Novel insights on the relative importance of clinical and gait measures for detecting fall risk in community-dwelling older adults

We sought to extend recent research that explored model-based approaches for combining clinical and gait measures (e.g., mobility, balance, strength, postural sway, difference scores between dual- and-single-task gait conditions) to determine the most sensitive grouping for retrospectively classifying fallers from non-fallers. In the present study, the clinical assessment battery was augmented by incorporating more challenging balance items while removing clinical measures characterized by ceiling effects and restricted range. The current analysis yielded improved specificity, but slightly lower sensitivity. Notably, in both analyses, gait variables were central in identifying fall risk, with single- vs. dual-task difference scores of particular predictive importance. The differences observed between the best-fitting models across the two cohorts implies that modelling methods should accommodate and harness individual differences (e.g., machine learning techniques).