Machine Learning Prediction of Fall Risk in Older Adults Using Timed Up and Go Test Kinematics.

Area of Science:

• Gerontology
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Falls in the elderly lead to severe physical, mental, and financial consequences, including mortality.
• Current fall-risk assessment tools like the Timed Up and Go (TUG) test are subjective and rely on clinician judgment.
• Wearable sensors and machine learning offer objective gait analysis but often require extensive data and complex feature engineering.

Purpose of the Study:

• To develop an objective, cost-effective fall-risk detection model for older adults.
• To utilize a sensor data-driven convolutional neural network (CNN) for predicting fall risk.
• To assess the model's sensitivity compared to geriatrician expert assessments.

Main Methods:

• A convolutional neural network (CNN) model was developed using data from three non-intrusive wearable sensors.
• Participants' gait kinematics were measured during the Timed Up and Go (TUG) test.
• The model analyzed sensor data to predict the fall-risk status of older adults.

Main Results:

• The developed CNN model demonstrated high sensitivity in predicting older adults' fall-risk status.
• The model's predictions showed good agreement with geriatrician expert assessments.
• The approach efficiently captured gait impairment aspects from various body locations.

Conclusions:

• A sensor data-driven CNN model provides an objective and efficient method for assessing fall risk in the elderly.
• This technology can aid clinicians in identifying at-risk individuals, potentially reducing fall-related injuries and costs.
• The use of non-intrusive wearable sensors during the TUG test offers a practical approach for fall risk assessment in clinical settings.