The determinants of walking velocity in the elderly. An evaluation using regression trees

To determine predictors of walking velocity in the elderly.

Five hundred thirty-two persons from 3 cohorts of elderly persons (retirement community, ambulatory care medical clinic, or chronically homebound population) performed a 10-foot (for the homebound subjects) or 50-foot (for all other subjects) walk time test and underwent a standardized interview, chart review, and clinical examination. The 73 independent variables that were evaluated included demographic, musculoskeletal, neurologic, psychologic, and other comorbidity items. Least-squares and least-absolute-deviation regression tree analyses were performed to determine the strongest predictive factors associated with walking velocity.

Sampling cohort (homebound versus non-homebound), quadriceps strength, hip flexion strength, lumbosacral spine impairment, lower joint impairment, and education were found to be associated with walking velocity. Joint pain measures were not associated with walking velocity.

Muscle strength variables are better predictors of walking velocity than are joint pain variables. Thus, clinical trials and observational studies using walking velocity as an outcome need to take into consideration the influence of muscle strength on this outcome variable.