| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a The Pennsylvania State University, University Park
b University College of Health Sciences, Jönköping, Sweden
Elia E. Femia, Department of Human Development and Family Studies, Henderson S-110, The Pennsylvania State University, University Park, PA 16802 E-mail: femia{at}erols.com.
Decision Editor: Toni C. Antonucci, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
OVER previous years, research has gone beyond estimating the rates of specific disease diagnoses and acute health events to examining the prevalence of disability in the population. This shift reflects the growing distinction between disease and disability. Unlike disease diagnoses, which focus on the biochemical and physiological condition of the body, disability taps into the problems an individual faces in daily life and the kinds of assistance one is likely to receive to manage the disability. This distinction is especially important when characterizing older adults, who are at greatest risk for disease and disability, dependent life, and consumption of costly health care services (Hayward, Crimmins, and Saito in press
). Recent U.S. estimates have shown 22% of those noninstitutionalized elderly aged 85 years and over to be disabled in terms of activity of daily living (ADL) functioning (U.S. National Center for Health Statistics 1999). Understanding the relationship between disease and disability and the factors that contribute to disability can move clinicians beyond the examination of physical impairments to the nonphysical and environmental contributors to morbidity.
The shift towards understanding disability processes comes from two main findings of disability research. One finding supports the idea that older adults can function successfully in spite of declining physical health and other age-related losses (Camacho, Strawbridge, Cohen, and Kaplan 1993; Palmore, Nowlin, and Wang 1985; Roos and Havens 1991; U.S. National Center for Health Statistics 1999). Support for this claim can be seen in recent estimates of disease and disability; whereas 80% of those over the age of 65 have one or more of nine common chronic conditions or impairments, only 40% of them are restricted in their activities (Siegel 1993). A second finding is of individuals who do not have a disease condition but who are nonetheless disabled in some way (Siegel 1993). Information from the 1984 Supplement on Aging of the National Health Interview Survey (NHIS) showed that 21% of their sample had none of the nine most common chronic conditions, but 17% of these individuals had difficulty performing some ADL tasks (U.S. National Center for Health Statistics 1989). Fried and colleagues also found an interesting though small proportion of their sample that had some disability in the absence of 15 disease conditions (Fried, Ettinger, Lind, Newman, and Gardin 1994).
In general, these findings suggest that although disease conditions and physical impairments are strongly related to an individual's ability to carry out activities of daily living, they are not sufficient causes for disability. In the case of individuals who are seemingly healthy, but who have difficulty with ADL tasks, factors other than physical impairments and limitations seem to exacerbate the disablement process. In an effort to identify these factors, Verbrugge and Jette 1994 have proposed the disablement model, which addresses the risk factors and internal resources that affect disability. The conceptual model adapted for the current study is illustrated in Fig. 1. This model serves as a heuristic, suggesting potential relationships between the various components of disability. It consists of a main pathway and other forces, namely risk factors and internal resources that can impact the main pathway.
|
, Nagi 1991 and consists of four interrelated components: pathology, functional impairments, functional limitations, and disability. The first component, pathology, refers to the biochemical and physiological abnormalities medically labeled as disease or injury. Pathology leads to functional impairments, defined as dysfunctions in bodily systems such as cardiovascular, neurological, musculoskeletal, and pulmonary systems. Functional impairments are directly linked to functional limitations, defined as restrictions in basic physical and mental actions such as basic mobility, discrete motions, and strength. The ultimate result is disability, that is, difficulty in performing the activities of every day life. The underlying assumption of disability is that personal capabilities interact with environmental conditions that lead to a higher or lower probability of being disabled. The main pathway is influenced by a variety of factors. One of these is risk factors, which refer to long standing behaviors or characteristics that elevate or reduce one's chances for disability. They are predisposing in that they exist before the outset of disability and are usually permanent features of the individual. Typically, these include demographic characteristics (i.e., gender, age, marital status) that affect the severity of impairments, functional limitations, and disability.
A study by Lawrence and Jette 1996 tested the main pathway and risk factors of the disablement model and found evidence for a causal ordering of the components. In the study, pathology and functional impairments were used as direct predictors for functional limitations and indirect predictors for disability. By imposing this causal ordering, the authors' hypothesis was that functional limitations were the main driving force for increased disability. For example, an upper body limitation that causes difficulty in being able to reach overhead increased the likelihood that an individual has difficulty in dressing him- or herself. In a longitudinal test of the model, controlling for other risk factors, the hypothesis was confirmed. In particular, lower body limitations were found to have the strongest relationship to disability.
Lawrence and Jette 1996 study is a fruitful first step in operationalizing the conceptual model of the process for disability. The biggest limitation of the study was that it tested only the efficiency of the main pathway and risk factors, but did not include some additional components of the disablement model. These are the psychological factors and other internal resources that could further impact the daily functioning of the individual. As shown in Fig. 1, these other components of the disablement model form resources that can potentially affect the main pathway of disability. Internal resources refer to various factors and psychological resources that can reduce the demands or increase the capability of an individual. In this way, having more internal resources can offset disability or conversely, having less can exacerbate it. Work by Kempen and colleagues support this notion, showing that the effects of psychological attributes (neurotocism, self-efficacy, and mastery) on disability are unique, even after controlling for functional limitations (Kempen, van Heuvelen, van Sonderen, van den Brink, Kooijman, and Ormel 1999). Lower levels of psychological resources put people at a higher risk for developing disability. For example, better coping mechanisms can help a person accept some of the symptoms and impairments related to a disease by encouraging a person to act aggressively and not "let" functional impairments spiral into disability. In the disablement model, these internal resources play an important role and reflect the idea that disease and physical health alone do not formulate disability. Internal resources can strongly influence the relationship between personal capability and environmental demand by either increasing personal capability, reducing the environmental demand, or both.
A number of studies have consistently identified the demographic risk factors and psychological and social factors that impact levels of daily function. Being female, being older, being non-White, having lower socioeconomic status (SES), and living in an institutional residence are among the risk factors shown to relate to higher levels of disability and greater dependence (Berkman, et al. 1993; Crimmins, Hayward, and Saito 1996; Parker, Thorslund and Lundberg 1994; Rogers 1995; Rogers, Rogers, and Belanger 1992; Samuelsson and Sundstrom 1988). These demographic risk factors seem to play a part even after controlling for a variety of health variables. It is thought, especially for SES, that these variables influence the choices individuals have in the purchasing of health care resources and the knowledge they have of health promotion and disease prevention practices.
Turning to other factors that have been shown to influence disability, it has been observed that lower cognitive skills and the presence of dementia place individuals at risk for higher disability (Fitz and Teri 1994; Gill, Richardson, and Tinetti 1995; Schaie 1990). The speculation is that cognitive resources allow individuals to respond and adapt to the demands and needs of the environment as well as remain engaged in higher levels of functioning. In addition to cognitive performance, a psychological factor, depression, has also been linked to both the onset of and changes in disability (Bruce, Seeman, Merrill, and Blazer 1994; Camacho et al. 1993; Smits, Deeg, and Jonker 1997). One explanation is that being depressed decreases the motivation necessary to engage in behaviors that sustain levels of functional ability.
Other internal resources, subjective health, locus of control, mastery, life satisfaction, and self-efficacy have been noted to affect the disability process (Berkman et al. 1993; Bruce et al. 1994; Camacho et al. 1993; Idler and Kasl 1995). Higher levels of these factors are shown to effectively mediate the relationship between health status and ADL task. One speculation is that these resources influence the types of activity individuals engage in, especially when confronted with physically or emotionally challenging situations. In the case of subjective health, one suggestion is that subjective health represents a sensitive reflection of the threat an individual feels about illnesses, which affects the motivation to perform daily tasks (Krause and Jay 1994).
Finally, aspects of social support have been demonstrated to affect disability outcomes. Social isolation, for example, has been shown to relate to a decrease in functional ability (Baltes, Wahl, and Schmid-Furstoss 1990). Conversely, group membership to organizations and social contact show positive correlations to ADL performance, even after adjusting for various health conditions (Camacho et al. 1993). These findings suggest that social support in the form of engaging in activity, affiliation, friendship, and intimacy can benefit an individual's functioning and form yet another protective barrier against disability processes.
The purpose of the present study is to explore the disability experience of the oldest old, taking into account not only physical impairments and functional limitations but also psychological and social functioning. Using the conceptual model of the disablement process proposed by Verbrugge and Jette 1994 as a rough guide, we first examine the efficiency of the main pathway of disability on data from 1991. To do this, we test a series of mediation models that regresses disability onto functional impairments and functional limitations. The advantage of using the Swedish data is the availability of variables that allow us to operationalize the concepts in the disablement model. We then engage in exploratory model development, constructing path models of disability that take into account risk factors and various aspects of psychological functioning and internal resources. Consistent with the conceptual model of disablement, these models explore whether internal factors mediate the relationship between physical functioning and disability. All of these models were tested using cross-sectional data. Although we would have liked to explore the use of longitudinal models, the amount of change in the dependent variable was very small (over a 2-year time interval of second wave data), which made it impossible to conduct change analyses or to construct comparison groups of sufficiently large sample sizes. Despite this limitation, we felt that the model-generating nature of the present study allowed us to test the conceptual ideas of the disablement model.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). One thousand four hundred seventy-four (1,474) same-sex twins were identified from the population-based Swedish twin registry with the birth years 1913 and earlier. Both members of the twin pair had to be alive to be eligible for the study. Of the 737 pairs identified, 188 twin pairs had one or both members of the twin-pair deceased at the time of initial contact. In addition, 198 individuals refused to participate. Because refusal of one twin resulted in the exclusion of the pair, the individual refusal rate was 198 out of 1,098 individuals, who were alive at the time of testing, or 18%. The final sample size was 351 twin pairs, or 702 individuals. This number represented an overall participation rate of 64%.
The extensive data as well as the low individual refusal rate among the participants (18%) made it possible to test the disablement model. Although the original data reflected a twin design, the present study used information provided by one randomly chosen member of the twin pair. The decision to use the data in a non-twin study had to do with the quality of the measures in the study, which included biomedical, disease history, cognitive, pychosocial, and demographic measures on all of the oldest old participants. An initial concern was whether the twin-based sample was biased in terms of demographic characteristics or health status when compared with a non-twin sample. These issues were addressed in a study that compared one randomly chosen member of the twin dyad with a population-based sample of non-twins in another Swedish study (Simmons et al. 1997). The comparison found significant differences on only four of nineteen variables. The twin sample was slightly younger, had slightly higher levels of education, had slightly more children, and was less likely to be living in an institution. Though these differences reached statistical significance, the actual differences between the two groups were quite small, with the exception of the proportions living in institutions. In terms of refusal rates, the effective refusal rate for the twin sample was 18% compared with 14% for singletons. As a result, the twin sample could represent a more highly select group in comparison with singletons the same age. Overall, however, Simmons and colleagues 1997 study suggested that the extent of a selection bias was relatively modest. Thus, the generalizability and representativeness of the sample for non-twin studies was supported.
From the 351 twin pairs, one member of the pair was randomly selected for the study. In addition, individuals who were too cognitively impaired to reliably report on the various measures were excluded (n = 42); however, those with mild impairment (Mini-Mental State Exam [MMSE] score near the cut-off point for probable dementia) were retained. Only those individuals with information at two waves were included in the study because the original intent was to explore changes in the disability experience. These criteria reduced the sample size to 266 individuals. Finally, the analyses planned for the study necessitated a listwise deletion of cases, which further reduced the sample size to 203 individuals who had complete data on all of the study's variables.
Table 1 shows the demographic characteristics of the sample. On average, the respondents were 83.4 years of age. Women represented two-thirds of the sample. The overwhelming majority (80%) had some basic schooling equivalent to an elementary education in the United States. This proportion is typical of older Swedes who grew up during the pre-war era. Most of the sample was unmarried (66%), lived alone (54%), and lived in the community in ordinary housing (92%). In terms of social status, half of the sample was categorized in the lowest social grouping (according to the Swedish Election Statistics, Social Research Institute), roughly indicating previous or current employment in manual or blue-collar work.
|
: (1) disability, (2) functional limitations, (3) functional impairments, (4) risk factors, and (5) internal resources. The pathology component of the disablement model was not included in the study because of data as well as theoretical limitations. Although Verbrugge and Jette simplify the concept of pathology through disease diagnoses, they acknowledge that the real notion underlying pathology is in terms of the biochemical processes that are congenital and developmental in nature. Because these processes are difficult to measure and not available in most data sets, disease diagnoses are often substituted. One problem with using strict disease diagnoses is that they are often confounded with functional performance and abilities, especially if a disease condition is managed by therapy or medication. In addition, disease diagnoses are difficult to assess in terms of their level of seriousness. For example, one individual with the same biochemical dysfunctions as another (e.g., diabetes) could have different manifestations and symptoms of the disease. Because of these potential confounding effects, the component of pathology was not included in the model testing procedures. We do not discount the impact of pathology on the disablement process; rather, we acknowledge that the available data do not suitably capture pathology in a way that would not confound the findings of the study.
Risk factors.
The risk factors for the study consisted of the demographic variables, age (0 = <85 years; 1 = 85+ years), gender (0 = male, 1 = female), education (0 = basic, 1 = extended), marital status (0 = not married, 1 = married), social status group (0 = professional, 1 = white collar, 2 = blue collar), and living arrangement (0 = lives alone, 1 = lives with others).
Functional impairments.
Three baseline measures of vitality and two indicators of sensory impairment were used to indicate functional impairments. These were assessed by the trained nurse interviewer and included (a) sitting and standing systolic and diastolic blood pressure, (b) grip strength as measured by a vigorimeter (average of right and left hands over three attempts), and (c) pulse per minute. The two baseline measures of sensory impairment were self-rated measures of hearing and vision rated along a 6-point Likert scale ranging from excellent (1) to very poor (6).
Functional limitations.
In Verbrugge and Jette 1994 scheme, functional limitations are defined as the fundamental physical and mental actions used in daily life, which reflect the body's overall ability to do purposeful work. Differentiated from disability, functional limitations refer to the individual's capability without regard to the environmental context and conditions that could affect one's performance. Three measures of functional limitations were included in the analysis. For two of these, upper and lower body limitations, the participants performed a series of functional tasks that were rated by the nurse interviewer as presenting no difficulty (0) or some difficulty (1). Construction of these two components came from a principal components analysis on 19 functional limitations items, which resulted in the retention of two main components (refer to Femia 1998, for a more thorough explanation). Three of the nineteen task items were dropped as they did not load strongly on either component. The first component was upper body flexibility/strength and included six tasks: lifting a glass, lifting a 1-kg weight, moving the right finger to the left earlobe, pouring water from a jug to a glass, and pouring water from glass to glass (dominant and nondominant hands). The second component was lower body strength and bending and included seven items: normal gait and turn, Romberg's test, side by side (looking), right finger to left toe, left finger to right toe, chair stand, and repeated chair stand. The internal consistencies (using Cronbach's alpha) of upper body flexibility/strength and lower body strength/bending were .93 and .91, respectively. For both measures, summary scores of the number of items that presented difficulty were calculated, and because to the skewness of the scores, the final measures were reduced to categorical variables as indicated in Table 2 .
|
), was also considered a functional limitation variable. Because our original criteria excluded those individuals suffering from dementia, the distribution of MMSE scores was highly skewed (i.e., showing a ceiling effect), with only about 10% of participants having possible cognitive impairment. Consequently, we decided to categorize participants as either not cognitively impaired (0) or possibly cognitively impaired (1). The cut-off score for categorizing participants into these groups was 23.
Disability.
In this study, the concept of disability reflected the difficulty in performing ADLs, that is, personal care, complex mobility tasks in the environment, and household management activities (e.g., instrumental activities of daily living [IADL]). The individual items were self-rated on the basis of whether the individual could perform the task without difficulty or with some difficulty. The basic ADL items were eating, bathing, dressing, toileting, getting up from bed, moving from bed to chair, and hygiene. The IADL tasks were cleaning, making the bed, cooking, shopping, transportation, using the bank, and using the telephone. The mobility tasks were walking indoors, walking outdoors, and walking stairs. A composite variable was constructed on the basis of the hierarchical relationships among the IADL, complex mobility, and ADL measures. The composite disability variable assumed that the more complex functions of mobility and instrumental ADL function were disrupted more easily and before the basic ADL functions (Lawton 1971; Lawton and Nahemow 1973; Spector, Katz, Murphy and Fulton 1987; Zarit, Johansson, and Malmberg 1995). That is, anyone with difficulties in basic ADLs were assumed to also have difficulties with IADL and mobility functioning, whereas the reverse (difficulties with basic ADLs without deficits in the more complex functions) was assumed not to occur. The composite variable of disability became a three-category variable of severity indicating no difficulty on any of the mobility, IADL, or ADL tasks (0 = no disability), difficulty with instrumental ADLs and mobility tasks only (1 = moderate disability), and difficulty with any basic ADL (2 = severe disability).
Psychosocial variables and internal resources.
Four psychosocial measures were used that measured subjective health, depression, social integration, and social isolation. For subjective health, four questions were asked that related to (a) how the participants appraised their general health (good, average, or bad), (b) how they rated their health compared with 2 years ago (better, same, worse), (c) how they would rate their health compared with others the same age (better, same, worse), and (d) whether they thought health prevented them from doing things they liked. These items were subjected to a principal components analysis that indicated the presence of one component that explained 51% of the total variance. A summary score was created by calculating the mean among the four items. Higher scores indicated better self-reported health. Cronbach's alpha for this measure was .68.
Depression was measured by the Center for Epidemiologic Studies—Depression Scale (CES-D; Radloff 1977). The instrument had 20 items rated along a 4-point Likert scale that measured the participant's current level of depressive symptomatology. A total depression score was calculated with higher scores indicating more depressive symptoms (Cronbach's 
= .84). Because the distribution of the scale was highly skewed, a dichotomous variable indicating depressed (1) or not depressed (0) was calculated using the cut-off score of 16.
Feelings of social isolation were measured by the sum of three items that asked the degree to which the participants felt lonely, lacked company, or felt abandoned. The scale was rated along a 4-point Likert scale ranging from not at all (0) to nearly always (3). A higher score indicated higher feelings of social isolation (Cronbach's = .66). Feelings of social integration consisted of the sum of two items measuring whether the respondent felt he or she had friends to talk with and felt a part of a set of friends. This scale was rated similarly to social isolation. A higher score reflected higher feelings of social integration (Cronbach's = .81).
Data Analysis
Because several of the variables were ordinal, including the dependent variable of disability, a method of analysis that used ordinal variables was used. This involved using a matrix of polychoric correlations obtained through the PRELIS program and analyzing it with the WLS (weighted least squares) method (Dolan 1994) and appropriate weight matrix (asymptotic covariance matrix) in LISREL (Joreskog and Sorbom 1993). This method of estimation required a correlation matrix based on complete data on all variables (203 cases).
A series of structural models on the baseline data were constructed to evaluate the main pathway of disability and a full model of disability (i.e., one that included the risk factors and psychosocial variables). Two issues with the structural models were the number of variables used for the analysis and the sample size, which would not only limit the number of estimated parameters but also the fit indices. These were notable concerns with the WLS method of estimation because the sample size for the present study was around 200 and the number of variables was potentially high (around 20). To address these concerns, at least three steps were taken. First, before constructing the LISREL models, bivariate analyses were conducted to identify the variables significantly related to the dependent variable. Variables not significantly correlated to disability at the .05 level were not included in the path models. Second, additional indicators of model fit other than chi square (i.e., GFI [Goodness of Fit Index], AGFI [Adjusted Goodness of Fit Index], NFI [Normed Fit Index], NNFI [Nonnormed Fit Index]) were used to evaluate the efficiency of the model. Third, the number of post hoc modifications to the structural model was limited so as to avoid the capitalization on chance (MacCallum, Roznowski, and Necowitz 1992). The initial construction of models was based on our conceptual understanding of the disablement process, and subsequent modifications were based not only on the modification indices but also on the substantive justification for freeing up the given parameters. We used Joreskog and Sorbom 1993 guidelines and recommendations on conducting model specification searches, which included eliminating parameters with small t values and adding parameters with large modification indices. We also used the recommendations of MacCallum 1995, who advocated fewer modifications and a more substantively meaningful final model over its absolute fit.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Bivariate correlations between sets of variables were calculated with the PRELIS program (Joreskog and Sorbom 1996) and represented three types of correlation coefficients: Pearson–product correlations, point–biserial correlations, and polychoric correlations, depending on the scaling of the variables. Table 3 shows the observed correlation matrix among the final study variables as obtained by PRELIS. Only those variables with a significant bivariate correlation to disability were retained for further analyses. The other variables (gender, marital status, education, living arrangement, social status, blood pressure, pulse, and hearing) are discussed later but not included in the subsequent path models because of their nonsignificant relationship with the dependent variable.
|
). Four models testing these requirements showed support for these assumptions. The fit of the full mediation model was excellent, 
2(16, N = 203) = 17.50, p = .35, GFI = .98, AGFI = .97, root mean squared error of approximation [RMSEA] = .02, CFI = .99, NFI = .94, NNFI = .99. The R2 value for the model was 0.19. Fig. 2 shows the significant standardized estimates for the path model of the main pathway.
|
Test of the Full Disablement Model
The mediation model for the main pathway of the disablement model was established as an initial model on which to generate other models that included risk factors and psychosocial characteristics. These models did not provide a strict test of the requirements for mediation as defined by Baron and Kenny 1986, rather, they examined plausible relationships among variables as guided by the theory of disablement.
Turning first to risk factors, only one variable demonstrated a significant bivariate relationship with disability: age. We treated age as an exogenous variable and hypothesized it to have an indirect relationship to disability whose effect was filtered by the functional impairment and functional limitations variables. The rationale for this strategy stemmed from research that suggested chronological age to be directly and negatively related to an individual's "vitality" (i.e., functional impairments; Anstey, Lord, and Smith 1996; Grant 1997). We specified direct paths from age to grip strength and vision impairment. Although this model fit reasonably well, our examination of the modification indices suggested that freeing up one path from age to cognition would significantly improve the fit of the model. This relationship between age and cognition was well supported by research showing age to be a risk factor for poor cognitive status (Bickel and Cooper 1994; Hogan and Ebly 2000); hence, we freed this path. With the rest of the paths constrained according to the results of the main pathway, we found age to have significant direct effects are grip strength and cognitive impairment (ß = -0.26 and 0.12, respectively; p < .05) and a significant indirect effect on disability (ß = .08; p < .05). The fit of this model was excellent, 2(14, N = 203) = 20.25, p = .12, GFI = .98, AGFI = .97, RMSEA = .05, CFI = .99, NFI = .96, NNFI = .98. Being over age 85 was associated with lower grip strength and with being cognitively impaired. Age also had indirect influences on the functional limitation variables and disability severity. These relations were mediated by grip strength, which as described previously, was both directly and indirectly related to disability severity.
In the next step, we introduced the psychosocial variables into the model. The variables included CES-D, subjective health, social integration, and social isolation. Previous research on the effects of these variables on disability has consistently identified them as predictors and mediators for disability outcomes (Berkman et al. 1993; Bruce et al. 1994; Camacho et al. 1993; Idler and Kasl 1995). On the basis of these findings, we hypothesized these internal resource variables to mediate the relation between functional limitations and disability. We tested an initial model by regressing disability on all four psychosocial variables and regressing the four psychosocial variables on the three functional limitations variables. We deemed the fit of this model to be relatively poor although the chi-square value was significant, 2(39, N = 203) = 268.93, p = .00, GFI = .96, AGFI = .93, RMSEA = .17, CFI = .96, NFI = .95, NNFI = .94. At this point, we initiated a specification search to find a more parsimonious model. Our examination of the fit and modification indices revealed four paths that if freed, would significantly improve model fit. These included a direct path from depression to cognition, depression to subjective health, depression to social isolation, and vision to subjective health. These paths, which are well supported by prior research (Buntix, Kester, Bergers, and Knottnerus 1996; Mulsant, Ganguli, and Seaberg 1997; Ritchie, Touchon, and Ledesert 1998; Wang, Mitchell, and Smith 2000) were freed in the next three series of models.
A similar examination of the standardized beta estimates revealed that social isolation contributed no unique variance to the model, either as a predictor of disability or as a mediator of it. To test whether this variable could be removed from the model, the paths leading from and to social isolation were constrained to be 0. The fit of this model was not significantly worse than the fit of the previous nested model (
2 = 1.02, df = 4, p = .90); therefore, this variable was removed from further model testing procedures. We examine this finding in the Discussion section.
After the model was refitted without social isolation, one additional path was freed to improve the fit and parsimony of the model. This was between vision and depression. Again, this step was justified on the basis of the literature that has shown vision impairment to be linked to negative outcomes such as depression and disability (Rovner, Zisselman, and Shmuely-Dulitzki 1996). This model presented a significant improvement in model fit (2 = 26.09, df = 1, p = .00). At this point, the model was deemed acceptable, and no further modifications were made. The results of the full model of disability are illustrated in Fig. 3. The full model presented a very good fit to the observed data, 2(28, N = 203) = 114.24, p = .00, GFI = .98, AGFI = .96, RMSEA = .12, CFI = .98, NFI = .98, NNFI = .97. The model R2 value was 0.43. Compared to the model of the main pathway (Fig. 2), the full model contributed 23% additional variance.
|
Fig. 3 also illustrates a bidirectional relationship between depression and cognitive impairment. Interestingly, this bidirectional relationship had opposite signs, meaning that being cognitively impaired was associated with not being depressed, whereas being depressed was associated with being cognitively impaired. This finding is discussed below. Finally, upper body limitations was associated with lower body limitations and being depressed was associated with lower subjective health.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Lawrence and Jette 1996) and work highlighting the effect of cognitive impairment (Gill et al. 1995). Turning to the results of the full model, our results suggest disability to be as much a function of an individual's psychosocial characteristics as the degree of functional limitations and functional impairments. Subjective health, depression, and social integration were shown to mediate some of the influence of the functional impairments and limitations. That is, after controlling for functional limitations, the psychosocial variables did retain their unique effects on disability severity. Again, these findings are consistent with previous research that has identified the unique contributions of psychosocial factors on disability (Camacho et al. 1993; Smits et al. 1997; Thomae 1993). One impairment variable, grip strength, retained a significant direct association with disability, even after controlling for the psychosocial variables.
These findings provide evidence of the usefulness of the disablement model in understanding the disability experience of the oldest old. In terms of prevalence, 36% of the sample suffered from some kind of basic ADL difficulty. This finding is slightly higher than U.S. estimates of ADL disability, which is likely the result of sampling frame differences (e.g., U.S. sample were community-dwelling elderly persons only). Whether in the United States or in Sweden, disabled elderly persons need regular, intensive, and sustained help, which is costly to both families and health care systems who must provide this basic level of assistance. However, knowing that some of the variance in disability can be explained by psychosocial variables sheds an optimistic light in terms of intervention possibilities. Previous research supports the idea that an individual's belief structures form resources that individuals can use to affect the likelihood of disability independent of underlying physical abilities (Seeman, Unger, McAvay, and Mendes de Leon 1999). The findings of the present study provide additional evidence that beliefs about one's health (e.g., subjective health) and the motivation to perform daily activities (e.g., depression) are potentially as important as the ability to perform them. These belief structures are potentially modifiable and therefore are excellent targets for interventions that are less costly and less invasive than biomedical treatments.
For example, these findings can inform clinicians who often focus on improving physical abilities (through therapy and medication), but who sometimes disregard psychosocial functioning. Assessing an individual's mood or his or her perception of health would be equally important tasks faced by a geriatric management team. In addition, knowing the quality of a person's social life and the environmental conditions in which he or she lives could provide additional insight to the disability experience. Feelings of being socially integrated can act as a buffer for declines in physical functioning, especially for someone who values recreational and leisure activities with other people. Work by Magaziner and his colleagues (Magaziner, Simonsick, Kashner, Hebel, and Kenzora 1990) has shown that contact with one's social network is associated with recovery of ADL functioning after hip fracture. In terms of vision impairment, our study verified not only that vision impairments affect functional limitations, but also that they have a strong relationship to depression, which in turn, impacts disability (Rovner and Ganguli 1998). On the basis of these analyses and previous work showing similar results, treating depression among those with vision impairments could likely reduce the severity of disability.
The consistently strong direct and indirect association of grip strength to disability demonstrates the power of this variable in predicting disability. Although studies have documented grip strength as declining linearly with age, research has also indicated that increasing physical activity can retard the loss of strength associated with aging (Cauley et al. 1987). Research has indeed documented the benefits of strength and exercise training on various abilities, including mood, physical impairments, and ADL tasks (Brill 1999; Singh, Clements, and Fiatarone 1997; Tsutsumi et al. 1998). Encouragingly, intervention trials have demonstrated that individuals over the age of 80 can reap benefits from strength training even among individuals who have been sedentary for the majority of their lives (Grimby et al. 1992).
One interesting result that emerged was the bidirectional and opposite relationship between depression and cognitive impairment. Being cognitively impaired was associated with not being depressed, whereas being depressed was associated with being cognitively impaired. Both these findings are consistent with research that has examined depression and cognitive status, which suggests a suppressor effect for dementia. Several studies have identified depression as a risk factor for subsequent onset of dementia (LaRue, Swan, and Carmelli 1995; Palsson, Aevarsson, and Skoog 1999). One suggestion is that depression in old age reflects organic brain changes that might contribute to dementia, especially in the case of early onset major depression (Palsson et al. 1999). On the other hand, for persons who are already demented, the onset and prevalence of depression has been found to be less common than once reported and higher in vascular rather than in Alzheimer's dementia (Ballard, Bannister, Solis, Oyebode, and Wilcock 1996; Brodaty and Luscombe 1996). In sum, depression in the presence of dementia can strongly accelerate disability, whereas dementia in the presence of depression does not necessarily have the same effects. Work by Ritchie, Touchon and Ledesert 1998 supports the notion of depression as a vulnerability factor and stresses further the importance of treating depression in this older population.
Interestingly, although our findings regarding age effects are consistent with other studies, our findings are inconsistent with studies that showed significant gender and SES differences in disability (Berkman et al. 1993; Crimmins et al. 1996; Parker et al. 1994; Rogers 1995; Rogers et al. 1992; Samuelsson and Sundstrom 1988). This inconsistency suggests possible cultural differences between the United States and Sweden; however, there could be other explanations. Unlike the studies that include a full range of ages of older adults (65 to 85+), the present study focuses only on individuals age 80 and over. There could be cohort differences between the "young-old" and the "old-old" that translates into their disability experience. The old-old could be "selected" for survival on the basis of certain resistance characteristics to disability that have to do with factors other than gender, educational, and SES effects (Smith 1997). In other words, at very old ages, the gender and SES gap has simply gotten smaller, such that gender and SES disadvantages have lost their salience. Unfortunately, this study does not explore the potential crossover effects for these variables, but certainly suggests that future research should compare the differential experience of disability for young-old and old-old adults.
The variable of social isolation had initially shown a significant bivariate relationship to disability. However, this variable also had very high bivariate correlations to depression, social integration, and subjective health. As a consequence, social isolation provided no unique variance when the other variables were included in the model (i.e., collinearity). This finding does not diminish the importance of social isolation; rather, it continues to highlight the importance of understanding the individual's social environment as a way to motivate the individual and enhance ability.
Although we would have liked to examine the longitudinal change in disability severity and its associated predictors, the small cell sizes within different change categories (improved, declined, and stable) prevented this possibility. Over the 2-year transition, 60% of the sample remained stable; however, this stability was of three types: no disability at both waves (11%), moderate disability at both waves (10%), and severe disability at both waves (37%). The remaining 40% changed over the 2-year period by improving (7%), suffering moderate decline (16%), or suffering severe decline (19%). With the sample size of 203, the cell sizes for any comparisons between or among change categories were too small to conduct reliable change analyses (e.g., polytomous logistic regression). Therefore, we were limited to testing cross-sectional models and interpreting them on the basis of the foreknowledge that the paths represented associations among sets of variables, rather than causal relations. We acknowledge that the directionality of relationships (from risk factors to disability) could just as well be reversed (disability as a predictor for depression, etc.).
Another limitation to the study was the inability to include the pathology component of the disablement model. Future steps to understanding the components and correlates of disability should incorporate pathology into the model in a way that makes it possible to identify the specific organ-level conditions that put an individual at risk for developing a disability. Different disease conditions can impact the disablement process in different ways, ranging from their effects on functional impairments and limitations to their effects on the intraindividual factors. In the present study, we found cerebrovascular diseases, osteoporosis, diabetes, and visual disorders to have the strongest relations to disability. These conditions not only have physical and functional effects but also lifestyle components that could impact disability, both for the better and for the worse.
We also acknowledge that the complexity of the path models, the limited sample size, and the large number of variables could have masked relationships that are actually present. Research with an emphasis on prospective analyses would better enable us to define causal relations and directionality. With the small sample size and the number of post hoc modifications, we also faced the issue that as more variables were introduced in the path model, the likelihood increased that the results were obtained by chance. Future steps to address this issue include validating the modified model on data obtained from a new, independent sample. Despite these shortcomings, the nature of the study was suitable for exploratory model development without assuming causative relations among variables.
This work touches the surface of the issue of disability in the oldest old. Currently, the oldest old are still relatively unique in their survivorship; however, the 21st century will find more and more individuals entering their ninth and tenth decades faced with the challenge of living with chronic illnesses and disability. On the one hand, our findings imply that a relatively large proportion is severely disabled, requiring regular, sustained, and costly assistance. On the other hand, our findings identify one-quarter of the sample without any disabilities and suggest the areas of physical and psychosocial functioning that may have contributed to the successful lives of these very old adults. Future steps should include a thorough longitudinal testing of the disablement process with an emphasis on the compensatory mechanisms that older adults could use to offset severe disability. Many of the resources used by older adults are potentially modifiable, either by changes in the person or by changes in his or her environment. The goal would be to assess the needs of a person within his or her sociocultural environment so as to reach an optimal balance between personal resources and environmental demands.
|
Acknowledgments |
|---|
Received for publication September 23, 1998. Accepted for publication April 11, 2000.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. J. Phillips and A. K. Stuifbergen Structural Equation Modeling of Disability in Women with Fibromyalgia or Multiple Sclerosis West J Nurs Res, February 1, 2009; 31(1): 89 - 109. [Abstract] [PDF]
|
|
|
|
 |
|
 J. S. Wadsworth, N. Estrada-Hernandez, C. M. Kampfe, and S. M. Smith Economic Outcome of Employment for Older Participants in Rehabilitation Services Funded by the Rehabilitation Services Administration Rehabil Couns Bull, January 1, 2008; 51(2): 107 - 117. [Abstract] [PDF]
|
|
|
|
 |
|
 E. Braungart Fauth, S. H. Zarit, B. Malmberg, and B. Johansson Physical, Cognitive, and Psychosocial Variables From the Disablement Process Model Predict Patterns of Independence and the Transition Into Disability for the Oldest-Old Gerontologist, October 1, 2007; 47(5): 613 - 624. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Jang, W. E. Haley, B. J. Small, and J. A. Mortimer The Role of Mastery and Social Resources in the Associations Between Disability and Depression in Later Life Gerontologist, December 1, 2002; 42(6): 807 - 813. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
