting that this definition of unmet care need is relatively narrow, compared with measures that directly ask older adults whether they perceive the care they receive to be adequate (Allen et al., 2014). Focal Predictors The key independent variables were indicators for each of several care configurations. I used two sets of care configurations to conduct the analyses presented here. These configurations resulted from combining several dimensions which are potentially related to unmet need for assistance: whether an older adult received any caregiving assistance or not, whether s/he lived in an institution, whether the primary caregiver had help, and the relationship between a caregiver and his/her care recipient (formal/informal or spousal/non-spousal). I categorized caregivers as informal if they were identified by the NHATS participant as having some type of relationship other than 21

36 simply a caregiving relationship, 1 and categorized caregivers as formal if they were identified as paid aide/housekeeper/employees, staff person at the place the participant lives, or any other nonrelative who was paid for providing care. The first set of care configurations was very simple (no care, one caregiver, multiple caregivers, or residence in an institutional setting; or death of the care recipient after baseline), capturing only the dimensions of number of caregivers and institutional settings. The simplicity of these configurations allowed me to add a continuous variable consisting of the percent of care provided by formal caregivers. The second set of care configurations accounted for the unique role of spouses as default caregivers for married older adults. This set of care configurations did not distinguish between formal and informal care, but rather between spousal and non-spousal care. By applying this distinction to the simple configurations described above, I obtained the following categories: one spousal caregiver, one non-spousal caregiver, a spouse and other caregivers, multiple non-spousal caregivers, assisted living/group home, and nursing home. Covariates I categorized activity limitations by the level of difficulty reported by NHATS participants. This resulted in four counts of activity limitations a count of activities with which an individual reported a lot of difficulty, some difficulty, a little difficulty, 1 Informal caregiving relationships include spouse/partner, daughter, son, daughter-in-law, son-in-law, stepdaughter, stepson, sister, brother, sister-in-law, brother-in-law, mother, stepmother, mother-in-law, father, stepfather, father-in-law, granddaughter, grandson, niece, nephew, aunt, uncle, cousin, stepdaughter-s son/daughter, stepson s son/daughter, daughter-in-law s son/daughter, son-in-law s son/daughter, ex-wife/ex-husband, boyfriend/girlfriend, or other relative. 22

37 and activities the participant had not completed by herself or himself in the past month. In sensitivity analysis, I explored the use of alternative indicators. One alternative was a count of limitations regardless of the level of reported difficulty, and the other consisted of two separate counts of self-care limitations and mobility limitations. The inclusion of level of difficulty improved the predictive power of the model, and was therefore retained. However, separating self-care from mobility limitations was only useful when modeling unmet self-care needs and unmet mobility needs as separate outcomes, and those models are not presented here. NHATS participants answered an extensive series of questions designed to determine the likelihood of cognitive impairment, and NHATS researchers have developed a method and an accompanying computer program for classifying individuals cognitive functioning based on proxy responses and participants performance on cognitive assessment tests (Skehan & Spillman, 2013). I used that NHATS program to classify individuals into three groups: those with probable dementia, those with possible dementia, and those with no dementia. Ultimately, I chose to include only an indicator for probable dementia in the model for parsimony. I also included indicators for fair or poor self-rated health. Finally, I included several demographic variables (age category, gender, race/ethnicity, and educational attainment). Due to sample size and a relatively high amount of missing data, the race/ethnicity categories were collapsed to Black, White, Hispanic, and Other. Hours of care was derived from NHATS participants responses to questions on each helper s number of days per week or per month, and the number of helping hours per day of help. Participants were asked about the total hours of assistance provided, not 23