Medical outcomes study short form 36: testing and cross-validating a second-order factorial structure for health system employees.

OBJECTIVE: To test the factorial validity of the SF-36. DATA SOURCE: Sample data collected in 1995 and 1996 using telephone interviews with health system employees as part of a study of health status. METHODS OF ANALYSIS: Confirmatory factor analysis and structural equation modeling techniques were used to evaluate the data. PRINCIPAL
FINDINGS: The results of this study suggest that (1) Mental Health and Physical Health are not independent; (b) Mental Health cross-loads onto Physical Health; (c) general health loads onto Mental Health instead of Physical Health; (d) many of the error terms are correlated; (e) the physical function subscale is not reliable across the samples or the "age" or "education" subgroups; and (f) the mental health subscale path from Mental Health is not reliable across some subgroups. This hierarchical factor pattern was replicated across both samples.
CONCLUSIONS: This study supports the second-order factorial structure of the SF-36. Adding the covariance path between the variables Physical Health and Mental Health improved model fit. Two paths from the second-order latent variables to the first-order latent variables differ from the original hypothesized structure of the SF-36. Health perception was influenced by Mental Health rather than Physical Health, and mental health was influenced by both Mental Health and Physical Health. This cross-loading suggests that the perception of Physical Health greatly affects mental health. Scale instabilities in the SF-36 across subgroups suggest that a comparison of mean scores or summary scores is inappropriate. Data interpretation can be improved if multigroups structural equation modeling is used.