gerontgerontThe Gerontologist1758-53410016-9013Oxford University Press10.1093/geront/gnp170FUNCTIONAssistive Device Use as a Dynamic Acquisition Process in Later LifePresslerKaris A.MA12FerraroKenneth F.PhD22Center on Aging and the Life Course and Department of Sociology, Purdue University, West Lafayette, Indiana1Address correspondence to Karis A. Pressler, MA, Center on Aging and the Life Course, Purdue University, Ernest C. Young Hall, 155 South Grant Street, West Lafayette, IN 47907-2114. E-mail: kpressler@purdue.eduDecision Editor: William J. McAuley, PhD620102712010503371381281020098122009© The Author 2009. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.2010Purpose: This study identifies risk factors, including incident disability, for the use of assistive devices (ADs) among older people. Design and Methods: Three waves of data from the National Long-Term Care Survey (NLTCS) are used to examine whether upper and lower body disability lead to use of ADs (both number of devices used and number of activities of daily living domains for which ADs are used). Predictors of AD use include demographic variables, body mass index, and disability (both initial and incident). Relationships are estimated with negative binomial regression models. Results: Lower body disability, advanced age, and obesity are consistent predictors of the number of ADs used. An interaction between age and incident disability revealed the highest rate of adoption among the younger respondents who experienced increases in disability. Implications: Many older adults use ADs in response to the disablement process. In addition to need driven by rising disability, obese older adults use more ADs. Results from this study clarify who and why ADs are adopted by older persons and should facilitate effective intervention by health care personnel and caregivers.Assistive devicesDisabilityDisablement processAssistive technologyObesityAssistive devices (ADs), also referred to as assistive technology, are “items frequently used by people with functional deficits as alternative ways of performing actions, tasks, and activities” (LaPlante, Hendershot, & Moss, 1992, p. 2). Examples of ADs include walkers, wheelchairs, raised toilet seats, and grab bars. It is now widely recognized that ADs aid functional independence, enabling older adults to remain active when facing disability in activities of daily living (ADL; Agree & Freedman, 2003; Cornman, Freedman, & Agree, 2005). The use of ADs has risen substantially in the past 20 years, and current estimates reveal that approximately one third of all adults aged 65 years or older use at least one device (Freedman, Agree, Martin, & Cornman, 2006; Schoeni, Freedman, & Martin, 2008; Spillman, 2005). From another viewpoint, more than half of all AD users are older adults, and demand is expected to increase due to population aging and growing sales and acceptance of such aids (Centers for Disease Control and Prevention and The Merck Company Foundation, 2007).ADs can be an important resource for older adults, but timely adoption can mean the difference between maintaining and losing one’s functional independence. We conceptualize the adoption of ADs as a response to the disablement process, but find that many previous studies of the topic rely on interpretations from cross-sectional analyses or point estimates of disability. The present study makes use of 10 years of data from a longitudinal panel study to identify incident disability and whether it, in turn, increases the likelihood of using one or more ADs.Disability Dynamics and Use of ADsThe disablement process can be an abrupt transition for some people, such as for stroke victims (Gitlin, Luborsky, & Schemm, 1998), but it is a long and somewhat circuitous process for others. Research during the past decade has revealed that many functional limitations emerge during late middle age and that many older people experience spurts, plateaus, and declines in disability (Manton, 2008; Manton, Gu, & Lamb, 2006; Schoeni et al., 2008). Despite predictions of a compression of morbidity (and disability), most people spend years adjusting to growing disability, seeking help through environmental modifications, health care, and activity accommodations to manage their conditions (Verbrugge & Jette, 1994). ADs are one such resource that many people use effectively in the struggle to maintain functional independence (Verbrugge, Rennert, & Madans, 1997).Broadly speaking, previous research focused on ADs has had one of two main aims. First, several studies identify trends over time in the use of ADs, and most such studies use data from cross-sectional studies, including repeated cross sections (Freedman et al., 2006; Spillman, 2005). The overall trend is straightforward—increasing use of ADs—but especially for devices to aid walking, bathing, and toileting (Freedman et al., 2006; Russell, Hendershot, LeClere, Howie, & Adler, 1997).Second, a number of studies have been conducted to identify which older adults are most likely to adopt ADs to help maintain their functional independence. Findings from these studies generally concur that advanced age and disability are the most consistent predictors of use of ADs by older adults (Agree, Freedman, Cornman, Wolf, & Marcotte, 2005; Agree, Freedman, & Sengupta, 2004; Cornman et al., 2005; Hartke, Prohaska, & Furner, 1998; Tomita, Mann, Fraas, & Stanton, 2004). At the same time, there is some evidence of a nonlinear relationship between disability and AD use: a rise in AD use across low-to-moderate levels of disability but less so thereafter suggesting a threshold effect (Mathieson, Kronenfeld, & Keith, 2002). Research on predictors of AD use has also addressed the related question of whether such devices supplement or substitute for personal care. Most studies of these processes provide support for a process of supplementation (Agree, 1999; Hoenig, Taylor, & Sloan, 2003; Manton, Corder, & Stallard, 1993), although ADs may substitute for informal care on specific tasks (Agree & Freedman, 2000; Agree et al., 2005).Previous research has contributed much to our understanding of the magnitude of recent increases in AD use and who is likely to adopt them, but we identify four limitations of prior research in order to advance our knowledge and improve interventions. First, most studies examine overall disability, but Verbrugge and colleagues (1997) showed that lower body disability is more consequential to AD use. Accordingly, we differentiate upper and lower body disability as predictors of ADs.Second, beyond age and disability, there is little agreement in the literature as to other factors that predict AD use. For instance, some studies show that persons with less education are more likely to use ADs (e.g., Kaye, Kang, & LaPlante, 2000), whereas other studies show the opposite (e.g., Cornman et al., 2005). Similarly, there is inconsistency in reports of racial differences in AD use (cf. Cornman & Freedman, 2008; Kaye, Yeager, & Reed, 2008). Moreover, some variables that are theoretically relevant have received scant attention. One example is obesity, which many studies show is related to disability (Ferraro & Kelley-Moore, 2003; Guralnik, Fried, & Salive, 1996), but few studies actually examine whether obesity increases AD use (Cornman & Freedman; Verbrugge & Sevak, 2002). The current study considers these and other covariates while prospectively modeling use of ADs over time.Third, there is some disagreement as to what actually constitutes AD use. Some studies (including the present one) identify ADs used specifically for physical disability (Hoenig et al., 2003; Jette, 1994; Manton et al., 1993), but other studies use a more expansive definition that includes visual, cognitive, and hearing devices (Mann, Ottenbacher, Hurren, & Tomita, 1995; Mann et al., 2008; Tomita, Mann, Fraas, & Burns, 1997). Of course, with the more inclusive definition, nearly all older adults would be AD users. In addition, Cornman and colleagues (2005) show quite convincingly that some surveys underestimate AD use by posing the question to only those respondents who report difficulty with daily activities. This decision may contribute to inconsistency in findings on the topic and understate the importance of AD use in the daily lives of older people.Fourth, if AD use is presumed to be a resource that is activated in response to growing disability, longitudinal panel studies are needed to uncover how people respond to the disability episodes. Cross-sectional analyses might even give the appearance of less disability if AD use is part of the question tapping performance of activities. The extant literature is replete with cross-sectional analyses as well as calls for panel studies (e.g., Agree et al., 2004). Nevertheless, we were able to identify only one study that used a prospective two-wave panel design on a national sample (Cornman & Freedman, 2008) and a few others that are experiments, testing the effect of AD use on a select sample over time (Gosman-Hedström, Claesson, & Blomstrand, 2002; Mann, Ottenbacher, Fraas, Tomita, & Granger, 1999; Mann et al., 2008).There are two primary reasons why longitudinal panel designs make sense for research on ADs. First, illness episodes are the likely turning points in AD adoption. During the early stages of a disease, a person may be able to manage their condition without an AD, but the advanced stages of a disease may mean that the person needs to do more to maintain the same functional level. Tracking this over time is important. Indeed, Agree and colleagues (2004, p. 267) called for incorporating change in disability into the study of the “dynamic acquisition process” whereby older people adopt an AD—and this is precisely the aim of the present analysis.Second, as disability increases, it may diffuse to other domains of functioning (Ferraro & Shippee, 2009). Do most people use multiple devices within one domain of functioning? Or might the disablement process lead to AD use in multiple domains of functioning? To answer these questions, it would be helpful if research could not only identify who uses ADs—and how many—but also the number of functional domains affected. As the number of domains rises for which a person uses one or more ADs, this is likely an indicator of overall vulnerability.Three main research questions guide the present research. First, does incident disability lead to the use of one or more ADs? Second, if yes, do both upper and lower body disability increase the likelihood of adopting ADs? Third, does rising disability also lead to domain diffusion in the adoption of ADs? Answers to these questions are vital to effective intervention by health care personnel and caregivers who assist older adults with ADLs. Our aim is to capture important transitions in the disability process as it unfolds and test the effect that this change has on subsequent use of ADs.MethodsSampleData for this study come from the National Long Term Care Survey (NLTCS), a nationally representative sample of those aged 65 years and older in the United States, drawn from the Medicare beneficiary list, accounting for 97% of U.S. older adults (Manton, Corder, & Stallard, 1997). Even though the NLTCS began in 1984, the current study focuses on the three most recent waves of data: 1994, 1999, and 2004. The 1994 survey will be referred to as W1 (Wave 1), 1999 as W2, and 2004 as W3. Although the NLTCS tracks respondents who enter institutions, measures needed for this study vary across the institutional and community surveys. Therefore, only community-dwelling older adults were included here.A detailed questionnaire was administered during 1994 to 4,126 chronically disabled persons (i.e., difficulty performing an ADL or independent activity of daily living for at least 3 months) and a sample of 963 nondisabled older persons, resulting in 5,089 respondents. The inclusion of initially nondisabled persons is important given our aim to capture incident disability as it occurs naturally.Among the 5,089 W1 respondents, 1,706 (33.5%) died by W2; 320 (6.3%) were institutionalized; and 600 (11.8%) were nonrespondents. Among the 2,463 interviewed at both W1 and W2, 957 (38.9%) died by W3, 131 (5.3%) were institutionalized, and 328 (13.3%) were nonrespondents. Analyses proceeded with 2,436 respondents between W1 and W2 and 1,047 respondents across three waves.To account for sample attrition, multivariate analyses are adjusted with selection bias modeling (Heckman, 1979). A probit model predicting death was estimated with the following predictors: age, gender, lives alone, body mass index (BMI), personal care, upper body disability, lower body disability, incident lower body disability, self-rated health, smoker, and whether a proxy provided responses to the cognitive questions. A selection factor based on the inverse Mills ratio of the probit results was used to adjust for attrition in the final equations.MeasuresTwo dependent variables were created to assess use of ADs. (Of note, we follow Agree & Freedman’s, 2000, use of the term “assistive device” as opposed to “assistive technology” because of our focus on ADL disability; see also Cornman et al., 2005). The first is an overall count of ADs that respondents reported currently using, referred to as number of ADs used. When respondents indicated they used devices for both getting into or out of bed and getting around indoors, each device was counted only once. Respondents could report using as many as 29 ADs, but 88% reported using 5 or fewer ADs in 2004. A complete list of ADs used in the analyses is found in Table 1 (note a).Table 1.Descriptive Statistics of NLTCS Respondents Present at W1, W2, and W3 Community SurveysSurvey yearW1W2W3VariableRange199419992004Number of ADs useda0–161.26b,c (2.01)d1.64 (2.47)2.05 (2.83)Number of ADLs with ADse0–60.94 (1.31)1.04 (1.37)1.23 (1.44)Age65–98f78.67 (7.66)Male0–134.09%White0–187.74%Education1–73.35 (1.60)Lives alone0–137.91%42.27%Underweight (BMI <18.5)0–17.28%4.90%Normal weight (BMI 18.5–24.9)0–144.20%43.07%Overweight (BMI 25–29.9)0–131.49%34.02%Obese (BMI >30)0–117.03%18.01%Personal care0–124.37%21.84%Upper body disability (Nagi)0–121.93 (2.86)1.78 (2.87)Lower body disability (Nagi)0–123.76 (3.65)3.39 (3.60)Incident upper body disability, W1–W20–60.55 (1.32)Incident lower body disability, W1–W20–60.74 (1.37)Number of cases5,0892,4631,047Note: AD = assistive device; ADL = activity of daily living; BMI = body mass index (kg/m2); NLTCS = National Long Term Care Survey.aNumber of ADs used included the following, which are ranked in order from most to least frequently used at W3 among respondents who participated at all three waves: Shower seat/tub stool, grab bars/handle bars for bathing, walker, cane, raised toilet seat, rubber mat for bathing, hand-held shower, rail/grab bar for toileting, wheelchair, portable toilet/bedside commode, special underwear/diapers, indoor railing, furniture/walls to get around indoors, bedside railing, bedpan or urinal, lift to get in/out of bed, oxygen/respirator, orthopedic shoes, elevator/escalator, brace (leg or back), crutch, prosthesis, and chairlift on stairs. Respondents also indicated whether they used “any” AD for eating and dressing and were asked whether they used any “other” ADs for each of the ADL domains—both of these items were included in count of ADs at W2 and W3.bMean or percentage.cFor respondents available at all three waves, the mean values for number of ADs used were 0.47 at W1 and 0.89 at W2.dStandard deviation.eADLs include eating, dressing, getting in/out of bed, getting around indoors, bathing, and toileting.fAge at W1 (1994).The second dependent variable is the number of ADL domains for which respondents reported using one or more ADs. This measure, number of ADL domains with AD, taps AD use across domains of ADL (i.e., eating, dressing, getting in/out of bed, getting around indoors, bathing, and toileting). Use of these two outcomes enables us to compare the robustness of predictors by differentiating between people who use many devices within a single ADL domain and those who use few devices spread across several ADL domains.Disability was measured with the Nagi items and differentiated as upper and lower body (Nagi, 1976). Lower body disability questions asked respondents the degree of difficulty experienced climbing one flight of stairs, walking to the end of the room and back, bending to put on socks or stockings, and lifting a 10-pound package and holding it for a few minutes. Upper body disability captured the degree of difficulty respondents experienced reaching above their head, combing and brushing their hair, washing their hair, and using fingers to grasp and handle small objects (no Nagi items were used for both disability variables). Responses for each disability variable were captured on a 4-point scale (0 = not difficult, 1 = somewhat difficult, 2 = very difficult, and 3 = cannot do the task at all). Each disability index has an alpha reliability coefficient greater than .80 at each wave. To capture incident disability, persons who experienced an increase in disability were assigned the difference between waves (top-coded so that 6 = 6 or more). Persons who were followed at all three waves and had no change between W1 and W2 (n = 411) or experienced a disability decline (n = 266) were assigned a value of 0.Personal care was measured with a binary variable if the respondent reported someone helping them to perform at least one ADL during the previous week. To assess the potential influence of BMI, we used a categorical form of kilograms per meters squared. Binary variables were created for the categories outlined by the National Institutes of Health, National Heart, Lung, and Blood Institute (1998): underweight, BMI less than 18.5; normal weight, BMI of 18.5–24.9; overweight, BMI of 25.0–29.9; obese, BMI greater than or equal to 30. Control variables included age, gender (male), race (White or non-White), education, and lives alone. Several additional variables were included in preliminary multivariate analyses but omitted from the final analyses because they were nonsignificant in various specifications. These variables included marital status, comorbidity, and social isolation.Analytic PlanGiven the overdispersion of the outcome variables, negative binomial regression was selected as the statistical estimator (Long, 1997). All analyses were conducted with Stata 10 (StataCorp, 2007).The analysis was divided into two main stages. First, equations predicting number of ADs used at W2 and the number of ADL domains with ADs at W2 were estimated; two models were estimated for each outcome, where the W1 variable of the outcome was excluded and then included. The first model provides point estimates at W2 and the second reveals change in the outcome. Second, negative binomial regression models were estimated to examine change in the outcomes by W3. We capitalized on the three waves of data to isolate the influence of both initial disability (W1) and incident disability (between W1 and W2, using lagged models). In both stages, we tested for potential statistical interactions (e.g., age and disability) and polynomial terms for independent variables (e.g., disability).ResultsDescriptive statistics for the sample are presented in Table 1. AD use increased substantially over the decade, from 1.26 at W1 to 2.05 at W3 (p < .001). The number of ADL domains with AD also increased over time (p < .001). At W1, the majority of respondents were either normal weight (44%) or overweight (31%); about 17% were obese. Approximately one quarter of respondents reported receiving personal care at W1. Lower body disability was greater than upper body disability, with W2 respondents reporting a value of about 2 for upper body disability and about 4 for lower body disability.Table 2 displays the results of the negative binomial regression analyses for the two W2 dependent variables, and two models are presented for each outcome. Model 1 provides estimates of overall AD use at W2; Model 2 adds AD use at W1 in order to examine the adoption of ADs since W1.Table 2.Negative Binomial Regression Predicting Number of ADs Used at W2 (1999)Number of ADs usedNumber of ADLs with ADsW1 independent variablesModel 1Model 2Model 1Model 2Age0.065a,*** (0.010)b0.057*** (0.010)0.059*** (0.008)0.053*** (0.008)Male−0.073 (0.104)−0.128 (0.101)−0.033 (0.083)−0.054 (0.082)White−0.004 (0.108)−0.021 (0.105)−0.063 (0.085)−0.080 (0.083)Education0.087*** (0.022)0.061** (0.022)0.059** (0.018)0.041* (0.018)Lives alone0.064 (0.074)0.000 (0.073)0.078 (0.059)0.047 (0.059)Underweight (BMI <18.5)c0.133 (0.256)0.076 (0.246)0.066 (0.198)0.058 (0.195)Overweight (BMI 25–29.9)c−0.003 (0.088)0.032 (0.087)0.004 (0.072)0.030 (0.071)Obese (BMI >30)c0.343** (0.106)0.382*** (0.104)0.259** (0.084)0.267** (0.083)Personal care0.407** (0.130)0.145 (0.130)0.328** (0.099)0.138 (0.101)Upper body disability0.019 (0.019)0.022 (0.019)0.008 (0.014)0.009 (0.014)Lower body disability0.181*** (0.019)0.122*** (0.020)0.155*** (0.015)0.116*** (0.016)Number of ADs used, W10.190*** (0.023)0.127*** (0.017)Mortality λ−0.346 (0.300)−0.179 (0.290)−0.305 (0.221)−0.193 (0.218)Observations2,1612,1612,1612,161Pseudo R20.0600.0710.0870.096Likelihood ratio χ2437.45512.42529.60587.62Note: AD = assistive device; ADL = activity of daily living; BMI = body mass index (kg/m2).aUnstandardized coefficient.bStandard error.cCompared with those with a normal BMI (18.5–24.9).*p < .05. **p < .01. ***p < .001.Findings from Model 1 of Table 2 are quite similar across the W2 two outcomes. The number of ADs used and the number of ADL domains with ADs were higher for persons who were older, had more education, obese, received personal care, and had more lower body disability. When AD use at W1 was added in Model 2—predicting change in AD use from W1 to W2—most results are consistent with Model 1, except that personal care is no longer significant. Across all four equations, lower body disability predicts AD use, but upper body disability is not independently related to the outcomes.Supplementary analyses reveal that 42% of respondents present at W1 reported using ADs for bathing and/or indoor mobility difficulty at W2, which are devices that are more useful to those with compromised lower body function and require a certain degree of upper body strength and dexterity to operate. The percentage of respondents using bathing and/or mobility devices increased to 50% by W3.The results displayed in Table 3 carry the analyses forward to W3 and take advantage of data from the prior two waves. Both Models 1 and 2 include the W2 measure of AD use in order to focus on change in use of ADs. Model 1 for number of ADs reveals that age, obesity, and lower body disability at W2 are associated with an increase in AD use by W3. Again, upper body disability was not a significant predictor of AD use. In addition to initial level of lower body disability (W1), Model 1 also includes a variable for incident lower body disability (between W1 and W2), which is associated with the adoption of additional ADs (p < .01). (Incident upper body disability was not included in the equation due to the W2 level of upper body disability having no significant influence on AD use in W3.)Table 3.Negative Binomial Regression Predicting ADs Used at W3 (2004), NLTCSNumber of ADs usedNumber of ADLs with ADsIndependent variablesModel 1Model 2Model 1Model 2Number of ADs used at W20.154a,*** (0.031)b0.133*** (0.032)0.099*** (0.020)0.078*** (0.021)Age0.059*** (0.016)0.068*** (0.016)0.054*** (0.012)0.058*** (0.012)Male−0.219 (0.133)−0.261* (0.123)−0.138 (0.104)−0.179 (0.097)White0.147 (0.153)0.155 (0.152)0.082 (0.119)0.061 (0.119)Education0.026 (0.032)0.036 (0.032)0.034 (0.025)0.041 (0.025)Lives alone0.096 (0.097)0.103 (0.097)0.126 (0.075)0.138 (0.075)Underweight (BMI <18.5)c−0.433 (0.353)−0.506 (0.345)−0.437 (0.275)−0.534* (0.272)Overweight (BMI 25–29.9)c0.184 (0.131)0.261 * (0.126)0.082 (0.100)0.149 (0.096)Obese (BMI >30)c0.539*** (0.145)0.594*** (0.140)0.344** (0.109)0.394*** (0.105)Personal care0.011 (0.206)−0.098 (0.213)0.083 (0.143)−0.015 (0.148)Upper body disability at W20.004 (0.029)0.012 (0.029)−0.005 (0.020)−0.003 (0.020)Lower body disability at W10.067* (0.030)0.060* (0.026)0.079** (0.023)0.070*** (0.020)Incident lower body disability, W1–W20.135** (0.048)1.630** (0.488)0.111** (0.035)1.150*** (0.329)Interaction between age and incident lower body disability between W1–W2−0.019** (0.006)−0.013** (0.004)Mortality λ0.422 (0.387)0.659 * (0.326)0.263 (0.272)0.518 * (0.230)Observations933933933933Pseudo R20.0680.0710.1020.106Likelihood ratio χ2232.74244.07287.94300.84Note: AD = assistive device; ADL = activity of daily living; BMI = body mass index (kg/m2); NLTCS = National Long Term Care Survey.aUnstandardized coefficient.bStandard error.cCompared with those with a normal BMI (18.5–24.9).*p < .05. **p < .01. ***p < .001.Several interactions between incident lower body disability and predictors were tested in supplementary analyses, and the product term of incident disability and age was significant as shown in Model 2 of Table 3. The effect of incident lower body disability on AD use varies across age groups. Incident lower body disability increased AD use, but this relationship was moderated somewhat for the oldest respondents. Note also that with the interaction specified in Model 2, overweight emerged as a significant predictor of using more ADs by W3, and the mortality selection factor (λ) was also significant.The last two equations (columns) in Table 3 display similar models for the second outcome. Findings for the number of number of ADL domains with ADs are quite similar. The pattern of significant relationships for Model 1 is similar to the equation for number of ADs used, but the effect due to initial lower body disability is slightly stronger for number of ADLs with ADs.The findings from Model 2, which include the interaction between age and incident lower body disability, are similar to those for the first outcome, albeit overweight is not significant. The product term in Model 2 reveals the same pattern: Incident lower body disability increases AD use but this relationship is stronger for the younger members of the sample.To better understand the interaction between age and incident lower body disability that was observed for both outcomes, we provide a graphical representation for two different levels of AD use at W3 while controlling for the demographic, BMI, and disability variables. Using a logistic regression model, Figure 1 displays the predicted probability of using one or more ADs at W3 across four age groups while accounting for incident lower body disability between W1 and W2. As shown in Figure 1, the probability of using one or more ADs by W3 varies considerably by age even if there was no new disability between W1 and W2. Incident lower body disability, however, raised the likelihood of using ADs for all age groups, and the slope for persons aged 65–69 years was fairly linear. By contrast, for the oldest respondents, the slope rose faster across incident disability and then tapered off at the highest levels approaching 100% saturation.Figure 1.Predicted probability of using 1+ assistive devices (ADs) by W3 (2004) by incident lower body disability and age (N = 933).Figure 2 predicts the probability of a less frequent scenario—when a person used three or more ADs at W3. Again, there are age differences even with no incident disability. The slopes for the middling age categories are fairly linear. By contrast, the probability of using three or more ADs rose more slowly across incident disability for the youngest age group and rose faster for the oldest age group before tapering off at about 80%.Figure 2.Predicted probability of using 3+ assistive devices (ADs) by W3 (2004) by incident lower body disability and age (N = 933).DiscussionSeveral years ago, Agree and colleagues (2004, p. 267) called for research to examine use of ADs as a “dynamic acquisition process, with attention to age and disability severity.” We sought to answer that call by using three waves of panel data from a nationally representative sample of older adults to examine how ADs are adopted by persons facing ADL disability. The current study examines not only who is more likely to use ADs but also how incident disability and prior AD use influence number of ADs used and the number of ADL domains for which ADs are used.Three research questions were examined in the present study. First, we sought to determine whether incident disability leads to the use of one or more ADs. Although many gerontologists no doubt conceptualize AD use as a consequence of the disablement process, we are unaware of any nationally representative studies that used three or more waves of data to isolate whether incident disability accelerates the adoption of ADs. The present study found consistent evidence across two different outcome variables that incident disability is the precursor of AD adoption. This is not to say that baseline levels are unimportant; they are and result in AD adoption. Nevertheless, incident disability leads to additional use of ADs. For practice, therefore, it may be helpful to identify persons in the early stages of incident disability to optimize the effectiveness of ADs.Second, we sought to determine if both upper and lower body disability increase the likelihood of adopting ADs. Although some studies do not differentiate types of disability, findings from the current study are consistent with other investigations showing that lower body disability is the engine of AD adoption (Cornman & Freedman, 2008; Spillman, 2005; Verbrugge et al., 1997). In addition, the present research builds upon this finding by demonstrating that AD use is not only driven by baseline levels of lower body disability but also by incident lower body disability as well. We believe that there is now ample evidence to call for distinguishing between these two forms of disability in order to identify persons who are likely adopters of ADs.Third, we asked whether increasing disability also leads to domain diffusion in the adoption of ADs. This is an important question for intervention. Are older people in need of additional devices for one domain of ADL functioning or will the need spill over into other domains? Our results point to the latter. When incident lower body disability occurs, it leads not only to additional AD use within one domain, but across multiple domains. Underlying physical capacity is important for the performance of many ADLs. Persons who have difficulty with transfers to the toilet are probably also at risk for transfers involving getting in/out of bed. Future research could confirm this by investigating the sequence of AD adoption among older adults. These results from the NLTCS clarify that incident lower body disability diffuses to AD use across ADL domains.While answering these three research questions, two additional conclusions emerged as noteworthy. First, as suggested by Agree and colleagues (2004), age merits special attention. Based on this national sample of older adults, AD use varied by age in a nonlinear way. Although there are age differences in the likelihood of AD use—such that persons of advanced age (80+ years at W1) used more devices—we also observed a tapering of the rising probability of AD adoption for the oldest age groups (perhaps reflecting a saturation effect for the oldest respondents). Small changes in incident lower body disability were associated with rising AD use among the youngest respondents, an important finding for intervention efforts.Second, we were struck by the consistent predictive ability of obesity as a risk factor for AD use. Across all models, regardless of the outcome considered, obesity was a consistent, significant, and positive predictor. Obesity is clearly a risk factor for disability, but few previous studies have included it as a predictor of AD use. Of those that did, the results revealed either no influence on AD use (Verbrugge & Sevak, 2002) or few findings that merited discussion (Cornman & Freedman, 2008). The current study reveals the lasting impact that obesity has on the use of ADs: Those who were obese at either W1 or W2 were more likely to use a greater number of ADs at W2 and W3, respectively. Given the increased prevalence of adult obesity during recent decades, it should be recognized as a critical risk factor for both disability and AD use.This study’s contributions must be weighed against three limitations. First, although we devised an analytic plan to take advantage of the sequence of events observed in these data, the NLTCS 5-year time frame is fairly coarse when attempting to measure both incident disability and AD use (see also Freedman, Martin, & Schoeni, 2002; Spillman, 2005). Whereas the precise date of AD adoption was not known, we know only that it occurred during the 5-year period between surveys. Accordingly, we used a lagged analytic strategy to guard against conclusions that could be due to reverse causality (because both incident disability and AD adoption can occur within either of the 5-year periods). This strategy enabled us to isolate the influence of incident disability on adoption of ADs, but the 5-year periods are fairly long for studying these dynamic processes. There is clearly a need for research with shorter time frames to detect early stages of incident disability.Second, although we theorized that incident disability is a risk factor for AD use, there are other factors that merit attention, including illness episodes and hospitalizations (Wolff, Agree, & Kasper, 2005). Hospitalizations linked to specific illness episodes such as a stroke serve as a gateway to both incident disability and AD use (Gitlin et al., 1998). Now that we have documented that AD adoption is linked to incident disability, there is a need for future research to identify these “upstream” factors that may lead to both rising disability and AD use.Third, although we know that AD adoption does not occur in a social vacuum, the NLTCS does not provide information on which social network members facilitate the process. We envision caregivers as pivotal agents in the decision to use ADs, but this proposition merits systematic examination in future research.Implications for Policy and PracticeOur findings have implications for caregivers, clinicians, and policy makers, and we articulate two that are especially important. First, according to several studies, poor lower body function is a strong indicator of additional functional decline among older adults (e.g., Guralnik et al., 2000). We found that incident lower body disability triggers a response of increased reliance on ADs—and across multiple ADL domains—to counteract the slippery slope associated with functional decline. Moreover, those who begin experiencing this decline after age 70 are especially sensitive to respond to this change by using more ADs. Therefore, clinicians and caregivers should take note of increases in lower body disability among those benefitting from their care and be prepared to respond by suggesting the use of ADs to effectively bridge the gap between impairment and function. Medicare’s durable medical equipment benefit has helped millions of Americans obtain ADs, but many remain unaware of this benefit (Wolff et al., 2005).Second, obesity is more prevalent among the baby boom generation (born between 1946 and 1965) compared with those born in the prior decades (Leveille, Wee, & Iezzoni, 2005). Based on our finding that obese individuals are more likely to use more ADs—and for a wider number of ADLs—it should be anticipated that AD use will climb as the baby boom population ages. Beyond their sheer numbers, the higher prevalence of obesity among baby boomers and their higher levels of education will make ADs even more widely used in the future.We appreciate the comments on an earlier version of the manuscript by Kathy Abrahamson, Neale R. Chumbler, Ann Howell, Markus H. Schafer, Tetyana P. Shippee, and two anonymous reviewers.AgreeEMThe influence of personal care and assistive devices on the measurement of disabilitySocial Science & Medicine199948427443AgreeEMFreedmanVAIncorporating assistive devices into community-based long-term care: An analysis of the potential for substitution and supplementationJournal of Aging and Health200012426450AgreeEMFreedmanVAA comparison of assistive technology and personal care in alleviating disability and unmet needThe Gerontologist200343335344AgreeEMFreedmanVACornmanJCWolfDAMarcotteJEReconsidering substitution in long-term care: When does assistive technology take the place of personal care?Journal of Gerontology: Social Sciences200560BS272S280AgreeEMFreedmanVASenguptaMAFactors influencing the use of mobility technology in community-based long-term careJournal of Aging and Health200416267307Centers for Disease Control and Prevention and The Merck Company FoundationThe state of aging and health in America 20072007Whitehouse Station, NJMerck Company FoundationRetrieved October 13, 2009, from http://www.cdc.gov/aging/pdf/saha_2007.pdfCornmanJCFreedmanVARacial and ethnic disparities in mobility device use in late lifeJournal of Gerontology: Social Sciences200863S34S41CornmanJCFreedmanVAAgreeEMMeasurement of assistive device use: Implications for estimates of device use and disability in late lifeThe Gerontologist200545347358FerraroKFKelley-MooreJACumulative disadvantage and health: Long-term consequences of obesity?American Sociological Review200368707729FerraroKFShippeeTPAging and cumulative inequality: How does inequality get under the skin?The Gerontologist200949333343FreedmanVAAgreeEMMartinLGCornmanJCTrends in the use of assistive technology and personal care for late-life disability, 1992-2001The Gerontologist200646124127FreedmanVAMartinLGSchoeniRFRecent trends in disability and functioning among older adults in the United StatesJournal of the American Medical Association200228831373146GitlinLNLuborskyMRSchemmRLEmerging concerns of older stroke patients about assistive device useThe Gerontologist199838169180Gosman-HedströmGGClaessonLBlomstrandCAssistive devices in elderly people after stroke: A longitudinal, randomized study—The Göteborg 70+ stroke studyScandinavian Journal of Occupational Therapy20029109118GuralnikJMFerrucciLPieperCFLeveilleSGMarkidesKSOstirGVLower extremity function and subsequent disability: Consistency across studies, predictive models, and value of gait speed alone compared with the short physical performance batteryJournal of Gerontology: Medical Sciences200055AM221M231GuralnikJMFriedLPSaliveMEDisability as a public health outcome in the aging populationAnnual Review of Public Health1996172546HartkeRJProhaskaTRFurnerSEOlder adults and assistive devices: Use, multiple-device use, and needJournal of Aging and Health19981099116HeckmanJJSample selection bias as a specification errorEconometrica197947153161HoenigHTaylorDHSloanFADoes assistive technology substitute for personal assistance among the disabled elderly?American Journal of Public Health200393330337JetteAMHow measurement techniques influence estimates of disability in older populationsSocial Science & Medicine199438937942KayeHSKangTLaPlanteMPMobility device use in the United States. Disability Statistics Report, (14)2000Washington, DCU.S. Department of Education, National Institute on Disability and Rehabilitation ResearchKayeHSYeagerPReedMDisparities in usage of assistive technology among people with disabilitiesAssistive Technology200820194203LaPlanteMPHendershotGEMossAJAssistive technology devices and home accessibility features: Prevalence, payment, need, and trends (No. 217)1992Hyattsville, MDNational Center for Health StatisticsLeveilleSGWeeCCIezzoniLITrends in obesity and arthritis among baby boomers and their predecessors, 1971–2002American Journal of Public Health20059516071613LongJSRegression models for categorical and limited dependent variables1997Thousand Oaks, CASageMannWCJohnsonJLLynchLGJustissMDTomitaMWuSSChanges in impairment level, functional status, and use of assistive devices by older people with depressive symptomsAmerican Journal of Occupational Therapy200862917MannWCOttenbacherKJFraasLTomitaMGrangerCVEffectiveness of assistive technology and environmental interventions in maintaining independence and reducing home care costs for the frail elderly: A randomized controlled trialArchives of Family Medicine19998210217MannWCOttenbacherKJHurrenDTomitaMRelationship of severity of physical disability to pain, functional status, and assistive device use of home-based elderly clientsHome Health Care Management & Practice199587584MantonKGRecent declines in chronic disability in the elderly U.S. population: Risk factors and future dynamicsAnnual Review of Public Health20082991113MantonKGCorderLStallardEChanges in the use of personal assistance and special equipment from 1982 to 1989: Results from the 1982 and 1989 NLTCSThe Gerontologist199333168178MantonKGNational Long-Term Care Survey: 1994, 1999, and 2004 [Computer file]. ICPSR09681-v1Ann Arbor, MIInter-university Consortium for Political and Social Research [distributor]2007-02-07. doi:10.3886/ICPSR09681MantonKGCorderLStallardEChronic disability trends in elderly United States population: 1982-1994Proceedings of the National Academy of Sciences of the United States of America19979425932598MantonKGGuXLambVLChange in chronic disability from 1982 to 2004/2005 as measured by long-term changes in function and health in the U.S. elderly populationProceedings of the National Academy of Sciences of the United States of America20061031837418379MathiesonKMKronenfeldJJKeithVMMaintaining functional independence in elderly adults: The roles of health status and financial resources in predicting home modifications and use of mobility equipmentThe Gerontologist2002422431NagiSZAn epidemiology of disability among adults in the United StatesMillbank Memorial Fund Quarterly197654439467National Institutes of Health, National Heart, Lung, and Blood InstituteClinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults: The evidence report1998Hyattsville, MDNational Center for Health StatisticsRetrieved October 13, 2009, from http://www.nhlbi.nih.gov/guidelines/obesity/ob_gdlns.pdfRussellJNHendershotGELeClereFHowieLJAdlerMTrends and differential use of assistive technology devices: United States, 1994 (No. 292)1997Hyattsville, MDNational Center for Health StatisticsSchoeniRFFreedmanVAMartinLGWhy is late-life disability declining?Milbank Quarterly2008864789SpillmanBAssistive device use among the elderly: Trends, characteristics of users, and implications for modeling2005Report to the Department of Health and Human Services, Assistant Secretary for Planning and Evaluation, Office of Aging and Long-Term Care Policy. Washington, DC: The Urban Institute. Retrieved October 13, 2009, from http://www.urban.org/url.cfm?ID=1001277StataCorpStata Statistical Software: Release 102007College Station, TXStata CorporationTomitaMRMannWCFraasLFBurnsLLRacial differences of frail elders in assistive technologyAssistive Technology19979140151TomitaMRMannWCFraasLFStantonKMPredictors of the use of assistive devices that address physical impairments among community-based frail eldersJournal of Applied Gerontology200423141155VerbruggeLMJetteAMThe disablement processSocial Science & Medicine199438114VerbruggeLMRennertCMadansJHThe great efficacy of personal and equipment assistance in reducing disabilityAmerican Journal of Public Health199787384392VerbruggeLMSevakPUse, type, and efficacy of assistance for disabilityJournal of Gerontology: Social Sciences200257BS366S379WolffJLAgreeEMKasperJDWheelchairs, walkers, and canes: What does Medicare pay for, and who benefits?Health Affairs20052411401149