Predicting falls among patients in a rehabilitation hospital.

A retrospective case-control study within a cohort was carried out to identify factors associated with falling by patients in a rehabilitation hospital. The cohort comprised all patients admitted during a 2-year period from 1984 to 1986. Cases were those who fell one or more times during their hospitalization. For each case, one control matched on sex and date of admission, was chosen. Material from the first year of the cohort (201 cases and 201 controls) was used to define a model to estimate the risk of falling. Of the more than 100 variables considered in the conditional logistic regression models, stroke, incontinence, anticonvulsant medications, and topical eye preparations were significantly associated with the risk of falling. Analyses carried out on the second year of the cohort (155 cases and 155 controls) confirmed the predictive ability of these four variables. By combining the material from both subcohorts we were able to determine that a program to reduce falls directed towards stroke patients and incontinent patients alone could, theoretically, prevent 44% of first falls.