Predicting Geriatric Rehabilitation Stays of ≤4 Weeks After Hip Fracture Surgery: Machine Learning Approach Using...

Area of Science:

• Geriatric Medicine
• Rehabilitation Science
• Machine Learning in Healthcare

Background:

• Hip fractures are common in older adults (≥70 years), with 50% requiring geriatric rehabilitation.
• Increasing patient numbers and healthcare pressures necessitate efficient rehabilitation strategies, including home-based care.
• Early identification of patients suitable for discharge facilitates timely planning and smoother transitions home.

Purpose of the Study:

• To develop machine learning (ML) models for predicting geriatric rehabilitation stays of ≤4 weeks post-hip fracture surgery.
• Utilize continuously monitored physical activity data from the first week and patient characteristics for prediction.

Main Methods:

• Prospective cohort study of 100 patients undergoing hip fracture surgery.
• Collected patient characteristics and physical activity data using the MOX1 accelerometer.
• Employed principal component analysis and Bayesian hyperparameter optimization to develop and refine ML models.

Main Results:

• The Support Vector Machine (SVM) model achieved high predictive accuracy (0.95).
• Key predictors included physical activity data, emergency room time, functional ambulation, age, and cognitive status.
• Other important features were independence in daily activities, social support, surgery type, and comorbidities.

Conclusions:

• Developed accurate ML models, particularly SVM, for predicting short geriatric rehabilitation stays (≤4 weeks).
• These models can aid in optimizing discharge planning and resource allocation for hip fracture patients.
• Integrating physical activity monitoring with patient data offers a promising approach for future-proof rehabilitation.