Machine Learning Models Leveraging Smartphone-Based Patient Mobility Data Can Accurately Predict Functional Outcomes...

Area of Science:

• Spine surgery outcomes
• Machine learning in healthcare
• Patient recovery monitoring

Background:

• Adjacent segment disease and spondylosis can cause functional decline after spinal surgery.
• Early detection of recovery inflection points is crucial for timely intervention.
• Predicting post-operative mobility decline can improve patient management.

Purpose of the Study:

• To develop machine learning (ML) models for predicting post-operative decline in patient mobility.
• To assess the efficacy of different ML models in forecasting functional recovery after spine surgery.
• To identify key predictors of post-operative functional decline.

Main Methods:

• Retrospective enrollment of patients undergoing spine surgery for degenerative spinal stenosis or spondylolisthesis.
• Collection of peri-operative activity data (steps-per-day) and clinical/demographic variables.
• Development and validation of logistic regression, random forest, and extreme gradient boosting ML models.

Main Results:

• Random forest and XGBoost models demonstrated high accuracy (86.7% and 80%) in predicting post-operative functional decline.
• Random forest model achieved an area under the curve of 0.80.
• Logistic regression model showed lower predictive performance.

Conclusions:

• Machine learning models trained on smartphone-collected activity data can effectively forecast functional decline post-spine surgery.
• These ML tools can enhance surgical prognostication and planning.
• Integration of ML into clinical practice holds promise for improving patient care.