| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
a Department of Occupational Therapy, University of Illinois at Chicago
Correspondence: Marcia Finlayson, PhD,OT(C),OTR/L, Department of Occupational Therapy (MC 811), University of Illinois at Chicago, 1919 West Taylor Street, Chicago, IL 60612. E-mail: marciaf{at}uic.edu.
Decision Editor: Laurence G. Branch, PhD
 |
Abstract |
|---|
 TOP Abstract Guiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
Key Words: Longitudinal study • Home care predictors • Nursing home predictors
Current demographic projections for Canada's older population suggest that approximately 22% of Canada's population will be more than 65 years of age by the year 2031, and that people more than 85 years old will make up almost 4% of the total population (Minister of Industry 1997
). These projections mean that increasing numbers of Canadians will be long-term survivors, that is, the "oldest-old" (Suzman, Willis, and Manton 1992). Although this demographic shift may be considered relatively benign in and of itself, it becomes of increasing concern when one realizes that these long-term survivors are the largest users of long-term care services (Havens 1996). Long-term care is care that is provided over a sustained period of time by trained caregivers, such as health and social service professionals, paraprofessionals and nonprofessionals, and volunteers working for organized health and social service agencies and programs (e.g., Meals on Wheels). Long-term care is provided through nursing homes and home care to individuals with physical, mental, or social disabilities and limitations (Havens 1996).
Although the proportion of the provincial health care budget spent on long-term care in Manitoba is relatively small, the growth of the older population is occurring simultaneously with dramatic changes in the way that health is conceptualized and the way that health care is being delivered. Together, the shift in the age structure, the disportionate use of long-term care services by the oldest-old, and the changes in the health care delivery system have created a great need to understand what leads people over the age of 85 to use long-term care services. Currently, little information is available to facilitate health-related planning for the oldest-old, because relatively little is known about this population. It is critical to understand what changes people experience as they survive past the age of 85 and what changes influence their need for and use of long-term care services. Information in these areas will increase the likelihood that appropriate decisions about health-related programs and policies for older people can be made.
To address these identified program and policy needs, and to address the gaps in the literature, the purpose of this study is to identify changes occurring between 1983 and 1990 that characterize and differentiate 1996 long-term care outcomes among a cohort of people aged 85 and older. These outcomes are nonuse of long-term care services, use of home care, and use of nursing home. The specific research question addressed in this study is: What health-related changes between 1983 and 1990 predict 1996 long-term care outcomes among people over the age of 85?
|
Guiding Conceptual Framework |
|---|
|
TOPAbstractGuiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
and expanded on by Evans, Barer, and Marmor 1994. This framework provides a comprehensive way to think about the various factors that influence persons' health and functioning outcomes, including their use or nonuse of long-term care services. The framework is described by its authors as a "comprehensive and flexible framework" that provides "meaningful categories in which to insert various sorts of evidence that are now emerging as to the diverse determinants of health" (Evans and Stoddard 1994, p. 32). These categories include health and functioning, health care, social environment, disease, prosperity, physical environment, genetic endowment, individual responses, and well-being. Each of the nine categories within the framework has its own internal structure and is viewed as being captured by multiple rather than single variables (Evans et al. 1994). Through the use of this framework, variables falling within these nine categories are explored as potential factors influencing long-term care use. In addition, these nine categories were used to organize a review of existing literature. |
Predictors of Formal Long-Term Care Service Use |
|---|
|
TOPAbstractGuiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
; Johnson and Wolinsky 1996; Shapiro and Tate 1997) and nursing home use (e.g., Branch and Jette 1982; Cohen, Tell, and Wallack 1986; Severson et al. 1994). The greater the extent of limitation, the greater is the likelihood of use of both of these types of long-term care services. None of the studies reviewed used changes in functioning over time as a potential predictor of home care use or of nursing home use.
Prior health care system use is also a relatively consistent predictor of both home care and nursing home use. For both types of long-term care, greater prior use of the health care system tends to be predictive. Prior hospital or nursing home stays (Coughlin et al. 1992; Johnson and Wolinsky 1996), greater numbers of available institutional beds (Coughlin et al. 1992), and greater numbers of physician visits (Wan 1987) predict home care use. Variables that fairly consistently predict nursing home use include use of respite services (Kosloski and Montgomery 1995), prior home care visits (Liu, Coughlin, and McBride 1991; Liu, McBride, and Coughlin 1994), prior hospital use or nursing home use, a recent hospitalization, and a greater nursing home bed supply (Coughlin, McBride, and Liu 1990).
Measures of the social environment are more varied and greater in number in the home care literature, compared with the nursing home literature. Household size (including living alone; e.g., Chappell 1985; Grabbe et al. 1995; Penning 1995) and number or type of informal helpers (including number of daughters, sons, neighbors, marital status, etc.; e.g., Johnson and Wolinsky 1996; Kemper 1992) have been studied as predictors of home care use. Although there are some minor discrepancies in the findings across studies, overall it appears that people who have numerically fewer members in their social network are at greater risk of using home care services than people with a larger social network. Among studies focusing on nursing home use, the social environment factors considered as potential predictors tend to focus on marital status and household size (Newman, Struyk, Wright, and Rice 1990; Shapiro and Tate 1985, Shapiro and Tate 1988). Being widowed and living alone have been relatively consistent predictors of nursing home use.
Disease-related variables have been considered as potential predictors of both home care and nursing home use, but have had inconsistent results in that they predicted one form of long-term care, but not the other. Variables capturing cognitive impairment, lung disease, neurological conditions, and the number of health conditions were almost universally predictive of nursing home use among the studies that included them (e.g., Branch and Jette 1982; Coughlin et al. 1990; Kosloski and Montgomery 1995; Liu et al. 1991, Liu et al. 1994; Newman et al. 1990; Severson et al. 1994; Shapiro and Tate 1985, Shapiro and Tate 1988; Temkin-Greener and Meiners 1995). In comparison, disease-related variables in the home care literature have tended to focus on dementia (e.g., Coughlin et al. 1992; Grabbe et al. 1995; Johnson and Wolinsky 1996; Kemper 1992; Penning 1995; Shapiro and Tate 1997; Wan 1987). Findings have been inconsistent, with results suggesting that it is not dementia per se that predicts home care use, but the behaviors associated with dementia instead.
In both the home care and nursing home literatures, variables measuring aspects of prosperity (such as poverty status, source of income, income amount, home equity, and Medicaid eligibility) have been used as potential predictors. The results have been inconsistent. A number of factors may be contributing to these inconsistencies (e.g., cost of care, source of payment for care, and the overall organization of care services). It is currently unclear the role that these factors play in the use of long-term care.
Variables relating to the physical environment have also been inconsistent in their ability to predict long-term care. In the home care literature, the only variables that have been considered include living in an apartment (Chappell 1985) and length of time living in the community or stability of residence (Kemper 1992; Shapiro 1986). Physical environment variables considered as potential predictors of nursing home use have included climate (Newman et al. 1990), home modifications (Newman et al. 1990), and type of housing (Shapiro and Tate 1985, Shapiro and Tate 1988). Even with this diversity, only living in seniors' housing was found to predict entry into a nursing home (Shapiro and Tate 1985).
Among studies focusing on home care use, only two (Coulton and Frost 1982; Shapiro 1986) measured the predictor variables before measuring the outcome (home care use). In all of the other home care studies, the independent and dependent variables were measured simultaneously (i.e., cross-sectional studies). Among the studies exploring predictors of nursing home use, the majority predicted nursing home use by measuring predictor (independent) variables at some baseline point and by measuring the outcome variable (nursing home use) at some later point. Four of these studies measured the outcome at more than one time (Branch and Ku 1989; Severson et al. 1994; Shapiro and Tate 1985, Shapiro and Tate 1988); however, the predictor variables for each of these four studies were measured only at the baseline.
In summary, previous studies of home care or nursing home use have not focused on changes over time in health and functioning, health care, social environment, disease, prosperity, physical environment, individual responses, or well-being as potential predictors of long-term care use. Instead, variables within these categories have been considered at one time, either within a cross-sectional study (typical in home care research) or within studies that investigate whether baseline status influences a later outcome (typical in nursing home research).
|
Methods |
|---|
|
TOPAbstractGuiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
). This program operates a single-entry system that provides access to all of the services within Manitoba's long-term care system (e.g., home care, nursing home care, respite, adult day care, etc.) through the same assessment process. Detailed descriptions of the system have been provided by Havens 1990 and others. As of June 1, 1996, there were 12,058 home care clients aged 65 years and over in Manitoba, representing approximately 8% of the older adult population (Manitoba Health 1998a). As of March 31, 1997, there were 120 nursing homes in the province providing a total of 8,953 licensed nursing home beds (Manitoba Health 1998b).
The AIM Study is a population-based, longitudinal panel study that includes three panels of participants who entered the study at three different times and then were followed over time. The entry points for the study were 1971, 1976, and 1983, with interviews being conducted in 1971, 1976, 1983, 1990, and 1996. Graphic depictions of the study design have been published by Chipperfield, Havens, and Doig 1997 and Finlayson and Havens 2001. The sixth wave of interviewing took place in the summer of 2001.
All participants in the AIM Study were randomly selected from the computerized records of Manitoba's universal health insurance system using an age and gender-stratified, area-probability sampling technique. These samples have been shown to compare favorably to both Manitoba's older population and to the older population of Canada as a whole (Chipperfield et al. 1997). The rate of nonresponse in each of the follow-up years has ranged from 4.6% in 1983 to 5.2% in 1996 (Hall and Havens 1997). Details regarding the study teams and their training have been described by Hall and Havens 1997.
All of the interview waves of AIM Study collected data in the areas of demographics, social networks and supports, leisure activities, ADLs, IADLs, use of services, income and expenses, and chronic illnesses. The core questions on the AIM Study interview guide have remained consistent throughout the course of the study, thereby allowing comparisons within as well as between participants over time. Chipperfield and colleagues 1997, Finlayson and Havens 2001, and others have published descriptions of the interview guide contents and the measurement properties of survey items.
Sample
All surviving AIM Study participants who were born in 1911 or earlier, and who had complete interview data from 1983, 1990, and 1996, composed the sample for this study. Only data from these three interview points were used for the study, regardless of when the participant joined the study. Six hundred and sixteen AIM Study participants met these criteria. Eligible individuals were slightly younger, viewed their health more positively, and were more likely to be living in a house in 1983, compared with other AIM Study participants who participated in the 1983 interviews, but who did not survive to be eligible for this study (Table 1 ).
|
Independent Variables
To select the variables for use in the current study, the interview guides from 1983 and 1990 were reviewed. All questions with 8% or less missing data that appeared in both interview guides and that corresponded to one of the components of the Evans and Stoddard framework were used to create 1983 to 1990 change variables using the concatenation function ("CONCAT") of SPSS. Concatenation captures the existence and direction of change, and may have the potential to capture magnitude of change, depending on the measure being used (Wolinsky, Armbrecht, and Wyrwich 2000). By concatenating the values of a series of variables across study years (e.g., dressing in 1983, dressing in 1990), the researcher is able to create a new, multicategory, nominal variable that retains the individual values of each of the variables that is used to create it. For example, if a subject reported that he/she was independent in dressing in 1983 (coded as 1), but required home care assistance with this task in 1990 (coded as 4), the concatenated variable representing change in dressing would be coded as 14. Because the initial concatenation process produced numerous categories of change for some variables (e.g., if there were 5 response categories, up to 25 patterns of change were possible; i.e., 5 x 5), the categories were collapsed for ease of analysis. A full list of the change variables created for this study is provided in Table 2 .
|
Because of the large number of potential predictors, the first stage of the modeling process was to model the variables within each of the nine categories of the theoretical framework (e.g., health and functioning, social environment, physical environment, etc.) against the outcome pair of interest. In each of these models, age and sex were entered into the model, and a backward stepwise process was used for the category-specific variables. The results of these within-category models are summarized in Table 3 . The variables listed here were used for the full predictive models (e.g., no services vs. home care; home care vs. nursing home; no services vs. nursing home). To run the final analytic models, age and sex were entered as a block, and then the variables identified in Table 3 for the outcome pair of interest were entered as a block and selected through a backward stepwise process. For all models, the criteria for removal was p = .10.
|
|
Results |
|---|
|
TOPAbstractGuiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
.
Model 1: Nonuse of Services Versus Home Care Use
Column 2 of Table 3 shows the independent variables that were significant in each of nine conceptual categories for the nonuse of services versus use of home care model. None of the individual response variables (e.g., participant education level) were significant, and therefore only 8 of the 9 components of the framework were included in the final model. As Table 4 shows, 4 of the 16 independent variables entered into the model were able to predict home care use relative to using no services: change in self-rated health (stable at fair, poor, or bad between 1983 and 1990), change in service use (no services in 1983 to using home care in 1990), change in current income adequacy (decline from 1983 to 1990), and change in railings outside the home (stable with maximum support or decline in support between 1983 and 1990).
|
2 = 109.26, df = 18, p > .001).
Model 2: Nonuse of Services Versus Nursing Home Use
Column 3 of Table 3 shows the independent variables that were significant in each of the conceptual categories for the nonuse of services versus use of nursing home care model. All nine of the components from the framework had at least one variable that was significant for inclusion in the final model. As Table 5 shows, 6 of the 15 independent variables entered into the model were able to predict nursing home use relative to using no services: age (younger age reduced the risk), change in service use (no services in 1983 to using home care in 1990), change in current income adequacy (decline between 1983 and 1990), short time in the community (3–5 years), change in interviewer-observed state of mind (decline between 1983 and 1990), and change in type of housing (remaining stable in an apartment or condominium or moving to a less supportive environment between 1983 and 1990). The interviewer-observed state of mind variable reflects how the interviewer perceived the participant's ability to follow and attend to the interview throughout its duration. As a result, this variable captures the interviewer's combined observations of concentration, attention, and mental fatigue.
|
2 = 174.16, df = 16, p > .001).
Model 3: Home Care Versus Nursing Home
Column 4 of Table 3 shows the independent variables that were significant in each of the conceptual categories for the home care use versus nursing home use model. None of the prosperity variables were significant, and therefore only 8 of the 9 components of the framework were included in the final model. As Table 6 shows, only 2 of the 11 predictors that entered into the model were able to predict nursing home use relative to home care use: age (younger age reduced risk) and changes in general life satisfaction (decline between 1983 and 1990). The two significant variables represent the genetic endowment and well-being components of the framework. Together, these variables were able to correctly classify 69.8% of cases in this model (2 = 52.39, df = 6, p > .001).
|
|
Discussion |
|---|
|
TOPAbstractGuiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
; Liu et al. 1991, Liu et al. 1994), or predictors over different lengths of time (e.g., Branch and Ku 1989; Shapiro and Tate 1985, Shapiro and Tate 1988).
Results indicated that different predictors were significant across the three comparisons, although some overlap did exist. Age was significant in predicting nursing home use, both relative to using no services as well as to using home care services. In both models, being younger reduced the risk of nursing home use. This finding is consistent with findings presented in previous literature (Branch and Ku 1989; Liu et al. 1991; Temkin-Greener and Meiners 1995). For the two models that examined using no services versus using long-term care services (either home care or nursing home), two variables overlapped. These included a change in service use between 1983 and 1990, specifically changing from using no services to using home care, and experiencing a decline in current income adequacy between 1983 and 1990. Neither of these variables have specifically been used in similar previous studies, but they do capture similar concepts that have been predictive in the past, specifically previous health care service use (Johnson and Wolinsky 1996; Kemper 1992; Shapiro and Tate 1985, Shapiro and Tate 1988; Temkin-Greener and Meiners 1995) and lower socioeconomic status (Branch and Ku 1989).
Finding that a decline in current income adequacy predicts the use of both home care and nursing home care relative to using no services raises two important points for discussion. First, this study was done in a country that has universal health insurance and in a province that has provided subsidized long-term care services since 1973. Therefore, a drop in income adequacy, in and of itself, is not an easily explainable predictor of long-term care use. Second, self-reported income adequacy is not a commonly used measure in studies of older adults, but its ability to independently predict long-term care use in this study raises a question about its potential utility for future work. Many research subjects are reluctant to provide annual or monthly income information, and this has been true of the AIM Longitudinal Study participants. Nevertheless, these same individuals have been consistently willing to rate the adequacy of their current and future income, with the proportion of missing data (e.g., refusals) under 8% in most study years. Given the current interest in the role of socioeconomic factors in health outcomes, and the frequent difficulty obtaining income data, further investigation of the meaning and utility of self-reported income adequacy among older adults, and its relationship to health service use, may be warranted.
The primary inconsistency between the findings presented in this study and in previous work is the lack of significance in the measures related to ADLs and IADLs. Having these variables drop out of the models in this study may be an artifact of including only survivors in the sample, or it may be the result of looking at changes over time rather than simply baseline status. The finding may also reflect ongoing efforts to maintain frail elderly individuals in the community as long as possible, thereby creating a situation in which users of home care and users of nursing home are similar in their functional status profiles. A final alternative explanation is that the ADL and IADL measures available in this study may have a ceiling effect, making it not possible to detect subtle changes in ADL and IADL status at the upper end of the functional spectrum (i.e., declines were happening, but not being captured by the scale).
The findings of the study show that the factors that lead to home care use are different than the factors that lead to nursing home use. This is an important finding for those individuals who are responsible for preparing the long-term care system to meet future needs. In terms of developing policy, the findings of this study show that the three outcomes used here (i.e., home without home care, home with home care, nursing home) are not clearly "linear" in nature. Had the outcomes been clearly ordered, the factors predicting home care and nursing home would have been the same, and the odds ratios for the various predictive factors would have been greater for nursing homes than for home care use, both relative to being at home without home care. This means that home care and nursing home care are addressing different types of care needs, not the same needs in different degrees. Although future cohorts of the oldest-old are likely to experience long-term care in a different policy environment than the participants in this study, this lack of ordering in the outcomes would be expected to remain a relevant point for discussion among policy makers unless the basic nature of long-term care services and their eligibility are altered.
The limitations of this study include the lengths of the follow-up periods (7 years and 6 years, respectively) and the inclusion of only survivors. Nevertheless, the results showed that changes during these periods showed important associations with the outcomes under consideration. This adds new knowledge to the field, because none of the studies reviewed used changes over time as a potential predictor for use of formal long-term care. Instead, the temporal aspect of these studies was obtained by testing whether baseline measures were associated with outcomes at a later point. Finding that changes captured by concatenating responses across interview waves, and over relatively long follow-up periods, are associated with long-term care use is particularly important for future research. Concatenation is a simple procedure in SPSS and provides the researcher with a broad perspective on the amount of heterogeneity in the sample.
By using the categories of the determinants of health framework to organize the background knowledge for this study, and to organize the analytic process, it became apparent that there are important gaps in knowledge regarding the predictors of formal long-term care, regardless of whether studies are cross-sectional or longitudinal in nature. Currently, very little is known about the role of the physical environment, individual responses, well-being, and genetic endowment on formal long-term care use, regardless of the research design or sample population. Although the current study has made a small step in minimizing this knowledge gap, additional knowledge in these areas needs to be developed.
This research should leave readers with two primary messages. First, although long-term care is rightly considered a continuum of care, different factors played a role in predicting the use of home care than predicted the use of a nursing home. Second, the application of a different conceptual framework than has been used traditionally in this body of research provided new insights into the gaps in existing knowledge, permitted the exploration of systematic differences across the three outcomes, and generally provided a different perspective on the complexity of predicting formal long-term care use.
The Forum
Book Reviews
Practice Concepts
|
Acknowledgments |
|---|
Received for publication July 11, 2001. Accepted for publication December 18, 2001.
|
References |
|---|
|
TOPAbstractGuiding Conceptual FrameworkPredictors of Formal Long-Term...MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. A. Sargent-Cox, K. J. Anstey, and M. A. Luszcz Determinants of Self-Rated Health Items With Different Points of Reference: Implications for Health Measurement of Older Adults J Aging Health, September 1, 2008; 20(6): 739 - 761. [Abstract] [PDF]
|
|
|
|
 |
|
 J. Ness, A. Ahmed, and W. S. Aronow Demographics and Payment Characteristics of Nursing Home Residents in the United States: A 23-Year Trend J. Gerontol. A Biol. Sci. Med. Sci., November 1, 2004; 59(11): 1213 - 1217. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
