Stepping Beyond Assessment: Fall Risk Prediction Models Among Older Adults from Cumulative Change in Gait Parameter...

Area of Science:

• Gerontology
• Biomedical Engineering
• Data Science

Background:

• Falls are a major health concern for older adults, leading to loss of independence and increased mortality.
• Early fall risk detection enables timely interventions by healthcare providers and caregivers.
• Gait parameters, like walking speed, are vital signs indicative of fall risk.

Purpose of the Study:

• To evaluate supervised classification models for estimating fall risk in older adults.
• To assess the effectiveness of 3D depth sensor data for real-time gait parameter estimation.
• To determine the predictive accuracy of cumulative gait changes in the days preceding a fall.

Main Methods:

• Utilized supervised classification techniques including logistic regression, support vector machines, and tree-based models.
• Employed 3D depth sensors to capture gait parameters within participants' homes.
• Focused on recall as the primary metric due to the critical nature of fall injuries.

Main Results:

• Multivariate logistic regression, support vector, and tree-based models improved fall risk assessment by 18.80%, 31.78%, and 33.94% respectively, in the 14 days before a fall.
• Random Forest and XGBoost models achieved recall and precision scores of 0.805.
• These advanced models outperformed the best univariate regression model (Y-Entropy) with a recall of 0.639 and precision of 0.527.

Conclusions:

• Supervised classification of gait parameters from 3D depth sensors effectively estimates fall risk in older adults.
• Machine learning models significantly enhance predictive accuracy for falls in the crucial 14-day window prior to an event.
• This technology offers a promising avenue for proactive fall prevention strategies in home environments.