Sex-specific prediction of one-year fall risk using self-reported measures: Results from the European DO-HEALTH trial

Introduction: Despite extensive research, no robust and clinically implementable tool for predicting one-year fall risk in older adults is currently available. Existing prediction models are rarely sex-specific and frequently rely on predictors that are difficult to apply in routine practice. Moreover, many have been developed using retrospective data increasing the risk of recall bias. This study addresses these limitations by developing sex-specific fall risk prediction models based on simple, self-reported measures using prospectively collected data from European community-dwelling older adults.

Method: This is a secondary analysis of DO-HEALTH, a multicenter randomized controlled trial enrolling 2157 adults aged 70+ across seven European sites. The outcome was time to first fall within one year. Sociodemographic, lifestyle, physical and mental health self-reported candidate risk factors were measured at baseline. Sex-specific Cox proportional hazards models were fitted, with censoring at the time of first fall, death, or one-year follow-up, whichever occurred first. Predictor selection was performed using LASSO-penalized regression, with trial design variables forced into the models. Model discrimination was assessed using Harrell’s concordance index (C-index).

Results: A total of 2060 participants were included (792 men and 1268 women; mean age 74.9 ± 4.4 years). Among men, 245 fall events occurred and penalized Cox models retained age and prior falls as the main contributors to one-year fall risk prediction, with no additional self-reported predictors selected (C-index = 0.63). Among women, 510 fall events occurred and penalized models identified prior falls, living alone, self-reported memory problems, and poorer self-rated health as key predictors of one-year fall risk (C-index = 0.61).

Conclusion: One-year fall risk prediction differed markedly by sex, with parsimonious prediction after accounting for prior falls in men and added prognostic value from simple self-reported measures in women. These findings highlight sex-specific opportunities for fall risk prediction, with further refinement and external validation planned.