| The purpose of this study was to determine whether the Functioning and Activity Model, a conceptual model for studying physical functioning in elders, fit the data from the Community Healthy Activities Model Program for Seniors (CHAMPS) (Stewart et al., 1998). The data were analyzed using structural equation modeling (SEM), a technique that permits the simultaneous analysis of many hypothesized associations among variables, including indirect effects and multiple outcomes.; A random sample of members of two Medicare HMOs in northern California was screened for eligibility in the CHAMPS study. Baseline data were collected in 1996 from 249 medically stable participants, aged 65--90, who lived independently in the community. 173 of the participants were physically underactive and 76 were active. The SEM analysis reported in this paper used baseline data from individuals in both these groups (N = 225) who had no missing data on the measures used in this analysis.; The SEM analysis of the fit of the CHAMPS data to the conceptual model was conducted using the AMOS 4 computer program for structural equation modeling (Arbuckle & Wothke, 1999). The model contained six hypothesized predictor variables measured by self-report: pathology, obesity, age, gender, socioeconomic status, and symptoms. One predictor, fitness, was measured by a timed walk test. Three outcome variables in the model represented separate hypothesized levels of functioning: (1) physical functioning measured by performance tests, (2) physical functioning measured by the physical functioning scale of the SF-36, and (3) activity participation measured by the SF-36 role functioning scale and social functioning scale.; Results showed that the physical functioning and activity outcomes were each predicted by different groups of variables, supporting the hypothesis that the physical functioning and activity variables in the model represent different concepts. The goodness-of-fit index (GFI) for the model was .98, indicating that approximately 98% of the variance in the data was explained by the model. The strongest predictor variable in the model was presence of symptoms, which predicted self-report of difficulty in physical functioning (standardized path coefficient = --.49, p < .001) and activity participation (standardized path coefficient = --.76, p < .001). |
