Predicting fallers in a community-based sample of people with Parkinson's disease

Background. The risk of people with Parkinson's disease (PD) failing is gre ater than that of the general population but to date, disease-specific pred ictors of failing have not been identified. Objectives: To identify one or more features, which would predict individuals at risk of falling during a 3-month prospective follow-up study. Method: A battery of standardised test s administered in the home and the laboratory with a 3-month follow-up tele phone interview. Results: Sixty-three people with PD were recruited from GP practices. Eleven interview variables and six gait laboratory variables we re used with subsamples (55 and 44 subjects, respectively) to fit predictiv e models for identifying future fallers. The number of falls in the previou s year was the most important variable, without exception, to be selected a s a predictor in various logistic regression models. A history of two or mo re falls had a sensitivity of 86.4% (95% Cl 67.3-96.2%) and a specificity o f 85.7% (95% Cl 71.2-94.2%) in predicting falling in the next 3 months. Con clusion: Healthcare workers should be asking their patients with PD regular ly and carefully about falling, and should consider instigating programmes of fall management for patients with PD who have fallen two or more times i n the previous 12 months. Copyright (C) 2001 S. Karger AG, Basel.