Modeling recovery from stumbles: preliminary data on variable selection and classification efficacy.

Abstract

The primary purpose of this preliminary investigation was to determine the functional relationship between selected information processing time and response execution variables and measures of postural stability in elderly women. A secondary purpose was to explore the efficacy of a neuromotor model using selected variables to retrospectively identify subjects with a self-reported history of falling. Descriptive, retrospective, cohort. General community. Convenience sample of 17 community-dwelling females with a mean age of 72.2 years. Postural stability variables included the amplitude and frequency of postural sway during static vision-aided no-vision conditions. Information processing and response execution variables were collected using a simple-choice reaction time paradigm for an isometric knee extension task. Postural stability and information processing variables were functionally independent. Based upon significant intergroup differences, simple and choice pre-motor reaction time and non-vision aided anterior posterior sway amplitude were selected for inclusion in a discriminant analysis. The resulting discriminant function was significant (P = 0.01), correctly categorizing all of the subjects with a self-reported history of falling and identifying six out of seven of the non-fallers. These preliminary results suggest that it is feasible to identify a predisposition to falling by detecting an inability to respond successfully to a postural disturbance.