Identifying fallers with Parkinson's disease using home-based tests: who is at risk?

Abstract

The objective of this work is to determine risk factors for falling in patients with Parkinson's disease (PD) using home-based assessments and develop a prediction model. Data on falls, balance, gait-related activities, and nonmotor symptoms were obtained from 153 PD patients (Hoehn-Yahr 2-4) in their home. Fifty-one candidate determinants for falling were independently tested using bivariate logistic regression analysis. A multivariate logistic regression model was developed to identify patients susceptible to falls. Sixty-six subjects (43%) were classified as fallers. Eighteen determinants for falling were selected. The final multivariate model showed an accuracy of 74% and included: (1) Freezing of Gait Questionnaire, (2) Timed Get Up and Go (TGUG) score, (3) disease duration, (4) item 15 of the Unified Parkinson's Disease Rating Scale. Based on disease duration, freezing symptoms, walking problems, and a prolonged TGUG duration, assessed in the home situation, it was possible to accurately identify 74% of PD patients as fallers.