Models to Predict Fall History and Fall Risk for Community-Dwelling Elderly

The objective of this study was to compare fall-risk models for the prediction of 1-year fall history in community-dwelling elderly persons. The study design was a descriptive analysis of factors associated with retrospective fall history for individuals living in community-based independent living facilities. Thirty-three older adults (10 men and 23 women, mean age +/- standard deviation, 82.6 +/- 5.5 years) volunteered to participate. The main outcome measure was multivariate logistic regression models using a minimal set of predictor variables for predicting 1-year fall history and fall-risk status. The results showed that a fall history prediction model using age, gait velocity, and time to complete Trails Making Test Part B yielded 76% of overall predictive accuracy (75% sensitivity, 76% specificity). A second logistic regression with gait speed eliminated was used to identify fall-risk status (fall history plus a positive score on the Timed Up and Go test) with similar results. These findings suggest that these variables are critical for identifying elderly fallers and those at risk for falls.