Prediction of Clinically Significant Improvements During the Interdisciplinary Intensive Outpatient Program for...

Area of Science:

• Neuroscience
• Medical Informatics
• Machine Learning

Background:

• Traumatic brain injury (TBI) poses significant challenges for active duty service members.
• Interdisciplinary Intensive Outpatient Programs (IOPs) are utilized for TBI rehabilitation.
• Predicting treatment success is crucial for optimizing patient care and resource allocation.

Purpose of the Study:

• To evaluate machine learning (ML) models for predicting clinically significant patient improvement.
• To identify key factors associated with treatment success in an IOP for TBI.
• To assess the utility of ML in precision medicine for TBI care.

Main Methods:

• Utilized Extreme Gradient Boosting (XGBoost) models on data from 790 active duty service members.
• Included demographic, posttraumatic stress, depression, anxiety, post-concussion, and sleep measures.
• Compared models using total self-report scores versus symptom cluster scores.

Main Results:

• The model incorporating symptom clusters achieved higher predictive accuracy (79% AUC, 72% accuracy) than the model with total scores (75% AUC, 68% accuracy).
• Top predictors included posttraumatic stress arousal, avoidance, and reexperiencing sub-scores, education, and cognitive post-concussion sub-scores.
• Severity of posttraumatic stress symptoms at admission was a primary predictor of improvement.

Conclusions:

• XGBoost models effectively predict clinically significant improvement in TBI patients undergoing IOP.
• ML integration into clinical practice offers a precision medicine approach for predicting treatment efficacy.
• Accurate prediction of treatment outcomes can enhance healthcare resource allocation and patient outcomes.