Preprint · Mar 17, 2026
Development and Validation of a Fall Risk Prediction Model for Slope Walking in Older Adults Using Multimodal Biomechanical Data (Preprint)
By Sr JX, Li S
BACKGROUND Falls are a leading cause of injury and mortality among older adults, with slopes posing particular risk due to elevated biomechanical demands. However, objective tools for slope-specific fall risk assessment remain lacking. OBJECTIVE This study aimed to develop and validate a fall risk prediction model for slope walking in older adults using multimodal biomechanical data acquired through wearable devices, with the goal of enhancing the objectivity and precision of fall risk assessment. METHODS Eighty-six community-dwelling older adults aged ≥60 years were recruited and classified into faller and non-faller groups based on fall history in the previous 12 months. Plantar pressure, hip-knee-ankle joint kinematics, and lower limb muscle electromyographic signals were collected during walking on a 10° slope. LASSO regression (λ1se criterion) was employed to identify independent predictors, and multivariate logistic regression was used to construct the prediction model with a corresponding nomogram. Internal validation was performed using Bootstrap resampling (1,000 iterations), and external validation was conducted with an independent sample of 37 participants. Model performance and clinical utility were comprehensively evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). RESULTS Compared with non-fallers, fallers demonstrated significantly different sEMG root mean square (RMS) amplitudes, joint range of motion, and plantar pressure parameters during both uphill and downhill walking (P < 0.05). Through sequential univariate screening, LASSO regression, and multivariable logistic regression analysis, three independent predictors were retained: uphill vastus lateralis RMS (OR = 1.076), downhill knee range of motion (OR = 0.952), and downhill heel-medial (H-M) peak force (OR = 0.891). The model achieved an AUC of 0.859 (95% CI: 0.777–0.941) in the training cohort and 0.838 (95% CI: 0.703–0.974) in the external validation cohort. Calibration curves and DCA demonstrated satisfactory model calibration and clinical utility. CONCLUSIONS Uphill vastus lateralis RMS, downhill knee range of motion, and downhill H-M peak force constitute independent risk factors for slope-related falls among older adults. The multimodal biomechanical prediction model exhibited favorable discriminative ability and calibration, providing an evidence-based foundation for early screening and targeted intervention strategies in this population.