Development and validation of an algorithm to assess risk of first-time falling among home care clients

Abstract

Abstract Background The falls literature focuses on individuals with previous falls, so little is known about individuals who have not experienced a fall in the past. Predicting falls in those without a prior event is critical for primary prevention of injuries. Identifying and intervening before the first fall may be an effective strategy for reducing the high personal and economic costs of falls among older adults. The purpose of this study was to derive and validate a prediction algorithm for first-time falls (1stFall) among home care clients who had not fallen in the past 90 days. Methods Decision tree analysis was used to develop a prediction algorithm for the occurrence of a first fall from a cohort of home care clients who had not fallen in the last 90 days, and who were prospectively followed over 6 months. Ontario home care clients who were assessed with the Resident Assessment Instrument-Home Care (RAI-HC) between 2002 and 2014 (n = 88,690) were included in the analysis. The dependent variable was falls in the past 90 days in follow-up assessments. The independent variables were taken from the RAI-HC. The validity of the 1stFall algorithm was tested among home care clients in 4 Canadian provinces: Ontario (n = 38,013), Manitoba (n = 2738), Alberta (n = 1226) and British Columbia (n = 9566). Results The 1stFall algorithm includes the utilization of assistive devices, unsteady gait, age, cognition, pain and incontinence to identify 6 categories from low to high risk. In the validation samples, fall rates and odds ratios increased with risk levels in the algorithm in all provinces examined. Conclusions The 1stFall algorithm predicts future falls in persons who had not fallen in the past 90 days. Six distinct risk categories demonstrated predictive validity in 4 independent samples.

Keywords

• Falls
• Home care
• Older adults
• Machine learning
• interRAI