| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Rush Institute for Healthy Aging, Rush-Presbyterian-St. Luke's Medical Center, Chicago, Illinois.
2 Epidemiology, Demography, and Biometry Program, National Institute on Aging, Bethesda, Maryland.
3 Department of Biostatistics, Johns Hopkins University Bloomberg School of Public Health, Baltimore, Maryland.
Address correspondence to Carlos F. Mendes de Leon, PhD, Rush Institute for Healthy Aging, Rush-Presbyterian-St. Luke's Medical Center, 1645 W. Jackson Blvd., Suite 675, Chicago, IL 60612. E-mail: cmendes{at}rush.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Data from 93 of the 102 women who participated in the Weekly Substudy of the Women's Health and Aging Study (WHAS) were used to explore the association of changes in physical function with disability. The WHAS Substudy included 24 weekly assessments of three standard performance tests and self-reported disability in activities of daily living (ADLs) and basic mobility.
Results. Using random-effects models, we found small but significant (ps < .01) changes in ADL and mobility disability during weekly follow-up. Baseline performance scores were significantly associated with both ADL and mobility disability (ps < .001), accounting for 27% and 36% of the between-person variability in each type of disability, respectively. After adjustment for baseline scores, change in performance scores was significantly associated with ADL disability (ß = 0.08, p < .01) and mobility disability (ß = 0.12, p < .001), but accounted only for a small proportion (<10%) of the variability in the rate of change in disability outcomes. There was no evidence for an additional effect on either type of disability because of having a single episode of a higher or lower than usual performance score, or because of periods of at least 4 consecutive higher or lower than usual performance test scores.
Discussion. Basic physical functions account for a substantial proportion of the heterogeneity in ADL and mobility disability among older disabled women, but have a relatively small impact on short-term changes in either type of disability. Effective prevention of disability may require attention to a wider array of risk factors than just limitations in basic physical functions.
Disability is generally conceptualized as the diminished ability to perform specific tasks and activities that are considered essential for self-care and independent living (Pope & Tarlov, 1991). Among older persons, this disability usually results from the cumulative damage of the multiple chronic disease processes that become more common and severe with increasing age. These disease processes cause impairment in physical, cognitive, and sensory domains that, in turn, lead to loss in specific functions and disability (Fried & Guralnik, 1997). These pathways are explicated in prevailing disability models in which development of disability follows a progressive biological process, which begins at the level of disease pathology, including biochemical or physiological abnormalities. These abnormalities will, in turn, lead to impairment in specific body systems and functional limitations in basic physical and mental functions, which are eventually expressed in increasing difficulties to perform basic social roles and routine daily tasks (Nagi, 1976; Pope & Tarlov, 1991; Verbrugge & Jette, 1994).
According to general disability models, impairments in specific physical functions, such as muscle strength and balance, form an essential causal influence in the development of disability, consistent with the notion that these functions constitute the basic physical capacities necessary to perform basic or routine daily tasks successfully, such as activities of daily living (ADLs). A considerable amount of research has been devoted to establish the validity of this pathway by examining the nature of the interrelationship between measures of physical function and disability. Basic physical functions are typically assessed using performance-based tests, whereas assessment of disability usually relies on self-report information. In cross-sectional studies, generally strong and consistent relationships have been found between performance-based measures of physical function and self-reported disability (Bruce, Seeman, Merrill, & Blazer, 1994; Cress et al., 1995; Daltroy, Larson, Eaton, Phillips, & Liang, 1999; Guralnik et al., 1994; Judge, Schektman, Cress, & the FICSIT Group, 1996; Kelly-Hayes, Jette, Wolf, D'Agostino, & Odell, 1992; Kempen, Steverink, Ormel, & Deeg, 1996; Kempen, van Heuvelen, et al., 1996; Myers, Holliday, Harvey, & Hutchinson, 1993; Reuben, Valle, Hays, & Siu, 1995). For example, Kempen and colleagues have shown that a summary measure of performance-based tests of function explains about 38% of the variance in self-reported ADL levels (Kempen, Steverink, et al., 1996). Daltroy and colleagues (1999) found that about 50% of the variance in self-reported disability was accounted for by a measure of observed physical function. Similarly robust, although somewhat more modest, cross-sectional associations have been found in a number of other studies, including the Established Populations for the Epidemiologic Studies of the Elderly (EPESE) study (Guralnik et al., 1994), the six FICSIT sites (Judge et al., 1996), and the Framingham study (Kelly-Hayes et al., 1992).
Although these findings are generally consistent with the theoretical formulation of the disablement process, they also indicate that there is a considerable amount of discrepancy between the observed capability to perform physical functions and the self-perceived ability to complete routine daily tasks. There are a number of well-recognized reasons for this apparent discrepancy. First, both measures suffer from varying degrees of measurement error, which will reduce the magnitude of the association between physical functions and disability (Hoeymans, Wouters, Feskens, van den Bos, & Kromhout, 1997; Jette, Jette, Ng, Plotkin, & Bach, 1999; Smith et al., 1990; Tager, Swanson, & Satariano, 1998). Second, assessment of self-assessments of disability may be subject to influences other than underlying physical capacity, such as aspects of affective functioning, personality attributes, environmental factors, and cognitive status (Gill, Robison, Williams, & Tinetti, 1999; Kempen, Steverink, et al., 1996; Kempen, Sullivan, van Sonderen, & Ormel, 1999; Satariano, 1997). Although some of these characteristics conceivably could affect performance-based functional tests as well, their influence on basic functional capacities is thought to be smaller than on the self-perceived ability to perform routine daily tasks (Glass, 1998). Third, performance-based measures of physical function usually assess limitations in basic physical functions. According to prevailing disability models, these limitations form an intermediate stage in the development of disability. In other words, tests of physical function and measures of disability represent conceptually distinct aspects of the disablement process (Jette, Assmann, Rooks, Harris, & Crawford, 1998; Kempen, Steverink, et al., 1996; Lawrence & Jette, 1996).
An important limitation of cross-sectional research is that it cannot adequately characterize the progression of the disablement process, or establish the degree to which functional limitations are associated with changes in disability status over time. However, a series of prospective studies have provided substantial empirical evidence for this link by showing that functional limitations are significantly associated with subsequent disability (Gill, Williams, & Tinetti, 1995; Gill, Williams, Mendes de Leon, & Tinetti, 1997; Guralnik et al., 2000; Guralnik, Ferrucci, Simonsick, Salive, & Wallace, 1995; Ostir, Markides, Black, & Goodwin, 1998). For example, Gill and colleagues reported that community-dwelling older adults who perform in the lowest quarter of performance tests, such as chair stands and gait speed, have a twofold increased risk of developing ADL disability, compared with the upper three quarters (Gill et al., 1995). In a prospective analysis of several EPESE data sets, Guralnik and colleagues divided subjects who were free of self-reported disability at baseline into those at roughly the bottom 10%, the intermediate 45%, and upper 45% of a summary measure of three commonly used performance tests: walking speed, chair stands, and tandem stand. The group with the lowest level of performance had a greater than fourfold increased risk for ADL and mobility disability 4 years later, compared with the highest performance group, whereas the risk among the intermediate group was between 1.5-fold and twofold higher (Guralnik, Ferrucci, et al., 1995). The predictive ability of this summary measure of physical performance was later found to be remarkably consistent across different populations and different lengths of follow-up, ranging from 1 to 6 years (Guralnik et al., 2000).
These robust, long-term predictions of disability are congruent with the causal specifications outlined in the disability model. However, they reveal relatively little about the actual strength of association between limitations in physical functions and disability. Despite the high estimates of relative risk, the predictive power of physical function tests for long-term changes in disability status tends to be rather modest. For example, in an analysis of several different population-based studies, it was found that 40% or more of those scoring in the bottom 10% of a summary measure of physical performance tests did not develop mobility disability during follow-up (Guralnik et al., 2000). For ADL disability, this percentage was much higher, being in the range of 70% or greater. Thus, this summary measure of physical performance provides only relatively limited predictive power for the identification of those who will develop future mobility or ADL disability. Other studies have reported similarly modest predictive power of physical performance measures, in particular with regard to the prediction of ADL disability (Gill et al., 1995; Gill et al., 1997; Ostir et al., 1998).
Although prospective research has provided some insight into the effect of functional limitations on disability, much of the exact nature of this interrelationship remains unknown. This may be caused, in part, by methodological limitations of previous studies in this area, which have relied mostly on single assessments of physical function. As a result, the magnitude of the estimated association of physical function with changes in disability may be biased to the extent that measures of physical function are influenced by other aspects of health or of other, non-health-related factors, which are also associated with disability. Although most studies attempt to control for such factors, the ability of observational studies to eliminate completely all sources of confounding is usually fairly limited, in particular when confounders are either unknown or poorly measured. Using multiple assessment of physical function and disability offers a much more stringent control of the influence of extraneous factors, because this will permit analyses in which individuals serve as their own controls. Thus, the degree to which extraneous influences are stable within persons will lead to unbiased estimates of the physical function–disability relationship.
Another limitation of previous research is that they have mostly focused on long-term changes in disability, usually 1 year or longer. Consequently, little is known about the more immediate impact of changes in physical function on disability. A better understanding of the more immediate, or short-term, association of changes in physical function and disability is important for two reasons. First, a greater immediate impact of changes in physical function on disability would strengthen confidence in the causality of the relationship. Second, a better insight into the short-term association between change in physical function and disability may be important for the prevention of disability. Interventions designed to reduce or delay disability need to be informed by an understanding of the factors that are likely to result in immediate benefits, especially because most interventions tend to be limited in duration.
In this article, we will use data from the Women's Health and Aging Study (WHAS), a study of the determinants and course of disability in older women, to examine the association of weekly changes in physical function and disability. Because this is one of the first attempts to examine the correspondence between short-term changes in physical function and disability, we will explore several different types of associations between these two processes. In the main analysis, we will examine the overall association of short-term changes in physical function with concurrent changes in self-reported ADL and mobility disability. In additional analyses, we will explore the degree to which disability levels change as a result of episodes or periods of unusual physical function levels, independent of the overall association between physical function and disability. The analysis will focus on two separate, but related, measures of disability: self-reported difficulties in ADL and mobility difficulties.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
85. Of the 5,285 women identified, 4,108 agreed to participate in a screening interview to establish their current disability level. Eligibility for WHAS was based on cognitive functioning and disability status. Overall, 1,409 women met eligibility criteria of scoring 18 on the Mini-Mental State Examination and reporting difficulty in two or more of the following domains: basic self-care (bathing, dressing, eating, using the toilet); upper extremity ability (raising arms up over head, using fingers to grasp or handle, lifting and carrying 10 lb); mobility/exercise tolerance (walking G mile, walking up 10 steps without resting, getting in and out of bed or chairs, doing heavy housework); and higher functioning tasks of independent living (using telephone, doing light housework, preparing meals, shopping for personal items). Of these, 1,002 (71%) agreed to participate in the full WHAS baseline interview and examination. A convenience sample of 102 women who agreed to weekly assessment participated in the Weekly Disability Substudy. Women were selected consecutively until an approximately equal number of women were obtained in each of nine groups, defined by age group (65–74, 75–84, 85+) and initial disability level (disabled in 2, 3, or 4 domains): 34 came from each age group, 32 had disability in two domains, 33 in three domains, and 37 in four domains.
Data Collection and Measures
Weekly Substudy assessments began 1–2 weeks after baseline assessment in the larger WHAS was completed and were repeated for 24 consecutive weeks. All assessments were administered by trained lay interviewers and conducted in the respondents' homes. Where possible, participants were tested by the same interviewer, and assessments were conducted on the same day of the week, and the same time of the day, to maximize control for extraneous influences on assessment of physical functions and disability. Of the 2,448 (24 x 102) possible assessments, 2,279 (93.1%) were completed. An average of 1.7 visits were missed per subject (range 0–14; Guralnik et al., 1999). Information collected each week included measures of self-reported disability status and a brief set of physical performance tests.
The first measure was constructed using six commonly used questions on ADL tasks (walking across room, bathing, dressing, getting in and out of bed, eating, using a toilet). Each item was scored according to the degree of difficulty performing a task independently (i.e., without help from another person or special equipment). Responses were scored on a scale from 0 to 3 (3 = no difficulty; 2 = little or some difficulty; 1 = a lot of difficulty; 0 = unable to do). These scores were summed across items, yielding a total score ranging from 0 to 18. A second measure of disability was constructed that specifically focused on mobility disability, which combined two questions of the ADL disability measure (walking across a small room, getting in and out of bed) with two questions about higher level mobility (walking a G mile, walking up 10 steps without resting). Scoring was similar to the measure of ADL disability, yielding a total score with a range from 0 to 12. Higher scores on each measure indicate less disability.
The scoring of degree of difficulty for these tasks, rather than the more commonly used scoring in terms of dependence (needing help or unable to do), was designed to capture finer gradations in underlying disability. An important aspect of this approach is that the resulting measure is more sensitive to short-term fluctuations in disability. Although dependency scores are more suitable to determine service needs, difficulty scores are thought to be more useful to estimate the consequences of disease and impairments (Jette, 1994). The use of difficulty scoring in the assessment of ADL function has been validated previously, both in terms of their psychometric characteristics and their predictive validity with regard to hospitalizations, home care needs, nursing home admission, and survival (Gill, Robison, & Tinetti, 1998; Langlois et al., 1996; Suurmeijer et al., 1994).
Assessment of limitations in physical function was based on three commonly used physical performance tests, including a 4-meter walk, standing balance, and rising from a chair. Strictly speaking, these tests combine assessments at the level of physiological impairment (e.g., muscle strength) and the level of functional limitations (e.g., walking speed) according to the Nagi model of disability (Nagi, 1976). For ease of presentation, we will refer to them together as tests of physical function. Walking speed was evaluated on a 4-meter course, with subjects being instructed to walk at their usual pace. A cane was allowed, but not help from another person. Each subject was given two trials, of which the best time was retained. Standing balance was tested by observing if the subject could hold three increasingly difficult positions for 10 seconds: side-by-side stand, semitandem stand, and full-tandem stand. Rising from a chair was timed for five repeated chair stands from an armless, straight back chair, without the use of arms. Using previously established cutoffs, performance scores on each test were recoded on a 5-point ordinal scale (0–4) and summarized across tests, yielding a total score ranging from 0 to 12 (Guralnik et al., 1994). Higher scores indicate better performance. Summary performance scores were set to missing if at least one of the individual tests was missing. This measure of limitations in physical function has been shown to be significantly predictive of nursing home admissions and overall mortality (Guralnik et al., 1994; Guralnik, Ferrucci, et al., 1995). In addition, its predictive validity for ADL disability has been validated in different populations and across varying lengths of follow-up (Guralnik et al., 2000).
Data Analysis
To describe the overall levels of physical performance and disability, we first summarized each measure by averaging the scores at each weekly assessment for all subjects. Thus, these data show the average, or marginal, levels of disability and physical function across weekly assessments, and provide an estimate of average change in the entire sample (see Figure 1). However, this method may produce a biased estimate of the average, within-person change. This would occur, for example, if sicker subjects were to drop out more frequently during follow-up and no longer contribute information to the marginal means, which might lead to an underestimation of the decline in function. An alternative way of summarizing change over time is to fit an ordinary least-squares regression line through the weekly data that are available for each person. A regression line, or slope, is estimated for each person separately, after which these "person-specific slopes" are averaged across subjects (see Table 1).
|
|
and
At Level 1, 
0j represents the intercept for person i, that is, the baseline score for ADL or mobility disability. The parameter 1j represents the rate of change (slope) for person i in each disability outcome during follow-up, and rij represents the within-person residual (error term). The Level 2 model expresses variation in individual intercept and slope parameters as a function of covariates (fixed effects) and two random variables, u0j and u1j. These two random variables or components account for two sources of between-person heterogeneity: (1) a subject's tendency to have an initial disability score above or below the average level predicted for subjects with the same characteristics (random intercept), and (2) a subject's tendency to decline faster or slower than predicted for similar subjects (random slope). The random variables, rij, u0j, and u1j are assumed to be normally distributed, with a zero mean and variances of 
2, 
00, and 11, respectively.
We first fitted a base model, modeling disability scores as a function of follow-up time only. This model estimates the initial level (intercept) and the magnitude of weekly change (slope) in ADL and mobility disability in this sample. Next, we examined the effect of baseline physical performance scores, after centering this variable at the mean value at baseline, to facilitate interpretation of the intercept and change regression coefficients. We then tested the effect of change in performance since baseline on the individual intercept and slope parameters. At the next step, we added age group and initial number of disability domains to examine the effect of these sampling variables on the previously estimated parameters. Two additional models were fitted to test the effect of unusual patterns of performance (i.e., scores that differ substantially from the person-specific weekly performance levels). We explored two different types of such unusual patterns of performance. The first type consists of single episodes of an unusually high or low performance level. Such episodes were represented by an indicator variable when a score at a particular week is 1.5 standard deviations (SDs) above or below the person-specific weekly average score. The second type consists of periods of performance levels that are better or worse than usual. Indicator variables were constructed for the fourth and any additional consecutive week that a performance score was either above or below the person-specific weekly average. In other words, such periods can be interpreted as representing a "good month" and "bad month" respectively. It is of note that the definitions of these departure scores were chosen arbitrarily and not based on any a priori criteria. We conducted some preliminary analysis to ensure that there was a sufficient number of each type of departure score to be included in the multivariate models. Variations of these criteria were explored in additional analyses, but these did not reveal any other insights into the effect of these departure scores on disability outcomes.
Model assumptions were examined by plotting residuals against predicted values, and by inspection of the distribution of the random intercept and slope variables. Overall, few serious departures from normality were evident, although these occurred slightly more frequently in the models of ADL disability than of mobility disability, likely because of its more skewed distribution. We refitted the models after removing observations that contributed the most to model violations, but this yielded mostly the same results. All random-effects models were fitted in SAS PROC MIXED (SAS Institute, 2000).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The average physical performance score at baseline was 6.72 (SD 3.1), whereas the average baseline levels of ADL and mobility disability were 15.54 (SD 3.0) and 8.12 (SD 3.09), respectively (see Table 1). The person-specific slopes estimated from the individual regression lines show that the average change in physical performance scores during follow-up was virtually zero. The average of the person-specific slopes showed a decline in both ADL (-0.026) and mobility (-0.030) function during follow-up. These changes amount to about 0.62 (24 x 0.026) units of decline in ADL scores during the 24-week follow-up period and about 0.72 units of decline in mobility scores. Although the average declines in function were relatively small, there was considerable variation between persons in terms of rate of change during follow-up, with some women showing much greater declines and others exhibiting improvement in ADL or mobility function (see Table 1).
The results for the random-effects models for ADL disability are presented in Table 2. The base model shows that the estimate for the intercept is 15.70, which is close to the observed average baseline ADL disability score as reported in Table 1. The regression estimate for weekly change is -0.024, which is also close to the average of the person-specific regression slopes presented in Table 1. The variance components indicate that there was significant between-person variability in terms of the initial ADL disability score (intercept), with a variance 00 = 7.37 ( p < .001). In fact, this variability between persons was considerably greater than the variability within persons over time (2 = 1.24). The small, within-subject residual variance suggests that, on average, women exhibited fairly stable ADL disability levels across weekly follow-up. The random component for slope indicates significant between-person variability in terms of change over time in ADL disability scores, with a variance estimate of 11 = 0.00553 ( p < .001).
|
00 = 7.37 to 5.38, or a decrease of 27%. In other words, a single physical performance score can be said to account for about a quarter of the variability in ADL function (measured weekly over a 24-week period). However, it is also clear that this single score accounted for very little variability between persons in terms of their rate of change, because the variance of this random variable decreased by only 1%. In Model 3, we added the change in physical performance scores to the previous model. This change score is associated with a relatively small, but significant, effect (ß = 0.08, p < .001) on ADL disability. In other words, each unit change in physical performance relative to a person's baseline score is associated with a 0.08 unit change in ADL disability scores measured the same week. The change in performance scores accounted for 3.5% (from 5.38 to 5.19) of the between-person variability in initial ADL disability scores, and about 5.5% (from 0.00547 to 0.00519) of the between-person variability in the rate of change in ADL scores over time. Model 4 indicates that age group was not significantly associated with ADL disability scores, but that women who had four disability domains at screening had significantly lower ADL scores (ß = -2.30, p < .001), compared with women who had only two disability domains. Inspection of the variance components reveals that these two variables accounted for an additional 12% of the between-person variability in initial ADL level, but less than 1% of the variability in the rate of change in ADL function.
In the final two random effects of ADL disability scores, we explored the effect of the two types of unusual performance scores. There were 79 (4.2%) episodes that were defined as higher than usual, and 136 (7.2%) episodes that were lower than usual performance scores. After adjustment for baseline and weekly change in performance score, a global test of the overall association of such episodes with ADL disability was not significant ( p = 0.57; data not shown). As shown in Model 5 of Table 2, separate indicator variables for having either a higher than usual performance score (ß = 0.05, p = .76) or lower than usual performance score (ß = -0.16, p = .29) were not significantly associated with ADL disability scores. There were also a total 158 (7.1%) performance tests with a score that represented being the fourth or higher consecutive week above the person-specific mean, and 146 (6.5%) tests with scores being the fourth or higher consecutive week below the person-specific mean. The overall test of the association of these performance scores with ADL disability was not significant (p = .89), after adjustment for baseline and weekly change in performance scores. Again, the separate indicator variables for these unusual departure scores were not significantly associated with ADL disability scores (see Model 6, Table 2). It can also be seen that adding unusual performance scores to the model did not lead to a change in the between-person variability in either initial ADL disability scores or the rate of change, as indicated by the variance estimates of the intercept and slope random variables.
The results for the random-effects models for mobility disability are presented in Table 3. The base model estimates of initial value (intercept) and rate of change were 8.34 ( p < .001) and -0.032 ( p < .001), respectively, which are very similar to the values reported in Table 1. There was significant between-person variability in terms of the initial mobility disability scores (00 = 7.92; p < .001), as well as in terms of the rate of change (slope) over time (11 = 0.00501; p < .001). As for ADL disability, the magnitude of the between-person variability in initial mobility disability was greater than the residual or within-person variability over time (2 = 1.19). Baseline performance scores were significantly associated mobility disability scores (ß = 0.54, p < 0.001), and accounted for about 36% [7.92 - 5.13)/7.92 x 100%] of the variability in initial mobility disability scores (see Model 2, Table 3). Change in performance scores was also significantly associated with mobility disability (ß = 0.12, p < 0.001), after adjustment for baseline scores. This term accounted for an additional 7% of the variability in initial mobility disability and 9% of the variability in the rate of change in mobility disability, as indicated by the change in the variance estimates of the random variables between Models 2 and 3 in Table 3.
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Our findings provide a somewhat mixed picture about the relationship between basic physical functions and disability. On the one hand, the results indicate that physical functions are strongly associated with disability, explaining about 27% and 36% of the variability in, respectively, ADL and mobility disability that exists in this group of disabled women, with an additional 3.5% and 7% of variance explained when follow-up information on physical functions was added to the models. The magnitude of these associations is generally consistent with those reported in previous studies (Cress et al., 1995; Daltroy et al., 1999; Guralnik et al., 1994; Judge et al., 1996; Kelly-Hayes et al., 1992; Kempen, van Heuvelen, et al., 1996; Kempen, Steverink, et al., 1996; Myers et al., 1993; Reuben et al., 1995) and are further evidence that basic physical functions (such as walking speed, balance, and lower-extremity strength) are related to a person's ability to carry out self-care tasks and overall mobility. Because these results are based on tests of concurrent association, other explanations cannot be ruled out entirely. Similar patterns of association might arise from reverse causation, whereby changes in disability affect performance on physical function tests, or even more complex relationships, such as reciprocal causation between physical function and disability. Also, other factors may cause simultaneous changes in physical functions and disability. For example, depression, which has been found to predict changes in both physical function and disability (Bruce et al., 1994; Cronin-Stubbs et al., 1997; Kempen, Sullivan, et al., 1999; Penninx, Deeg, van Eijk, Beekman, & Guralnik, 2000; Penninx, Leveille, Ferrucci, van Eijk, & Guralnik, 1999; Penninx et al., 2000), may be a common underlying cause of the associations between physical function and disability. In general, however, current conceptualizations of the disability process tend to favor the interpretation of a direct effect leading from limitations in basic physical functions to increasing disability.
At the same time, however, our models indicate that a single performance score accounts for only a minimal portion, of the order of 1%, of the variability in short-term changes in disability levels observed in this group of older, disabled women. Moreover, information on change in physical functions did not add much additional information on changes in disability, explaining only 5.5% and 9% of the variability in the rate of change in ADL and mobility disability, respectively, during weekly follow-up. Also, a unit change in basic physical functions was associated with small changes in each disability measure, even if these associations were statistically significant. These findings suggest that short-term fluctuations in physical functions show only weak associations with concurrent changes in disability, when measured across a period of about half a year.
There may be several explanations for the lack of a stronger correspondence between changes in physical function and disability levels. As described previously, this may be caused, in part, by a lack of reliability of performance tests and disability self-reports (Hoeymans et al., 1997; Jette et al., 1999; Tager et al., 1998). For example, the extent to which performance tests are subject to situational and motivational factors may have attenuated their true relationship with mobility and ADL disability. However, the degree to which these factors remain constant over time should have reduced their influence on the observed relationships. In addition, the performance tests used in this analysis do not assess the entire spectrum of physical functions that may be required to perform basic self-care tasks or basic mobility functions without difficulty (Kempen, Steverink, et al., 1996; Kempen, Sullivan, et al., 1999). Thus, it is possible that limitations in other, unobserved physical functions would have explained a larger proportion of the short-term changes in disability. However, we utilized commonly used tests of physical function that have been found to be consistently predictive of long-term changes in disability levels (Guralnik et al., 2000). Finally, weekly assessment of physical functions may have led to training and other effects related to repeated testing, which may attenuate the correspondence between short-term changes in physical functions and disability levels.
Apart from these considerations, our findings raise the possibility that some of the variability in disability and rate of change in disability observed in these frail, older women is unrelated to changes in basic physical functions. This suggests that other factors are involved in this process as well. Others have recognized that the ability to perform basic self-care tasks and social roles is shaped by the psychological characteristics of the individual, as well as the physical and social environments in which they are performed (Satariano, 1997; Verbrugge et al., 1994). In addition, ambulation and self-care function, even at a basic level, may require a minimum level of cognitive ability to complete these tasks successfully. Both cognitive function (Aguero-Torres et al., 1998; Gill, Williams, Richardson, & Tinetti, 1996; Moritz, Kasl, & Berkman, 1995) and psychosocial characteristics (Kempen, van Sonderen, & Ormel, 1999; Mendes de Leon, Gold, Glass, Kaplan, & George, 2001; Mendes de Leon, Seeman, Baker, Richardson, & Tinetti, 1996; Mendes de Leon et al., 1999; Penninx et al., 1999; Seeman, Bruce, & McAvay, 1996; Unger, McAvay, Bruce, Berkman, & Seeman, 1999) have been shown to predict long-term changes in disability in community-dwelling older adults. In general, cognitive functions are unlikely to show sufficient change across short periods of time to have an immediate impact on self-perceived disability levels. It may be possible, however, that short-term fluctuations in ADL and mobility disability depend to a greater extent on the emotional and social conditions of older adults, such as their level of social engagement and social support, emotional vitality and positive affect, and depression (Mendes de Leon et al., 1999, 2001; O'Brien et al., 1999; Ostir, Markides, Black, & Goodwin, 2000; Penninx et al., 1998, 1999).
There was little evidence in our data that unusual episodes or patterns of change in physical function affected disability independently from the concurrent association. It is possible that our ability to identify such associations may have been limited because of relatively high levels of within-person stability of performance scores and disability across weekly follow-ups. However, our models were able to identify significant between-person variability in both ADL and mobility disability, even if the magnitude of the changes over the 24 weeks were relatively small. In a separate random-effects analysis, we also found significant variability in the rate of change in performance scores during follow-up (data not shown). Nevertheless, longer follow-up periods, and possibly more sensitive measures of physical function and disability, may be required to explore the correspondence between changes in these variables with greater accuracy. For example, despite all women having at least some degree of disability at baseline, our measure of ADL disability may have been hampered by a ceiling effect, with many women reporting little difficulty in any of the ADL tasks. This may have adversely affected our ability to examine changes in ADL disability over time.
In analyses not shown, we also explored other types of associations. For example, others have reported that the pattern of association between limitations in physical functions and reported disability levels is more consistent with a threshold effect than a linear association, with overt disability occurring only after a certain magnitude of functional limitations has been reached (Buchner, Larson, Wagner, Koepsell, & de Lateur, 1996; Jette et al., 1998). However, we found no evidence for significant departures from a linear pattern of association. We also explored the possibility of a lag effect between changes in physical functions and disability. Specifically, we fitted models to test a lag effect of 3 and 4 weeks, respectively, to see if changes in physical function would be significantly associated with changes in self-reported disability levels 3 or 4 weeks later. Again, no evidence for these lag effects was found. It is possible that we had only limited ability to identify these patterns of association from the relative lack of change in disability levels.
Although it may be assumed that basic physical functions have a greater or more immediate impact on mobility disability than ADL disability, we cannot directly compare the magnitude of these associations based on the present analysis. We did find, however, that performance scores accounted for a greater portion of the variability in mobility difficulty than in ADL disability. Measures of ADL function often include mobility-related items, as did ours, which shared two mobility-related items with the mobility disability measure. It seems likely that tests of physical function that emphasize lower extremity functions are associated with ADL disability primarily, or at least initially, via their effect on tasks that require considerable lower extremity involvement, such as ambulation and transfer. Indeed, mobility-related disability may well play an intermediary role in the pathway from decline in physical functions to ADL disability. Recent research has started to focus on mobility difficulties in the disability process, suggesting that more severe disability, such as limitations in ADL function, are often preceded by mobility difficulties (Jette et al., 1998). This is consistent with the literature on the hierarchical structure of disability patterns, which has shown that disability in tasks involving general mobility and lower-extremity strength tend to develop prior to onset of disability in more basic ADLs, such as self-feeding and dressing (Ferrucci et al., 1998; Foldvari et al., 2000; Kempen, Miedema, Ormel, & Molenaar, 1996). In other words, limitations in lower extremity capacities, such as strength and balance, may lead first to difficulties in ambulation, but may require more prolonged duration or increasing severity before they will affect self-care tasks less directly related to mobility (such as dressing or bathing). Longitudinal studies with repeated testing over longer periods of time may be required to disentangle the sequence of such relationships in greater detail.
Various types of interventions have been designed to improve physical function and overall physical activity in community-dwelling older adults (Cress et al., 1999; Schlicht, Camaione, & Owen, 2001; Sharpe et al., 1997; Stewart et al., 1998; Tinetti, McAvay, & Claus, 1996). In general, such approaches have shown significant, albeit modest, improvements in basic physical functions, such as strength, standing balance, and walking speed. However, whether such interventions have a clear benefit for disability-related outcomes remains uncertain (Keysor & Jette, 2001). Although our findings may be somewhat limited because of a lack of change in ADL disability, they are consistent with the notion that improvements in basic physical functions alone may be insufficient to slow down the rate of decline in ADL disability, at least in the short-term. Future interventions may well have to consider a wider array of risk factors to prevent this type of disability more effectively. Given that treatments aimed at the slowing down or reversal of cognitive decline are still largely in a state of development, interventions for the prevention of ADL disability should consider including the psychological, social, and possibly environmental conditions associated with development of ADL disability.
Although the WHAS Weekly Substudy offered a unique opportunity to evaluate the short-term impact of physical function on disability, it also had some limitations. The WHAS study was restricted to women, and these findings may not necessarily generalize to men. Furthermore, the request to participate in weekly assessments may have led to a selection effect, although actual disability levels in the Weekly Substudy sample were not markedly different than those in the overall WHAS cohort. As indicated previously, another potential limitation of the study was the use of repeat assessments across short time intervals, which may have led to training and selective drop-out effects. For example, there appeared to be a considerable increase in performance on the physical function tests after the first week of testing, and there was a somewhat unusual improvement at the last weekly assessment, possibly from selective drop-out. We refitted all models after removing data from the first and last weekly assessments, but obtained essentially the same results. Another potential limitation was the relative lack of overall change in disability levels over time, as well as a relative lack of within-person change across weekly assessments. Thus, we may have had only limited power to reliably assess the underlying relationship of short-term effects of physical function on disability. Despite these limitations, the availability of 24 consecutive weekly assessments in a reasonably large cohort of older women allowed us to perform a detailed test of the short-term correspondence between changes in physical function and self-reported disability. The findings suggest that changes in physical functions are significantly associated with concurrent changes in ADL and mobility disability. At the same time, they seem to explain only a small portion of the variability in short-term fluctuations in ADL or mobility disability observed among older disabled women.
|
Acknowledgments |
|---|
We thank Dr. Scott Zeger for his helpful suggestions regarding data analysis.
Received for publication January 7, 2002. Accepted for publication May 24, 2002.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  B. Bartali, R. D. Semba, E. A. Frongillo, R. Varadhan, M. O. Ricks, C. S. Blaum, L. Ferrucci, J. M. Guralnik, and L. P. Fried Low Micronutrient Levels as a Predictor of Incident Disability in Older Women Arch Intern Med, November 27, 2006; 166(21): 2335 - 2340. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. W. Li Predictors of ADL Disability Trajectories Among Low-Income Frail Elders in the Community Research on Aging, November 1, 2005; 27(6): 615 - 642. [Abstract] [PDF]
|
|
|
|
 |
|
 S. Perera, S. Studenski, J. M. Chandler, and J. M. Guralnik Magnitude and Patterns of Decline in Health and Function in 1 Year Affect Subsequent 5-Year Survival J. Gerontol. A Biol. Sci. Med. Sci., June 1, 2005; 60(7): 894 - 900. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
