Explainable Fall Risk Prediction in Older Adults Using Gait and Geriatric Assessments

Area of Science:

• Gerontology and Health Informatics
• Machine Learning in Healthcare
• Geriatric Medicine

Background:

• Older adults (65+) face a high risk of falls, leading to injuries and reduced quality of life.
• Early prediction of fall risk is crucial for implementing timely interventions.
• Current methods may not fully capture the multifactorial nature of fall risk in this population.

Purpose of the Study:

• To develop and validate a machine learning model for predicting 6-month fall risk in older adults.
• To identify key predictors of fall risk using geriatric assessments and gait analysis.
• To enhance personalized interventions by understanding contributing factors to fall risk.

Main Methods:

• Utilized data from 92 older adults, including geriatric assessments (ADL, IADL, MMSE, GDS, SF12), GAITRite measurements (FAP scores, gait speed), and fall history.
• Employed machine learning models trained on this data to predict 6-month fall risk.
• Applied the SHAP (SHapley Additive exPlanations) approach to interpret model predictions and identify influential variables.

Main Results:

• The developed models achieved an AUC of 0.80, sensitivity of 0.82, specificity of 0.72, F1 score of 0.76, and accuracy of 0.75 in predicting 6-month fall risk.
• The SHAP analysis provided insights into individual predictor contributions to fall risk.
• Geriatric assessments and gait parameters were significant in fall risk prediction.

Conclusions:

• The study demonstrates a robust method for early fall risk prediction in older adults.
• Accurate prediction allows healthcare providers and caregivers valuable time for preventive actions.
• This approach supports personalized interventions, potentially reducing fall incidence and improving senior safety.