A classification tree for predicting recurrent falling in community-dwelling older persons

OBJECTIVES: To develop a classification tree for predicting the risk of recurrent falling in community-dwelling older persons using tree-structured survival analysis (TSSA). DESIGN: A prospective cohort study. SETTING: A community in the Netherlands. PARTICIPANTS: One thousand three hundred sixty-five community-dwelling older persons (greater than or equal to65) from the Longitudinal Aging Study Amsterdam (LASA). MEASUREMENTS: In 1995, physical, cognitive, emotional, and social aspects of functioning were assessed. Subsequently, a prospective fall follow-up, specifically on recurrent falls (two falls within 6 months) was conducted for 3 years. RESULTS: The classification tree included 11 end groups differing in risk of recurrent falling based on a minimum of two and a maximum of six predictors. The first split in the tree involved two or more falls versus fewer than two falls in the year preceding the interview. Respondents with two or more falls in the year preceding the interview (n = 193) and with at least two functional limitations (n = 98) had a 75% risk of becoming a recurrent faller, whereas respondents with fewer than two functional limitations were further divided into a group with regular dizziness (n = 11, risk of 68%) and a group with no regular dizziness (n = 84, risk of 30%). In respondents with fewer than two falls in the year preceding the interview (n = 1,172), the risk of becoming a recurrent faller varied between 9% and 70%. Predictors in this branch of the tree were low performance, low handgrip strength, alcohol use, pain, high level of education, and high level of physical activity. CONCLUSION: This classification tree included 11 end groups differing in the risk of recurrent falling based on specific combinations of a maximum of six easily measurable predictors. The classification tree can identify subjects who are eligible for preventive measures in public health strategies.