Regression-Based Balance Discordance Refines Prediction of Prospective Falls After Subacute Stroke: A Secondary Analysis of the LEAPS Trial

Regression-based multivariable balance discordance predicts prospective, recurrent falls after subacute stroke (HR 1.27, 95% CI 1.03–1.58), outperforming threshold-based indicators in the LEAPS cohort.

• Why it matters: Prior work using threshold classifications did not show significant prediction of recurrent falls, leaving uncertainty about whether more continuous, data-driven balance discordance better captures fall risk.
• What they did: In a secondary observational analysis of the LEAPS trial (384 adults; 2,357 person-intervals; 179 fall events), recurrent falls over 12 months were modeled with Andersen-Gill Cox regression, using residual-based discordance from regressing balance confidence on walking distance, Berg Balance Scale, or both jointly.
• The result: Single-variable discordance related to lower fall hazard (HR 0.79, 95% CI 0.63–0.99), while multivariable discordance related to higher hazard; threshold-based indicators were not significant, and multivariable discordance matched a three-component PCA composite (concordance 0.616 vs 0.618; R-squared 0.49).

AI summary of the abstract · may contain errors · not medical advice · read the paper.