36-Item Short Form Survey (SF-36) Versus Gait Speed As Predictor of Preclinical Mobility Disability in Older Women: The Women's Health Initiative.

OBJECTIVES To compare the value of clinically measured gait speed with that of the self-reported Medical Outcomes Study 36-item Short-Form Survey Physical Function Index (SF-36 PF) in predicting future preclinical mobility disability (PCMD) in older women. DESIGN Prospective cohort study. SETTING Forty clinical centers in the United States. PARTICIPANTS Women aged 65 to 79 enrolled in the Women's Health Initiative Clinical Trials with gait speed and SF-36 assessed at baseline (1993-1998) and follow-up Years 1, 3, and 6 (N = 3,587). MEASUREMENTS Women were categorized as nondecliners or decliners based on changes (from baseline to Year 1) in gait speed and SF-36 PF scores. Logistic regression models were used to estimate incident PCMD (gait speed <1.0 m/s) at Years 3 and 6. Area under the receiver operating characteristic curve (AUC) was used to compare the predictive value of SF-36 PF with that of measured gait speed. RESULTS Slower baseline gait speed and lower SF-36 PF scores were associated with higher adjusted odds of PCMD at Years 3 and 6 (all P < .001). For gait speed, decliners were 2.59 times as likely to have developed PCMD as nondecliners by Year 3 and 2.35 times as likely by Year 6. Likewise, for SF-36, decliners were 1.42 times as likely to have developed PCMD by Year 3 and 1.49 times as likely by Year 6. Baseline gait speed (AUC = 0.713) was nonsignificantly better than SF-36 (AUC = 0.705) at predicting PCMD over 6 years (P = .21); including measures at a second time point significantly improved model discrimination for predicting PCMD (all P < .001). CONCLUSION Gait speed identified PCMD risk in older women better than the SF-36 PF did, although the results may be limited given that gait speed served as a predictor and to define the PCMD outcome. Nonetheless, monitoring trajectories of change in mobility are better predictors of future mobility disability than single measures.