Resident Physician University of New Mexico Health and Sciences Center Lafayette, Louisiana, United States
Disclosure(s):
Evan Courville, MD: No financial relationships to disclose
Introduction: Traumatic brain injury (TBI) is a leading cause of death and disability worldwide. Use of machine learning (ML) has emerged as a key advancement in TBI management. This study aimed to synthesize machine learning models with demonstrated effectiveness in predicting TBI outcomes.
Methods: We conducted a systematic review in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) statement. There were 15 articles identified using the search strategy. Patient demographics, clinical status, ML outcome variables and predictive characteristics were extracted. Pooled analysis of mortality prediction was performed and meta-analysis of diagnostic accuracy was conducted for the ML algorithms used across studies.
Results: ML algorithms including Support Vector Machine (SVM), Artificial Neural Networks (ANN), Random Forest (RF), and Naiive Bayes (NB) were compared to Logistic Regression (LR). Significant improvement in prognostic capability using ML versus LR was found in 13 studies. The algorithm accuracy was consistently above 80% when predicting mortality and unfavorable outcome, as measured by Glasgow Outcome Scale (GOS). Receiver Operating Characteristic (ROC) curves analyzing sensitivity of ANN, SVM, Decision Tree (DT), and LR demonstrated consistent findings across the studies. Lower admission Glasgow Coma Scale (GCS), older age, elevated serum acid, and abnormal glucose were all associated with increased adverse outcomes and had the most significant impact on ML algorithms.
Conclusion : Machine learning algorithms were stronger than traditional regression models in predicting adverse outcomes. Admission GCS, age, and serum metabolites all have strong predictive power when used with ML and should be considered important components of TBI risk stratification.