Results are based on log-binomial regression models estimated on 737 fall events among 136 individuals. Robust 95% CIs that account for individual clustering and uncertainty from the imputation of missing data are shown in parentheses for the relative risk estimates. The other estimates are based on predicted probabilities from the models. The multivariable model includes adjustment for age, sex, body mass index, room type, time of day, activity type, cognitive ability, visual impairment, walking ability, history of falls, the use of sedatives and antidepressants, room type, time of day, activity type and hip protector use.