| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Jane N. Bolin, PhD, Department of Health Policy and Management, School of Rural Public Health, The Texas A&M University System Health Sciences Center, College Station, TX 77843-1266. E-mail: jbolin{at}srph.tamhsc.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Rural • Urban • Nursing home • Medicare • Case mix • Minimum Data Set
The nursing home industry also varies across locations differing in their rurality. Individuals in more rural areas are more likely than their metropolitan counterparts to be served by nursing homes operated by private not-for-profits or by governmental bodies, such as county-run nursing homes (Phillips et al., 2003). Rural homes are more likely to be smaller, receive fewer citations for severe deficiencies from state surveyors, and, somewhat paradoxically, be more poorly staffed than homes in metropolitan areas (Phillips et al.).
These data give us a relatively clear picture of the differences between more urban and rural areas in the structure of the nursing home industry. Much less is known about the differences between residents entering urban and rural nursing homes and receiving care in these different settings. This lack of comprehensive knowledge makes it difficult to consider policy options for older persons and for the long-term-care service structure in rural areas.
Some research is available to help us address these issues. Previous research has examined differences in urban and rural nursing home populations by examining differences in newly admitted residents. Greene's (1984) analysis of nursing home admission data from Arizona indicated that those individuals admitted to rural nursing homes were younger and less impaired than residents in metropolitan nursing homes. Duncan and colleagues (1997) largely replicated Greene's analysis with data from Florida, but they found a more complex picture of the differences between residents admitted to nursing homes in different locales. They found that the differences were relatively minor and in some instances implied higher acuity or care needs among those admitted to rural nursing homes. More recently, Penrod (2001) examined the association between rural residence and availability of nursing home and home health care to level of functional disability at the time of nursing home admission. Penrod's findings suggested that a person's level of functional disability at admission is more associated with specific diseases and medical conditions, as well as cognitive status, gender, and marital status, rather than urban or rural locale. However, like the studies of Greene and Duncan and associates, Penrod's study was limited to a single state; thus, sample size and restricted study populations limited the study's utility. Collectively, the studies present a conflicting picture of differences in resident case mix between rural and metropolitan nursing homes.
In this research we attempt to elaborate on the prior work by investigating the differences in admissions to urban and rural nursing homes, using a nationally representative sample of nursing home admissions. In addition, we investigate whether acuity differences between residents in urban and rural nursing homes simply reflect differences in the facilities' Medicare censuses. Prior studies have found that higher acuity residents were more often admitted with Medicare as a payer, particularly in swing-bed, for-profit, and larger nursing facilities (Bishop & Dubay, 1991; Holmes, 1996; McKay, 1991). Rural nursing homes, however, have significantly fewer Medicare admissions than do homes in metropolitan areas (Phillips et al., 2003). This payer-mix difference may account for some substantial proportion of the observed differences found in some prior studies in acuity among individuals admitted to nursing homes in urban and rural areas, because Medicare coverage of extended care services in skilled nursing facilities requires that a resident's condition be treated during a qualifying hospital stay or be a condition for which a resident was previously treated in a hospital (see Medicare Relief Act). Thus, it is expected that residents with Medicare as primary payer will require more complex nursing care or intensive rehabilitation and have higher acuity scores.
Determining whether there were acuity differences between residents admitted to rural and metropolitan nursing homes and whether any differences are associated with payer type will help clarify several policy and programmatic issues. Prior studies (Duncan et al, 1997; Greene, 1984; Penrod, 2001; Phillips, Holan, Sherman, Leyk Williams, & Hawes, 2004) that found lower case mix or acuity among rural nursing home residents have suggested a variety of potential explanations. These include lack of access to a range of long-term-care services that might provide an alternative to nursing homes; greater likelihood that rural elders aged 75 and older live alone; and higher levels of poverty and near-poverty among older persons living in rural areas, which makes them more dependent on Medicaid to help pay for long-term care. Because Medicaid expenditures have been more heavily slanted toward nursing home care, this greater reliance among rural elders could lead to higher use of nursing homes and thus lower acuity. Our hypothesis, however, suggests less use of Medicare-covered postacute or subacute care in rural nursing homes may account for any observed differences in resident acuity between rural and metropolitan residents, and that this may be due to the tendency of rural elders to be admitted from home or under circumstances disqualifying them from Medicare-covered services in skilled nursing facilities. Each of these explanations suggests a different problem and remedy. Thus, an analysis of one potential explanation using national data—not subject to potential state-only variations—can increase our understanding of any observed differences in acuity of new residents in our sample as a whole.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
We have chosen here to use only admission assessments. That choice follows the tradition of earlier research in this vein (Duncan et al., 1997; Greene, 1984). This approach also ensures that resident characteristics are not confounded with facility performance or activity or resident length of stay. By examining case-mix or acuity measures for an admission cohort, we avoid any changes in status that have occurred as a result of the home's care of the residents.
We constructed the analysis sample by randomly choosing 10% of the admissions for each nursing home in the MDS database. The sample is large enough (N = 197,589) that the likelihood of Type II error is minimized, but the processing time required is reduced dramatically from that for maximum likelihood modeling using the entire population.
Measurement
Rurality
We used rural–urban commuting areas codes to classify facilities, according to their zip code, into one of two categories of communities—urban–large town or small town–isolated (Morrill, Cromartie, & Hart, 1999). Urban areas refer to zip codes in an urban core area with a population greater than or equal to 50,000 or zip codes in which a significant proportion of the population commutes into such an urban center. Those areas defined as large town refer to zip codes in an urban place with a population between 10,000 and 49,999, and those surrounding zip codes where a substantial proportion of the population commutes into the large town. Those areas defined as small town refer to zip codes in a place with a population between 2,500 and 9,999 persons or an area in which a relatively large proportion of the population commutes into the small town. Isolated areas are the remaining rural areas that lack substantial commuting to urban centers, large towns, or small towns. In multivariate analyses, we combined isolated and small towns into one category; we combined large town and urban into another category. These two categories serve as the binary dependent variable for the logistic regressions.
Resident Characteristics
The characteristics we used to compare residents on admission to nursing homes located in different types of communities fall roughly into four categories: demographic characteristics, resident location prior to admission, functional status (physical and cognitive), and special care needs. We gave functional status and special care needs special attention in the analysis because they reflect resident acuity. We drew all of these data from MDS admission assessments.
We characterized resident payer status on the basis of the primary payer for care as designated on the MDS Section A8, Reasons for Assessment, and divided it into Medicare or non-Medicare based on whether a particular assessment was a Medicare-required assessment versus state-required or other required assessment. (A resident who is dually eligible for Medicare and Medicaid will show both as a payer on the MDS. However, we classified these individuals as Medicare, because that was the primary payer for care for that stay.) Non-Medicare payers include Medicaid, other government payers (such as the Department of Veterans Affairs), private insurance, and private-pay or out-of-pocket spending by individuals. Medicare, because of its coverage rules for nursing home care, is restricted to those residents who will essentially receive what might be termed postacute or subacute care, that is, skilled nursing care or intensive rehabilitation. As a result of these coverage rules, residents who have Medicare as a primary payer typically have short stays in the nursing home or, if a longer stay, convert to another payer status. Nevertheless, on admission, they will have higher acuity scores as measured by receipt of special care and therapies than will the average long-stay resident whose care is covered by another payer.
The demographic characteristics in the analyses included gender, race or ethnicity, martial status, and age. We derived location prior to admission from the MDS Admission Assessment, Section AB(2), and it reflects information about the setting from which the resident was admitted (i.e., a hospital, a psychiatric facility, an assisted living facility, home with no home-health services, home with home-health services, or another nursing home).
We measured resident care needs and functional status by a number of MDS-based scales. We measured residents' cognitive status by using the Cognitive Performance Scale (CPS), which includes six levels ranging from 0 (no impairment) to 6 (very severe impairment; see Morris et al., 1994). We measured residents' functioning in activities of daily living (ADLs) by using the ADL Hierarchy, which includes seven levels ranging from independent (0) to severely impaired (6) in late loss of ADL status (Morris, Fries, & Morris, 1999). We measured residents' behavioral status with a dichotomous variable, "0/1," reflecting whether they had or had not engaged in at least one disruptive or problem behavior (e.g., 1 = yes if documented wandering, resisting ADL assistance, or physical aggression) during the 7 days preceding the assessment. The MDS–Changes in Health, End-Stage Disease and Symptoms and Signs (MDS-CHESS) Scale measures residents' clinical instability and likelihood of death (0 = no symptoms to 5 = end stage; see Hirdes, Frijters, & Teare, 2003). In the logistic model, we also included the CPS and indicators of behavioral needs.
Analysis
Statistical analyses began with the estimation of percentages and means of key variables in the sample. These variables include marital status, race or ethnicity, age, prior living arrangements, resident behavior, and needs and function scales. Next, we conducted a multivariate logistic regression (Hosmer & Lemeshow, 2000), examining differences between admissions to rural–small town nursing homes and urban–large town nursing homes.
We elected a nontraditional use of the logistic regression model, using rural as the dependant variable with patient-level variables and using case-mix indicators as independent variables. The odds ratios (ORs) provide the likelihood that a given patient would be more or less likely to reside in a rural (vs urban) nursing-care facility given his or her unique characteristics. This does not imply that the characteristics that differentiated between groups in different locales were the cause of the patient's admission to a rural rather than a metropolitan nursing home. Instead, it simply indicates that a statistically significant difference existed in the prevalence of this characteristic among the resident populations in the two locales.
Because of the large sample size, relatively small effects can be statistically significant. To focus attention on effects that may be of greater substantive importance, we restrict the discussion of the logistic regression results to variables with statistically significant relative ORs that were equal to or below 0.90 or equal to or above 1.10. We adjusted variance estimates for all parameters throughout the analyses to account for the clustering of residents in nursing homes.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
We noted more striking differences in location prior to admission between residents with Medicare–non-Medicare payer status and across urban–rural geographic locations. Non-Medicare residents were more likely than Medicare residents to have been admitted from home, but the probability was significantly higher that, prior to admission, they were not receiving formal home-care services (e.g., nursing care, occupational, physical, or speech therapy, and home-health-aide services). Both married and nonmarried residents were more likely to be admitted from home when home-health services were not be used than when these services were in use immediately prior to admission (13% vs 5% in urban areas, 32% vs 13% in isolated areas; p <.01). The use of formal home-health services, whether the resident was married or not, appears to be a significant difference among individuals admitted with a payer other than Medicare (i.e., usually for chronic care or long stays) across all geographic locations—but especially in rural areas. For example, in isolated areas and small towns, residents admitted to a nursing home with a non-Medicare payer were two to three times less likely to have been receiving home-care services prior to admisison.
The majority of new nursing home residents across geographic locations and payer type were from acute-care hospitals. The percentage of nursing home admissions coming from hospitals was 20% to 30% higher for Medicare patients than it was for non-Medicare patients, across all locations. However, the percentage of admissions from acute-care hospitals for non-Medicare patients was significantly higher in urban areas than it was in small towns and isolated areas (i.e., 32% for isolated rural areas vs 57% major metropolitan areas; p <.01).
Comparisons of scales summarizing resident care needs and functional levels are provided in Table 3.
|
CHESS Scores
The mean of MDS-CHESS scores for those admitted under Medicare at admission was 1.96 (SD = 1.1), compared with 1.4 (SD = 1.2) for individuals admitted for chronic care. Generally, a higher CHESS score indicates a stronger relationship to overall mortality and a variety of medical treatments (Hirdes et al., 2003).
Cognitive Performance Scale
Overall, persons admitted for chronic care were more cognitively impaired than those admitted under Medicares (CPS = 2.08 ± 1.8, vs 1.67 for Medicare; see Morris et al., 1994).
ADL Hierarchy
Assessment of ability to perform ADL measures showed that the mean impairment levels for individuals admitted for short stays or for chronic care were relatively similar, though slightly higher for short-stay residents. The mean for non-Medicare was 3.12 ± 1.8, and the mean for Medicare was 3.59 ± 1.6 (Morris et al., 1999).
Logistic Regression Analyses
The multivariate model provides the relative odds of a rural–small town admission as a function of the covariates. We recognize that the direction of the model is nontraditional in that rurality is normally a predictor variable. However, given the flexibility offered by logistic regression in predicting discrete group membership from a set of variables, we considered this approach optimal for revealing differences in needs and treatment characteristics for new admissions in urban and rural areas. Using the entire sample of new admissions, we observed the relative odds of key characteristics and patient care needs. Demographically, new residents in rural areas were less likely to be Black (OR = 0.50), Asian (OR = 0.15), or Hispanic (OR = 0.31). In contrast, Native Americans were significantly more likely to be admitted to a rural facility than an urban facility (OR = 2.27). Patients admitted to any rural nursing facility were more likely to be admitted from home or from another nursing facility (OR = 1.22) than urban patients. Rural nursing home patients were significantly less likely to be admitted from an acute care facility (OR = 0.73) or from a rehabilitation facility (OR = 0.89).
New residents in rural facilities were much more likely to engage in some problem behaviors (be verbally abusive, disruptive, or resist care), but they were less likely than new residents in urban nursing homes to engage in physical abuse. The nursing case-mix index scores were lower for patients admitted to rural facilities (OR = 0.81). Generally, the MDS-CHESS, CPS, and ADL hierarchy scores did not differ dramatically between those entering rural and urban nursing homes. Mean scores on these scales did not differ by a great deal and the ORs for the variables were marginal (OR > 0.9 or OR < 1.1).
The results in Table 4, however, combine newly admitted residents for both Medicare and non-Medicare. Therefore, any differences observed in acuity could result from differences in rates of Medicare admissions, rather than the characteristics of all admissions. In order to understand the degree to which differences in the characteristics of nursing home residents derive from different levels of dependence on differing funding source, separate analyses must be done for residents whose stays were funded by different payers. Table 5 presents the results of the multivariate analysis separately for those admitted under Medicare, typically shorter stays, and those admitted under non-Medicare payer status, which is more typically chronic care. The basic question for these models was whether the same patterns of acuity differences would be seen for individuals admitted for short stays or chronic care.
|
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This research, on a national sample of newly admitted nursing home residents, supports the general conclusion that rural facilities have residents with acuity scores lower than those of individuals admitted to urban nursing homes. Persons admitted to rural and isolated homes are more likely to be clinically stable with a lower likelihood of death (CHESS; Hirdes et al., 2003), and have less ADL impairment than do persons admitted to urban facilities. However, rural facilities were somewhat more likely to have a newly admitted resident who was cognitively impaired and more likely to engage in problem behaviors than those admitted to more urban facilities.
One important issue to which the answer has been unclear historically is whether the observed lower acuity reflects a differing mix of postacute and chronic-care admissions between urban and rural homes or results the situation in which rural homes simply admit less impaired individuals in general. In essence, rural facilities could admit individuals covered by Medicare who are just as impaired as those admitted to urban facilities. They also could admit individuals whose stays are not covered by Medicare who are just as impaired as such individuals admitted in urban areas. The overall difference in acuity would then simply be the result of the fact that rural facilities have fewer Medicare postacute admissions. Alternatively, rural facilities could just admit individuals who are less impaired. The results in Table 5 offer an answer to this question.
When we examine newly admitted Medicare residents across the rural–urban facilities, it seems that rural Medicare patients are just as impaired as Medicare patients admitted in urban areas. They have slightly higher levels of cognitive impairment, slightly higher mortality risk, but a slightly lower ADL hierarchy score. All in all, that is a picture of similar acuity. For other types of admissions, acuity status is more diverse. For non-Medicare residents, we see greater cognitive impairment in rural nursing homes, but less frailty and ADL needs, suggesting lower acuity levels. These results indicate that the lower general acuity levels in more rural nursing homes are a function of both lower percentage of patients admitted with Medicare as the primary payer and different admission patterns for those admitted for chronic or long-term care.
The genesis of this pattern of admission for chronic care in rural nursing homes remains, at this point, somewhat obscure. Previous research offers a variety of hints concerning those factors that might generate these results. Rogers (1997) found that rural elderly people are more likely than other elderly people to be living alone by age 75, when the likelihood of a nursing home admission increases. Rural elderly individuals are also more likely than urban elderly individuals to be poor, making it more difficult to purchase services that might extend their time at home (Coburn & Bolda, 1999). Some research also indicates that individuals admitted to nursing homes in rural areas indicate more frequently than their urban counterparts that needed services were not available to them in their communities (Coward et al., 1994).
In addition, though we face something of a "chicken and egg" issue, the supply of nursing home beds in rural areas may be a factor. The earlier research by Greene and Ondrich (1990) from the National Long Term Care Channeling Demonstration found that new nursing home residents were more likely in areas with the larger supply of nursing home beds. It is in more rural areas that we find a much higher per capita nursing home bed supply and much higher per capita rates of nursing home utilization (Phillips et al., 2003). Any of the factors noted herein might, by operating in isolation or in concert, result in the pattern of admissions observed in these data.
Implications for Practice and Policy
Research has fairly consistently demonstrated that residents of rural nursing homes tend to be less impaired than residents of nursing homes in metropolitan areas. However, there has been no clear explanation for this phenomenon.
Using a 10% sample of all nursing home admissions in the nation, we found in our research that residents admitted to rural nursing homes differed in their acuity from residents admitted in urban homes. We found that part of the explanation was the lower rate of Medicare coverage for nursing home care in rural areas. This suggests the need for researchers and policy makers to determine whether this represents an access problem or reflects the rural elders' preferences or patterns of acute hospital use.
We also found, however, that the difference in Medicare coverage was only a partial explanation for the observed acuity differences. For non-Medicare residents, acuity was significantly lower in more rural homes. With that relationship now firmly established on a national level, researchers should turn to the development of studies aimed at determining the etiology of this relationship. Each of the previously suggested reasons—paucity of home-health-care services, scarcity of assisted living in rural areas, lack of Medicaid coverage for home- and community-based care, and individual preference—has different implications for policy makers who wish to reduce the use of nursing home care or offer a range of long-term-care services that more closely approximates the needs and preferences of rural elders. However, our research suggests that such future research would benefit from a closer examination of differences in types of nursing home residents, by payer type (Medicare and non-Medicare), along a continuum of rurality in order to better understand differences in the use of long-term care.
This study, though suggestive, is not without its limitations. Though national in scope, the observed relationship may not hold in some specific environments. In addition, these data cover only 1 year of nursing home admissions, though there is little reason to consider this year aberrant. Finally, we used a variety of measures of acuity and a nontraditional criterion for considering a relationship significant. Researchers using other acuity measures and some other criterion for the importance of relationships might reach different conclusions.
|
Footnotes |
|---|
1 School of Rural Public Health, Texas A&M University, College Station.
Decision Editor: Linda S. Noelker, PhD
Received for publication August 25, 2004. Accepted for publication August 23, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
404.30 et seq.This article has been cited by other articles:
 |
|
 |
|
  A. Gruneir, S. C. Miller, O. Intrator, and V. Mor Hospitalization of Nursing Home Residents With Cognitive Impairments: The Influence of Organizational Features and State Policies Gerontologist, August 1, 2007; 47(4): 447 - 456. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
