716
Hall is with the Dept of Geriatric Research, Education, and Clinical Center, Veterans Affairs Medical Center, Durham, NC. McAuley is with the Dept of Kinesiology and Community Health, University of Illinois at Urbana-Champaign, IL.
Examining Indirect Associations
Between Physical Activity, Function, and Disability
in Independent- and Assisted-Living Residents
Katherine S. Hall and Edward McAuley
Background: Few studies have examined physical activity behavior and its associated outcomes in older adults
living in retirement communities. Guided by the disablement model and social cognitive theory, we tested a cross-sectional model in which physical activity was hypothesized to influence disability indirectly through self-efficacy, functional performance, and functional limitations. Methods: One hundred six older men and women residing in independent-living (ILF) assisted-living (ALF) facilities completed self-report measures of self-efficacy, function, and disability. Objective assessments of physical activity and functional performance were conducted using waist-mounted accelerometers and the short physical performance battery (SPPB), respectively. Path analysis was used to examine the proposed associations among constructs. Results: Older adults who were more active were also more efficacious and had better physical function and fewer functional limitations. Only higher levels of self-efficacy were associated with less disability. The effects of individual-level covariates were also examined. Conclusions: This cross-sectional study is among the first to examine the associations between physical activity, function, and disability among older adults residing in ILFs and ALFs. Future research addressing the physical and psychological needs of this growing population is warranted.
Keywords: accelerometer, structural equation model, older adults, disablement
In a recent position paper, the Centers for Disease Control and Prevention’s Healthy Aging Network out-lined research priorities to address issues relative to aging, physical activity, and public health.1 Four research
priorities were identified, one of which reflected the need to measure physical activity patterns of older adults, par-ticularly those with varying levels of frailty. Prohaska et al further recommended that existing models and theories of physical activity behavior be expanded to incorporate factors associated with the aging process, including dis-ability and functional limitations, which have implica-tions for quality of life. Residents of independent (ILF) and assisted living facilities (ALF) are a functionally heterogeneous group,2 and represent a continually
grow-ing segment of the population.3 As the number of
indi-viduals entering these retirement communities continues to increase, so too does our need to better understand the associations between physical activity and health-related outcomes such as function and disability. To date, there have been relatively few attempts to examine outcomes associated with physical activity and function and deter-mine the possible underlying mechanisms.
Although physical activity has been demonstrated to positively influence functional performance,4–6 the
effects of individual behaviors (eg, physical activity) and psychosocial factors (eg, perceived abilities) on disability are less clear.4–6 Observational studies have demonstrated
a protective effect of physical activity on risk of dis-ability,7,8 whereas the results of randomized controlled
exercise interventions are mixed.5,9,10 Postponing the
onset and minimizing the severity of age-related disability is of importance to health providers and public health officials alike. Consequently, identifying those factors associated with disability, particularly those which are amenable to change, has implications for future interven-tions to minimize age-related disability and dependence on costly long-term care.
In an effort to more closely align physical activity research with disability research, Stewart11 proposed
modifications to Nagi’s12 disablement model. This
modified disablement model included physical activity as a determinant of functional decline and disability, and made a distinction between functional limitations and functional performance. Specifically, performance on functional tasks is identified as a distinct step in the disablement process, preceding functional limitations. The inclusion of performance as a distinct step in the disablement model would consequently allow health researchers to examine the extent to which physical activity is associated with subsequent performance on functional tasks, and in turn, how changes in functional performance influence self-reported limitations.
An important component of Nagi’s model which has received less attention in previous studies is the
recognition that not all limitations precipitate disability, and that similar impairments may result in very dif-ferent patterns of disability across individuals. These concessions by Nagi imply a role for extraindividual factors in the disablement model, however, where these factors lie along the disability pathways is not specified. Self-efficacy, or an individual’s belief in their ability to successfully complete a task, is the central component of social cognitive theory13,14 and has been shown to be
sig-nificantly associated with physical activity,15 functional
limitations,16,17 and functional performance.18
In a study of older women, McAuley and col-leagues16,17 report a mediating effect of self-efficacy
between physical activity, functional performance, and functional limitations, such that the effects of physical activity on functional limitations were indirect, operat-ing through self-efficacy and functional performance. To date, the role of self-efficacy in the pathway from physical activity to disability has not been examined. Previous research suggests that changes in physical activity are associated with changes in specific domains of self-efficacy, namely efficacy relative to balance, in older adults.19 The very definition of disability, as the
ratio of task demands to individual capabilities, suggests a potential role for balance self-efficacy beliefs.
In this cross-sectional study, we examined the relationships among physical activity, balance self-efficacy, physical function, and disability in ILF and ALF residents. Guided by the disablement model and social cognitive theory, we tested a model in which physical activity was hypothesized to influence disability indi-rectly through self-efficacy, functional performance, and functional limitations.
Methods
Participants and RecruitmentOlder adults residing in ILFs and ALFs were recruited to participate in a study of health and aging. Administrators of ILFs and ALFs located in the Midwestern region of the United States were contacted to provide initial consent for the distribution of flyers to residents. A recruitment orientation session was held at each site during which time the purpose of the study was presented, along with the inclusion criteria and testing procedures. In addi-tion, the project coordinator discussed the content of the questionnaire packets and explained the functional performance assessments. Recruitment strategies were identical at each site.
Interested individuals were required to be at least 65 years of age and pass a basic cognitive screening task20
to qualify. All study procedures were approved by a Uni-versity Institutional Review Board and each participant completed a written informed consent before study entry. Measures
Demographics. Basic demographic information including age, sex, race, education, marital status,
income, and ethnicity were collected during the intake interview. Whether individuals used an assistive walking device was also noted at this time.
Mental Status. Cognitive impairment was assessed verbally using the Short Portable Mental Status Ques-tionnaire (SPMSQ).20 Individual responses to the 10
items are scored as correct or incorrect, resulting in a total score ranging 0 to 10, with higher scores indica-tive of better cogniindica-tive function. Total scores ≤ 7 are considered low, and indicative of cognitive deficiency. Individuals who missed 3 or more items were excluded from the study.
Health Status. Health status was determined via self-report by asking participants to indicate which, if any, conditions they currently suffered from referencing a list of 17 chronic conditions. A health status score for use in these analyses was computed as the sum of responses to cardiovascular disease, pulmonary disease, peripheral vascular disease, arthritis, hypertension, and diabetes items.
Physical Activity. Participants were asked to wear an Actigraph accelerometer (Manufacturing Technology Inc. (MTI), Pensacola, FL), for a period of 7 days as a measure of objective physical activity. The minimum period of valid monitoring to be included in data analysis was 5 days, which is required to achieve 80% reliabil-ity.21 Participants were instructed to wear the monitor on
the nondominant hip, under clothing, and fastened to a belt worn around the waist. The accelerometer was to be worn during all waking hours, except for when bathing or swimming. Activity data were collected in 1-minute intervals (epochs). In comparisons with other activity monitors and self-report assessments of physical activity, the Actigraph accelerometer has demonstrated acceptable reliability and validity among young, middle-age, and older adults.22,23
Functional Limitation and Disability. Functional limi-tation and disability were assessed using the abbreviated versions of the Functional Limitations and Disability-Limitations subscales from the Late-Life Function and Disability Instrument (LL-FDI).24–26 The response scale
for the Functional Limitations subscale ranges from 0 (cannot do) to 5 (no difficulty), and the response scale for the Disability-Limitations subscale ranges from 1 (com-pletely limited) to 5 (not at all limited). Higher scores indicate fewer functional limitations and less disability. The internal consistencies of the abbreviated function subscale (α = .87) and abbreviated disability limitation subscale (α = .81) were acceptable in this sample.
Functional Performance. Lower-extremity physical performance was assessed using the Short Physical Performance Battery (SPPB).27,28 The SPPB score is
based on timed measures of gait speed, ability to rise from a chair, and standing balance. Each of the individual performance measures were assigned values of 0 to 4, with 4 indicating the highest level of performance and 0 the inability to complete the test. The overall physical performance score combined the results of the gait speed,
chair stands, and balance tests, with scores ranging from 0 to 12. Higher scores correspond to better performance.
Self-Efficacy. Self-efficacy for balance was assessed with the Activity-Specific Balance Confidence Scale (ABC).29 This 16-item scale asks participants to rate their
confidence to execute activities of daily living without losing their balance. Example items included “walking up and down a ramp,” “bend over to pick up an object from the floor.” The ABC scale uses a 100-point per-centage scale, ranging from 0% (not at all confident) to 100% (highly confident) and showed excellent internal consistency (α = .95).
Procedures
Measurement sessions were completed in 3 stages. First, participants who qualified for the study were asked to complete an informed consent form approved by a Uni-versity Institutional Review Board and questionnaires including demographics and general health information. Each participant was then fitted with an activity moni-tor and instructed to record the dates which marked the beginning and end of their 7-day monitoring period. Second, participants were mailed a second packet of questionnaires that included self-efficacy, functional limitations, and disability and asked to bring it with them to their scheduled follow-up session. Each participant was given approximately 2 days to complete the questionnaire packet. For those individuals who expressed difficulty with reading, the research coordinator completed the packets via interview in the participant’s residence (n = 6). Third, during the final testing session the activity monitor and record of use form were collected and the SPPB was conducted with each participant.
Data Analysis
Individual accelerometer data files were viewed for abnormal data (eg, unusually low counts, continuous data with the same counts). Such data files were excluded from data analysis (n = 1). Secondly, participants’ logs were checked for periods when the accelerometer was not worn and matched against the recorded accelerometer data. Days on which the monitor was not worn or was worn for fewer than 10 hours were excluded. Data files with fewer than 5 days of valid data, required to achieve 80% reliability,21 were excluded from the analysis (n =
3). Data files which included more than 5 days of valid data were trimmed such that activity measure included 3 week days and 2 weekend days.
Model Specification. All analyses were performed using Mplus structural equation modeling software (V. 5.1).30 The proposed structural model shown in Figure
1 was tested on the basis of the hypothesized relation-ships among observed variables. This model specified (a) direct effects of physical activity on self-efficacy and functional performance; (b) an indirect effect of physical activity on functional limitations through self-efficacy
and functional performance; and (c) direct effects of self-efficacy, functional performance, and functional limitations on disability.
Covariates. An additional test of the hypothesized model was conducted controlling for age, race, gender, education, the type of residential facility (independent or assisted), number of chronic conditions, and use of a walking aid. This analysis allowed for the assessment of whether the fit of the model and the proposed relation-ships were differentially influenced by these important health and demographic variables.
Model Fit. Goodness of fit was assessed using the chi-square statistic, root mean chi-square error of approximation (RMSEA), and standardized root mean square residual (SRMR), in combination with comparative fit (CFI) and nonnormed fit indices. RMSEA values ≤ 0.06 and SRMR values ≤ 0.08 demonstrate close fit of the model to the data.31 Values approximating 0.95 or greater for the CFI
indicate good model-data fit.31
Results
Descriptive StatisticsOf the 196 individuals who attended an on-site informa-tion session, 120 expressed interest for participating in the study, of which 3 were excluded from participation due to cognitive deficiencies assessed in the intake interview. Over the course of the study, 11 individuals ceased partici-pation because loss of interest (n = 5), unwilling to wear the activity monitor (n = 5), or personal health or illness (n = 1). The final sample consisted of 106 older men (n = 27) and women (n = 79) who completed all testing; the majority of whom were recruited from ALFs (75.5%). Participants in this study ranged between 68 and 99 years of age, with an average age of 85 years. As expected given the demographics of this age group and the geographic location of this study, the sample was primarily Caucasian (99.1%), female (74.5%), widowed (67.9%), and well educated (62.3% college/university educated), with a household income of greater than $40,000 (36.7%). The most prevalent chronic health conditions reported in this sample were cardiovascular disease (25.5%), arthritis (51.9%), and hypertension (48.1%). 67% (n = 71) of the sample indicated that they use an assistive walking device. The average length of stay in a residential facility was less than 3 years (mean = 33 months), which is similar to that of the typical ALF resident.32
Of the 30 facilities contacted, 13 facilities agreed to participate. Among those who declined participation, 12 were not interested in participating and 5 were will-ing but unable to participate within the timeframe of the study. Study participants were recruited largely from assisted living facilities and those residential care facili-ties which offered tier-style care, in which independent and assisted living facilities were offered within the same residential campus (ie, continuous care). The majority of these facilities was located in rural communities (92%)
and were owned and operated by corporate shareholders. Participants were recruited equally across small, medium, and large facilities.
Descriptive statistics for all variables included in the hypothesized model are shown in Table 1. As is evident in the table, study participants were active at a level of 73,000 activity counts per day indicating very low levels of activity and consistent with values in other reports.33,34 This sample was moderately efficacious,
but demonstrated restricted functional capability and limitations as well as disability. All model variables demonstrated significant bivariate associations with each other (P < .05), with the exception of physical activity and functional limitations.
Step 1: Testing the Hypothesized Model The next step of data analysis involved testing the hypothesized pattern of relationships. This model pro-vided an excellent fit to the data (χ2 = 1.00, df = 2, P =
.61, RMSEA [90% CI] = 0.00 [0.00–0.16], CFI = 0.99, SRMR = 0.02). Standardized parameter estimates indi-cated that physical activity had significant direct effects on self-efficacy (β = .22). Both physical activity (β = .24) and self-efficacy (β = .51) were significantly associated
with functional performance. In turn, self-efficacy was significantly related to functional limitations (β = .40) as was functional performance (β = .21). Self-efficacy (β = .39), but not functional performance (β = .04) or func-tional limitations (β = .12) was significantly associated with disability. Thus, older adults who were more active were also more efficacious and had better physical func-tion. Higher levels of self-efficacy were associated with fewer limitations, and contributed to less disability. This cross-sectional model accounted for 24% of the variance in disability. These relationships are shown in Figure 1.
Although the hypothesized model demonstrated a good fit to the data, this model only partially supports the pattern of relationships originally predicted. Thus, a series of alternative, theoretically viable, models were tested next in an effort to identify a more parsimonious model. Step 2: Testing Alternative Models
An exploratory model in which a direct path was specified from physical activity to disability was tested. This model was designed to reflect the line of thinking prominent in observational studies and randomized controlled trials of physical activity and disability. Although this model provided a good fit to the data (χ2 = 0.01, df = 1, P = .91,
Table 1 Descriptive Statistics for All Measures
Variable Mean SD Participants’ range Possible range
Accelerometer (activity counts) 73,301 38,960 4016–221,269 N/A Activities specific balance scale 53.3 25.4 0.0–97.5 0.0–100.0
SPPB score 5.4 2.5 1.0–12.0 0.0–12.0
LL-FDI functional limitation 45.4 11.6 19.0–72.0 15.0–75.0 LL-FDI disability limitation 28.0 6.7 13.0–40.0 8.0–40.0
Abbreviations: SPPB, Short Physical Performance Battery; LL-FDI, Late Life Function and Disability Instrument.
Note. Higher scores on the LL-FDI functional limitation and disability measures denote fewer functional limitations and less disability.
Figure 1 — Hypothesized path model of relationships between physical activity, self-efficacy, functional performance, functional limitations, and disability. Note: All standardized path coefficients are statistically significant (P < .05). Dotted lines indicate paths that were tested but were not significant. Higher scores on the functional limitations and disability measures reflect fewer limita-tions/disability. Fit indices for the hypothesized model: χ2 = 1.00, df = 2, P = .61, RMSEA [90% CI] = 0.00 [0.00–0.16], CFI =
RMSEA [90% CI] = 0.00 [0.00–0.11], CFI = 0.99, SRMR = 0.00), there was no significant association between physical activity and disability (β = .09).
Next, a model was examined in which the path from self-efficacy to disability was not specified. This model was designed to test an assumption that self-efficacy was a major correlate of disability perceptions, thus negat-ing any associations between disability and functional performance and functional limitations. This model resulted in significant standardized parameter estimates for functional limitations (β = .26) and an effect that approached significance for functional performance (β = .17) on disability but physical activity (β = .10) was not directly associated with disability. In addition, the fit of this alternative model was not acceptable (χ2 = 11.93,
df = 2, P < .01, RMSEA [90% CI] = 0.22 [0.11–0.34], CFI = 0.91, SRMR = 0.05); suggesting a pivotal role of self-efficacy in perceptions of disability.
The results of these exploratory analyses suggest that despite the lack of associations observed between functional limitations and disability and functional performance and disability, the original model proposed here is indeed the best fitting solution for these cross-sectional data.
Step 3: Testing the Effects
of Demographics, Chronic Conditions, and Walking Aids on Structural Relations The next set of analyses examined the extent to which age, gender, education, type of residential facility, and health status differentially influenced the fit of the hypothesized model or altered any of the original rela-tionships. Once again, the model fit was not substantially changed (χ2 = 0.45, df = 2, P = 0.80, RMSEA [90% CI]
= 0.00[0.00–0.12], CFI = 0.99, SRMR = 0.01) and all of the hypothesized paths in the original model retained their significance level, with the exception of the path from functional performance to functional limitations, which now approached significance (β = .19, P = .06). This model accounted for 30% of the variance in disability. Several significant relationships between the covariates and model components were observed, however. As expected, residing in an ALF was associated with greater disability (β = .23) than residing in an ILF. Men were more efficacious (β = .26) than women, and higher level of education (β = –.20) was associated with less physi-cal activity. The use of an assistive walking device was associated with lower levels of self-efficacy (β = –.31).
Discussion
In this cross-sectional study, we examined the relation-ships among physical activity, self-efficacy, physical function, and disability in ILF and ALF residents. Guided by the disablement model and social cognitive theory, we tested a model in physical activity was hypothesized
to influence disability through the direct effects of self-efficacy, functional performance, and functional limitations. In this study we found that individuals who were more physically active were more efficacious and demonstrated better functional capacity, both of which were associated with fewer functional limitations; thus lending further support to the model proposed by McAu-ley and colleagues.17
Contrary to models of the disablement process in which functional performance influences disability, only self-efficacy was significantly associated with disability in this study. Although functional performance (r = .32), functional limitations (r = .35), and physical activity (r = .20) demonstrated significant bivariate correlations with disability initially, the association between self-efficacy (r = .48) and disability was somewhat stronger. Thus, once included in the model, the path from self-efficacy to disability appears to eclipse any other associations, rendering them nonsignificant. Indeed, in a post hoc model in which the path from self-efficacy to disability was excluded, the standardized parameter estimates for functional limitations were significant, though the model fit was not acceptable. Our results suggest that interven-tions targeting self-efficacy warrant investigation in this population.
Despite being included in numerous models of dis-ablement, very little research has investigated the path-ways leading from physical activity to disability.5 Those
few studies which have included disability as an outcome have largely operated under the assumption that physical activity lessens disability by improving impairments.4
Consequently, the role of important individual-level vari-ables, such as self-efficacy, on disability, has largely gone unstudied. The inconsistent findings of previous research on the physical activity-disability relationship,5,9,10
coupled with our results, suggest that self-efficacy plays a pivotal role in formulating perceptions of disability. The patterns of associations reported here challenge the long-held assumptions that have shaped physical activity interventions and suggest a complex series of interactions between individual characteristics and task performance demands underlie disability.
Indeed, the lack of association between disability and measures of functional performance, functional limitations, and physical activity suggests that improving physiologic indicators (eg, muscle strength) via physical activity may not be sufficient in and of itself to impact disability. Although cross-sectional, our results lend fur-ther support to previous studies of physical activity and disability7,9,35 which demonstrate the ability of physical
activity to attenuate incident disability, but report small or nonsignificant effects of physical activity on recovery from disability. These results are important when one considers the motivation for the oldest-old to participate in physical activity programs and the exercise intensity which is prescribed within these programs. Indeed, it appears as though activity interventions in this segment
of the population would be best served by targeting the physical activity threshold needed for preserving function on daily tasks as opposed to disability recovery.
In this sample of frail older adults, however, perhaps the intensity of activity is not as critical as doing some activity. It seems reasonable to expect that activity inter-ventions aimed at increasing walking behavior, regardless of the intensity of that activity, can be expected to reap dividends in the way of self-perceptions and functional capability among ILF and ALF residents, which may carry implications for health-related quality of life. How-ever, the implications of such a design on program adher-ence and willingness to participate are unclear. Clearly, future longitudinal and intervention studies are needed to examine this pattern of relationships more thoroughly and to further delineate these patterns of associations. Indeed, although the model tested here was informed by both theory and previous research, the cross-sectional nature of the study precludes us from drawing any conclusions relative to causality. Further examination of how these relationships are maintained across time is warranted.
The effects of health status and demographic vari-ables on the hypothesized pattern of relationships were also examined. Controlling for these variables had no significant effect on the overall fit of the model; however, doing so rendered the previously significant path between functional performance and functional limitations non-significant. No significant associations were observed between individual health status and demographic factors and functional performance or functional limitations. Although significant associations between education and gender and select model variables were observed, these associations are likely driven by this sample being com-prised predominantly of well-educated, women; a sample which is reflective of ILFs and ALFs as a whole.36,37
Finally, a negative association between the use of an assistive walking device and balance self-efficacy was observed. This association was to be expected, consider-ing the reason that assistive walkconsider-ing devices are sought/ prescribed: to help with poor balance. Thus, although the use of a walking aid was not associated with significant differences in functional performance or limitations, these individuals report still being wary of their ability to per-form activities that rely on balance. It is important to note that although this model provided a good fit to the data, the inclusion of covariates dramatically decreased the power of our model. Future studies with larger samples are needed to clarify whether these reflect true differences or low statistical power.
This study has several limitations, including small sample size, cross-sectional design, and the homogeneity of the sample, which limit the conclusions that can be drawn. Indeed, we know very little about the 17 facili-ties that were contacted but declined participation and how they differ from the facilities used here. Similarly, the characteristics of those ILF and ALF residents who chose not to participate and how they differed from the
sample studied here in terms of functional status or level of disability are not known. Obtaining such information will be important for establishing the generalizability of our findings to other facilities and other ILF and ALF residents in subsequent studies.
Despite efforts to recruit a representative sample that included both for-profit and nonprofit, private pay and insurance-supplemented, and urban and rural facilities, the participants in this study were predominantly white women of higher socioeconomic status. Although the demographic characteristics of this sample are largely reflective of those who reside in care facilities, the results of this study may not be generalizable to lower-income or ethnic minority residents. Research suggests that low-income and ethnic minority older adults experience greater levels of functional impairment than affluent and white older adults.38,39 However, studies have
dem-onstrated that the use of assisted and independent living facilities among ethnic minorities is disproportionate to their increased need and proportions in the national population.36,37,40
Although we view the use of an objective measure of physical activity as a strength of the study, this method of assessment is not without limitation. Indeed, previous studies that have used objective measures of physical activity, such as pedometers, have discussed that the mere act of receiving such a device may prompt some individuals to increase their activity behavior.41 Although
we recognize that such an effect may be present in our sample, it is important to note that unlike a pedometer, which visually displays the cumulative number of steps taken, the Actigraph provided no visual display of the data collected. As such, participants in this study were unable to monitor their behavior patterns and perhaps less likely to make substantive changes in their behavior to achieve some benchmark from the monitor.
Finally, we acknowledge that the influences of other social cognitive factors such as affective status and social support, which may also serve as mediators, were not included in this study. Although we chose to assess self-efficacy for balance given that balance is likely to be implicated in function, other domains of self-efficacy need to be examined in future studies. The models tested here only explained 24 to 30% of the variance in disability, suggesting that other factors not included in this model may have potential for affecting disability in older adults.
Despite these limitations, this study is the first to our knowledge to examine the relationships between physical activity, function, and disability among older adults resid-ing in ILFs and ALFs. Consequently, very little is known about residents’ levels of activity, physical function, and the implications such variables have on disability. The use of accelerometers in this population is also a novel contribution to the literature, and provides a preliminary, objective, look at the physical activity behavior of these individuals. As expected, given the level of care and the
structural characteristics of these facilities, participants in this sample were relatively inactive. Future work examin-ing the energy demands of daily activities is needed in frail older adults to allow quantification of the time spent in light, moderate, and vigorous activity. Currently, valid activity monitor cut points for use in the oldest-old have not been identified, precluding us from analyzing and commenting on the intensity of physical activity in our sample. It is also important to consider that for these indi-viduals, physical activity levels may largely be dependent on the programs and exercise facilities available and not strictly a reflection of individual functional limitations. Thus, examining characteristics of the built environment as well as the programs and facilities available in ILF and ALF residences and their impact of resident activity levels are needed.
Acknowledgments
This material is based upon work KSH conducted at the Uni-versity of Illinois and was supported by a research grant from the American College of Sports Medicine.
References
1. Prohaska T, Belansky E, Belza B, et al. Physical activity, public health, and aging: critical issues and research priori-ties. J Gerontol B Psychol Sci Soc Sci. 2006;61(5):267–273. 2. Fonda SJ, Clipp EC, Maddox GL. Patterns in functioning
among residents of an affordable assisted living housing facility. Gerontologist. 2002;42:178–187.
3. Cummings SM. Predictors of psychological well-being among assisted-living residents. Health Soc Work. 2002;27(4):293–302.
4. Jette AM, Keysor JJ. Disability models: implications for arthritis exercise and physical activity interventions.
Arthritis Rheum. 2003;49:114–120.
5. Keysor JJ. Does late-life physical activity or exercise prevent or minimize disablement?: a critical review of the scientific evidence. Am J Prev Med. 2003;25:129–136. 6. Keysor JJ, Jette AM. Have we oversold the benefit of
late-life exercise? J Gerontol A Biol Sci Med Sci. 2001;56:412– 423.
7. Boyle PA, Buchman AS, Wilson RS, Bienias JL, Bennett DA. Physical activity is associated with incident disability in community-based older persons. J Am Geriatr Soc. 2007;55(2):195–201.
8. Landi F, Onder G, Carpenter I, Cesari M, Soldato M, Bernabei R. Physical activity prevented functional decline among frail community-living elderly subjects in an international observational study. J Clin Epidemiol. 2007;60(5):518–524.
9. von Bonsdorff MB, Leinonen R, Kujala UM, et al. Effect of physical activity counseling on disability in older people: a 2-year randomized controlled trial. J Am Geriatr Soc. 2008;56(12):2188–2194.
10. Latham NK, Bennett DA, Stretton CM, Anderson CS. Systematic review of progressive resistance strength training in older adults. J Gerontol A Biol Sci Med Sci. 2004;59(1):48–61.
11. Stewart AL. Conceptual challenges linking physical activ-ity and disabilactiv-ity research. Am J Prev Med. 2003;25:137– 140.
12. Nagi SZ. An epidemiology of disability among adults in the United States. Milbank Mem Fund Q Health Soc. 1976;54(4):439–467.
13. Bandura A. Social foundations of thought and action: a Social Cognitive Theory. Englewood Cliffs, NJ: Prentice Hall; 1986.
14. Bandura A. Self-efficacy: the exercise of control. New York: W.H. Freeman; 1997.
15. McAuley E, Elavsky S, Motl RW, Konopack JF, Hu L, Mar-quez DX. Physical activity, self-efficacy, and self-esteem: longitudinal relationships in older adults. J Gerontol B Psychol Sci Soc Sci. 2005;60(5):268–275.
16. McAuley E, Konopack JF, Morris KS, et al. Physical activity and functional limitations in older women: influ-ence of self-efficacy. J Gerontol B Psychol Sci Soc Sci. 2006;61(5):270–277.
17. McAuley E, Morris KS, Doerksen SE, et al. Effects of change in physical activity on functional limitations in older women: mediating roles of physical func-tion performance and self-efficacy. J Am Geriatr Soc. 2007;55(12):1967–1973.
18. Seeman T, Chen XG. Risk and protective factors for physi-cal functioning in older adults with and without chronic conditions: MacArthur Studies of successful aging. J Gerontol B Psychol Sci Soc Sci. 2002;57:135–144. 19. McAuley E, Doerksen SE, Morris KS, et al. Pathways
from physical activity to quality of life in older women.
Ann Behav Med. 2008;36(1):13–20.
20. Pfeiffer E. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients.
J Am Geriatr Soc. 1975;23:433–441.
21. Matthews CE, Ainsworth BE, Thompson RW, Bassett DR. Sources of variance in daily physical activity levels as measured by an accelerometer. Med Sci Sports Exerc. 2002;34:1376–1381.
22. Swartz AM, Strath SJ, Bassett DR, O’Brien WL, King GA, Ainsworth BE. Estimation of energy expenditure using CSA accelerometers and hip and wrist sites. Med Sci Sports Exerc. 2000;32(9, Suppl):450–456.
23. Focht BC, Sanders WM, Brubaker PH, Rejeski WJ. Initial validation of the CSA activity monitor during rehabilita-tion exercise among older adults with chronic disease. J Aging Phys Act. 2003;11(3):293–304.
24. Haley SM, Jette AM, Coster WJ, et al. Late life function and disability instrument: II. Development and evaluation of the function component. J Gerontol A Biol Sci Med Sci. 2002;57(4):217–222.
25. Jette AM, Haley SM, Coster WJ, et al. Late life function and disability instrument: I. Development and evaluation of the disability component. J Gerontol A Biol Sci Med Sci. 2002;57(4):209–216.
26. McAuley E, Konopack JF, Motl RW, Rosengren K, Morris KS. Measuring disability and function in older women: psychometric properties of the late life function and disability instrument. J Gerontol A Biol Sci Med Sci. 2005;60:901–909.
27. Guralnik JM, Ferrucci L, Simonsick EM, Salive ME, Wal-lace RB. Lower-extremity function in persons over the age of 70 years as a predictor of subsequent disability. N Engl J Med. 1995;332(9):556–561.
28. Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol. 1994;49(2):85–94.
29. Powell LE, Myers AM. The Activities-Specific Balance Confidence Scale (ABC). J Gerontol A Biol Sci Med Sci. 1995;50:28–34.
30. Muthén LK, Muthén BO. Mplus user’s guide. Los Angeles: Muthén & Muthén; 2007.
31. Hu L, Bentler PM. Cutoff criteria for fit indices in covari-ance structure analysis: conventional versus new alterna-tives. Struct Equ Modeling. 1999;6:1–55.
32. American Association of Retired Persons. Assisted living in the United States. 2004.
33. Bergman RJ, Bassett DRJ, Muthukrishnan S, Klein DA. Validity of 2 devices for measuring steps taken by older adults in assisted-living facilities. J Phys Act Health. 2008;5(1, Suppl):166–175.
34. Pruitt LA, Glynn NW, King AC, et al. Use of accelerom-etry to measure physical activity in older adults at risk for mobility disability. J Aging Phys Act. 2008;16:416–434. 35. von Bonsdorff MB, Leinonen R, Kujala UM, et al. Effect
of physical activity counseling on home care use in older people. J Am Geriatr Soc. 2009;57(3):571–573.
36. Hawes C, Phillips CD, Rose MS. High service or high privacy assisted living facilities, their residents and staff: results of a national survey of facilities. Washington, D.C.: U.S. Department of Health and Human Services, Office of Disability, Aging, and Long-Term Care Policy; 2000. 37. Spillman BC, Liu K, McGilliard C. Trends in residential
long-term care: use of nursing home and assisted living and characteristics of facilities and residents. Washington, D.C.: U.S. Department of Health and Human Services, Office of Disability, Aging, and Long-Term Care Policy; 2002.
38. Liao Y, McGee DL, Cao G, Cooper S. Black-white differ-ences in disability and morbidity in the last years of life.
Am J Epidemiol. 1999;149(12):1097–1103.
39. Mendes de Leon CF, Glass TA. The role of social and personal resources in ethnic disparities in late-life health. In: Anderson NB, Bulatao RA, Cohen B, eds. Critical perspectives on racial and ethnic differences in health in late life. Washington, D.C.: The National Academies Press; 2004.
40. U.S. Census Bureau. The older population in the United States: March 2000. Detailed tables (PPL-147). 2000. 41. Bravata DM, Smith-Spangler C, Sundaram V, et al. Using
pedometers to increase physical activity and improve health. JAMA. 2007;298(21):2296–2304.
