Predicting Pain Response in Remote MSK Care

Study Overview

This study developed an AI tool to predict pain relief outcomes in a remote musculoskeletal (MSK) care program. Analyzing data from 6,125 patients, researchers used machine learning models (RNNs, LightGBM) to assess pain response over seven sessions and guide treatment adjustments.

Key findings:

• Prediction accuracy improved over time, reaching an AUC of 0.70-0.71 by session 7.
• Pain levels, exercise performance, and adherence were the strongest predictors of pain relief.
• High anxiety and depression scores were linked to poorer pain outcomes.
• Longer gaps between sessions (>3 days) correlated with lower treatment success.

By leveraging real-time patient data, this AI-driven approach enables early intervention and personalized treatment adjustments, helping physical therapists improve outcomes for chronic low back pain patients.

Read the full study