geronb      J Gerontol B Psychol Sci Soc Scigeronb      The Journals of Gerontology Series B: Psychological Sciences and Social Sciences      J Gerontol B Psychol Sci Soc Sci      1079-5014      1758-5368              Oxford University Press                    22010.1093/geronb/59.4.S220                        Journal of Gerontology: Social Sciences                            Is There Equity in the Home Health Care Market? Understanding Racial Patterns in the Use of Formal Home Health Care                                          White-Means            Shelley I.                                                                    Rubin            Rose M.                                    Department of Economics, Fogelman College of Business and Economics, University of Memphis, Tennessee.                    Address correspondence to Dr. Shelley I. White-Means, University of Tennessee Health Science Center, 847 Monroe Avenue, Room 205N, Memphis, TN 38113. E-mail: kingram.utmem.edu                    7        2004            59      4      S220      S229                        17          2          2004                          11          4          2003                            The Gerontological Society of America        2004                    Objective. This article explores whether the formal home health care (HHC) market is equitable or manifests unexplained racial disparities in use.        Methods. The database is the 1994 National Long Term Care Survey. We estimate logit regression models with a race dummy variable, race interaction terms, and stratification by race. We apply the Oaxaca decomposition technique to quantify whether the observed racial gap in formal HHC use is explained by racial differences in predisposing, enabling, need, and environmental characteristics.        Results. We find numerous unique racial patterns in HHC use. Blacks with diabetes and low income have higher probabilities of HHC use than their White counterparts. Black older persons have a 25% higher chance of using HHC than Whites. Our Oaxaca analysis indicates that racial differences in predisposing, enabling, need, and environmental characteristics account for the racial gap in use of HHC.        Discussion. We find that the HHC market is equitable, enhancing availability, aceptability, and accessibility of care for older Black persons. Thus, the racial differences that we find are not racial disparities.                              hwp-legacy-fpage          S220                          hwp-legacy-dochead          RESEARCH ARTICLE                                      Distinct racial patterns in older persons' use of both institutional and community-based health care services are well documented (Wallace, Levy-Storms, Kington, & Andersen, 1998). In this article, we explore whether formal home health care (HHC) use among older persons varies by race, and if so, then we seek to explain why this differential use occurs. We assess the use of HHC as an integral part of community health care rather than as one component of the spectrum of long-term care, as is done by Wallace and colleagues (1995, 1998), and we delineate the use of HHC in the context of the broader literature of racial differences in community care to highlight that this market differs from other alternatives. Thus, discussion of racial patterns of use for alternative types of community health care provides important contextual background for our focus on HHC.      Owing partly to limited financial access to nursing homes and cultural preferences (Falcone & Broyles, 1994), older Blacks are less likely to use nursing homes and more likely to live in the community with chronic conditions and disabilities that exceed those of older Whites (Wallace et al., 1998; White-Means, 2000). However, lower use of skilled nursing facilities by Blacks is not counterbalanced by higher use of HHC services (Cagney & Agree, 1999). Further, noninstitutionalized older disabled Blacks are less likely to use other community-based services such physician services and prescription medications (White-Means, 2000). Differential use of in-home assistive devices is also found, with Whites more likely to use home modification devices and Blacks more likely to use portable devices (Rubin & White-Means, 2001).      What explains these distinct racial patterns in health services use? Racial differences in physician services use and prescription medication are not fully explained by racial differences in average income or socioeconomic or health characteristics (White-Means, 2000). Researchers (Wallace, 1990; Wallace et al., 1998; White-Means, 2000) note that racial differences are explained by three factors: availability, accessibility, and acceptability. Problems of availability arise from the lack of close geographic proximity to services and/or segregation in the use of services. Lack of knowledge about available services and limited referrals to auxiliary services lead to gaps in accessibility. Patient dissatisfaction, culturally insensitive care, and unfavorable provider–patient interactions lead to problems of acceptability.      The Institute of Medicine (2002) distinguishes between racial differences and disparities. It reports that some racial differences in health care use can be expected and are acceptable, including differences due to clinical need, patient preferences, and the appropriateness of particular interventions. But if differences are due to characteristics of the health care system, provider discrimination, or incorrect or deficient patient information, then these differences are considered racial disparities in care. Thus, differences in use, reflecting such factors as stereotyping, biases, and uncertainty regarding treatment efficacy by race, limited supplies of racial/ethnic minority health care providers, and poor patient–provider communication and trust, are seen as disparities. We seek to determine whether the racial differences in HHC use are acceptable differences (and thus equitable) or actually reflect disparities, as defined by the Institute of Medicine. If the differences found are disparities, then we would expect to find divergences in availability, accessibility, and acceptability in HHC use. So, disparities are seen as differentials that are not explained by either socioeconomic or health characteristics.      We define formal HHC as market-purchased services provided in the patient's home, including both Medicare-financed care associated with subacute care and community-based care that meets needs associated with activities of daily living (ADLs) and instrumental activities of daily living (IADLs). Our comparison focuses on use versus nonuse of community-based HHC and whether the comparative use is equitable by race. The dichotomous dependent variable indicates whether an older Black or White person who lives in the community uses HHC or not. Nonusers of HHC may use informal care support or various alternate formal (purchased or market-supplied) health care services or no care, as distinguished by Wallace et al. (1998), who also explore nursing home users.      We hypothesize that HHC is an equitable market. If it achieves the three criteria of availability, accessibility, and acceptability, then this would suggest that HHC represents an equitable market. HHC is more available to the home bound than other community-based care because it does not require transportation (Rubin, White-Means, & Feagin, 2003). Accessibility is broadened by Medicare and by the relatively low price of HHC. Medicare is the largest single HHC payer (about 40%; National Association of Home Care [NAHC], 2000), covering HHC under Part A with no out-of-pocket charges and under Part B with a $100 deductible. Other federal programs and Medicaid also pay for some HHC for the poor or medically indigent (NAHC, 2000; Shi & Singh, 1998). Also, the culturally diverse workforce of HHC (35% are Black and 10% Hispanic) may be more acceptable to diverse groups of older persons than nonhome community-based care and may be perceived to be more culturally sensitive (Stone & Wiener, 2001). As this low-paid workforce is economically disadvantaged, cultural and class barriers tend to be minimized. Additionally, because of the relatively low start-up cost of a home health agency (approximately $150,000), many agencies are owned by ethnic minorities and are more likely to serve communities where a higher proportion of ethnic minorities reside (Brown, 1998).      Data from national surveys tend to support the hypothesis of HHC as an equitable market. Indeed, Blacks are more likely to use HHC than Whites (Mui & Burnette, 1994; Wallace et al., 1998). In contrast, regression analysis provides equivocal support for the role of race in increasing home health use. Freedman (1999) reports that non-Whites are significantly more likely to use more HHC services and also to use them for longer terms. This higher intensity of use is associated with the greater likelihood that Blacks postpone use and remain in the community longer with chronic conditions (Cagney & Agree, 1999). Although the coefficients are insignificant, Picone and Wilson (1999) find higher Medicare home health agency use rates for Blacks, but they conclude that socioeconomic factors are not strong determinants compared to health status. Similarly, Miller and colleagues (1996), who do not find racial differences, find that overall health and functional status, as well as Medicaid use, are primary predictors of service use. In contrast, Wallace and colleagues (1998) note that African Americans are less likely than non-Latino Whites to use HHC, with HHC use more likely for those of advanced age, women and single women, and those with more formal education, higher ADLs, stroke or heart conditions, low income, and Medicaid.              Research Questions        This study makes important contributions to a growing literature on racial disparities in health care. The broad research objective is to discover whether HHC provides an equitable health care market for Blacks. We determine empirically based answers to two specific questions: (a) Are there distinct racial patterns in the use and determinants of use of HHC? (b) What is the role of racial differences in need, predisposing conditions, enabling factors, and environmental factors in explaining observed racial differences in HHC use?                    Methods              Data        The database is the 1994 National Long-Term Care Survey (NLTCS), a repeated panel survey designed to estimate chronic disability status, functional problems, and institutionalization rates of older persons and their use of community-based in-home long-term care. The NLTCS provides nationally representative data on the prevalence and patterns of functional limitations, medical conditions, and recent medical problems for Medicare-enrolled Americans aged 65 and older. It also includes data on health services use, out-of-pocket expenditures, financial resources, and sociodemographic characteristics of older persons and their families. The 1994 NLTCS sample is the fourth panel and includes living respondents from previous panels, despite constancy or changes in their disability level (Manton, Corder, & Stallard, 1997). The 1994 dataset is supplemented with additional persons meeting the impaired status screening criteria from those who were previously screened but not interviewed because at that time they did not meet the impaired status criteria or reached age 65 after the 1989 cutoff. The addition of an oversample of persons aged 95 and older and a “healthy” sample not meeting the “impaired” status criteria is unique to the 1994 NLTCS to enhance the national representativeness of the data.        Detailed questions are asked if persons meet screening criteria for chronic ADL and IADL disability. However, many carried over were previously chronically disabled but in 1994 had no ADL and/or IADL disabilities so that the sample shows many people with no current ADLs or IADLs (Liu, Manton, & Aragon, 2000). In contrast to some analyses (e.g., Liu et al., 2000), we include going outside as an ADL, because it is grouped as an ADL in the NLTCS screener definitions. The NLTCS screener includes limitations in walking around inside and also going outside as ADLs; this ADL definition may result in relatively high prevalence of ADLs compared with IADLs.        The NLTCS includes 5,089 noninstitutionalized older persons in the detailed community interviews, of whom 4,740 are non-Hispanic Whites and non-Hispanic Blacks; our study sample includes 3,448 non-Hispanic White and non-Hispanic Black complete reporters (participants who provided data for all variables). Similar to other survey research data, it has missing data on family income (25%) and education (5%). Typically, persons at upper and lower incomes are least likely to report income or education or both, resulting in patterns of data loss that are nonrandom or systematic. Thus, the factors explaining the likelihood of missing data are unrelated to the current study's exploration of the use of home care.        With a sample of complete reporters, we lose some statistical power, but this does not lead to inaccurate parameter estimates. First, statistical tests (Kromrey & Hines, 1994; Switzer, Roth, & Switzer, 1998) indicate that in cases where no more than 30% of the data are missing (as is true for our sample), accurate parameter estimates are generated. Second, standard estimation procedures restricted to the available observations yield valid inferences when the process that generates the missing data is independent of the observed and unobserved data values (Rotnitzky & Wypij, 1994). As the data indicate that it is equally likely that Blacks and Whites have missing income data (27.19% vs 24.3%) and that home health users and nonusers have missing income data (26.81% vs 24.11%), we have further confirmation of lack of bias in the parameter estimates using the available observations. Thus, our analysis uses a sample of 3,448 persons, with 3,136 White non-Hispanic and 312 Black non-Hispanic participants, as described in Table 1.                    Analytic Design        To examine our first research question, we use three regression strategies: one with a race dummy variable, a second with race interaction terms, and a third stratified by race. For the second research question, we apply the Oaxaca decomposition technique (Oaxaca, 1973), quantifying the percentage of the observed racial gap in HHC use explained by racial differences. This approach can contribute to our understanding because it questions whether an identified racial difference in home health use is what we should observe. As the health status of community-residing older Blacks is significantly lower than that of Whites, the Oaxaca approach can help to answer whether we should similarly find differences in home health use, with significantly higher use by Blacks, and if not, then why not?        We first estimate logistic regression models of HHC use, including a race dummy variable (Model I) and also race interaction terms (Model II). Next, we estimate logistic regressions of HHC use, stratified by race. The measure of HHC use is a dichtomous variable that equals 1 if paid home services are provided by a helping organization (i.e., a home health agency or other business provider) and 0 otherwise. The NLTCS allows respondents to define a helping organization as any organization that provides in-home assistance, by describing their relationship to each in-home caregiver (or helper). If any caregiver was from a helping organization, the respondent was coded as having HHC.        Our choice of explanatory variables and conceptual framework are guided by the Andersen model (1995) and previous research on HHC use. The Anderson model suggests that medical use patterns are influenced by (a) predisposing conditions, (b) enabling factors, (c) environmental factors, and (d) need. Our measures of predisposing conditions are gender, age, marital status, education, and three quantifiable behavioral characteristics (whether the respondent currently smokes, exercises or plays sports regularly, and/or drinks alcoholic beverages one to three times a month or less frequently; Cutler & Sheiner, 1994; Newman, Struyk, Wright, & Rice, 1990).        Our measures of enabling factors are income, income squared, health insurance status (Medicare only, Medicare plus supplements, or Medicare and Medicaid dual eligibility), number of adults living in the home, and use of home-delivered meals. We utilize income squared because previous research (White-Means, 1997) documents a nonmonotonic relationship between home health use and income. Home-delivered meals, part of community-based nutrition programs in Title VII of the Older Americans Act and part of community-based in-home care, may enhance nutrition and forestall institutionalization (Shi & Singh, 1998). As home-delivered meals are not part of standard home care and are not covered by Medicare, we view use of home-delivered meals as an enabling factor that facilitates living in the community.        Similar to previous research (Picone & Wilson, 1999), we include two proxy measures for geographic environment: urban (respondents living in cities of 50,000 or more residents or their suburbs) and South. We also include measures of (a) whether the state has a large Medicaid home- and community-based care waiver program to reflect financial support for HHC services (Sloan, Picone, & Hoerger, 1997) and (b) the number of primary care physicians per 1,000 residents by state to measure underservice by physicians. State Medicaid waiver program funding of HHC greater than the median of $142.4 million indicates a high level of funding, whereas funding less than or equal to $142.4 million indicates a low level of funding.        We examine HHC need with the five most prevalent chronic health conditions in the NLTCS and the ADLs and IADLs used as screening criteria in the NLTCS. The five chronic conditions are (a) joint diseases (rheumatism, paralysis, and arthritis), (b) diabetes, (c) heart conditions, (d) hypertension, and (e) breathing disorders (bronchitis, pneumonia, flu, emphysema, and asthma). Although persons in the sample may have one or more of these conditions, we enter each condition as a separate variable in our regressions. The simultaneous impact of multiple conditions can be explored by summing the impacts of individual conditions. We include the number of ADLs and IADLs as measures of disability (Cutler & Sheiner, 1994; Newman et al., 1990; Picone & Wilson, 1999; Sloan et al., 1997). As data on severity of health conditions are unavailable in the NLTCS database, use of continuous measures of ADLs and IADLs provides an approach to a proxy for severity. The nine ADLs are limitations with eating, getting in and out of bed, getting in and out of chairs, walking around inside, going outside, dressing, bathing, toileting, and controlling bowel movement/urination; and the seven IADLs are limitations with meal preparation, laundry, light housework, grocery shopping, managing money, taking medicine, and making telephone calls.        The general form of our logit regression models is as follows:where DG = a vector of demograhic and sociostructural variables (gender, age, marital status, education), PH = a vector of health behaviors (smoking, exercise, and alcohol consumption practices), EN = a vector of enabling factors (income, income squared, health insurance status, number of potential caregivers living in the home, and home-delivered meal use), EF = a vector of environmental factors (urban, South, underservice by physicians, and residence in a state with high- or low-funded Medicaid waiver programs), HS = a vector of chronic health conditions (joint diseases, diabetes, heart conditions, hypertension, and breathing disorders), ADL = number of ADLs, IADL = number of IADLs, and ε = error term.        We use three regression strategies to examine our first research question. We begin with regression models that include race dummy variables (Model I) and race interaction terms (Model II), because they are most directly comparable with previous research. If the race dummy variable is statistically significant, then race has a unique and independent influence on HHC use, given the role of other regression variables. The interactive model explores whether the impact of other independent variables on HHC use varies by race. Finally, using models stratified by race, we explore differences in estimated β coefficients and marginals (to measure impact of change in the independent variable on the probability of HHC use) for Blacks and Whites. We seek to determine whether independent variables impact HHC use differently for those of different races. We explore whether factors found significant in the unstratified data are significant in the stratified models and whether they are significant for both Blacks and Whites.        For our second research question, we explore the role of racial differences in need, predisposing conditions, enabling factors, and environmental factors in explaining the observed racial difference in HHC use. If the variables included in our model do not fully account for racial differences in HHC use, then other nonmeasured factors are assumed to account for these differences. A priori, we expect Blacks' relatively lower health levels to be associated with greater HHC use compared with Whites. We also predict that the magnitude of the racial gap in HHC use directly corresponds with the size of the racial gap in variables included in the model.        Our methodology for the second research question is application of the Oaxaca decomposition technique. The Oaxaca technique is used to quantify the percentage of the observed racial gap in HHC use explained by racial differences in the predisposing, enabling, need, and measured environmental characteristics. This technique is a standard approach in labor economics to determine the extent of wage discrimination. It was originally applied to log-linear models of wage compensation and later to nonlinear logit, probit, and bivariate probit discrimination models (Hinks & Watson, 2001; Mohanty, 2002). With use of this technique, the total racial gap in HHC use is the difference in the probability that Whites and Blacks use HHC services: (PW − PB). PW − PB is equal to F(Xw, ßw) − F(Xb, ßb), where ßw and ßb = regression coefficients for Whites and Blacks, Xw and Xb = characteristics of Whites and Blacks, and F = the cumulative logistic probability function: 1/[1 + e−(α+βXi)]. The total gap emphasized in the Oaxaca technique can also be expressed asThe first term of the above equation measures the gap in the probability of HHC use that is explained by racial differences in characteristics, that is, acceptable differences in use. The gap explained by racial differences uses the regression coefficients for Whites and examines how the probability of HHC use varies when these regression coefficients are paired with the characteristics of Whites versus the characteristics of Blacks. The unexplained gap (i.e., disparity) is the difference in the probability of use of HHC by Blacks assuming their own characteristics (Xb) and using the regression coefficients of Blacks (ßb) and the probability of use of HHC by Blacks assuming their own characteristics and the coefficients (ßw) of Whites.        If unexplained differences in service use are found, they may be due to either unmeasured demand side or supply side factors, including racial differences in the knowledge about or preferences for HHC use, barriers to formal medical care, or possibly discrimination in the provision of HHC (i.e., factors in the health delivery system that influence accessibility, availability, and acceptablity of services). If the Oaxaca technique reveals unexplained differences in services use, then an important next step of investigation would be to seek data sources that allow identification of the causal factors among the unmeasured variables. However, if application of the Oaxaca technique reveals no unexplained differences in HHC service use, then an equitable market with acceptable racial differences is indicated. An additional caveat in the use of the Oaxaca decomposition approach is that measurement errors may also generate unexplained racial variation in the use of HHC. If regression variables are systematically measured with error by race, for example, self-reported diseases, then the systematic differences would appear as components of the unexplained gap in service use.                    Results              Sample Characteristics        Table 1 presents the study sample characteristics by race. About 16.2% of Whites and 20.2% of Blacks used HHC, or Blacks had a 25% greater chance of utilizing HHC. The sample includes 91% non-Hispanic Whites and 9% non-Hispanic Blacks, with similar mean ages of 78 and gender distribution (63% female). Whereas 40% of Whites live in medium or large cities and 35% are in the South, more than half of Blacks live in medium or large cities, and two thirds live in the South. Fully 70% of Whites, but only 29% of Blacks, have Medicare plus supplemental health insurance coverage, a striking racial differential. Blacks are much more likely than Whites to have only Medicare coverage (30.7% vs 17.3%) and 3.5 times more likely than Whites to be dually eligible (36.7% vs 10.4%). These racial insurance status differences correspond to differences in education and income. The average income is slightly over $21,000 for Whites and only $13,000 for Blacks; about 70% of Whites have at least a ninth grade education compared with 39% for Blacks. Preferences for preventive health and wellness differ for the two racial groups. Whites are more likely than Blacks to report drinking one to two times a week (15% vs 5%) and also getting exercise or playing sports regularly (23% vs 14%), but Blacks are slightly more likely to smoke (13% vs 10%).        The ability of older persons with chronic illnesses to remain in the community is affected by the availability of caregivers (measured by the number of other adults in the household) and nonhousehold support services (measured by use of home-delivered meals). The number of other adults living in the home differs by race, with Blacks having an average of slightly over two, while Whites average only one and a half. The data indicate differences in the use of home-delivered meals (7.3% of Blacks vs 4.8% of Whites). Blacks are more likely than Whites to live in states with high physician/population ratios (6.63 vs. 5.00), with high Medicaid waiver programs (25% vs 8%), and also with small Medicaid waiver programs (12% vs 3%).        Blacks report the presence of chronic health conditions more frequently than do Whites, with the exception of breathing disorders. The presence of heart conditions is reported by Blacks and Whites with similar frequency (28%). However, Blacks report joint disease, diabetes, and hypertension more frequently. Blacks average more ADLs than Whites (1.78 vs 1.38) and report more IADLs (averaging 1.08 vs 0.81). About 56% of Blacks report four or more ADLs (compared with 47% of Whites), and 29.4% report one or more IADLs (compared with 26.5% of Whites). Note that for both Blacks and Whites, ADL prevalence is higher than IADL prevalence. Based on the NLTCS screener criteria, this is not unexpected, as suggested in the data description section above. The prevalence for difficulty in getting around inside is 33.1%, and the prevalence for difficulty in getting around outside is only 3.9% (3.9% for Whites and 3.8% for Blacks). Thus, the results might be slightly skewed by including difficulty getting around outside as an ADL rather than an IADL.                    Racial Patterns in Use of Formal HHC        Table 2 reports two models analyzing the logistic regression results for HHC use by Blacks and Whites. Model I includes a race dummy variable, and Model II also includes race interaction terms. Chronic medicial conditions, ADLs, IADLs, and enabling, environmental, and sociodemographic factors influence HHC use.        In both Model I and Model II, heart condition is the chronic condition most likely to lead to HHC use. Older persons with more ADLs and IADLs have greater HHC use, and both income and income squared are statistically significant. Interestingly, there is a nonmonotonic relationship between income and HHC use. First, HHC use decreases as income increases; then, for those with higher incomes, increases in income are associated with increases in the probability of HHC use. It appears that HHC use is U-shaped with respect to income, as has been found for nursing home care (Headen, 1993; Morrow-Howell & Proctor, 1994). Those participating in meals programs are more likely to use HHC. However, as the number of persons living in the household increases, the probability of HHC use decreases. Dual eligibility increases the likelihood of HHC use, a likelihood that varies with the extent of state funding available for Medicaid community waiver programs. Women and persons who are older and have more formal years of education are more likely to use HHC, but those who are married are less likely to use HHC. The regression indicates that the variable of race is statistically insignificant in influencing HHC use.        The role of race is further explored in Model II with race interaction terms. Each variable that was significant in Model I remains significant in the full interaction model, and the variables interacting race and diabetes and race and income are also significant. Thus, if Blacks have diabetes, they are more likely to use HHC than White diabetic individuals and also more likely than Blacks who have other major chronic conditions. Further, when we examine Blacks and Whites who have the same income, Blacks are more likely than Whites to use HHC.        Table 3 presents the results of our stratified logistic regressions of the use of formal care by race and the marginals for significant coefficients. The influence of race on some independent variables, as reported in Table 2, allows us to estimate the regressions separately, although a Chow test indicates that race does not affect the impact of all independent variables on the dependent variable. With the exception of the marginals for the income and income-squared terms, in the reported logistic regressions, the marginal is exp(−α − βx)*bi/[1 + exp(−α − βx)]2, where bi = the regression coefficient for the variable for which a marginal is reported. Thus, Table 3 reports the marginal effects of the significant independent variables on the probabilities of HHC use, that is, the percentage point differences in the probability of HHC use when the regression variable changes by 1 unit. For regression variables that are dichotomous, the marginal is the percentage point difference in the probability of HHC use when the regression variable has a value of 1.        Computing the marginals for the income and income-squared terms is complicated because we use an interactive term (income squared) in a nonlinear regression model, that is, logistic regression (Ai & Norton, 2003; Norton, Wang, & Ai, 2003). Thus, the marginal for income is (b2 + 2b4x2) * exp(−α − βx)/[1 + exp(−α − βx)]2, where b2 = the estimated coefficient of the income term, b4 = the coefficient of the squared income term, and x2 = the mean of the income variable.        Marginals are comparable with reporting odds ratios, but they are more straightforwardly comprehended. Whereas odds ratios provide data on the relative increased probability of use, marginals provide data on the absolute increased probability of use. Odds ratios indicate the effect of a unit increase of the independent variable on the odds of an event [(probability of HHC use)/(1 − probability of HHC use)], but marginals report the effect of a unit increase in the independent variable on the probability of HHC use.        The results in Table 3 indicate that despite the difference in sample size for the Black and White samples, there are unique racial patterns in HHC use. As suggested in the interactive model, among Blacks, diabetes increases the probability of using HHC by 10%; among Whites, heart conditions are associated with higher probabilities (4.5%) of HHC use. ADL disability affects HHC use for both Blacks and Whites. The direct effect of each additional ADL is to increase the probability of HHC use by 4.3% for Whites and by 4.6% for Blacks. For both racial groups, participation in delivered meals programs is associated with a greater likelihood of using HHC, with greater probability of 24% for Blacks and 12% for Whites.        Several impacts are unique among Whites. More IADLs are associated with greater HHC use, with each additional IADL increasing the probability of HHC use by slightly less than 1%; each year of age increases the probability of HHC use by 0.25%; women have a slightly (3%) higher probability of HHC use; more education increases the likelihood of HHC use by 3%; and married status reduces the likelihood of HHC use by 6%. Three enabling/environmental characteristics also influence HHC use among Whites. Dual eligibility increases the probability of use by 8% in states with high Medicaid community waiver programs and by 7% in states with low Medicaid community waiver programs. There is a nonmonotonic relationship between income and HHC use. At lower income levels, each $1,000 increase in income decreases the probability of HHC use by 1.7%, but higher income (above $49,500) increases the probability of using HHC. For each person living in the house, the probability of HHC use decreases 3.5%.                    Explained and Unexplained Racial Patterns in the Use of HHC        We apply the Oaxaca technique to determine how much of the racial differential in the probability of using HHC is explained by racial differences in medical and socioeconomic characteristics. We calculate the difference between the probability that Whites would use HHC assuming their own actual characteristics (0.1554) and the probability that Whites would use HHC assuming they have the characteristics of Blacks (0.1984), as measured by [F(Xw, ßw) − F(Xb, ßw)]. This calculated differential probability is (0.1554 − 0.1984) or −0.0430, which is approximately the actual racial gap in HHC use (0.162 − 0.202). Thus, the Oaxaca decomposition suggests that differences solely in the medical, socioeconomic, enabling, and environmental characteristics of Blacks and Whites should lead to the observed racial gap in HHC use. Expressed differently, we find that HHC is indeed an equitable health care market.                    Discussion      Consistent with our prediction that HHC provides an equitable community-based health care market, the raw data indicate that in 1994, Blacks were more likely to use HHC than Whites. In this study, we find that predisposing, enabling, environmental, and need factors influence the use of HHC by both Blacks and Whites. Blacks have a 25% higher probability of using HHC than Whites. We find that racial differences in chronic conditions and socioeconomic factors fully predict this greater likelihood of HHC use among Blacks. This finding supports our hypothesis that HHC use is equitable by race. Older Black patients who live in the community with health conditions that are more extensive than those of White patients (possibly due to lower use of nursing home care and earlier-than-average release from hospitals) are absorbed by the HHC market, which presents few barriers to access.      What do these results mean when other research (e.g., Wallace and colleagues 1998) finds that racial differences persist? The highlight of our findings is that HHC provides a uniquely equitable market. Similar to other research, our descriptive statistics (see Table 1) note a racial difference in HHC use, with Blacks having greater use. However, using our regression model and an expanded set of variables relative to those of Wallace and colleagues (1998), we find that racial differences in these sample characteristics explain racial differences in use of HHC among community-based older persons. Thus, the racial differences we identify in HHC are not racial disparities (Institute of Medicine, 2002) and do not reflect divergences in availability, accessibility, and acceptability.      As described in previous research (Hodgson & Cohen, 1999; Wallace et al., 1998), chronic conditions (e.g., need) significantly influence HHC use, and those with the most impact vary by race. Diabetes is most reported by black HHC users, whereas a heart condition is most reported by Whites. These findings indicate that health policies or policy changes that differentially affect HHC use by chronic condition may also have unintended differential racial impacts. For example, the 1997 Balanced Budget Act, disallowing venipuncture as a sole qualifier for HHC coverage by Medicare, possibly imposed a greater burden on Blacks than Whites, because venipuncture patients are primarily diabetic. Research indicates that loss of HHC among disqualified venipuncture patients reduced their bathing and dressing assistance, requiring them to do more for themselves (Rubin et al., 2003). Thus, an unintended effect of the Balanced Budget Act, decreasing access to a service more heavily used by Blacks, possibly reduced the racial differential in HHC use. This legislation may even have reversed the prior trend of relatively higher HHC use among Blacks, and its impact is worthy of further investigation.      ADL limitations increase HHC use for Blacks and Whites, although by different percentages, which is consistent with previous research (Kemper, 1992; Wallace et al., 1998; White-Means, 1997) noting the central role of disability level in affecting HHC use. The unique role that meals programs play in increasing use of HHC needs further exploration. A remaining question is why HHC differs from other health care services for older persons. Examining the differential impacts of race and chronic conditions on older persons' use of assistive devices, Rubin and White-Means (2001) find a 50% unexplained underuse gap for Blacks. White-Means (2000) finds unexplained underutilization of medical services and prescription drugs for Blacks. For HHC, we find higher use among Blacks and no unexplained use differential.      It is important to note that whereas the NLTCS involves a complex sampling method, it is nonetheless nationally representative. The NLTCS has only a minimal research design effect of sampling (maybe 10%). Therefore, use of SUDAAN in regression analysis to adjust standard errors for underestimation of the true variation is considered unnecessary. Further, putting the SUDAAN model on top of the regression would possibly portray an inaccurate degree of precision (K. Manton, personal communication, 2004). This is in contrast to the procedures necessary for numerous other complex survey design data sources where the sampling technique leads to substantive correlation among observations, necessitating the use of weights.      We predict use of HHC at one point in time and show that this cross-sectional view of HHC use indicates equity. Analyzing HHC use within a dynamic modeling framework remains an interesting area for future research. For example, it is unclear whether Blacks are more likely than Whites to use HHC for extended periods or with more or less continuity than Whites. These are also additional components of equity of care that are beyond the scope of this article.      Another area for future research is the importance of including detailed measures of variation in support at the community level, comparable with the enabling variables that Bradley and colleagues (2002) label “availability of support.” We are unable to incorporate these variables analytically, because although we have individual data, the NLTCS lacks detailed community-level variables, giving our research imperfect measures of the community variables of Bradley and colleagues. Inclusion of these community support variables would address significant differences at the small area level.      Although our data suggest there are not unexplained racial differences in HHC use, we are unable to determine whether there are racial differences in the types and quality of HHC received, because different types of HHC are not delineated in the data. Thus, further research is still warranted to understand whether there is equity in terms of quality of care provided by different types of helping organizations. In addition, questions of how racial differences in household and family formation affect HHC use and whether the actual intrahousehold relationships are important deserve future investigation. Our results indicate that being married and also the number of other adults in the household are significant for Whites but not for Blacks.      The relationship between HHC use and cultural differences is another area for prospective research. Culture, expressed in group values and beliefs, may affect the selection among long-term care options and between institutional and noninstitutional choices (Thornton & White-Means, 2000; Wallace et. al., 1998). Thus, measured racial differences may embody cultural influences that are implicit rather than explicit. Explicit measures of knowledge and cultural preferences for receipt of HHC versus other health care services, personal experiences of discriminaton when interacting with the health care system or with the HHC services market, and charactersistics of patients' community health care systems are further issues needing investigation. These analyses would provide input for effective public policy measures to address racial disparities in use of health care services and enable replication of the equity in use that our data indicate is provided by the HHC market.                                      Decision Editor: Charles F. Longino Jr., PHD                          Table 1.                      Characteristics of the Sample.                                                                              Total                                                Non-Hispanic White                                                Non-Hispanic Black                                                                            Sample Characteristics                % or Mean                No.                                  SD                                % or Mean                No.                                  SD                                % or Mean                No.                                  SD                                                                                    N, sample size                100%                3,448                                90.4%                3,136                                9.6%                312                                                            Home health care usea                16.6%                571                0.37178                16.2%                508                0.3685                20.2%                63                0.40208                                            Av. age (mean), yr                77.8                                7.3624                77.8                                7.2659                78                                8.28181                                            Gender                                                                                                                                                                                                Male                36.4%                1,255                                36.4%                1,142                                36.7%                115                                                                Female                63.6%                2,192                0.48113                63.6%                1,994                0.48109                63.3%                198                0.4823                                            Marital status                                                                                                                                                                                                Marriedb                46.1%                1,590                0.49851                47.6%                1,493                0.49949                31.0%                97                0.46228                                                Nonmarried                53.9%                1,858                                52.4%                1,643                                69.0%                216                                                            Location                                                                                                                                                                                                Medium or large cityb                41.2%                1,420                0.49229                39.9%                1,251                0.48942                54.0%                169                0.49906                                                Small city, town or rural                58.8%                2,027                                60.1%                1,885                                46.0%                144                                                            Region                                                                                                                                                                                                Southb                37.2%                1,282                0.4834                34.7%                1,088                0.47597                62.9%                197                0.48406                                                Non-South                62.8%                2,165                                65.3%                2,048                                37.1%                116                                                            Insurance status                                                                                                                                                                                                Medicare only                18.5%                638                                17.3%                543                                30.7%                96                                                                Medicare plus supplementb                66.7%                2,300                0.4712                70.4%                2,208                0.45638                29.4%                92                0.45672                                                Medicare and Medicaid                12.7%                438                                10.4%                326                                36.7%                115                                                            Education                                                                                                                                                                                                9th grade and aboveb                66.9%                2,307                0.47082                69.6%                2,183                0.46001                39.1%                122                0.48876                                                K through 8th grade                33.1%                1,141                                30.4%                953                                60.9%                190                                                            Income (mean $)b                $20,451                                18,422.8                $21,182                                18,821.7                $13,096                                11,448.1                                            Other adults (mean)b                1.59                                1.12943                1.55                                1.06706                2.06                                1.55599                                            Delivered meals                                                                                                                                                                                                Yes                5.0%                172                                4.8%                151                                7.4%                23                                                                No                95.0%                3,276                0.21893                95.2%                2,985                0.21412                92.6%                289                0.26173                                            Behavioral characteristics                                                                                                                                                                                                Smokinga                10.3%                355                0.30395                10.0%                314                0.29979                13.5%                42                0.34186                                                Drinkingb                14.1%                486                0.34772                15.0%                470                0.35669                5.1%                16                0.22093                                                Exerciseb                21.9%                755                0.41361                22.7%                712                0.41899                13.8%                43                0.34527                                            Chronic health conditions                                                                                                                                                                                                Joint diseasesb                69.1%                2,383                0.46198                68.2%                2,139                0.46549                77.9%                243                0.41569                                                Diabetesb                16.4%                565                0.37047                15.5%                486                0.36194                25.6%                80                0.43735                                                Heart conditions                27.7%                955                0.44757                27.7%                869                0.44764                27.6%                86                0.44755                                                Hypertensionb                43.6%                1,503                0.49591                42.3%                1,327                0.49409                56.4%                176                0.49667                                                Breathing disordersb                30.3%                1,045                0.45953                30.7%                963                0.46149                25.6%                80                0.43735                                            ADL frequency distribution                                                                                                                                                                                                0 ADLs                51.8%                1,786                0.49973                52.6%                1,649                0.49941                44.2%                138                0.49746                                                1 or more ADLsb                48.2%                1,662                                47.4%                1,487                                55.8%                174                                                            ADL (mean)b                1.41                                1.92609                1.38                                1.91197                1.75                                2.03585                                            IADL frequency distribution                                                                                                                                                                                                0 IADLs                73.3%                2,527                0.44267                73.5%                2,306                0.44123                70.5%                220                0.45672                                                1 or more LADLsb                26.7%                921                                26.5%                830                                29.5%                92                                                            IADL (mean)b                0.84                                1.77436                0.81                                1.74772                1.08                                2.00949                                            Underserved by physician (mean)                5.15                                2.7202                5.00                                2.58481                6.63                                3.49789                                            High Medicaid waiverb                9.3%                321                0.29102                7.8%                245                0.26791                25.0%                78                0.43371                                            Low Medicaid waiverb                3.4%                117                0.18108                2.6%                81                0.15865                11.5%                36                0.32                                                                        Notes: ADL = activity of daily living; IADL = instrumental activity of daily living. Source: 1994 National Long-Term Care Study, Center for Demographic Studies, Duke University.                                      aIndicates racial difference significant for variable at.10 or better, based on t test for proportions.                                      bIndicates racial difference significant for variable at.05 or better, based on t test for proportions.                                                Table 2.                      Results of Logistic Regressions of the Use of Formal Care by Race for Non-Hispanic Black and White Older Persons.                                                                              Model I                                Model II                                                                            With Dummy Race Variable                                With Interactive Termsa                                                            Variables                Coefficient                t Statistic                Coefficient                t Statistic                                                                    Intercept                −4.747                −6.70764***                −4.656                −6.0807***                                            Joint diseases                0.0799                0.6277                0.1508                1.1187                                            Diabetes                0.0895                0.6303                −0.00669                −0.0430                                            Heart condition                0.4267                3.7496***                0.4555                3.7738***                                            Hypertension                −0.0728                −0.6535                −0.1012                −0.8584                                            Breathing disorders                0.1098                0.9498                0.1571                1.2941                                            ADL sum                0.4298                15.1873***                0.439                14.5364***                                            IADL sum                0.0802                3.0037***                0.0967                3.3693***                                            Underserved                0.000251                0.0111                −0.00561                −0.2262                                            South                0.0709                0.5645                0.0356                0.2681                                            Gender                0.3057                2.2882**                0.3088                2.1746**                                            Age                0.0262                3.4026***                0.0256                3.0955***                                            Married                −0.564                −4.0286***                −0.5959                −3.9993***                                            Education 9th grade+                0.3204                2.6155***                0.3331                2.5235**                                            Urban                −0.0707                −0.6284                −0.1099                −0.9235                                            Medicare plus                −0.00896                −0.0634                0.0493                0.3237                                            Dual eligible/high waiver                0.7514                4.0182***                0.8195                3.8950***                                            Dual eligible/low waiver                0.5834                2.1599**                0.7364                2.4057**                                            Alcohol                0.0939                0.4798                0.1278                0.6330                                            Smokes                0.1224                0.6405                0.0901                0.4458                                            Exercises                −0.094                −0.6196                −0.0475                −0.3006                                            Income ($10,000 units)                −0.2                −2.1423**                −0.03                −3.0374***                                            Income squared                0.0000026                3.1257***                0.000003                3.4788***                                            Other adults                −0.2983                −4.9800***                −0.3545                −5.0861***                                            Delivered meals                1.2842                7.0599***                1.1943                6.1689***                                            Black                −0.0487                −0.2542                −1.3254                −0.5915                                            B Joints                                                −0.6644                −1.4397                                            B Diabetes                                                0.7688                1.7946*                                            B Heart condition                                                −0.0963                −0.2402                                            B Hypertension                                                0.3958                0.9633                                            B Breathing disorders                                                −0.6389                −1.4141                                            B ADL sum                                                −0.0773                −0.7761                                            B IADL sum                                                −0.1134                −1.3065                                            B Underserved                                                0.0694                1.0206                                            B South                                                0.4699                1.0021                                            B Gender                                                −0.2683                −0.5492                                            B Age                                                0.00607                0.2409                                            B Married                                                −0.1268                −0.2498                                            B Education                                                0.2349                0.5696                                            B Urban                                                0.4882                1.1231                                            B Medicare plus                                                −0.5015                −0.9734                                            B Dual eligible/high waiver                                                −0.1561                −0.3228                                            B Dual eligible/low waiver                                                −0.7839                −1.1313                                            B Alcohol                                                −0.6826                −0.7285                                            B Smokes                                                0.213                0.3208                                            B Exercise                                                −0.3984                −0.6171                                            B Income ($10,000 units)                                                0.97                1.6724*                                            B Income squared                                                −0.000015                −1.299                                            B Other adult                                                0.2093                1.3788                                            B Delivered meals                                                0.6548                1.0868                                            Sample size (N)                3,448                3,448                                            −2 log L                2,368.28                2,340.741                                            χ2                726.962***                754.498***                                                                        Notes: ADL = activity of daily living; IADL = instrumental activity of daily living.                                      aRace interaction terms begin with “B” preceding a variable name.                                      *Significant at p =.10; **Significant at p =.05; ***Significant at p =.01.                                                Table 3.                      Results of Logistic Regressions of the Use of Formal Care Stratified by Race.                                                                              Non-Hispanic White Older Persons                                                Non-Hispanic Black Older Persons                                                                            Variables                Coefficient                t Statistic                Marginals                Coefficient                t Statistic                Marginals                                                                    Intercept                −4.656                −6.08071***                                −5.9814                −2.84017***                                                            Joint diseases                0.1508                1.11869                                −0.5136                −1.16383                                                            Diabetes                −0.00669                −0.04299                                0.7621                1.90955*                0.0975                                            Heart condition                0.4555                3.77382***                0.0450231                0.3592                0.93958                                                            Hypertension                −0.1012                −0.85835                                0.2946                0.74848                                                            Breathing disorders                0.1571                1.29407                                −0.4818                −1.10708                                                            Underserved                −0.00561                −0.22621                                0.0638                1.0079                                                            ADL sum                0.439                14.5364***                0.0433922                0.3617                3.81138***                0.04627                                            IADL sum                0.0967                3.36934***                0.0095581                −0.0168                −0.20513                                                            South                0.0356                0.26807                                0.5056                1.12431                                                            Gender                0.3088                2.17465**                0.0305228                0.0405                0.08665                                                            Age                0.0256                3.09553***                0.0025304                0.0317                1.33193                                                            Married                −0.5959                −3.99933***                −0.058901                −0.7227                −1.48949                                                            Education 9th grade+                0.3331                2.52348**                0.0329247                0.568                1.4538                                                            Urban                −0.1099                −0.92353                                0.3783                0.90481                                                            Medicare plus                0.0493                0.3237                                −0.4522                −0.91873                                                            Dual eligible/high waiver                0.8195                3.89496***                0.0810021                0.6634                1.52366                                                            Dual eligible/low waiver                0.7364                2.40575**                0.0727882                −0.0475                −0.07642                                                            Alcohol                0.1278                0.63299                                0.5549                0.60645                                                            Smokes                0.0901                0.44582                                0.303                0.47913                                                            Exercises                −0.0475                −0.30063                                −0.4459                −0.7123                                                            Income ($10,000 units)                −0.3                −3.03736***                −0.017                0.71                1.2456                                                            Income squared                0.000003                3.47876***                0.051                −0.00001                −1.0392                                                            Other adults                −0.3545                −5.08608***                −0.03504                −0.1452                −1.07715                                                            Delivered meals                1.1943                6.1689***                0.1180486                1.8492                3.24137***                0.23658                                            Sample size (N)                3,136                                                312                                                                            −2 log L                2,109.02                                                231.722                                                                            χ2                669.194***                                                82.185***                                                                                                        Notes: ADL = activity of daily living; IADL = instrumental activity of daily living.                                      *Significant at p =.10; **Significant at p =.05; ***Significant at p =.01.                                                      References              Ai, C., Norton, E. (2003). Interaction terms in logit and probit models. Economics Letters, 80,123-129.                    Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior, 36,1-10.                    Bradley, E. H., McGraw, S. A., Curry, L., Buckser, A., King, K., Kasl, S., et al. (2002). Expanding the Andersen model: The role of psychosocial factors in long-term care use. Health Services Research, 37,1221-1243.                    Brown, A. (1998). No place like home care. Black Enterprise, 29,33.                    Cagney, K., Agree, E. (1999). Racial differences in skilled nursing care and home health use: The mediating effects of family structure and social class. Journal of Gerontology: Social Sciences, 54B,S223-S236.                    Cutler, D. M., Sheiner, L. M. (1994). Policy options for long term care. In D. A. Wise (Ed.), Studies in the Economics of Aging (pp. 395–434). Chicago: University of Chicago Press.                    Falcone, D., Broyles, R. (1994). Access to long term care: Race as a barrier. Journal of Health Politics, Policy, and Law, 19,583-595.                    Freedman, V. A. (1999). Long-term admissions to home health agencies: A life table analysis. The Gerontologist, 39,16-24.                    Headen, A. (1993). Economic disability and health determinants of the hazard of nursing home entry. Journal of Human Resources, 28,80-110.                    Hinks, T., Watson, D. (2001). A multinomial logit nondiscrimination approach to estimating racial wage and occupational discrimination. Applied Economics, 33,605-612.                    Hodgson, T. A., Cohen, A. J. (1999). Medical care expenditures for diabetes, its chronic complications, and its comorbidities. Preventive Medicine, 29,173-186.                    Institute of Medicine. (2002). Unequal treatment: Confronting racial and ethnic disparities in health care. Washington, DC: National Academy of Sciences.                    Kemper, P. (1992). The use of formal and informal home care by the disabled elderly. Health Services Research, 27,421-451.                    Kromrey, J., Hines, C. (1994). Nonrandomly missing data in multiple regression: An empirical comparison of common missing-data treatments. Educational and Psychological Measurement, 54,573-593.                    Liu, K., Manton, K. G., Aragon, C. (2000). Changes in home care use by older people with disabilities: 1982–1994. AARP Public Policy Institute, 2000–02, January.                    Manton, K., Corder, L., Stallard, E. (1997). Chronic disability trends in the elderly United States populations: 1982–1994. Proceedings of the National Academy of Science, 94,2593-2598.                    Miller, B., Campbell, R. T., Davis, L., Furner, S., Giachello, A., Prohaska, T., et al. (1996). Minority use of community long-term care services: A comparative analysis. Journal of Gerontology: Social Sciences, 51B,S70-S81.                    Mohanty, M. (2002). A bivariate probit approach to the determination of employment: A study of teen employment differentials in Los Angeles County. Applied Economics, 34,143-156.                    Morrow-Howell, N., Proctor, E. (1994). Discharge destinations of Medicare patients receiving discharge planning: Who goes where? Medical Care, 32,486-497.                    Mui, A., Burnette, D. (1994). Long term care service use by frail elders: Is ethnicity a factor? The Gerontologist, 34,190-198.                    National Association of Home Care. (2000). Basic statistics about home care. Retrieved October 2, 2002, from http://www.nahc.org/consumer/hcstats.html.                    Newman, S. J., Struyk, R., Wright, P., Rice, M. (1990). Overwhelming odds: Caregiving and the risk of institutionalization. Journal of Gerontology: Social Sciences, 45B,S173-S183.                    Norton, E., Wang, H., Ai, C. (2003). Computing interaction effects and standard errors in logit and probit models. Working paper. September 9, 2003.                    Oaxaca, R. (1973). Male–female wage differentials in urban labor markets. International Economic Review, 14,693-709.                    Picone, G., Wilson, R. M. (1999). Medicare Home Health Agency utilization, 1984–1994. Inquiry, 36,291-303.                    Rotnitzky, A., Wypij, D. (1994). A note on the bias of estimators with missing data. Biometrics, 50,1163-1170.                    Rubin, R. M., White-Means, S. I. (2001). Race, disability, and assistive devices: Sociodemographics or discrimination? International Journal of Social Economics, 28,927-941.                    Rubin, R. M., White-Means, S. I., Feagin, T. (2003). Impacts of the 1997 Balanced Budget Act on venipuncture beneficiaries. Journal of Applied Gerontology, 22,252-268.                    Shi, L., Singh, D. A. (1998). Long term care. In Delivering health care in America: A systems approach (pp. 344–391). Gaithersburg, MD: Aspen.                    Sloan, F., Picone, G., Hoerger, T. J. (1997). The supply of children's time to disabled elderly parents. Economic Inquiry, 35,295-308.                    Stone, R., Wiener, J. (2001). Who will care for us? Addressing the long-term care workforce crisis. Washington, DC: Urban Institute and American Association of Homes and Services for the Aging.                    Switzer, F., III, Roth, P., Switzer, D. (1998). Systematic data loss in HRM settings: A Monte Carlo analysis. Journal of Management, 24,763-774.                    Thornton, M. C., White-Means, S. I. (2000). Race versus ethnic heritage in models of family economic decisions. Journal of Family and Economic Issues, 21,65-86.                    Wallace, S. (1990). The no-care zone: Availability, accessibillity, and acceptability in community-based long-term care. The Gerontologist, 30,254-261.                    Wallace, S. P., Levy-Storms, L., Ferguson, L. (1995). Access to paid in-home help among disabled elderly people: Do Latinos differ from non-Latino whites? American Journal of Public Health, 85,970-975.                    Wallace, S., Levy-Storms, L., Kington, R., Andersen, R. (1998). The persistence of race and ethnicity in the use of long term care. Journal of Gerontology: Social Sciences, 53B,S104-S112.                    White-Means, S. I. (1997). The demands of persons with disabilities for home health care and the economic consequences for informal caregivers. Social Science Quarterly, 78,955-972.                    White-Means, S. I. (2000). Racial patterns in disabled elderly persons' use of medical services. Journal of Gerontology: Social Sciences, 55B,S1-S14.