| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Institute of Gerontology, University of Michigan, Ann Arbor.
2 Department of Economics and Survey Research Center, University of Michigan, Ann Arbor.
Address correspondence to Dr. Lois M. Verbrugge, Institute of Gerontology, 300 North Ingalls, University of Michigan, Ann Arbor, MI 48109-2007. E-mail: verbrugg{at}umich.edu
 |
Abstract |
|---|
 TOP Abstract BackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
Methods. U.S. community-dwellers aged 55+ are studied using the 1994–1995 National Health Interview Survey Disability Supplement. Three types of assistance are considered: Personal Only, Equipment Only, and Both. Efficacy is measured by comparing the degree of difficulty doing a task with versus without assistance.
Results. Severe disability in a task and poor overall health/disability status increase use of assistance for the task, and especially both types rather than one. For people using one type of assistance, poor health/disability status is linked with personal help, but high severity is linked with equipment use. These results reflect high needs for assistance and limited potential for physiological improvement, joined possibly by a strong desire for self-sufficiency among persons who are severely disabled. Controlling for factors that route people to different types of assistance, equipment is more efficacious than personal assistance. Equipment may have distinctive technical and psychological advantages; for example, it can be tailored to a person's specific needs, is available when needed, and maintains self-sufficiency.
Discussion. The results about equipment give impetus to policies that promote development and dissemination of assistive technology.
Two common adaptations for disability are personal assistance and equipment assistance. They reduce task demand so people can accomplish their activities more easily. This study analyzes use of assistance for disabilities in personal care (activities of daily living [ADLs]) and household management (instrumental ADLs [IADLs]) tasks, type of assistance used (Personal Only, Equipment Only, Both), and the efficacy of assistance for relieving disability. Efficacy is measured by comparing a person's difficulty doing a task assisted versus unassisted. We are especially interested in whether personal and equipment assistance differ in their efficacy.
A fully integrated analysis of assistance use, type, and efficacy is not previously available in the literature. Our models of type of assistance are innovative, identifying factors that route people toward personal versus equipment assistance, and toward both types rather than one. They are crucial precursors of the efficacy models. Knowing and controlling for user-group differences is essential to have confidence that types of assistance differ in efficacy. The results suggest that equipment is more beneficial than personal assistance for relieving disability. We suggest reasons based on equipment's technical and psychological features; the reasons cannot be tested directly in the data set. The study gives impetus to policy and programs for disability technology.
|
Background |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
Personal and equipment assistance are buffers for disability. Although some buffers increase capability by augmenting an individual's physical or mental abilities, others reduce demand by diminishing a task's physical or mental requirements (Brandt & Pope, 1997; Verbrugge, 1990, 1994; Verbrugge & Jette, 1994). Personal and equipment assistance reduce task demand.
Rates of personal and equipment assistance for ADL/IADL disabilities are available from large-scale surveys (Cornoni-Huntley, Brock, Ostfeld, Taylor, & Wallace, 1986; Dawson, Hendershot, & Fulton, 1987; Forbes, Sturgeon, Hayward, Agwani, & Dobbins, 1992; Fulton, Katz, Jack, & Hendershot, 1989; Kennedy, Walls, & Owens-Nicholson, 1999; LaPlante, 1988; LaPlante, Hendershot, & Moss, 1992; Macken, 1986; Manton, Corder, & Stallard, 1993; Norburn et al., 1995; Russell, Hendershot, LeClere, Howie, & Adler, 1997). Differentials show greater use of assistance with older age, female gender, and higher education (Guralnik, Fried, Simonsick, Kasper, & Lafferty, 1995; LaPlante et al., 1992; Manton et al., 1993; McAuley & Arling, 1984). Multivariate analyses find the following: Use of personal assistance, particularly formal services, increases with older age, female gender, living with someone, and poorer functional status (Avery, Speare, & Lawton, 1989; Chappell & Blandford, 1991; McAuley & Arling, 1984; Penning, 1990; Stoller & Earl, 1983; Tennstedt & Chang, 1998). Use of equipment assistance is linked with high numbers of disabilities/functional problems and poor overall health (de Klerk, Huijsman, & McDonnell, 1997; Forbes, Hayward, & Agwani, 1993; Hartke, Prohaska, & Furner, 1998; Mann, Ottenbacher, Fraas, Tomita, & Granger, 1995; Zimmer & Chappell, 1994). Equipment use is positively correlated with personal assistance (Norburn et al., 1995). A recent analysis specifies this, showing that complex devices are linked with formal care (Agree & Freedman, 2000). Prior studies have not contrasted types of assistance, that is, identifying factors that route people to personal rather than equipment assistance, or both types instead of one.
How much does assistance help persons with disability? Two analyses of large-scale national surveys address this with cross-sectional data. Verbrugge, Rennert, and Madans (1997) found that, for U.S. adults aged 35+ with moderate/severe disability in a task, having assistance relieves difficulty for most people and eliminates it for many. Equipment by itself is more efficacious than personal assistance by itself, and even better than using both types. Agree (1999) found that for U.S. adults aged 70+ with inside-home mobility problems, using equipment by itself is linked with least difficulty and using both types with greatest difficulty. Both studies estimate models among assistance users, but have some fundamental differences. For the first study, the dependent variable is the difference between assisted and unassisted difficulty; it measures improvement with assistance. Coefficients show if a predictor is linked with high chances of improvement (positive) or low chances (negative). Thus, this study reviews efficacy directly, using individuals' judgments of difference between the assisted and unassisted states. For the second study, the dependent variable is the level of difficulty when using assistance, called "residual difficulty." Coefficients show if a predictor is linked with high difficulty doing a task (positive) or low difficulty (negative). Thus, this study investigates efficacy indirectly, comparing degrees of difficulty across different user groups. Several smaller scale studies also show the benefits of assistance. In an intervention study, assistive devices and home modifications slowed frail elders' functional declines over time, compared with control subjects (Mann, Ottenbacher, Hurren, & Tomita, 1999). In a longitudinal study, baseline use of assistive devices was related to better function 6 years later (Sonn, 1996). A cross-sectional study found that devices and procedural changes ease many daily tasks for women with rheumatoid arthritis (Nordenskiold, Grimby, Hedberg, Wright, & Linacre, 1996). In sum, the results about equipment are intriguing and merit further study.
Drawing on this background, we develop hypotheses for use, type, and efficacy of assistance. They assert the importance of levels of illness and disability for all three outcomes. We also propose that, controlling for illness and disability levels, the type of assistance people use is related to chances of disability improvement.
|
Hypotheses |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
Hypothesis 2: Personal versus equipment assistance—Severe difficulty doing a task and poor health/disability status increase the chances of personal assistance for the task, and also the chances of using both types rather than just one.
This assumes that people use equipment if they can, turning to others for help when their health/disability situation is poor. Using both equipment and personal assistance occurs in the most severe circumstances.
Hypothesis 3: Efficacy of assistance—Equipment assistance reduces and resolves disability more than personal assistance. People with severe disability benefit from assistance less than those with mild disability.
Prior literature has shown the relative benefit of equipment; possible reasons are in the Discussion. Persons with severe disability have less potential for functional improvement even when assisted.
|
Methods |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
Disability and Assistance Items
Phase 1 asks about health-related difficulty in six personal care activities (ADLs) and six household management tasks (IADLs). The ADLs are bathing or showering; dressing; eating; getting in and out of bed or chairs; using the toilet, including getting to the toilet; and getting around inside the home. The IADLs are preparing one's own meals; shopping for personal items, such as toilet items or medicine; managing money, such as keeping track of expenses or paying bills; using the telephone; doing heavy work around the house like scrubbing floors, washing windows, and doing heavy yard work; doing light work around the house like doing dishes, straightening up, light cleaning, or taking out the trash.
The ADL and IADL sections start with questions about assistance because of a physical, mental, or emotional problem in each task, then branch assisted and unassisted subgroups to other questions. The sequence is complex, so we summarize the data obtained. For each ADL, three kinds of assistance are asked of everyone: "get help from another person," "need to be reminded to do the task or need to have someone close by when doing it," and "use any special equipment to do" the task. The first two categories are personal assistance. Assistance users are asked a question on assisted difficulty, that is, how much difficulty they have doing the task with [personal and/or equipment] assistance. The response categories are No difficulty, Some, A Lot, Unable, and Don't know. Everyone is asked unassisted difficulty, that is, how much difficulty they [would] have doing the task on their own. For people who always use assistance, the answer is hypothetical. The response categories are the same as above (but assisted people must score Some, A Lot, or Unable). For each IADL, a question about personal assistance is asked of everyone: "get help or supervision from another person." Use of special equipment for health reasons is not queried. Assistance users are asked assisted difficulty, and everyone is asked unassisted difficulty. The response categories for the two items are No difficulty, Some, A Lot, Unable, Don't know.
Analysis Subsamples
This analysis uses 1994–1995 Phase 1 data for adults aged 55+ (N = 41,225). The hypotheses are tested for each ADL/IADL. For a given task, persons with a disability (difficulty doing the task on one's own because of health) are selected. Number of persons with disability range from 431 to 7,128 for the specific ADL/IADLs (Table 1). A person with several disabilities is in each pertinent subsample. Number of persons who use assistance range from 360 to 6,113 for the specific ADL/IADLs.
|
Predictors
Andersen's behavioral model (Aday & Andersen, 1974; Andersen & Newman, 1973; original 1968 paper cited therein) is suitable for thinking about factors that influence the use and efficacy of assistance. The model includes three groups of predictors: predisposing (sociodemographic and psychosocial features), enabling (resources and access), and need (levels of health and disability). All three groups can affect whether people use assistance for disability, what type of assistance they use, and possibly even their chances of benefiting from assistance. Specifically, predisposing factors reflect attitudes about assistance, enabling factors reflect access to and qualities of assistance, and need reflects urgency for assistance and physiological potential to benefit from it. The predictors for our analyses are described herein. Some are used in categorical form, and their dummy codes for regressions are indicated.
Predisposing
The items are age, gender, race/ethnicity, and disability identity. Age at last birthday is in single years (M = 68.2). Gender is coded Female (55.9%) and Male (44.1%; reference group). Race/ethnicity is coded White non-Hispanic (83.2%) and Other (Black, White Hispanic, Others; 16.8%; reference). Disability identity is based on two questions about whether the person considers himself/herself to have a disability or believes others consider him/her that way; scored 1 if yes to either (18.0%), else 0 (82.0%).
Enabling
The items are living arrangement, education, family income, family assets, and type of assistance. Living arrangement has three categories: Live with spouse (63.3%), and Live with others (12.0%), Live alone (24.7%; reference). Education is completed years of schooling, collapsed to <12 years (31.9%; reference), 12 years (38.0%), and >12 years (30.1%). Family income is from job/business, transfer payments, interest/dividends, and other sources, in the calendar month before interview date (in 000s; M = $2.803). Family assets are values of owned vehicles, houses, property, business, and stocks/bonds (in 000s; M = $62.842). Type of assistance has three categories: Personal Only, Equipment Only, and Both. It is used as a predictor just in the efficacy models. Distributions are in Table 2, and coding details in the Models section.
|
The disability-specific variables are severity and duration of a disability. Severity is degree of difficulty doing an ADL/IADL task on one's own: Some, A Lot, Unable. Because all persons have disability, the response None is inapplicable. Across the ADL/IADLs, Some ranges 12.6%–43.2% (low for Shop, high for Eat; reference), A Lot ranges 19.1%–34.7% (low for Heavy Housework, high for Dress), and Unable ranges 27.0%–67.8% (low for Dress, high for Heavy Housework); full distributions available on request. Duration is years since first onset of a disability, calculated by subtracting reported age of first onset from current age. Periods of remission are not queried in NHIS-D, so this is an indicator of lifetime experience plus concern for a disability. Mean durations among persons with disability are: Bathe 5.9 years, Dress 6.2, Eat 6.0, Transfer 6.5, Toilet 5.6, Inside 6.0, Meals 6.2, Shop 6.6, Money 8.1, Phone 9.0, Heavy Housework 7.5, and Light Housework 6.3. Duration-squared is included because duration effects may be nonlinear.
All predictors are included for substantive reasons. The three hypotheses focus on those we anticipate will have strongest effects.
Analysis Procedures
Logistic regressions are estimated because all of the dependent variables are dichotomous. Logistic coefficients are transformed from log odds into probabilities (marginal effects). This is because in social science analyses with observational data, probabilities are more readily interpretable than odds ratios. Each coefficient now reflects change in the probability of Y associated with a unit change in the predictor. For each equation, we calculate a pseudo-R2 to represent variance explained (Menard, 1995). Calculation procedures for marginal effects and pseudo-R2 are in our Table 3 footnotes.
|
Missing data (Not ascertained, Don't know, Refused) are handled in several ways, depending on amount. First, several predictors (disability identity, body mass index, mental symptoms limitation) had few cases with missing data (<.5%); the cases were coded to the mode for persons aged 55+. Second, NCHS commissioned imputations for missing data on income and assets, and we obtained the special files for those variables. Third, age of first onset of an ADL/IADL (used to compute disability duration) was difficult for some people to answer. They stated instead an age bracket (<18, 18–22, >22) or Don't know. Bracketed and Don't know responses are substantial, 12%–19% for the ADL disabilities and 14–22% for the IADLs. To obtain imputed values, we applied a detailed matching procedure. It assigns to each vague response the mean observed onset age of people with the same single year of age, gender, and race/ethnicity who have the disability. Fourth, severity questions are at the end of the ADL/IADL sections and were often skipped during interviewing, generating many Not ascertained codes. Across the ADL/IADLs, 10%–23% of unassisted severity responses and 11%–23% of assisted severity responses are missing. We applied a hot-decking imputation procedure, a stochastic approach that preserves observed distributions for the items. Rationales and technical details for our imputation work are available on request.
Models
For each hypothesis, models are estimated in an incremental, or staged, manner. For a given ADL/IADL, we begin with regressions for each set of predictors on its own: f{Predisposing}, f{Enabling}, f{Need}. Then, we have a regression with two sets: f{Predisposing, Enabling}. Lastly, we have a regression with all three sets: f{Predisposing, Enabling, Need}. The staged approach shows how a given predictor's coefficients change across models and how explained variances (R2) change. We find that, from initial to full models, predictor effects typically retain their sign, with some becoming smaller in size and significance. Explained variance differs greatly across models.
For Hypothesis 1, the multivariate models predict use of any assistance for a task (1 = assistance, 0 = no assistance). The models are for persons with disability in the task.
For Hypothesis 2, we study factors that route people to different types of assistance: Personal Only, Equipment Only, or Both. The models are for persons who use assistance for disability in a task. Four ADLs (Bathe, Transfer, Toilet, Inside) are analyzed. Sample sizes for Dress and Eat are too small to use (Table 1). IADLs are not studied because just personal assistance is queried. Three Y's are used: Personal Only (= l) versus Equipment Only (= 0); Both (= 1) versus Equipment Only (= 0); and Both (= 1) versus Personal Only (= 0). Each regression compares two user groups, and the third is excluded.
For Hypothesis 3, the models predict efficacy, that is, improvement with assistance. Coefficients show which predictors are associated with greater probability of improvement (+) with assistance, or lesser (-). The models are for persons who use assistance for disability in a task. For the ADLs, we begin with a baseline model f{Type of Assistance}. We continue with the five staged models; all include type of assistance, as well as the usual predictor set(s). To learn how each type of assistance compares with the others for efficacy, we run the models with different formats for the type of assistance predictor. One format has two dummies for Equipment Only and Both, with Personal Only as the reference group; this gives two of the desired contrasts. The other has dummies for Personal Only and Both, with Equipment Only as the reference group; this gives the remaining contrast. The Personal Only versus Equipment Only effect equals the first format's Equipment Only versus Personal Only effect, with opposite sign. For the IADLs, the models show how all predictors (except type of assistance) are linked to efficacy.
In tables herein, we report effects for the full models, f{Predisposing, Enabling, Need}. Effects are shown for predictors that regularly produce statistically significant effects; effects for other predictors are available on request. Explained variances for all stages are shown to evaluate the relative importance of predictor sets.
Analytic Issues
NHIS-D is a cross-sectional survey that provides associations between predictors and dependent variables. With theoretically plausible models, the results can yield suggestive evidence of causal ties. We approach the data with that premise, using cautious prose and inferences. If data were longitudinal, we would have firmer evidence about factors that influence startups of assistance use and how assistance prompts functional improvement over time.
Our measure of improvement with assistance is genuine, based on people's own judgments of their degree of difficulty doing a task in two situations: unassisted and assisted. It compares two states at a given time, rather than change over time. Subjective evaluations are not better or worse than objective ones made by professionals; they are simply different. We actually prefer the subjective evaluation because it stays close to individuals' disability experience.
Proxy response was allowed in Phase 1 if an adult household member was not present at the time of interview. For adults aged 55+, 64% were self-respondents, 25% proxy, and 11% unknown if self or proxy. ADL/IADL disability prevalences are only slightly higher for proxy respondents (sample persons who had proxies) than self-respondents. Proxy respondents are more likely than self-respondents to report high severity for disabilities. They are equally likely to have missing data for disability items. These are distribution differences. Prior research shows that such differences affect disability prevalence rates (Todorov & Kirchner, 2000), but do not routinely cause coefficients to differ for models with, versus without, a proxy predictor (Moore, 1988; Mosely & Wolinsky, 1986).
|
Results |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
A disability's severity is the pre-eminent factor linked to assistance use (Table 3). For ADLs, Unable greatly increases use by 19%–50%, compared with Some difficulty. A Lot of difficulty increases use 13%–22% relative to Some. For IADLs, the increases are 7%–20% and 12%–19%, respectively. Second, overall volume of IADL disability boosts assistance use. For each increment in IADLs, assistance rises 2%–3% for ADL tasks and 1%–4% for IADL tasks. Note that IADL disabilities affect assistance not just in their own domain, but in the other as well. Significant effects for counts of ADLs and physical limitations are smaller and inconsistent in sign (not shown). Third, use of assistance for ADLs is higher when people live with spouse or live with others, compared with living alone. Coresidence with spouse increases use of assistance by 6%–16%, and coresidence with others by 3%–10%. Lastly, Other race/ethnicity persons use assistance for disabilities 5%–11% more than do White non-Hispanic adults.
For explained variance, need characteristics explain use of assistance more than predisposing or enabling ones. R2 values for f{Need} range from .159 to .321 for ADLs, and .027 to .166 for IADLs. By contrast, f{Predisposing} values are .012–.044 for ADLs and .006–.002 for IADLs; f{Enabling} values are .011–.052 and .002–.034, respectively. The full model has values of .193–.387 for ADLs and .046–.189 for IADLs. This represents good prediction for ADLs.
In sum, Hypothesis 1 is supported. Need is foremost in explaining use of assistance by persons with ADL/IADL disability. Specific factors that increase the probability of assistance are disability severity and disability volumes. Coresidence with spouse or others (enabling) and Other race/ethnicity (predisposing) also have effects.
Hypothesis 2: Type of Assistance
Among assistance users, Personal Only is virtually the sole type used for Dress (94%) and Eat (89%) disabilities, and it is also the most common type for Transfer (65%) and Bathe (55%) (Table 2). Equipment Only is the most common type of assistance for Toilet (40%) and Inside (42%). Using Both types has sizable percentages for Toilet (26%), Inside (25%), Bathe (22%), and Transfer (18%). More generally, the distributions reflect how much tasks emphasize upper and lower extremity actions. Personal help by itself or with equipment dominates for tasks focusing on upper extremity movements, Dress (98%) and Eat (96%). It is central but less dominant for tasks involving both upper and lower extremity actions, Transfer (82%) and Bathe (78%). Equipment by itself or with personal help is typical for tasks focusing on lower extremities, Toilet (65%) and Inside (68%).
The multivariate models show factors linked with using each type of assistance for ADL disabilities (see Models). We discuss predictors of Personal Only versus Equipment Only, of Both versus Personal Only, and of Both versus Equipment Only. Results are in Table 4.
|
Both versus equipment only
Coresidence is the most important factor for using Both rather than Equipment Only. Living with spouse increases chances of Both by 13%–28%, and living with others by 11%–13%. Second, high numbers of ADLs, IADLs, and physical limitations strongly increase Both, compared with Equipment Only. For each ADL increment, the probability of using Both rises 5%–12%; for IADLs, 2%–6%; and for physical limitations, 2%–4%. Lastly, disability severity is consistently linked to Both; but most effects are below the significance threshold (not shown).
Both versus personal only
Disability severity is the main factor prompting use of Both rather than Personal Only. Compared with persons with mild disability, Unable boosts Both by 11%–21%. Effects for A Lot are also positive, but seldom significant (not shown). Second, high disability volumes increase Both over Personal Only. The probability of Both rises 1%–3% with each increment in physical limitations, and 3%–4% for ADL increments. Contrary to hypothesis, there is evidence that poor health and mental symptoms limitation decrease Both over Personal Only; the effects are significant at marginal levels. Third, persons with disability identity are more likely to use Both than Personal Only, by 11%–17%. Lastly, Other race/ethnicity persons use Both less than Personal Only, by 12%–15%, compared with White non-Hispanics adults.
For explained variance, need characteristics produce higher R2 values than do predisposing or enabling ones in all three sets of models. The f{Need} values are .144–.393 for Personal Only versus Equipment Only, .309–.418 for Both versus Equipment Only, and .045–.098 for Both versus Personal Only. Enabling factors rank second because of coresidence effects: .066–.117 for Personal Only versus Equipment Only and .043–.119 for Both versus Equipment Only, but negligible for Both versus Personal Only. For full models, R2 values are sizable for Both versus Equipment Only (.347–.460) and for Personal Only versus Equipment Only (.227–.446), and modest for Both versus Personal Only (.083–.131). This comes about because the first two models contrast having versus not having personal help, a demarcation in which need and coresidence figure strongly. The third model contrasts having versus not having equipment. Each side of the contrast includes personal help, and the predictors are less able to distinguish them.
In sum, Hypothesis 2 is supported with just one exception (high severity's link to Equipment Only rather than Personal Only). Need is central to explaining the type of assistance people use. Coresidence (enabling) and disability identity and Other race/ethnicity (predisposing) figure as well.
Hypothesis 3: Efficacy
Among people with ADL disabilities, 68%–78% have less difficulty in a task with assistance than unassisted, and 21%–36% resolve it completely with assistance (Table 5). Efficacy is less for IADLs but still substantial; 33%–56% of people with IADL disabilities have less difficulty when assisted, and 13%–28% resolve the disability.
|
|
Resolved difficulty
Severity of disability is foremost in explaining resolution. Effects are in the hypothesized direction. People who are Unable to do a task on their own are much less likely to resolve disability with assistance, than those with Some difficulty. The difference is 19%–41% for ADLs and 26%–40% for IADLs. Similarly, people with A Lot of difficulty resolve less often than those with Some, by 18%–29% for ADLs and 11%–20% for IADLs. Second, poor health/disability status decreases chances of resolution. This is pronounced for physical limitation and mental symptoms limitation; the probability of resolution decreases 1%–4% for each increment of the first and by 3%–12% for presence of the second. High numbers of IADLs and ADLs, and poor overall health also decrease resolution; signs are mostly negative, but seldom significant (not shown). Third, living with spouse is linked with less resolution than living alone; virtually all effects are negative and some are statistically significant. Living with others has no relationship to resolution. Fourth, effects for type of assistance weaken in the Resolved Difficulty models compared with the Reduced Difficulty models. Directions of effects are similar: largely positive for Equipment Only versus Personal Only, mixed for Both versus Personal Only, and largely negative for Both versus Equipment Only. Statistical significance is occasional (not shown).
For explained variance, need characteristics produce higher R2 values than do predisposing or enabling factors. For Reduced Difficulty, f{Need} values are .048–.190 for ADLs and .093–.152 for IADLs. For Resolved Difficulty, they are .132–.289 for ADLs and .159–.251 for IADLs. Full models for Reduced Difficulty have R2 values of .068–.199 for ADLs and .118–.167 for IADLs. Full models for Resolved Difficulty have values of .138–.329 for ADLs and .166–.262 for IADLs. These are similar to the f{Need} values, further indication of the small contribution predisposing and enabling factors make to efficacy. Note that the Resolve models have higher values than the Reduce ones. In statistical modeling, it is generally easier to explain large demarcations. For Resolve, people with intrinsic difficulty say they have no trouble at all with assistance. Reduce is a more moderate situation, including people who reduce but do not eliminate difficulty.
In sum, Hypothesis 3 is supported for type-of-assistance effects. Severity effects are mixed, contrary to the hypothesis for Reduced Difficulty and supporting it for Resolved Difficulty. Need is foremost, but type of assistance and residence with spouse (enabling) also figure.
|
Discussion |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
(1) Severity of Disability. A disability's severity is the most important predictor in our models. Severity increases chances of using assistance, and particularly both types rather than one. For persons with one type of assistance, severity boosts the chances of Equipment Only rather than Personal Only. These results form a use hierarchy of Both > Equipment Only > Personal Only. The surprising aspect of the hierarchy is Equipment Only > Personal Only. We hypothesized the reverse, presuming that people with severe problems have no option but to get personal help. That premise concentrates on need, which gets some support because Both is at the hierarchy's apex. But the link of high severity with Equipment Only suggests two other reasons. First, persons with disability often strive for control over their situation, variously labeled autonomy, self-determination, self-efficacy, or self-regulation (Baltes & Baltes, 1990; Gignac & Cott, 1998; Heckhausen & Schulz, 1995). Using equipment entails more control than personal assistance. If people with severe disability greatly desire control, this might explain their preference for Equipment Only rather than Personal Only. Second, equipment design for persons with severe disability may be improving, more than for other groups.
People with severe/moderate disability have better chances of reducing difficulty with assistance than those with mild disability, but their chances of resolution are worse. This is true for ADLs and (one exception) IADLs. Apparently, persons with severe disability can make modest gains in function when assisted, and they may be very perceptive of gains achieved. But they have low objective chances of complete relief. By contrast, for persons with mild disability, there is little room to improve other than resolving disability. This is a "floor effect," often viewed as a measurement artifact. Here, it should be viewed instead as an aspect of disability experience. People with severe functional problems do indeed have a large zone for gain, whereas those with mild ones have a small zone. This has both objective and subjective consequences for people, and those with severe disability seem to benefit in both ways for disability reduction.
(2) Overall Health/Disability Status. A person's overall health/disability status ranks second in importance. The predictors we chose measure health and function for the person as a whole, and they may capture aspects of health vulnerability, resilience, and frailty. They have strong effects on assistance use, type, and efficacy. Poor overall health/disability status increases probabilities of assistance use, of both types rather than one, and (for those using one type) Personal Only rather than Equipment Only. This is a use hierarchy of Both > Personal Only > Equipment Only. Poor overall health/disability diminishes people's chances of reducing and resolving difficulty with assistance. All of these results are straightforward reflections of high need for assistance and low potential for improvement.
The efficacy results for disability severity and overall health/disability are not the same. How can this come about? They are actually two different features. One pertains to a specific disability, and the other is global aspects of the whole person. High severity of a particular disability is often reparable, but poor overall health is a larger context that compromises someone's chances of functional improvement.
(3) Type of Assistance. From results, we know that disability severity and overall health/disability status route people differentially to assistance types, and they also affect efficacy. Such variables must be well-controlled statistically to detect type-of-assistance effects on efficacy. Our efficacy models include many indicators of health/disability status, namely, the "need" set of predictors.
With need and other predictors controlled, we find that equipment alone or combined with personal assistance favors reduction of disability. The strongest results show that Equipment Only is more efficacious than Personal Only. On average, chances of reduction are 14% greater for Equipment Only. There is additional evidence that Equipment Only exceeds Both, and that Both exceeds Personal Only. This forms an efficacy hierarchy of Equipment Only > Both > Personal Only. The same pattern of results appears for resolution of disability, but statistically weaker.
Type-of-assistance effects are remarkably robust. For reduced disability, baseline models f{Type of Assistance} already show the hierarchy. As other predictors enter, type-of-assistance effects become only a bit smaller, and if significant at the outset, they are usually significant in the full model. Thus, the hierarchy persists from baseline to full models. For resolution, baseline models show a slightly different hierarchy, Equipment Only > Personal Only > Both. This transforms to the usual hierarchy in full models.
The pre-eminence of Equipment Only concurs with our prior analysis of the NHANES I Epidemiologic Followup Studies (NHEFS; Verbrugge et al., 1997). The efficacy hierarchy of Equipment Only > Both > Personal Only appeared for reduced difficulty. Results for resolved difficulty were more complex, varying by type of task. But, in average terms (numerical averages of effects across tasks), the hierarchy Equipment Only > Personal Only > Both also existed for resolution.
The NHIS-D analyses are superior to the NHEFS analyses in three ways. First, NHIS-D has more information about health and disabilities, so efficacy models have deeper statistical controls for subjects' levels of illness and dysfunction. For NHEFS, just four key predictors were available (age, gender, severity, and type of assistance), and they could not all appear together in a model because of sample sizes. Second, here we compare effects of severe, moderate, and mild disability on assistance use and efficacy. In NHEFS, persons with mild disability were not asked the assistance questions, so only severe and moderate levels could be compared. Third, NHIS-D sample sizes are much larger.
In sum, both analyses suggest that equipment, especially alone and also combined with personal assistance, has distinctive aspects that facilitate disability relief. With simple models for NHEFS, the effects were strong. With more statistical controls for NHIS-D, the effects are abated, but still evident. A short comparison of numerical results for NHIS-D and NHEFS models is available on request.
What may explain equipment's greater benefit? First, equipment is designed for the task, can be modified to suit the individual, and is generally on hand when needed. It has "tailored and timely" technical characteristics. Second, equipment maintains an individual's self-sufficiency. This can foster pride and keen perception of task improvements. By contrast, personal assistance is less standardized (its fit to individual needs can be variable or poor) and may not be readily available when needed. When people give away some or all of task performance, their perceptions of difficulty and ease, improvement and stasis, may also change.
These possible reasons cannot be tested in NHIS-D itself; it has no variables on technical features of assistance or users' feelings about assistance. To buttress or reject the reasons, we sought research comparing qualities of equipment and personal assistance. Literature searches on their relative "technical" aspects yielded nothing. On psychological aspects, one study shows that helplessness and low confidence about coping are common for people with personal assistance, but also for those using compensatory strategies (which includes equipment; Gignac, Cott, & Badley, 2000). We conducted a pilot study, interviewing persons with a disability about their views of equipment and personal help being used (report available from first author). Feelings of mastery and self-efficacy are higher for people using only or mainly equipment, than for people using only or mainly personal assistance. Feelings of being dependent are higher for the latter group. Equipment users routinely said that equipment increases their confidence, level of control, and independence; by contrast, personal assistance users reported decreases or no change in these attributes from personal help. The disability literature shows that high self-efficacy/mastery is beneficial for people, reducing functional declines and hastening functional recoveries (Femia, Zarit, & Johansson, 1997; Kempen, van Sonderen, & Ormel, 1999; Mendes de Leon, Seeman, Baker, Richardson, & Tinetti, 1996; Partridge & Johnston, 1989; Schiaffino, Revenson, & Gibofsky, 1991).
(4) Coresidence. Severe disability prompts people to live with others to get assistance (Freedman, Wolf, Soldo, & Stephen, 1991; Pezzin & Schone, 1999). Contemporary research tends to see residence as endogenous with personal help (thus, both are joint outcomes of disability), but it can also be stated as a two-step process (disability influences residence and residence influences receipt of personal help). Living with spouse is appropriately viewed as exogenous to disability and personal help. Our models use living arrangement as a predictor. Coresidence is positively associated with actual receipt of personal help, but negatively associated with efficacy.
Our analyses reveal a novel aspect of coresidence. Living with spouse is associated with lower efficacy, whereas living with others has no association with efficacy. Spousal help entails giving tasks away, but it may seem like sharing task responsibility in the household rather than "assistance for disability"—that could soften perceptions of gains actually being made in accomplishing a task.
(5) Disability Identity and Race/Ethnicity. Two other predictors have limited effects that deserve comment. First, political momentum in U.S. society is to make disability identity a positive and socially powerful attitude. The independent living and disability rights movements were impetus for the Americans with Disabilities Act of 1990, and they have increased since then in strength and public awareness. How this has influenced middle-aged and older persons with disabilities is not known. Our empirical results, however, have an intriguing signal. Disability identity is linked with using equipment by itself or together with personal assistance. This result ties self-definition and self-sufficiency. The link may be joint (desire for control leads to disability identity and equipment use) or causal (disability identity prompts equipment use or equipment use alters identity). In any case, disability identity may give extra impetus to achieving the independence so strongly valued by U.S. adults. Second, persons with Other race/ethnicity use assistance for disability, more often than White non-Hispanics adults. Type-of-assistance models continue this theme with higher probabilities for Other race/ethnicity persons of personal help (by itself or with equipment) rather than equipment by itself. The results may reflect cultural differences in willingness or availability of kin to help.
Finally, some predictors we thought would figure in assistance use, type, and efficacy do not: age, gender, education, family income, family assets, body mass index, and duration of disability. Body mass index was expected to have a strong role like other measures of overall health/disability status; the other predictors were expected to have secondary roles. Effects for all of them are weak and inconsistent.
Disability is a gap between personal capability and task demand. The gap can be narrowed by increasing a person's capabilities or by reducing a task's demand. When capabilities are low and not amenable to improvement, then reducing demand is the main option to relieve disability. Personal and equipment assistance are strategies that reduce task demand.
This analysis explains use of assistance, type of assistance, and improvement with assistance for U.S. community-dwellers aged 55+ with disabilities. Our hypotheses took a medical stance, emphasizing how poor health and disability severity are central to explaining the outcomes. But psychological aspects of disability are possible explanations for: (a) why persons with severe disability use Equipment Only more than Personal Only, (b) why chances of disability reduction are greater for people with severe rather than mild difficulty, and (c) why disability reduction and resolution are smaller for persons living with spouse than alone, and greater for those using Equipment Only rather than personal help. This reminds us that disability is not a medical phenomenon, but a human experience that must be understood from the individual's perspective. Large-scale surveys obtain data that measure the medical side of disability. They have few if any variables to address the social psychology of disability. The limitation leads researchers and results in myopic directions—pertinent, but narrow. When large-scale surveys introduce more items about disability attitudes and goals, the scientific literature will reflect the disability experience better.
Our analyses were motivated by special interest in the relative efficacy of equipment and personal assistance, and we close on that topic. The efficacy hierarchy of Equipment Only > Both > Personal Only appears for reduced difficulty, and is suggestive for resolved difficulty. It arises with few statistical controls and persists with many. These cross-sectional results need to be confirmed in longitudinal data about startups and terminations of assistance and changes in disability, and in intervention studies that contrast types of assistance for functional outcomes. Technical and psychological aspects of assistance also need to be investigated.
Federal health insurance programs now emphasize personal assistance for functional problems, with assistive devices as a minor theme. Better balance should be introduced into policy discussions (Brandt & Pope, 1997; Lesnoff-Caravaglia, 1988; Mann, 1997). Eventually, programs may change to foster design, dissemination, and reimbursement for assistive devices.
|
Acknowledgments |
|---|
Statistical advice was provided by Dr. James M. Lepkowski, Institute for Social Research and Department of Biostatistics, University of Michigan. Ms. Lindsay-Rose Boynton (medical student, University of Michigan) and Dr. Michelle Putnam (Washington University, St. Louis) helped in discussions of psychosocial aspects of disability.
The paper was presented at the 52nd Annual Scientific Meeting of The Gerontological Society of America (November 1999, San Francisco, CA) and the Conference on Promoting Independence and Quality of Life for Older Persons (sponsored by the Research Center on Aging, University of Buffalo, and the American Society on Aging, December 1999, Washington, DC). A prior version of this paper, dated January 2001, was issued as a Research Report of the Population Studies Center, University of Michigan (http://www.psc.isr.umich.edu/pubs/papers/rr01-470.pdf).
Purvi Sevak is now with the Department of Economics, Hunter College, New York, NY.
Received for publication February 21, 2001. Accepted for publication January 7, 2002.
|
References |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  P. Clarke, P. Chan, P. L. Santaguida, and A. Colantonio The Use of Mobility Devices Among Institutionalized Older Adults J Aging Health, June 1, 2009; 21(4): 611 - 626. [Abstract] [PDF]
|
|
|
|
 |
|
 J. C. Cornman and V. A. Freedman Racial and Ethnic Disparities in Mobility Device Use in Late Life J. Gerontol. B. Psychol. Sci. Soc. Sci., January 1, 2008; 63(1): S34 - S41. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Horowitz, M. Brennan, J. P. Reinhardt, and T. MacMillan The impact of assistive device use on disability and depression among older adults with age-related vision impairments. J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2006; 61(5): S274 - S280. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Meng, B. Friedman, A. W. Dick, B. R. Wamsley, G. M. Eggert, and D. Mukamel Effect of a Voucher Benefit on the Demand for Paid Personal Assistance Gerontologist, April 1, 2006; 46(2): 183 - 192. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. W. Williams, C. S. Williams, S. Zimmerman, P. D. Sloane, J. S. Preisser, M. Boustani, and P. S. Reed Characteristics Associated With Mobility Limitation in Long-Term Care Residents With Dementia Gerontologist, October 1, 2005; 45(suppl_1): 62 - 67. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. M. Agree, V. A. Freedman, J. C. Cornman, D. A. Wolf, and J. E. Marcotte Reconsidering Substitution in Long-Term Care: When Does Assistive Technology Take the Place of Personal Care? J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2005; 60(5): S272 - S280. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. M. Agree, V. A. Freedman, and M. Sengupta Factors Influencing the Use of Mobility Technology in Community-Based Long-Term Care J Aging Health, April 1, 2004; 16(2): 267 - 307. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
