| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Population and Family Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland.
2 Polisher Research Institute, Madlyn and Leonard Abramson Center for Jewish Life (formerly the Philadelphia Geriatric Center), North Wales, Pennsylvania.
3 Department of Public Administration, Maxwell School of Syracuse University, Syracuse, New York.
4 Department of Social Science Computing, University of Pennsylvania, Philadelphia.
Address correspondence to Dr. Emily M. Agree, Department of Population and Family Health Sciences, 615 North Wolfe Street, Room E4646, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD 21205. E-mail: eagree{at}jhsph.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. We jointly modeled hours of formal and informal care with use of AT in order to address the interdependence of these outcomes in ways not taken into account in previous studies. We analyzed a national sample of older persons with difficulty in activities of daily living drawn from Phase 2 of the 1994–1995 National Health Interview Survey (NHIS) Disability Supplement.
Results. Our findings show that the use of AT was associated with reductions in informal care hours, especially for those who were unmarried, better educated, or had better cognitive abilities, but appeared to supplement formal care services for these groups. Individuals with cognitive impairment were less likely than others to substitute AT with either type of personal care.
Discussion. These models raise the possibility that reductions of informal care hours may be accomplished with a combination of formal care and assistive devices, rather than from either alternative alone.
SEVERAL concurrent population trends, such as declines in disability and increases in education among the older population, as well as decreases in the availability of informal caregivers and a shrinking long-term care workforce (Freedman, Martin, & Schoeni, 2002
; Wolf, 2001) have led to discussions of the potential of assistive technology (AT) to promote independent aging in place. The latter half of the 20th century saw dramatic growth in AT use by older persons with disabilities (Manton, Corder, & Stallard, 1993; Norburn et al., 1995). During the 1980s, the number using any equipment rose from 3.3 to 4.1 million, and the number relying only on equipment more than doubled (Manton et al.). By the mid-1990s, the majority of those with activity of daily living (ADL) limitations used some form of technology, and almost a third (31%) used only AT to accommodate their needs (Agree & Freedman, 2000).
It is hoped that recent increases in the use of AT can improve quality of life and alleviate pressures on the existing long-term care system. Recent national studies suggest that technology may confer unique benefits in reducing difficulty with daily tasks and unmet need (Agree, 1999; Agree & Freedman, 2003; Taylor & Hoenig, 2004; Verbrugge, Rennert, & Madans, 1997; Verbrugge & Sevak, 2002). Yet it remains unclear whether AT replaces or supplements personal care. Where the use of assistive technology reduces the amount of personal care needed, public expenditures on home health care could be reduced (e.g., Mann, Ottenbacher, Fraas, Tomita, & Granger, 1999) and the burdens of informal care alleviated. Alternatively, the use of AT in addition to personal services could potentially improve the quality of care, and thus defer functional declines and institutionalization, which also would reduce public and private expenditures.
Prior research on the substitutability of devices for personal care has been limited in several ways. First, with few exceptions, studies have used a single metric of care to reflect substitution. Second, analyses often are limited to mobility devices (e.g., canes, wheelchairs, walkers). Third, most research has implicitly assumed that the use of AT precedes and can predict use of personal care, while in practice such decisions are not separable.
In this article, we expand upon prior studies in several ways. First, we consider multiple dimensions of personal care (e.g, hours, number of caregivers, and number of tasks) for both informal and formal sources of care. Second, rather than focusing on mobility alone, we examine equipment use for all ADL activities. Third, we adopt a modeling strategy that allows us to examine the use of AT and formal and informal care as a unified set of decisions, without assuming one necessarily precedes the other. We also focus not only on whether substitution occurs, but also on which groups have a greater propensity to replace personal care with technology.
Background
The literature on substitution in long-term care for the older population has primarily focused on tradeoffs among different types of personal care in order to address the "woodwork effect," the concern that public coverage for home care could cause a reduction in informal care. These studies have been limited because, although substitution of one type of care for another is in part a function of the relative cost of each type of care, information on the price of formal care and assistive technologies is difficult to obtain and the implicit price of informal care is difficult to define or measure. A small number of studies have addressed these issues with experimental data or state-level variation in home care prices (e.g., Ettner, 1994; Greene, Lovely, & Ondrich 1993), but the majority define substitution as a simple replacement of one type of care for another (e.g., Logan & Spitze, 1994; Soldo, Agree, & Wolf 1989; Tennstedt, Crawford, & McKinlay, 1993), and we followed this definition in the present analysis. These studies generally have found that formal home care does not substitute for (or crowd out) informal care and may, in many cases, supplement informal care (Kemper, Applebaum, & Harrigan, 1987; Tennstedt et al.; Tennstedt, Harrow, & Crawford, 1996).
Recently, studies of community-based long-term care have considered substitution between AT and personal care (Agree & Freedman, 2000; Allen, Foster, & Berg, 2001; de Klerk & Huijsman, 1996; Hoenig, Taylor, & Sloan, 2003; Mann et al., 1999). Using an experimental design that randomized about 100 older individuals to assistive devices, Mann and colleagues found that AT use was associated with lower costs for paid home care. The study did not address the role of informal care nor did it explore the mechanism through which costs were reduced (e.g., fewer visits, shorter visits, different types of visits). The remaining four studies were based on observational designs, with three of the four focusing on mobility limitations. In descriptive analyses, Agree and Freedman (2000) found that in the United States, simple assistive technology, such as canes, had the potential to substitute for informal care, while more complex devices, such as wheelchairs, appeared to supplement formal care. These relationships held after controlling for the underlying degree of disability severity. Similarly, de Klerk and Huijsman found in a sample of older persons from the Netherlands that the use of mobility aids was positively associated with the use of home care. Both studies defined substitution narrowly in terms of a discrete tradeoff between the use of AT and the receipt of any informal or formal care. In their analysis of mobility aids, Allen and colleagues considered a more extensive array of definitions including any formal or informal care, hours of formal or informal care, number of tasks for which help was received, and yearly out-of-pocket costs for formal services. In a series of logistic regression models, they found the use of canes and crutches inversely related with the number of tasks and hours of informal and formal care received; however, no evidence of substitution was found for wheelchairs and walkers. A fourth study by Hoenig, Taylor, and Sloan found that AT use for ADL limitations was associated with fewer hours of help.
Though it has been growing, the existing literature ignores the highly interdependent nature of decisions about AT and informal and formal care. This leads to two serious limitations. First, with the exception of Mann and colleagues (1999), studies that have modeled the relationship between AT use and personal care have attempted to predict informal and/or formal care use as a function of AT use. Such models by definition assume that decisions about AT precede decisions about personal care, which is rarely the case in practice. For example, a recent widow may use adaptive equipment to compensate for the loss of her informal caregiver. Additionally, formal providers, such as nursing aides or rehabilitation therapists, often will bring AT into an older individual's home as a part of their service package (Gitlin & Levine, 1992). Second, these studies have focused on whether substitution is occurring, with no attention to differences in the ability to substitute AT for personal care. For example, those with cognitive impairment may not be able to trade off technology for human assistance.
Moreover, AT can substitute for formal or informal care in a number of ways. For example, a cane or walker might alleviate the need for any personal care by allowing an individual to ambulate independently. Adding grab bars to the shower may enable older persons to shower safely on their own, thereby reducing the hours of care or the number of caregivers needed, or freeing caregivers to use the extra time to turn their attention to other tasks, such as paying medical bills, shopping, or cooking. In the last case, the total amount of care (in terms of personnel or hours) may remain the same, but the tasks with which they assist may change.
In this article we expand upon prior studies in two ways: First, rather than limiting our analysis to a single dimension of personal care, we consider four distinct dimensions: 1) whether any informal and formal care is received; 2) the number of informal caregivers; 3) the hours of informal and formal care received; and 4) the number of ADL and instrumental ADL (IADL) tasks for which informal and formal care is received. Second, unlike most previous studies that have been limited to mobility limitations, we consider the care options (informal and formal care and the use of AT) for any of six ADL activities. Third, we employ a statistical method that allows us to examine the use of formal and informal care services and AT without assuming the order in which they are acquired. The article is organized as follows: After providing a description of the data, we compare AT users and nonusers to examine descriptively how assistive technology may substitute for personal care use across the four dimensions described above. We then present models that consider the use of AT and hours of formal and informal care jointly in order to identify who is most likely to use AT in place of or in addition to personal care services. Finally, we discuss implications of our findings for future research.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Sample Selection
Our analytic sample comprised 4,006 persons aged 65 years and older who reported underlying difficulty with one or more of seven ADLs (bathing or showering, dressing, eating, getting in and out of bed or chairs, walking, getting outside, and using the toilet). For each activity respondents were asked how well they were able to do them by themselves and without special equipment. We limit our focus to the use of care for ADLs because these tasks have been shown to be more responsive than IADLs to the use of assistive devices (LaPlante, Hendershot, & Moss, 1992; Norburn et al., 1995; Sohn & Grimby, 1994; Zimmer & Chappell, 1994). They also are far less likely to be subject to the lifelong abilities and expectations built into social roles, especially by gender (Allen, 1994). We do not focus on sensory impairments such as hearing and vision limitations for two reasons. First, while these abilities are important in accomplishing ADLs and using equipment, they are not activities in and of themselves. Second, the NHIS-D collected no information on personal care for them.
Dependent Variables: AT and Personal Care Use
AT use
For each ADL, participants were asked if they used special equipment and to name the specific item(s) used. The most common items mentioned were canes, bath seats, walkers, bath rails, and wheelchairs (see Table 1). Altogether 63.8% of respondents reported using at least one of these five devices. The majority of AT was used for mobility-related activities (walking, transferring, and going outside), and mobility devices often were used for more than one activity, including toileting, which has a mobility component. AT was also commonly used for bathing; almost one quarter of the sample reported using rails or grab bars. In order to maintain the power of the models, and because not every AT user identified their devices by name, our measure of assistive technology use is a dichotomous indicator of whether a respondent uses any special equipment.
|
Drawing upon our multiple definitions of substitution, we created indicators for the following dimensions of personal care: First, we created a dichotomous indicator for whether or not the respondent used any informal or formal care in the last two weeks. Second, we represented the size of the care network by a dichotomous variable indicating whether respondents used two or more informal caregivers. We used the number of informal rather than formal caregivers because most secondary caregivers were informal. Third, we created a set of indicators for the number of hours of care provided by informal and formal caregivers in the past two weeks. In a number of cases, hours of care were incomplete or missing for one or more caregivers. For 1,139 respondents, we imputed information on hours of care for one or more helpers. We assigned informal caregivers with missing information the average number of hours of care provided by caregivers matched by sex (male, female, or unknown), network size (1, 2+), and position (primary, secondary). For formal care providers, we used network size and position only. Finally, to explore whether care hours were shifted to alternative tasks, we counted the number of ADL and IADL tasks for which formal and informal help was received.
Independent Variables
We included four sets of variables, which have been shown in previous research to be related to use of AT and informal and/or formal care, into the models: health needs, resources, access, and demographic characteristics.
Health Needs
The primary factor determining the nature and amount of care used is the extent of underlying health needs. In this study, health needs were measured by 1) the number of ADL limitations with which the respondent reported severe underlying difficulty and 2) the presence of extensive cognitive impairment. The first measure was obtained from questions about the amount of difficulty respondents had with each ADL when they were not using any special equipment or personal assistance. Those who reported that they had a lot of difficulty or were unable to do the task without assistance were considered to have a "severe" limitation with that activity. The reference group in our models consisted of respondents who reported some difficulty with one or more ADL tasks but who did not report "severe" difficulty with any ADLs.
Because measures of cognitive impairment in the NHIS-D were extremely limited we employed an indicator of whether a proxy respondent was used due to "poor memory, senility, and confusion" or because the respondent "had Alzheimer's disease." This measure identified only the most severely cognitively impaired respondents and excluded more mildly cognitively impaired individuals who were able to respond to the interview themselves.
Resources
The choice of care arrangement depends in part upon the economic and familial resources available to the older person. In this analysis we included measures of insurance, poverty, and kin availability. To represent insurance coverage, we included a three-category indicator: Medicare only, dually eligible for Medicare and Medicaid, and Medicare and private supplemental insurance. Those with other health insurance, typically government veterans insurance, were included with the Medicare only group as fewer than 3% of respondents were in this category. We represent poverty status with a three-category variable: below poverty (< 100% of the poverty line), near poverty (100–200% of poverty), or above poverty (> 200% of poverty). In addition to these categories, we used a separate category to indicate those individuals whose income was unknown (about 20% of our sample). With respect to kin availability, we included indicators representing marital status (currently married vs not currently married) and the number of living daughters both in and outside the household. We also examined number of living sons, brothers, and sisters, but found in preliminary analyses that these factors had little effect on either AT use or personal care.
Access
Access to technology and formal care can be measured by availability and by price. Because information on the price of AT is not widely available, here we focused on two indicators: acute care use and urban location. Many older persons are first introduced to the use of a device in the hospital or through prescription by a physician (Gitlin, 1995), and arrangements for home health care are commonly made upon hospital discharge, when Medicare will pay for these services. Contact with medical providers also may lead to use of a formal care manager, who would be more knowledgeable about both formal home care and appropriate AT. Thus, although imperfect, we represented the availability of these care options in our models by hospitalization and physician visits in the prior year.
Geographic location often is used as a proxy for availability of and accessibility to services such as health care (e.g., Krout 1983), but there are inconsistencies in the literature related to rural–urban residence and service utilization. We examined the contrasts between nonurban and urban residential location, where urban areas are cities with more than 250,000 residents, and nonurban areas are localities with fewer than 250,000 residents.
Demographic characteristics
We also included four demographic characteristics previously shown to be related to care arrangements: age, gender, education, and race and ethnicity. We classified age categorically as a distinction between those aged 65–74 and those aged 75 and older. Race and ethnicity were represented by a single indicator of whether the respondent was White or non-White because few respondents from other backgrounds are included in the sample.
Modeling Approach and Interpretation
We first examine the bivariate relationships between AT use and the four measures of personal care to determine where potential substitution of AT for personal care exists. Based on these results, multivariate analyses focus on the relationship between AT use and the hours of informal and formal care received in the last two weeks.
Following prior analyses of substitution between formal and informal personal care (e.g., Greene, 1983), we assume that formal and informal care and AT are not acquired in any particular order, as decisions regarding care tend to be idiosyncratic and interdependent. That is, we assume that accommodating one's care needs is not necessarily done sequentially (e.g., first choosing a piece of technology and then determining the type of caregiver and hours needed, or vice versa) but rather may occur in any possible order, or simultaneously (for example, upon discharge from hospital). The lack of recognition of this interdependence in previous research has often led to misspecification in other analyses with cross-sectional data. There is a body of statistical methods that have been developed to simultaneously estimate equations for interdependent outcomes (e.g., indirect regression, two [and three] stage least squares, instrumental variables, and full and limited information maximum likelihood models). Economists most commonly apply instrumental variable techniques to identify the set of equations. However, in the absence of plausible instruments, we have opted for a reduced-form modeling approach in which the three outcomes are estimated as functions of the same combination of exogenous health, resource, access, and demographic factors.
The dependent variables for the equations representing our three outcomes included a dichotomous (0, 1) variable for AT use and continuous variables truncated at 0 for both informal care and formal care. We assume that the error terms for the three equations come from a correlated trivariate normal distribution. Therefore, we jointly estimated a Probit equation for AT use along with two Tobit equations for the two hours-of-care variables. Correlations between pairs of the three error terms (designated by the Greek symbol 
[rho]) represents the relationships between the relevant pairs of care alternatives net of characteristics included in the model. We obtained parameter estimates with aML software (Lillard and Panis, 2003).
Although this approach does not allow comparisons across specific items of AT (because each would require its own equation), it offers the advantage of identifying characteristics associated with the substitution of AT for formal and informal hours of care. In other words, we were able to identify factors that increased the likelihood of using AT while decreasing the hours of either informal or formal services used. Consequently, we consider the combination of a positive (and statistically significant) coefficient in the AT equation with a negative (and statistically significant) coefficient in the informal care and/or the formal care equations to indicate substitution. The parameters also may be interpreted as indicating substitution between a pair of outcomes when the parameter estimate is negative (and statistically significant); in this case, however, the factors associated with the substitution remain unmeasured. In contrast, significantly positive coefficients in the AT equation and the formal hours or informal hours equations, along with positive parameters, indicate complementarity of the two types of assistance. We estimated one-tailed significance tests to test for a specific effect direction.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
Results presented in Table 5 provide some initial insight into which groups of respondents were most likely to use assistive technology either instead of or in addition to formal and informal care. The basic complementarity of AT and personal care is confirmed for those of advancing age and with more severe ADL disabilities, which were associated with increases in all three types of care. In addition, those with two or more doctor visits in the past 3 months showed increases in both use of AT and hours of informal care, and those that had any hospitalizations in the last year were both more likely to use AT and more formal care hours.
|
Several other variables included in the model produced statistically significant results for formal or informal hours, or both, but insignificant results for AT (and vice-versa). For example, individuals living in nonurban areas were significantly less likely to use AT but were not statistically different from those dwelling in urban areas with respect to either formal or informal hours of help. Such findings indicate a reallocation of assistive services, compared to the reference group, but do not meet our criterion for service substitution because the increase (or decrease) in the likelihood of using AT was not offset by a decrease (or increase, respectively) in use of help from others.
Net of the effects of the variables included in the models, the parameters indicate that AT use had a small positive correlation with informal care hours ( =.05) and a moderate positive association with formal care hours ( =.15). Informal and formal care hours were negatively related ( = –.13), which is consistent with prior research on personal long-term care services.
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
We also found that for some groups (e.g., unmarried persons and those with some college education) AT and formal care together seem to offset informal care hours. However, because the coefficients for both AT and formal care hours are positive, it is possible that formal care and AT substitute for different tasks or at different times and that informal caregivers cannot reduce their hours unless both are present in the care arrangement. This finding also underscores the importance of jointly considering AT with both informal and formal care sources in order to understand the long-term care decision process.
Results also indicate that those with worse cognitive capacity are less likely to substitute the use of AT for both formal and informal personal care hours. It is not surprising that these respondents would be less able to substitute AT for caregiver time, as the use of devices for physical disabilities requires some cognitive ability. This result may be a conservative estimate of the effect of cognitive abilities on the propensity to substitute, however, as our measure of cognition distinguishes only between those with the most severe cognitive impairment (who were unable to be interviewed and for whom proxy responses were obtained) and all others. As many as 5–7% of those who self-report could have had severe cognitive impairment (Freedman, Aykan, & Martin, 2001), and an unknown additional number might have had mild to moderate impairments in cognitive function. Use of a more refined measure should, in theory, yield even greater distinctions by cognitive status, but this relationship requires further investigation.
These findings substantiate prior research showing that AT substitutes for informal care, at least for some groups (Agree & Freedman 2000; Allen et al. 2001; Hoenig, Taylor, and Sloan 2003), and that AT supplements formal care (Agree and Freedman 2000; de Klerk and Huijsman 1996). However, our results differ from Allen and colleagues (2001), who concluded (using the same data) that use of canes substituted for hours of formal care. Reconciling the inconsistency in findings will require further analysis, as the two studies employ different inclusion criteria, modeling strategies, and treatment of the cognitively impaired.
Our study is limited in several respects. First, we use cross-sectional data and thus are unable to directly examine the order in which each type of assistance was adopted. Nevertheless, we improve upon past studies because our analytic models do not impose a priori any temporal order between AT and personal care as previous research has done. A second limitation is that we were unable in this study to distinguish among specific types of devices or between simple and complex technologies, though other research has shown these distinctions to be important (Agree & Freedman, 2000; Allen, 2001). Using a dichotomous indicator for AT may limit our ability to recognize task-specific cases of substitution or supplementation because we summarize care at the person level instead of the activity level. Further research at a more refined level of detail would give greater meaning to our findings and better suggest clinically relevant interventions to improve home-based long-term care. Finally, although we control for reported limitations in health and underlying severity in functioning, we cannot fully capture all of the effects of declining health on long-term care decisions. Consequently, there may remain some residual relationship among the three outcomes that is due to unobserved health factors.
Despite these limitations, this study provides important insights into the potential for technology to improve the daily lives of older disabled persons. We find, for example, that the reduced informal care hours cannot be attributed solely to AT, but rather may be due to the combination of formal care and assistive devices. The association between these types of care suggests that potential improvements in care management could be made through increased use of care managers or provision of information about AT to health care providers. This finding also underscores the need for caution in pointing to AT as a means of offsetting care costs for the older disabled population. Ultimately such devices may improve the quality of life of older persons and promote independence; however, their ability to substitute for more expensive formal care is not yet established.
Finally, our results point to three groups—those who are unmarried, those with at least some college education, and those with little or no cognitive impairment—as being most likely to substitute AT for personal care. Better education may be associated with increased receptivity to AT and thus facilitate the adoption of it into one's care network. Unmarried persons, because they most often live alone, may by necessity use AT to carry out their daily activities. While it is difficult to predict trends in marital status, over the next decades the educational attainment of older persons will increase dramatically (Freedman & Martin, 1999). All else equal, our analysis suggests that this trend will promote substitution of technology for hands-on informal care. Why these groups use AT in place of informal care, whether doing so confers a benefit, and if so, how best to encourage use for other groups are fruitful areas for further research on care for the elderly.
|
Acknowledgments |
|---|
Dr. Freedman is now at the University of Medicine and Dentistry of New Jersey School of Public Health.
|
Footnotes |
|---|
Received for publication July 29, 2004. Accepted for publication March 2, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S.-I. Hong Understanding Patterns of Service Utilization Among Informal Caregivers of Community Older Adults Gerontologist, July 2, 2009; (2009) gnp105v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Kitchener, T. Ng, H. Y. Lee, and C. Harrington Assistive Technology in Medicaid Home- and Community-Based Waiver Programs Gerontologist, April 1, 2008; 48(2): 181 - 189. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
