Biopsychosocial based machine learning models predict patient improvement after total knee arthroplasty

Area of Science:

• Orthopedics and Rehabilitation Science
• Artificial Intelligence in Healthcare
• Biopsychosocial Medicine

Background:

• Total knee arthroplasty (TKA) is a common procedure for end-stage osteoarthritis, but patient recovery is complex.
• Current TKA decision-making often overlooks crucial biopsychosocial factors influencing outcomes.
• Integrating these factors can lead to more personalized and effective patient care strategies.

Purpose of the Study:

• To develop patient-centered predictive models for TKA outcomes using machine learning and Bayesian inference.
• To incorporate a comprehensive set of biopsychosocial features into predictive models for enhanced accuracy.
• To predict improvements in quality of life and knee symptomology three months post-TKA.

Main Methods:

• Utilized data from 863 TKA patients across four Australian hospitals (2019-2022).
• Employed machine learning and Bayesian inference to build predictive models.
• Assessed outcomes using Short Form-12 Physical Composite Score (PCS) for quality of life and Western Ontario and McMasters Universities Osteoarthritis Index (WOMAC) for knee symptomology.

Main Results:

• Identified key predictive variables for quality of life, including pre-surgery PCS, nutrition, employment, and hand grip strength.
• Identified key predictive variables for knee symptomology, including pre-surgery WOMAC, pain catastrophizing, and exhaustion.
• Bayesian methods provided individual outcome predictions with associated uncertainty estimates.

Conclusions:

• A holistic model incorporating biopsychosocial features significantly enhances TKA outcome prediction.
• Machine learning and Bayesian inference offer a powerful approach for personalized patient care in TKA.
• Individualized predictions can better inform clinical decision-making and patient management strategies.