A multivariate regression model predicted falls in residents living in intermediate hostel care

Background and Objective: To evaluate whether individual falls risk could be predicted in a frail elderly population. Study Design and Setting: We developed and tested an assessment tool and falls risk score for predicting falls based on a multivariate regression model in a prospective cohort study of intermediate care residents. Results: During the follow-up period, 1,736 falls by 1,107 subjects were recorded with an average of 170 falls per 100 person-years. Fifty percent of the study population had at least one fall within a year. Significant independent risk factors were poor balance, cognitive impairment, incontinence, higher illness severity rating, and older age. Twenty-two percent of participants with a falls risk score ¡Ý7 accounted for 42% of the total falls, with a falls rate of 317 per 100 person-years. This rate was a sixfold increase from the falls rate of 52 per 100 person-years observed in participants with a score < 3. A high score (¡Ý7) indicated almost a 2 in 3 chance of falling, while a low score (<3) indicated approximately a 1 in 7 chance of falling within 6 months. Conclusion: The assessment tool and falls risk score could identify individuals in this frail elderly population at high risk of falls.