Early Prediction of Standing at Discharge in Moderate-to-Severe Traumatic Brain Injury: A Clinical Machine Learning...

Area of Science:

• Neuroscience
• Medical Informatics
• Rehabilitation Medicine

Background:

• Moderate-to-severe traumatic brain injury (TBI) poses significant challenges for functional recovery.
• Predicting standing ability at discharge is crucial for rehabilitation planning and patient outcomes.
• Existing prediction methods may not fully incorporate dynamic clinical factors.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting favorable standing ability at hospital discharge in TBI patients.
• To incorporate both modifiable and nonmodifiable clinical factors into the prediction model.
• To enhance early and individualized rehabilitation planning for TBI survivors.

Main Methods:

• Retrospective cohort study of 248 adult patients with moderate-to-severe TBI.
• Utilized logistic regression, extreme gradient boosting (XGBoost), random forest, and support vector machine models.
• Assessed model performance using area under the receiver operating characteristic curve (AUC), Brier score, and decision curve analysis (DCA).

Main Results:

• The XGBoost model demonstrated superior performance with an AUC of 0.85 and 78% accuracy.
• Logistic regression also showed strong results (AUC=0.82, 80% accuracy).
• Age, intubation duration, and early mobilization were identified as the most influential predictors.

Conclusions:

• An ML-based model can effectively predict discharge standing ability in TBI patients.
• The model's inclusion of modifiable factors improves clinical utility for rehabilitation.
• Supports data-driven quality improvement and highlights the need for external validation.