| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
Correspondence: Address correspondence to Susan Allen, PhD, Associate Professor, Deputy Director, Center for Gerontology and Health Care Research, Brown University, 2 Stimson Ave., Providence, RI 02912. E-mail: Susan Allen{at}Brown.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Long-term care • Architectural accessibility • Disability • Wheelchairs • Caregivers
Wheelchair users report high levels of limitation in both activities of daily living (ADLs) and instrumental ADLs (IADLs). Previous research has found that wheelchair users report approximately 9 ADL–IADL limitations, a level of disability that exceeds that of users of walkers (8 ADL–IADL limitations), crutches (6 ADL–IADL limitations), and canes (5.5 ADL–IADL limitations; Allen et al., 2001). Given the high level of disease and trauma-associated impairment characteristic of wheelchair users, it is not surprising that use of a wheelchair is predictive of receiving substantially more hours of both formal and informal care, as compared to use of any other type of mobility device (Allen et al.). What is surprising is the lack of research attention given to this high-need, high-use population to date.
Wheelchair Users Who Live Alone
Wheelchair users who live alone are a particularly vulnerable population given the combination of absence of in-home support and high need for assistance. Although living alone is associated with greater use of formal services (Tennstedt & Chang, 1998; Tennstedt, Sullivan, McKinlay, & D'Agostino, 1990), people who live alone have been reported to receive only half as many total hours of help with personal assistance as compared to people who live with others, independent of their need for assistance (LaPlante, Harrington, & Kang, 2002). This discrepancy is likely to be because most people who live alone do not have access to the same amount of unpaid help (from spouses and other family members) that is available to those who live with others. Thus, it is not surprising that living alone with a disability has been linked to higher levels of unmet need for personal assistance (Kennedy, 2001; Laplante, Kaye, Kang, & Harrington, 2004; Lima & Allen, 2001) and greater risk for adverse consequences associated with inadequate help. Such consequences include worsened prognosis in chronic disease (Case, Moss, Case, McDermott, & Eberly, 1992; Crockett, Cranston, Moss, & Alpers, 2002), falls, injuries due to falls, bedsores, and contractures (Desai, Lentzner, & Weeks, 2001; Laplante et al., 2003).
In summary, the high need for assistance and associated high utilization of personal care services that is typical of people with disability who use wheelchairs suggest that this population is an appropriate target for interventions designed to increase the efficiency and efficacy of assistance with personal care. Such interventions may be particularly useful for people who use wheelchairs and live alone, given that they are more likely than people who live with others to have extended periods of time when human assistance is not available. One promising direction for intervention development is assistive technology, and its potential to reduce the amount of help required for people who use wheelchairs, thus increasing their capacity for independent community living.
The Promise of Assistive Technology
Given the substantial expense associated with paid personal assistance services and the opportunity costs of family assistance, the notion that assistive technology may reduce need for human help and/or reduce time and physical burden for caregivers is compelling. Several studies have examined the relationship between human assistance and assistive technology, such as mobility equipment, and have demonstrated savings in hours of care (Allen et al., 2001; Hoenig, Taylor, & Sloan, 2003), reduction in unmet need for personal care (Agree & Freedman, 2003), as well as reduction in difficulty of task performance (Verbrugge, Rennert, & Madans, 1997). Such studies are consistent in reporting that reduction in levels of human help with use of assistive technology is limited to less severely disabled people (Agree, 1999). For example, Allen and colleagues found that the use of canes and crutches reduced hours of care received by people with mobility impairment; in contrast, use of a wheelchair predicted more rather than fewer hours of help. Thus, wheelchairs appear to be indicative of greater need for human assistance. However, there is an alternative explanation for the positive rather than negative association between wheelchair use and hours of human help, that is, the role of the home environment.
Wheelchair Use and the Home Environment
It has long been known in the more applied fields of occupational therapy and architectural design that physical environmental factors are crucial to the ability of people with functional impairment to perform daily activities, and that a poor person–environment fit can compromise independence (Iwarsson, 1999; Lawton, 1970, 1983; Steinfeld & Danford, 2000). Environmental barriers may prevent or limit wheelchair use without the assistance of a helper (Meyers, Anderson, Miller, Shipp, & Hoenig, 2002). Half of wheelchair users (49.2%) must use steps to enter or exit their homes (Kaye, Kang, & LaPlante, 2000). Small bathrooms prevent wheelchair users from maneuvering to transfer to the toilet or to use the sink without a helper. Narrow doorways and hallways may prevent entry into certain rooms. High kitchen counters and cupboards are inaccessible to those wheelchair users who cannot stand.
Whereas environmental barriers prevent or limit wheelchair use, home adaptations enable it. A strong predictor of wheelchair use in the kitchen and bathroom is home adaptation for the wheelchair (Hoenig, Pieper, Zolkewitz, Schenkman, & Branch, 2002). Home adaptations, such as widened doors, ramps, bathroom adaptations, and lower kitchen counters, have been shown to enable access to and from homes, empowering people with disabilities to gain control over their functional activities (Fawcett et al., 1994) and to promote greater independence in daily living activities (Fox, 1995). In other words, in the absence of home accommodations for disability, wheelchairs may actually increase the level of human help required to negotiate the home environment successfully. It is therefore likely that the focus of prior investigations has been misdirected, failing to recognize the interdependence of wheelchairs and home accommodations. This relationship merits further exploration and is the focus of this study.
We hypothesized that wheelchair users who have home modifications would receive fewer hours of human assistance than wheelchair users without such modifications. Further, we hypothesized that this effect would be greater for those wheelchair users who live alone because this population may have fewer opportunities for human assistance, and thus higher motivation to utilize alternatives, than wheelchair users who live with others.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Analytic Sample
As discussed above, selection into the DFS was based on a wide variety of indicators of disability; hence, the sample is heterogeneous in its health-related characteristics. Because this study's focus is specifically on the effect of home accommodations on hours of care provided to wheelchair users, the sample was limited to the 899 adults who reported using a wheelchair in the previous 2 weeks.
Measure
Dependent Variables
We used four dependent variables in regression analyses to test study hypotheses regarding an inverse relationship between home accommodations and the receipt of human assistance, as well as the amount of assistance received by those who received help. We created these variables based on responses to a series of questions asked in the DFS. Following questions designed to ascertain difficulty in the performance of ADLs and IADLs and receipt of help for those activities, respondents who received help were asked to name their helpers. Information was then obtained for a maximum of four helpers. Specifically, respondents were asked how each helper was related to them (e.g., spouse, daughter, agency-based help, privately hired help, etc.); which ADLs and IADLs each helper provided helped for; the number of days in the 2 weeks preceding the interview that each helper provided help; and the number of hours of help provided per day, on average.
From these data, we created two dichotomous indicators of receipt of any help, one indicating any formal (paid) help and the second any informal help. Next, we calculated interval variables representing total hours of formal help received per week and, separately, total hours of informal help received per week as follows. We calculated the average total hours of formal care received per week by adding together the hours of care associated with each formal helper during the 2 weeks preceding the interview and dividing by 2. We performed a similar calculation to derive average total hours of informal care received per week. We used the logarithm of hours for these dependent variables in order to eliminate right skewness.
The number of hours of help per day provided by each helper was missing for a substantial number of observations (n = 161; 17.9% of sample). Discarding these cases with missing values from the analysis would lead to a nontrivial reduction in efficiency of the estimators and potentially introduce bias into the sample. To avoid these problems, we used the following imputation method. First, we imputed hours per day only for observations where information on the number of days of help provided by each helper in the 2 weeks before interview was available. We then stratified the data based on the helper number (first, second, third, fourth) and the type of help (informal or formal) received. We imputed missing hours by drawing a simple random sample without replacement from the set of observed hours in each stratum. For example, we imputed missing hours per day for Informal Helper 1 among respondents who had valid values on days of help by taking a random draw from the stratum containing the set of observed hours per day from Informal Helper 1. This is a nonparametric imputation method known as hot deck (Little & Rubin, 2002). This method has an advantage over mean or median imputation in that it keeps the same amount of variability in the imputed sample as in the observed sample. Analyses revealed no significant differences between respondents with valid versus imputed values on hours of help, with the exception that cases with imputed values on hours of care were less likely to have private insurance than cases with valid values on this measure.
Independent Variables
The key independent variables used in these analyses were the following eight types of home accommodations: (a) widened doorways or hallways, (b) ramps or street-level entrances, (c) railings, (d) automatic or easy-to-open doors, (e) bathroom modifications, (f) kitchen modifications, (g) elevator, chair lift, or stair glide, and (h) accessible parking or drop-off site. We created dichotomous (1, 0) indicators of the presence of these accommodations in the residence. Additionally, we calculated and used a count of all accommodations as a continuous variable (0–8).
Of particular interest to this study was the living arrangement of the respondent, given its implications for access to human help. In prior studies, researchers have reported that formal service use is higher for people living alone, while informal help is lower (Allen et al., 2001; LaPlante et al., 2002). People who live alone are likely to have greater need for home accommodations to enable activity because human help is not as available, or is not as continuously present, as is the case for people who live with others. We included a dichotomous indicator of living arrangement (1 = alone, 0 = with others) in all models. We also included an indicator of whether or not an individual with disability was married (1, 0), because caregiving spouses typically provide more comprehensive care than other formal and informal caregivers, regardless of the individual's living arrangement. The correlation between living arrangements and marital status was not sufficiently high to suggest problems with multicollinearity (.33).
Because we know that the use of paid help is not independent from the use of unpaid help, we included indicators of paid help in modeling unpaid help, and vice versa. Additionally, as part of preliminary studies for this research, we conducted regression analyses to identify factors related to the presence of accommodations in the home to control for endogeneity in modeling receipt of help and hours of care (data not shown). Factors positively associated (p <.05) with the presence of home accommodations included duration of wheelchair use (at least 1 year vs less than 1 year), residence in elderly housing, at least some college or higher level of education, and private insurance coverage. Inverse associations included Hispanic ethnicity and use of other types of mobility equipment (e.g., a cane, crutches, or a walker) in addition to a wheelchair. We included these factors in all models to control for selection bias.
We included age of the respondent as a continuous variable. Because severity of disability increases with age, we expected to see increased hours for older wheelchair users. We grouped respondent's race into three categories: Black, non-Hispanic (1, 0), Hispanic (1, 0), and White, non-Hispanic (1, 0). Because of the small percentages of Asians, Native Americans, and other groups participating in the survey (a combined total of 2.5% of the sample), we did not extract additional minority subgroups from this general population sample. White, non-Hispanic was the reference group in all models.
We operationalized education using three categories: less than high school, high school graduate (reference), and at least some college or a college degree. People educated beyond high school generally have greater social support as well as greater knowledge regarding the availability of sources of paid care. Conversely, those with less than a high school education may have less access to care from either formal or informal sources (Turner & Marino, 1994). Similarly, indictors of income above and below the federal poverty line were included to control for ability to purchase supplementary care. Given the relatively large proportion of the sample (13%) that had missing values on income information, an indicator of missing poverty status also was included to avoid losing these cases in the multivariate analysis.
We included insurance status because reimbursement for certain long-term-care services may be provided by an insurance plan. For instance, Medicare reimburses home health care after discharge from the hospital for a 30-day period and may pay for some bathroom accommodations. For our analysis, Medicaid, though limited, was included as the sole provider of long-term-care services in general. Private insurance was the reference group because it does not generally cover long-term-care services. Thus, we included dichotomous indicators of Medicare, Medicaid, and private insurance coverage in all models. In addition, we combined those who had other types of public insurance (e.g., Veteran's Affairs) and the small number of people in this sample without insurance (n = 17) into a fourth insurance category, "other."
Finally, we included three measures to control for severity of respondents' impairment, given the known association between impairment severity and hours of care received. Specifically, we included in all models a count of ADL difficulties (range 0–7; bathing or showering, dressing, eating, getting in or out of bed and chairs, walking, getting outside, and using or getting to the toilet), and receipt of assistance with eating (1, 0), which were intended to capture the small minority of individuals with extensive care needs. We also included use of a proxy respondent because of cognitive impairment (1, 0), accounting for 52% of all proxy respondents, in recognition of the extensive time demands associated with dementia, Alzheimer's disease, and, in a small minority of cases, mental illness.
Analytic Approach
We conducted logistic regression analyses to determine whether the presence of home accommodations was inversely related to the receipt of any paid and/or unpaid human help among this national sample of wheelchair users, controlling for factors related to receipt of help (as described previously). We also controlled for these factors in multiple regression analyses conducted to determine whether home accommodations were related to reduced hours of paid and unpaid help among those who received any. As discussed earlier, we retained cases with data missing on hours of care through hot deck imputation methods. We lost an additional 168 cases to analysis due to missing data on a variety of variables in our models; thus, 735 of the full sample of 899 cases were available for multivariate analyses.
To test the first hypothesis regarding an inverse relationship between home accommodations and human help, we separately tested the effect of the total number of home accommodations, and then each type of home accommodation, on receipt of any paid and unpaid help as well as on hours of paid and unpaid help received. We tested individual types of accommodations to determine whether savings in hours of care was strictly cumulative or if the effect of any one accommodation could also be captured. We included an interaction of number of home accommodations (and for each type of accommodation, separately) with living arrangement (alone vs with others) in each model to test the second hypothesis that the effect of home accommodations on human help is stronger for people who live alone than for people who live with others. We used sampling weights to account for the core NHIS sampling design, for nonresponse to the NHIS core and the NHIS-D Phase I survey, and for nonresponse to the DFS. We analyzed data using SUDAAN (Research Triangle Institute, 2004). Thus, findings from this study can be generalized to community-dwelling adult wheelchair users, aged 18 and older.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
Receipt of Any Paid or Unpaid Help
Results of logistic regressions testing the relationship between home accommodations and the receipt of any help, and controlling for selection bias and other covariates, are presented in Table 3. Relative to having no home accommodations, the presence of each additional home accommodation decreased the odds of having unpaid help by 14% (OR =.86; 95% CI =.76,.97). In contrast, there was no association between home accommodation and paid help, and, in fact, trends indicated a positive rather than an inverse relationship (OR = 1.08; 95% CI =.99, 1.18).
|
In addition to the number of accommodations present in the home, other factors inversely related to receiving any unpaid help were living alone (OR =.18; 95% CI =.09,.40), receiving help with eating (OR =.53; 95% CI =.28,.99), using nonwheelchair mobility devices (OR =.61; 95% CI =.39,.97), and having paid help (OR =.31; 95% CI =.17,.57). Women were twice as likely as men to have unpaid help (OR = 2.06; 95% CI = 1.28, 3.32), and each ADL difficulty increased the likelihood of having unpaid help by nearly one third (OR = 1.32; 95% CI = 1.15, 1.53).
In order to test the influence of individual types of accommodations on receipt of any paid or unpaid help, we added each type to the models described above, testing the effect of one accommodation at a time, while controlling for total number of accommodations and other covariates. The presence of ramps nearly doubled the odds of having paid help (OR = 1.96; 95% CI = 1.18, 3.25). However, there was no other effect of individual types of accommodations on the odds of having paid help when they were tested separately through logistic regression (data not shown).
Presence of railings in the home was associated with an increased likelihood of having unpaid help (OR = 1.69; 95% CI = 1.01, 2.83). In addition, we found several individual types of home accommodations to decrease the odds of receiving unpaid help when tested separately. Specifically, people with widened doors (OR =.31; 95% CI =.15,.63) and people with automatic doors (OR =.39; 95% CI =.18,.84) had reduced odds of receiving any unpaid help at all, relative to people who did not have these types of accommodations (data not shown).
Hours of Paid and Unpaid Help
We conducted ordinary least squares (OLS) regressions to determine whether home accommodations were related to the amount of help received; these indicated that having more accommodations were related to fewer hours of unpaid help received from family and friends (p <.01) but not fewer hours of paid help. Significant and positive predictors of hours of paid help for the full sample included use of a proxy respondent due to cognitive impairment (p <.001) and help with eating (p <.05), both of which were measures of impairment severity (see Table 4).
|
We reran the OLS models to separately test the effect of each type of accommodation on hours of help, controlling for total number of accommodations and other covariates. These analyses revealed that there were no individual types of accommodations related to paid or unpaid hours of help as main effects (data not shown).
Wheelchair Users Who Live Alone: Receipt of Any Paid or Unpaid Help
Logistic models testing the interaction of number of accommodations and living arrangement, and, separately, individual types of home accommodations and living arrangement, indicated no higher likelihood of having any paid or unpaid help among those who live alone with accommodations versus those who live with others and also had accommodations, with one exception. For those who live alone (but not those who lived with others), bathroom modifications were associated with an increased likelihood of having paid help (OR = 2.52; 95% CI = 1.07, 5.92; data not shown).
Wheelchair Users Who Live Alone: Hours of Paid and Unpaid Help
We examined OLS regressions for people who live alone versus people who live with others by adding the interaction between accommodations and living arrangement into all models. Results indicated that effects were much more evident for people living alone (see Table 5).
|
Finally, interactions of living alone and four types of home accommodations were significantly and inversely associated with hours of unpaid help when tested separately. Specifically, results suggested that people who live alone (but not people who live with others) and have ramps (p <.05), railings (p <.01), and bathroom modifications (p <.05) received fewer hours of unpaid care than people who have these accommodations and live with others (see Table 5).
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The notion that home accommodations can substitute for human help is compelling. It is highly likely that the findings do reflect a substitution effect for many respondents in this study. However, other possible scenarios must be considered, given the cross-sectional nature of this data. For example, it is entirely possible that home accommodations are purchased by, or provided to, some wheelchair users because human help is in short supply. In other words, home accommodations may not reduce help for all users, but may be a deliberate response to need in the absence of available resources. It is improbable that accommodations are adequate to eliminate unmet need in all cases, although work by Agree and Freeman (2003) suggests that technology users do report less unmet need for personal care.
Home accommodations and paid help appear to go hand in hand, perhaps because individuals seeking services are assessed simultaneously for the need for home help as well as home accommodations. In fact, it is likely that caseworkers are aware that a combined strategy, or "package" of helping resources, may be most effective and efficient at the same time. In some situations, accommodations may supplement human help, while in other situations, they may be facilitative, reducing the physical challenge of caregiving, although not necessarily saving time.
Based on the findings regarding the interaction of home accommodations with living arrangement, it appears that wheelchair users who live alone make maximum use of home accommodations to accomplish tasks and activities, given that the option of human help is less available, or is available in limited quantities. In some cases, this may be a preference, an indication of the importance of autonomy to wheelchair users who live alone, and in other cases, it may be necessity, a strategy to avoid unmet need.
Particular types of accommodations appear to reduce the quantity of human help needed by people who live alone, namely ramps, railings, accessible parking, and bathroom modifications. Any one of these accommodations reduces dependence on another person to perform daily living activities and, thus, increases the likelihood that people who use wheelchairs and live alone can continue to live independently despite the severity of their impairment and absence of continuous social support. Ramps can provide entry to the outside world that would otherwise be denied to wheelchair users lacking substantial assistance from another person or persons. Bathroom modifications can restore privacy in the conduct of the most personal of daily activities.
The most important limitation of this study is the cross-sectional nature of the DFS, which greatly limits our ability to understand causality underlying observed associations. An additional limitation is that the survey asked respondents about the presence of accommodations in their homes and not about use. However, the implication of the latter caveat is that the findings may actually underestimate the potential of home accommodations to substitute for human help. Given the relatively strong results reported here, we are convinced that this is a fruitful line of research worthy of further investigations regarding the causality underlying our assumptions. While the traditional policy focus is on tradeoffs between family help and publicly financed home-care services, the results of this study suggest greater emphasis on provision of a valuable third home-care resource, home and environmental accommodations. Others have reported that the primary reason for lack of environmental modification is cost (Bayer, 2000; Pynoos, 2001). More than three quarters of home modifications are paid for out of pocket. Conventional Medicare plans only pay for assistive devices (such as canes, walkers, and wheelchairs) that are considered medically necessary and do not pay for the structural changes that may be needed to use these devices in the home environment. Reimbursement for structural modifications is more generous under Medicaid home- and community-based waivers, though still limited.
We found an inverse relationship between the presence of home accommodations and other mobility-equipment use, suggesting that some wheelchair users who are able to ambulate use walkers and canes rather than their wheelchairs to negotiate their environments. As evidence of the efficacy of technological aids accumulates, it is clear that there is no single solution for those who require help at home. Greater implementation of "cash payment" home-care-benefit programs that allow individuals to purchase the blend of home-care resources that is appropriate to their needs, circumstances, and preferences is one avenue to expand the use of home accommodations, as well as other types of assistive technology (Foster, Brown, Phillips, Schore, & Carlson, 2003). As the population ages, adoption of such policies and programs may be an efficient response to the growing demand for home-care support, with the additional advantage of facilitating independence and community participation for this vulnerable subgroup of our population.
|
Footnotes |
|---|
1 Center for Gerontology and Health Care Research, Brown University, Providence, RI.
2 Providence Veterans Administration Medical Center, RI.
3 Department of Biostatistics and Computational Biology, University of Rochester, NY.
Decision Editor: Linda S. Noelker, PhD
Received for publication November 2, 2004. Accepted for publication August 4, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
