Predictive model for early functional outcomes following acute care after traumatic brain injuries: A machine...

Area of Science:

• Neurology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Limited data exists on early functional outcomes post-acute care for traumatic brain injury (TBI).
• Developing predictive models is crucial for effective TBI patient management and resource allocation.

Purpose of the Study:

• To develop and validate a machine learning model for predicting functional outcomes at discharge in TBI patients.
• To identify key predictors of functional recovery following acute TBI care.

Main Methods:

• A retrospective analysis of 5281 TBI patients from a Beijing hospital discharge database.
• Model derivation and internal validation using 4181 patients; external validation with 1100 patients.
• Comparison of logistic regression, XGBoost, random forest, decision tree, and neural network models, evaluating performance metrics like AUC and F1-score.

Main Results:

• The random forest model demonstrated superior predictive performance in both internal (AUC 0.856) and external (AUC 0.779) validation.
• Key predictors identified include admission Barthel Index score, age, non-surgical treatment, neurosurgery status, and Charlson Comorbidity Index.
• The model achieved an F1-score of 0.724 in internal and 0.604 in external validation.

Conclusions:

• A robust random forest model was established for predicting early functional outcomes in TBI patients post-acute care.
• The model can inform clinical decision-making, patient management, discharge planning, and healthcare resource allocation for TBI.
• This predictive tool enhances healthcare quality assessment and resource optimization in TBI treatment.