| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to William J. McAuley, Department of Health Administration and Policy, Colvard Building, UNC Charlotte, 9201 University City Blvd., Charlotte, NC 28223. E-mail: wjmcaule{at}uncc.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Balanced Budget Act • Medicare home health • Medicaid coverage • Metropolitan • Nonmetropolitan
Medicaid coverage may play a powerful role in the receipt of home care by older people. In addition to making it possible for individuals to receive formal home-care services funded directly through Medicaid, Medicaid programs often provide access to case management (Silberman, Poley, James, & Slifkin, 2002), which may facilitate access to formal home care from other payment sources.
In addition to Medicaid coverage, rural–urban variations in populations and services may lead to differences in the use of formal home care. However, rural and urban population characteristics and service supply factors do not consistently point toward higher utilization rates in one or the other setting. Poverty and lower job earnings are more common in rural areas (Ghelfi & Parker, 1997; Reeder & Calhoun, 2002). Nonmetropolitan counties have older populations (Reeder & Calhoun), and elders in rural areas are somewhat more likely to have chronic conditions (such as arthritis, diabetes, hypertension, and heart disease) that would suggest a greater need for long-term care in rural than in urban areas (Nelson, 1994). When compared with residents of large metropolitan counties, residents of totally rural counties are more likely to report having a usual source of health care (i.e., a doctor's office, clinic, or health care center), but they tend to have fewer visits to health care providers during the year (Larson & Fleishman, 2003). There is evidence that rural residents use more informal assistance than their urban counterparts, although the higher informal assistance rates may be more the result of limited access to formal services than any advantage in availability of informal support (Coward & Cutler, 1989). Rural residents are more dependent on Social Security for income and are also more likely to use Medicare and Medicaid for their acute and chronic care needs (Coburn, 2002; Reeder & Calhoun, 2002).
The formal service systems also differ in rural and urban counties. Hospitals and advanced medical services, which might in some cases substitute for home care and in other cases generate formal home-care events, tend to be less available in rural counties (Dalton, Van Houtven, Slifkin, Poley, & Howard, 2002; Dansky & Dirani, 1998; Kenney, 1993a; Kenney & Dubay, 1992). Nursing home admissions and home care may substitute for one another, and national figures suggest that there are more nursing home beds per capita in rural counties (Shaughnessy, 1994). However, nonmetropolitan counties tend to have much greater variation in beds and bed rates than metropolitan counties, because larger percentages of them have no facilities at all (McAuley, Pecchioni, & Grant, 2002). The supply and types of home health agencies also differ across rural and urban counties, with rural counties having fewer agencies per county and per square mile (Kenney & Dubay). Furthermore, 48% of home health agencies located in nonmetropolitan areas are hospital based, compared with 23% in metropolitan areas, and 64% of nonmetropolitan home health agencies are not for profit (government or nonprofit), versus 33% for metropolitan counties (Franco & Leon, 2000).
Much of our understanding of rural–urban variations in formal home care is based on Medicare home health utilization patterns. Investigations of Medicare data during the 1980s (Kenney, 1993a, 1993b; Kenney & Dubay, 1992), as well as more recent assessments (General Accounting Office, 2000; McCall, Kosimar, Petersons, & Moore, 2001), have identified higher rates of utilization of Medicare home health care in urban than in rural areas. These Medicare-based investigations have generally been limited to fee-for-service recipients.
Other studies examining rural–urban differences in formal home care have used such varied data sources as the Asset and Health Dynamics Among the Oldest Old (AHEAD) study (Norgard & Rodgers, 1997), the Longitudinal Study on Aging (Johnson & Wolinsky, 1996), and the 1987 Medical Expenditure Survey, a precursor to the Medical Expenditure Panel Survey that forms the basis for this analysis (Altman & Walden, 1993). These investigations have used very different operational definitions of formal home care and of rural and urban residence, but none identified significant rural–urban differences in home-care use. As Coburn (2002) recently noted, findings of extant investigations regarding differences in the likelihood of formal home-care utilization in rural versus urban areas are mixed and inconsistent, and they do little to clarify the factors that account for any identified differences.
The conceptual framework for this research draws on prior investigations of the factors associated with formal home-care utilization, including both hospital postdischarge referral studies and utilization patterns in the general noninstitutionalized population. Much prior research has used the Andersen–Newman Model of health care use (Andersen, 1968; Chappel, 1994). A number of significant person-level variables have been identified in the literature, and these are presented in the paragraphs that follow according to the predisposing, need, and enabling categories of the Andersen–Newman model.
Among the predisposing characteristics that have been identified as being associated with home-care use are age (Bowles, Naylor, & Foust, 2002; Dansky, Brannon, Shea, Vasey, & Dirani, 1998; Norgard & Rodgers, 1997), gender (Dansky et al., 1998; Johnson & Wolinsky, 1996), racial or ethnic status (Altman & Walden, 1993; Johnson & Wolinsky, 1996), widowhood (Altman & Walden), educational attainment (Johnson & Wolinsky; Norgard & Rodgers), and region of the country (Johnson & Wolinsky; Norgard & Rodgers). Need-related factors have been among the most significant predictors of home-care use and have included activities of daily living (ADL) impairment (Altman & Walden; Bowles et al.; Dansky et al.; R. L. Kane et al., 1996; Norgard & Rodgers; Slivinske, Fitch, & Wingerson, 1998), instrumental activities of daily living (IADL) impairment (Altman & Walden; Johnson & Wolinsky; Slivinske et al.), cognitive or memory problems (Slivinske et al.), and health status or diagnoses (Bowles et al.; Dansky et al.; Johnson & Wolinsky; R. L. Kane et al.). Among the enabling variables found to be related to home-care use are financial resources (Slivinske et al.), Medicaid coverage (Dansky et al.; Johnson & Wolinsky), potential caregiving resources such as living arrangement (Altman & Walden; Dansky et al.; R. L. Kane et al.; Norgard & Rodgers), having a living child (Norgard & Rodgers), contact with family members (Johnson & Wolinsky), and limited social resources (Slivinske et al.). The MEPS-HC includes measures that reflect most of these characteristics.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
With the application of appropriate weights, the MEPS-HC is a nationally representative sample of the noninstitutionalized civilian population. Information from the Area Resource File (ARF; Bureau of Health Professions, 2001) was used to establish rural and urban categories for each respondent's county of residence. The 1998 MEPS-HC data set includes information on 24,072 individuals. Because of our interest in formal home-care use among older people, we limited the analysis to the 2,584 people aged 65 or older who were identified as alive and in the sample at the initial 1998 data-collection point.
Variables Used
Table 1 describes the dependent and independent variables we used in the analysis. At each of the three MEPS-HC survey points during 1998, determination of formal home-care use was based on a series of questions aimed at assessing whether each individual in the sample household received any home visits as a result of a health problem. Respondents were prompted by a showcard listing a broad range of home-care service and provider types, including skilled medical care, personal care, household chore, companionship, and other types of home care. We use information from the Medical Provider component to verify use and source of payment.
|
There have been many approaches to categorizing areas into rural and urban groupings (Ricketts, Johnson-Webb, & Taylor, 1998). We used the urban influence coding system (Ghelfi & Parker, 1997) as the foundation for classifying county of residence. The urban influence codes comprise a nine-category scheme incorporating two major factors as the basis for grouping counties—the county's largest city and its proximity to other counties with metropolitan populations. This coding system is valuable when one is examining health care issues, because the codes are likely to track economic threshold levels and system complexities that are related to the population size of the largest city (Ricketts et al.). Unfortunately, the number of elderly respondents in the 1998 MEPS-HC data set did not allow us to use the entire nine-category system. Therefore, we sought to identify an approach based on this system, but with fewer categories. Ghelfi and Parker compared numerous area characteristics across the nine urban influence categories, such as earnings, number of higher education institutions, and physician supply. Our assessment of these comparisons demonstrated a general concordance within three county types: (a) metropolitan, (b) nonmetropolitan counties with a city of 10,000 or more (which we call nonmetro large town counties), and (c) nonmetropolitan counties having no city of at least 10,000 residents (which we call nonmetro rural counties). We examined cross-tabulations of these three categories by demographic characteristics and functional status within our analytical sample (results not shown). There were no significant rural–urban differences for the functional measures. Significant differences that were found indicated that older metropolitan residents had higher incomes than the other two categories, whereas nonmetro rural residents had larger households and lower educational levels and were less likely to be Hispanic than residents of nonmetro large town and metropolitan locations. Although not significant, there was also a trend toward higher percentages having Medicaid coverage as county type moved from metropolitan to nonmetro rural. In addition to empirical validation of the three rural–urban categories, our operationalization retains major threshold and complexity attributes of the urban influence coding system useful for studies of health care (Ricketts et al.) by focusing on the population size of the largest city. Therefore, we concluded that this three-category "nonmetropolitan county and town size" coding was appropriate for the analyses.
We measured all person-level characteristics prior to April 1998 or, if a specific date was not provided, at the first data-collection round of the year, which generally occurred in the first quarter of 1998. Exceptions to this approach are the utilization of any formal home care and Medicare home care (both measured as use at any point in 1998 but based on multiple interviews), family income (annual income measured in relation to federal poverty level during 1998), and replacements for missing information for a small number of variables. We replaced this missing information, where feasible, by information gathered in a round completed later in the year. Furthermore, in order to obtain a complete set of health-condition measures, we drew some medical-condition information for one of the two survey panels from interviews of the panel that took place in the 1997 survey year. After these approaches were used, 19 cases remained with missing data on education. We ran the analyses with and without a dummy variable for missing education, and there were very minor differences in parameter estimates across the models. Therefore, we omitted the missing education place-holder variable in the final analysis.
As we noted previously, prior research has demonstrated the importance of medical diagnoses and medically related conditions in the utilization of formal home care. The medical-condition variables are based on self-reports by the respondents during each interview. These self-reports were recorded verbatim by the interviewer, and they were subsequently coded into ICD-9-M codes by professional coders (AHRQ, 2002). Our categorization of medical conditions is based on the Clinical Classification Codes (CCC), a system which partitions ICD-9-M diagnosis codes into sets of similar, clinically meaningful conditions (Elixhauser, Steiner, Whittington, & McCarthy, 1998). We used the CCC to produce a set of dummy variables reflecting acute and chronic conditions that were most likely to be associated with use of home care, according to prior investigations and our own understanding of the nature of formal home care and the qualifications for Medicare home care.
There is considerable state-level diversity in Medicaid and other policies that can influence the availability and utilization of formal home care (LeBlanc, Tonner, & Harrington, 2000, 2001; Miller, Harrington, Ramsland, & Goldstein, 2002). In addition, recent figures from the General Accounting Office (2002) demonstrate substantial state-to-state variations in Medicare home-care utilization that cannot be readily explained solely by policy differences. Therefore, we generated dummy variables for each state to serve as fixed effects controlling for this unmeasured state-level diversity. In our sample, 22 states had no or only one formal home-care or Medicare home-care user. We combined each of these states with the neighboring state that was closest in ranking on Medicaid home- and community-based care expenditures per capita (AARP Public Policy Institute, 1998). We adjusted the final models for these fixed-effects measures.
Although Coward and Cutler (1989) have indicated that there is little evidence regarding rural–urban differences in informal care, it is important to account for informal support in models that address formal home care. Our approach was to incorporate several household variables to capture the potential for informal care while limiting the endogeneity problems associated with direct measures of informal home care (e.g., living with spouse, widowed, and number of people in the household).
Finally, Medicaid coverage is an issue of special concern in this analysis. As we noted earlier, past research suggests that there are notable rural–urban differences in Medicaid coverage. Because of its potentially differential impact on access and care management in rural and urban environments, Medicaid coverage is an important factor for one to consider when examining formal care use in different residential settings. Therefore, we generated interaction terms for Medicaid coverage and the two nonmetropolitan residence types.
Analytical Approach
We used logistic regression with two dependent dummy variables: formal home-care use through any reimbursement source and Medicare-reimbursed home health care use. We initially examined correlations among potential predictor variables because of the potential for multicollinearity among independent variables within and across categories of the Andersen–Newman Model that can limit the efficacy and stability of logistic regression analysis (Long, 1997). We eliminated one variable, use of a mobility aid, from consideration because of its high correlation with the ADL variable.
Our analysis began with logistic regression of the dependent variables on the county location variables without adjustments for other factors. We then used hierarchical logistic regression with only main effects and fixed effects in the initial models, followed by models that added the Medicaid x Residence interaction terms. We used a significance level of.10 for the retention of interaction-term blocks. Finally, we examined the adjusted relative risks of using any formal home care and Medicare home health care for Medicaid x Residence subgroups, under controls for all personal characteristics and state fixed effects. Because of the complex sample design of the MEPS-HC, we utilized the SUDAAN statistical software package (Research Triangle Institute, 2002), with all analyses adjusting for the survey design factors to ensure appropriate significance levels. We also included population weights so that results would be nationally representative.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
These limitations notwithstanding, the findings indicate that there are significant rural–urban differences in formal home-care use, and that Medicaid coverage may well play a role in these differences. The results suggest that, for older people in rural counties, Medicaid coverage may facilitate access to acute and chronic-care services, especially Medicare home health services. It has been suggested that Medicaid primary-care case-management programs are especially useful entry points in rural environments, because recipients benefit from access to clinicians, other individualized care, and referrals (Poley, Silberman, & Slifkin, 2003). Perhaps in nonmetropolitan counties the case-management services that are associated with Medicaid serve to link elders with formal home care or with other services, such as hospital stays, that generate subsequent formal home-care utilization reimbursed by various sources, including Medicare. Another possible explanation for the finding that Medicaid-covered individuals in nonmetropolitan counties have higher risks of home-care utilization is that rural seniors more readily become eligible for Medicaid by virtue of their heavy utilization of health care services and their limited overall financial circumstances. Although we controlled for income in the analyses and measured Medicaid coverage early in the target year, it is nevertheless possible that nonmetropolitan seniors more readily meet the income-based or medically needy criteria for Medicaid coverage than urban seniors as a result of the extent or level of care they require and their generally lower financial status. This potential explanation is supported to a degree by a separate analysis (results not shown) indicating that older nonmetropolitan MEPS-HC respondents with incomes below the poverty level are significantly more likely than their metropolitan counterparts to be covered by Medicaid.
Our results differ from studies examining fee-for-service Medicare beneficiaries of all ages in approximately the same time period (General Accounting Office, 2000; McCall et al., 2001) and most prior studies addressing rural–urban home care utilization patterns among Medicare enrollees (General Accounting Office, 1999; Kenney, 1993a, 1993b; Kenney & Dubay, 1992; McCall et al.), which have identified higher utilization rates in metropolitan than nonmetropolitan areas. The difference between our findings and those of other studies may be due to the fact that the MEPS-HC data set provides access to a generalizable sample of all Medicare beneficiaries aged 65 and older, rather than data from fee-for-service Medicare beneficiaries that, in some cases, may have included disabled individuals under the age of 65.
The results demonstrate that elderly residents in the most rural areas (nonmetropolitan counties with no town of 10,000) have higher risks for Medicare home care, whether or not they are covered by Medicaid. Within these very rural counties, Medicare home care may sometimes be a necessary substitution for other forms of health and long-term care that would normally be reimbursed through Medicare but that are less available, such as hospitals or advanced medical services (Dalton et al., 2002; Dansky & Dirani, 1998; Kenney, 1993a; Kenney & Dubay, 1992), or for sources of care, such as nursing homes, are very unevenly distributed across the most rural areas (McAuley et al., 2002). Because, by definition, Medicare home care is delivered in the client's home, it may be the only readily available care option for older people in these communities.
The data for this investigation were from 1998, the year after the passage of the Balanced Budget Act (BBA) of 1997 but before prospective payment for home care was implemented. Prior research suggests that the overlap of Medicare and Medicaid and other state programs gives states an opportunity to substitute Medicare for state home-care dollars (Cohen & Tumlinson, 1997). The recent BBA policy changes have led to reduced Medicare funding of home care. This reduction was partially offset by increases in Medicaid and state programs, although there was an overall decline in home-care services after the BBA (Spector et al., 2004). As their fiscal environment has deteriorated, states have experienced difficulty funding both Medicaid and state long-term-care programs (Kaiser Commission on Medicaid and the Uninsured, 2003). How these changes may affect the use of home care in rural areas is not known. To the extent that states continue to find ways to garner federal funds for home care, they may be able to cushion the effects of these cutbacks. It remains to be seen whether these changes will disproportionately affect rural families who are poorer than urban families and who may be using home care as a substitute for other services that may be in lower supply. However, if some older residents receive formal home care because other long-term and acute-care services are less available or so unevenly distributed that accessibility is limited, any state and federal policies and funding decisions that restrict the availability of formal home care may lead to new care-related vulnerabilities for this group.
Finally, although the county has become a standard geographical unit for the analysis of rural–urban differences in health care and social-service use, and long-term-care markets are also often studied at the county level, there would be advantages to using finer residential categories, such as census tracts or zip codes, in studying urban–rural differences in home care. There is a growing body of census-tract-level data that should permit refinements in the definition of residence as research in this area proceeds (Morrill, Cromartie, & Hart, 1999). In addition, new methods have been developed to define nursing home markets that are smaller than counties (Zwanziger, Mukamel, & Indridason, 2002), and these may be applied to better understand the dynamics of long-term-care decision making in rural and urban areas.
|
Footnotes |
|---|
1 Department of Health Administration and Policy, University of North Carolina, Charlotte.
2 Center for Delivery, Organization, and Markets, Agency for Healthcare Research and Quality, Rockville, MD.
3 Office of Rural Health Policy, Health Resources and Services Administration, Rockville, MD.
Decision Editor: Linda S. Noelker, PhD
Received for publication December 12, 2003. Accepted for publication April 7, 2004.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  F. J. Hellinger and W. E. Encinosa The Impact of State Laws Limiting Malpractice Damage Awards on Health Care Expenditures Am J Public Health, August 1, 2006; 96(8): 1375 - 1381. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
