| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
a Health Analysis and Modelling Group, Statistics Canada, Ottawa, Ontario
b Manitoba Centre for Health Policy and Evaluation, Canada.
Jean-Marie Berthelot, Social & Economic Studies Division, Statistics Canada, R. H. Coats Building, 24th Floor, Tunney's Pasture, Ottawa, Ontario K1A 0T6, Canada E-mail: jean-marie.berthelot{at}statcan.ca.
Decision Editor: William B. Ershler, MD
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Proportional hazard models, incorporating changes in needs over time, are used to estimate the hazard of nursing-home entry over a 5-year period, using health and sociodemographic characteristics of a representative sample of elderly residents from Manitoba, Canada.
Results. After age, need factors have the greatest impact on nursing-home entry. Specific medical conditions have at least as great a contribution as functional limitations. The presence of a spouse significantly reduces the hazard of entry for males only.
Conclusions. The results suggest that the greatest gains in preventing or delaying nursing-home entry can be achieved through intervention programs targeted at specific medical conditions such as Alzheimer's disease, musculoskeletal disorders, and stroke.
AS the population ages, a greater demand for long-term care services is expected. Nursing homes have traditionally been the most commonly used form of long-term care. Policy analysts continue to search for alternative, less costly forms of care for older adults, and much research attention has been devoted to developing risk profiles for nursing-home admission (1)(2)(3). Risk profiles are useful for projecting future demands for nursing-home care and for developing and targeting programs to delay or prevent nursing-home entry. This study contributes to the ongoing research by assessing the relative importance of health and socio–demographic factors in predicting and understanding nursing-home entry for a representative sample of elderly residents from the Canadian province of Manitoba.
Many of the previous studies examining factors related to nursing-home entry (4)(5)(6) have used Andersen's conceptual framework, which considers the use of health services to be a function of an individual's predisposing, enabling, and need characteristics (7). Predisposing factors include demographics, social structure, and health beliefs. Enabling factors are those influencing an individual's ability to gain access to health services and they include family and community resources. Need factors refer to the functional and health problems that generate the need for health care services.
Studies have consistently shown that age, functional status and mental status are the best predictors of nursing-home entry (8)(9)(10). Inconsistent results have been observed for "enabling factors," with some studies showing a reduction in the hazard of nursing-home entry for individuals reporting certain types of income, such as private pension income or rental income (11), whereas other studies have not found any significant association between income and nursing-home entry (4)(6)(8)(12). An inverse relationship between income and risk of nursing-home placement has been observed for a variety of income measures, including monthly household income (13), and for private pay versus Medicaid patients (14). More consistent results have been observed for factors reflecting market conditions such as the availability of nursing-home beds, which has been shown to increase the likelihood of nursing-home entry (4)(12)(13)(15).
A small number of studies have focused on the role played by social networks in reducing the likelihood of institutionalization, arguing that interventions cannot be made to change a person's age, marital status, or health status, but the ability of families, friends, neighbors and other community-based caregivers to provide support can be strengthened (16). The size and composition of an individual's social support network (9)(12)(17) and caregiver burden (18)(19) have all been shown to be important factors influencing nursing-home entry.
Compared with the functional disability measures, which are strong predictors of nursing-home entry, the other "need factors" that reflect the presence of specific medical conditions have been shown to have either a less important influence on nursing-home entry (11)(12)(15)(20) or, in some cases, no impact at all (6).
The inconsistency of results of previous research can be attributed to a number of methodological differences, including the specialized nature of some of the study samples, the type of statistical analysis used, the explanatory variables included in or omitted from the analysis, and the quality of the data used.
The present study addresses many of the methodological weaknesses found in previous studies. Demographic, socioeconomic, and health care utilization data are available for a representative sample of all noninstitutionalized individuals aged 65 and above residing in Manitoba on June 3, 1986, as well as information concerning their mortality and nursing-home experience until June 1991. This made it possible to include both disabled and nondisabled individuals, unlike previous studies (4)(11)(15). The availability of the date of death and date of entry to a nursing home enabled the use of proportional hazards regression models instead of logistic regression models. Finally, the database contains many variables that are suitable for inclusion as predisposing, enabling, and need factors. In particular, the "need factor" component can make use of the diagnoses available in the health care utilization data to define the presence of a variety of specific medical conditions at any point in time for the period under study. Most previous studies have been limited to the major conditions present at the start of the study period, which are typically asked about in health interviews, and they have relied on the accuracy of self-reported data (6)(11)(12)(15)(20).
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
The sample has been shown to provide accurate estimates of mortality rates, number and costs of medical services, and number and duration of short hospital stays, in comparison with those obtained from the entire MHSIP population file (22). As a result of an incompatibility of definitions between the Census and the Manitoba data, the sample is known to underestimate the duration of hospital stays of 60 or more days (24). The sample studied in this analysis is representative of all individuals aged 65 and above residing in private households in Manitoba on June 3, 1986.
Proportional hazard regression models (25) were used to estimate the hazard of nursing-home entry for the 5-year time period from June 1986 to June 1991, based on various health and sociodemographic characteristics. Such models are often used to examine the relationship between explanatory variables and survival times, without having to specify any particular form for the distribution of survival times. Instead, the model assumes that the hazard hi(t;x) at time t, for an individual i, with covariate values x, is a constant multiple of the baseline hazard h0(t) at all times:
 |
The hazard of entering a nursing home is the conditional probability that an individual will enter a nursing home at time t, given that the individual has not entered a nursing home before time t. For each covariate xi, exp{ßi} provides an estimate of the risk ratio, which, in the case of a dichotomous variable, is the ratio of the risk for an individual having 
compared with the risk when 
For a continuous variable, exp{ßi} provides an estimate of the risk associated with a one-unit increase in xi, relative to the average risk. Proportional hazards models and other methods of survival analysis are useful in analyzing survival data, because they account for the censoring of observations that takes place when either the end of the study period or death occurs before entry into a nursing home is observed.
Multivariate proportional hazards regression models were fit to the data to estimate the hazard of nursing-home entry. The dependent variable used in this analysis was the time between the start of the study and entry into a nursing home. Observations were censored if either death or the end of the study period occurred before nursing-home entry was observed. Two regression models were fitted. The first model, called the Base-Year model, examines the hazard of entry into a nursing home over a 5-year period based on predisposing, enabling, and need characteristics observed at the start of the study period (June 3, 1986). This first model is prospective in that the baseline characteristics are used to predict future nursing-home entry. This model, in combination with population risk profiles, can be used for planning purposes. The second model takes into account changes over time in the need characteristics. Because health status in an elderly population can evolve quite rapidly, it can be argued that when need characteristics are observed only at the start of a 5-year study period—as in the Base-Year model—some characteristics might wrongly be associated (or not) with nursing-home entry. The Time-Varying Needs model is assumed to give a more accurate evaluation of associations between contemporaneous need factors and nursing-home entry. This model can provide more accurate information to policy analysts in their search for means of delaying or preventing entry into nursing homes. For the Time-Varying Needs model, needs variables were updated on a yearly basis, using the health care records from fiscal year 1986–1987 to fiscal year 1989–1990. Separate models were fitted for men and women.
Selection of Independent Variables
In keeping with Andersen's framework, variables were chosen to represent each of the predisposing, enabling, and need characteristics groups. Age, marital status, and education were included as predisposing characteristics. Enabling factors included household size, an urban/rural indicator, various income measures, home ownership, and the supply of nursing-home beds, hospital beds, and physicians in the health region of residence. Need factors included functional disability, hospital and physician use during the previous year, excessive comorbidity, and 18 specific medical conditions. Because functional disability was only observed on June 3, 1986, it was the only need variable not subject to change over time. With the use of Verbrugge and colleague's classification (26), medical conditions were classified into three categories: low, moderate and high disability impact.
Summary statistics for all independent variables are found in Table 1 . There are 5,153 elderly persons in the sample, of which 55% are women and over 60% belong to the age group of 65–74 years. There were a total of 295 nursing-home entries for the entire 5-year period under study.
|
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
.05) with entry to nursing homes within each of the three categories of variables.
|
and high education for women 
. Several of the "enabling" factors are significantly associated with nursing-home entry for women. Although home ownership reduces the hazard (
), living in an urban area increases it 
. The supply of nursing-home beds and the number of physicians in the health care region of residence are also significantly associated with nursing-home entry, for women. For every additional available nursing-home bed (per 1,000 elderly population), the hazard for women increases by 1% and for every additional physician (per 10,000 population), it decreases by 8%. The supply of services appears to play a more important role in determining nursing-home entry for women than for men. This may be due to the absence of spousal support among women, who tend to experience greater longevity and are more likely to be widowed compared with males. The only "enabling" factor associated with entry into a nursing-home for men is home ownership 
Aside from age, the most important variables that are significantly associated with nursing-home entry all belong to the "need" factor component. Presence of Alzheimer's disease or dementia increases the hazard of nursing-home entry by 20.2 times for men and 10.0 times for women. Other specific medical conditions that significantly increase the hazard of nursing-home entry are musculoskeletal disorders 
and stroke 
for men, and other mental disorders for both men 
and women 
. The hazard is significantly decreased for men that have arthritis or rheumatism or disorders of the digestive system . For women, the hazard is significantly decreased if they have a history of disorders of the eye or ear 
or of the genitourinary system 
. In terms of more general measures of needs, functional disability is an important predictor variable 
. In terms of health care utilization, hospital stays of fewer than 45 days for women 
and the use of more than five physician services for men (
) are significant.
Time-Varying Needs Model
Most predisposing and enabling variables identified with the Base-Year model as predictors of nursing-home entry remain as such with the Time-Varying Needs model: age, education, being a home owner, and availability of nursing-home beds and physicians. Except for advanced age, risk ratios are of the same magnitude in both models. Marital status for men and the urban/rural indicator for women lost their statistical significance, whereas income became significant for women. Women in the third income quartile showed a hazard of nursing-home entry 2.1 times greater than that of women in the lowest income quartile.
The need characteristics associated with nursing-home entry are generally the same for both models, but their corresponding risk ratios are in some instances very different. The risk ratio for functionally disabled men declined from 3.2 to 2.8. For women, the previously significant association between disability and nursing-home entry disappeared 
Hospitalization in the previous year became a much stronger predictor. The risk ratio for hospital stays of fewer than 45 days increased from 1.9 to 2.7 for women, but it remained not significant for males. Hospital stays of 45 days or more became significant, resulting in a higher risk of nursing-home entry for men (
) and women 
. Physician consultation, which was only significant for men, disappeared. Presence of Alzheimer's disease and dementia have a much lower risk ratio when incorporating changes to the health status of the elderly population. However, these medical conditions remain some of the most important predictors of nursing-home entry 
Three additional specific medical conditions became significant: musculoskeletal disorders 
for women, and ischemic heart disease 
and all other chronic conditions for men.
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
The findings for some of the variables warrant further discussion. Marital status results are consistent with previous findings (9), which showed that for men, regardless of the size of their support network, the presence of a spouse was the most important factor in reducing the risk of nursing-home admission. The absence of a similar significant association for females may be due to the fact that in a couple, the female generally outlives her husband. For the population under study, 83% of women aged 85 and over are widows, compared with 33% of men. No association was found between the size of the support network and entry to a nursing home for either men or women. The absence of an association may be caused by multicollinearity between marital status and size of the support network. Greene and Ondrich (4) suggest that some of the significant effects of marital status previously reported in other studies may have been spurious, because living arrangements were not controlled for. In order to assess this effect, another model was fitted, replacing the two separate marital status and household size variables with a single variable combining information from both. The results are summarized in Table 3 .
|
, might reduce the risk. For men, having a spouse present in the household significantly reduces the hazard of nursing-home entry 
, and having another person in addition to the spouse reduces it even further 
. In comparing the results for the wealth variables within the "enabling" characteristics category with previous findings, one should take care in interpreting the results in the Canadian context, where health care is publicly funded. Although income does not play the same role in purchasing nursing-home care services as it does in the United States, increased wealth can provide the opportunity for purchasing alternative forms of care and for making special adaptations to housing in order to enable individuals to continue residing at home. Although income, per se, was not found to be significant, home ownership significantly reduced the hazard of nursing-home entry for both men and women. The lack of association between income and nursing-home entry for a population aged 65 and over is not surprising, because most individuals have retired by age 65 and income alone should not be used as a sole marker of the availability of material resources.
Higher risks of entry into a nursing home for both men and women with Alzheimer's disease, dementia, and other mental disorders or musculoskeletal disorders, and for men who suffered a stroke, were also observed in previous studies. These disorders all belong to Verbrugge's (26) category of conditions having the highest disability impact; that is, the conditions that are expected a priori to be the most likely to lead to institutionalization.
For women, having a disorder of the eye or ear is associated with a significantly decreased risk of nursing-home entry (RR = .71). A risk ratio of less than one is consistent with Verbrugge's (26) classification of eye and ear disorders among conditions having a low disability impact.
Among the conditions with moderate disability impact, the conditions that are significantly associated with nursing-home entry mostly have risk ratios that are less than one. These findings are somewhat counterintuitive. Two broad explanations come to mind. First, it is possible that in cases of comorbidity, the less severe medical conditions might be slightly underreported through the medical care system. The consequence of this underreporting is that the compared subpopulations (e.g., population with arthritis and population without arthritis) are not mutually exclusive, therefore causing errors in the calculation of the risk ratios. Our results are nonetheless consistent with the findings of previous studies. A study done by Houle and colleagues (28) using the 1994–1995 Canadian National Population Health Survey found that arthritis was negatively associated with the presence of an individual in a nursing home. Yelin and Katz (29), using the Longitudinal Study on Age from the United States, found that "persons with arthritis alone had the same rate of institutionalization in a nursing home as did persons without chronic conditions." Other research has provided mixed results: either no significant association between arthritis/rheumatism and nursing-home admission (10), or a positive relationship when institutional residency was examined (3).
A second possible explanation is that some medical conditions, such as ischemic heart disease (IHD), do not necessarily lead to functional disability, which is associated with nursing-home entry. Verbrugge and colleagues (26) suggest that the nature of the disease process for IHD is such that patients cut down on their physical and role activities but have little difficulty in activities of daily living. Previous researchers who have found no relationship or an inverse one between the presence of a specific medical condition and institutionalization—for example, heart disease (14), cancer (12)(14), and respiratory disease (12)(14)—have attributed their findings to the fact that such patients are more likely to die in the community without ever being admitted to a nursing home. Conversely, dementia (14), stroke (14)(15), and hip fractures (14)(15) (medical conditions requiring skilled nursing care after an acute episode) are strongly predictive of nursing-home entry.
Because comorbidity has been documented to have an important impact on disability (26), the model presented in this study included this variable in addition to specific medical conditions. The observed lack of association might be caused by an underreporting of the less severe medical conditions. It could also be the result of the inclusion of a large number of specific medical conditions in the model. Because no significant association was found for the number of chronic conditions, the model was refitted to remove these variables to assess the possibility of collinearity effects. The regression coefficients for the 18 medical conditions did not change appreciably, indicating the stability of the results.
Time-Varying Needs Model
The Time-Varying Needs model allows us to better understand the impact of changes in health needs on entry into nursing homes. The impacts of predisposing and enabling factors were marginally affected by the use of the Time-Varying Needs model. This is an interesting result, because it supports the assumption of Andersen's model that needs factors are independent of enabling and predisposing factors. The changes observed in need characteristics are quite important. The effect of functional disability and age at the beginning of the follow-up are reduced, probably as a result of the changes over time in the chronic conditions. Although it was not possible to examine the impact of changes in functional status over time in this study, other research has shown a small increase in the risk ratio for high levels of disability when time-varying functional disability is included compared with a static model (30).
Chronic conditions with long latency before institutionalization have their risk ratio reduced when time-varying covariates are included. This is because cross-sectional data measure the impact of the prevalence of disease and include individuals at later stages of disease, whereas time-varying covariate models measure the impact of incidence of disease. This was observed to be the case for Alzheimer's disease and dementia, which are slowly progressing irreversible degenerative diseases.
Hospital utilization in the year prior to entry to a nursing home became much more important.
Conclusions
The Base-Year model presented in this paper can be quite useful for health care planners. This model provides risk ratios of entry to a nursing home for a 5-year period. This information can be used in conjunction with population profiles and rates of entry into a nursing home for planning purposes (e.g., number of beds). The Time-Varying Needs model provides a more accurate representation of the impact of changes in needs that trigger institutionalization. The findings that certain medical conditions are associated with an increased hazard of nursing-home entry can be useful for prevention purposes. They can allow the targeting of interventions to reduce the risk of acquiring specific chronic conditions and for minimizing their disability impact when present. The impact of social support provided by families and friends seems to play a smaller role in entry to a nursing home. This may be because marital status and household composition serve only as proxies for the actual provision of support. Previous studies have shown that network composition and other characteristics of the support relationship are more important than the size of the support network (31)(32). This study reinforces the need for collecting comprehensive, longitudinal data on an ongoing basis, in order to better understand the impact of changes in predisposing, enabling, and needs factors that may occur after baseline.
|
|
|
Acknowledgments |
|---|
Received June 2, 1997
Accepted October 2, 1999
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  E. Nihtila and P. Martikainen Institutionalization of Older Adults After the Death of a Spouse Am J Public Health, July 1, 2008; 98(7): 1228 - 1234. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Martikainen, E. Nihtila, and H. Moustgaard The Effects of Socioeconomic Status and Health on Transitions in Living Arrangements and Mortality: A Longitudinal Analysis of Elderly Finnish Men and Women From 1997 to 2002 J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2008; 63(2): S99 - S109. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Grundy and M. Jitlal Socio-demographic variations in moves to institutional care 1991 2001: a record linkage study from England and Wales Age Ageing, July 1, 2007; 36(4): 424 - 430. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. K. Nihtila, P. T. Martikainen, S. V.P. Koskinen, A. R. Reunanen, A. M. Noro, and U. T. Hakkinen Chronic conditions and the risk of long-term institutionalization among older people Eur J Public Health, June 11, 2007; (2007) ckm025v1. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. E. Freeman, S. J. Gange, B. Munoz, and S. K. West Driving Status and Risk of Entry Into Long-Term Care in Older Adults Am J Public Health, July 1, 2006; 96(7): 1254 - 1259. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. H. Flaherty, B. Stalvey, and L. Rubenstein Guest Editorial: A Consensus Statement on Nonemergent Medical Transportation Services for Older Persons J. Gerontol. A Biol. Sci. Med. Sci., September 1, 2003; 58(9): M826 - 831. [Full Text] [PDF]
|
|
|
|
|
|
R. Hancock, A. Arthur, C. Jagger, and R. Matthews The Effect of Older People's Economic Resources on Care Home Entry Under the United Kingdom's Long-Term Care Financing System J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2002; 57(5): S285 - 293. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
