Prediction of Mortality Among Patients With Isolated Traumatic Brain Injury Using Machine Learning Models in Asian Countries: An International Multicenter Cohort Study

Abstract

Study objectives Traumatic brain injury (TBI) is a significant health care concern in many countries, accounting for a major burden of morbidity, mortality, disability, and socioeconomic losses. Although conventional prognostic models for patients with TBI have been validated, their performance has been limited. Therefore, we aimed to construct machine learning (ML) models to predict clinical outcomes in adult patients with isolated TBI in Asian countries.

Methods This study used the Pan-Asian Trauma Outcome Study registry, prospectively collected from January 1, 2015, to December 31, 2020. Among 6,540 patients (≥ 15 years) with isolated moderate and severe TBI, 3,276 (50.1%) were included for model evaluation, and 3,264 (49.9%) for model training and validation. Logistic regression as a baseline, and ML models were constructed and evaluated using the area under the precision-recall curve (AUPRC) as the primary outcome metric, area under the receiver operating characteristic curve (AUROC), and precision at fixed levels of recall. The contribution of the variables to the model prediction was measured using the Shapley Additive ExPlanations method.

Results The ML models outperformed logistic regression in predicting in-hospital mortality. Among the tested models, the gradient-boosted decision tree had the best performance (AUPRC, 0.746 [0.700–0.789]; AUROC, 0.940 [0.929–0.952]). The most powerful contributors to model prediction were the Glasgow Coma Scale, O2 saturation, transfusion, systolic and diastolic blood pressure, body temperature, and age.

Conclusion Our study demonstrated that ML techniques perform better than conventional multivariate models in predicting outcomes among adult patients with isolated moderate and severe TBI.