Identifying interacting predictors of falling among hospitalized elderly in Japan: A signal detection approach

Falling is a complex phenomenon that involves interaction of multiple risk factors. The authors analyzed factors related to falls in a geriatric hospital to elucidate interaction of multiple risk factors for falls in elderly inpatients. Subjects were 364 patients (mean age, 81.7; women 76.7%) who were aged 60 years and over and had been hospitalized for more than 6 months between April 2000 and March 2001. A signal detection model was used to identify baseline variables that best divided the sample into subgroups using incidence of falling as an outcome variable. During a follow-up period, 91 patients (25%) had at least one incident of fall. Out of 14 independent variables, a higher-order interaction consisting of six significant variables was identified. Consequently, the subjects were categorized into seven subgroups whose fall rate varied 5.7–80.9%. We found that the combination of non-bedridden state, dementia, and medication of tranquilizers or sleeping drugs was the highest fall rate (80.9%). Signal detection analysis is useful to identify the combination of multiple risk factors of falling, and applicable to develop prevention programs for each subgroups.