Development of artificial intelligence model for predicting in-hospital mortality derived from international classification of diseases, triage scales, procedure codes, and other clinical features in trauma patients: A nationwide population-based study (Preprint)

BACKGROUND Trauma-related mortality is a heavy burden. Estimating the mortality risk in trauma patients is crucial to enhance treatment efficiency and reduce the burden. The most popular and accurate model is the injury severity score based on the abbreviated injury scale (AIS), which is an anatomical injury severity scoring system. However, the AIS requires specialists to code the injury scale by reviewing a patient's medical record; therefore, applying the model to every hospital is impossible. OBJECTIVE We aimed to develop an artificial intelligence (AI) model to predict in-hospital mortality in trauma patients using the international classification of disease (ICD)-10, triage scale, procedure codes, and other clinical features. METHODS We used the Korean National Emergency Department Information System (NEDIS) dataset (n=778,111) from over 400 hospitals from 2016 to 2019. To predict in-hospital mortality, we used ICD-10; patient’s age; gender; intentionality; injury mechanism; emergent symptom; AVPU scale; Korean triage and acuity scale (KTAS); and procedure codes as input features. We proposed the ensemble of deep neural networks (EDNN) via five-fold cross-validation, and compared with other state-of-the-art machine learning models, including traditional prediction models. We further investigated the effect of features. RESULTS Our proposed EDNN with all features provided the highest AUROC of 0.9507, which outperformed other state-of-the-art models, including traditional prediction models: AdaBoost (AUROC of 0.9433), XGBoost (AUROC of 0.9331), ICD-based injury severity score (AUROC of 0.8699 an inclusive model and AUROC of 0.8224 an exclusive model), and KTAS (AUROC of 0.1841). In addition, using all features provided higher AUROC than any other partial features: EDNN with the features of ICD-10 only (AUROC of 0.8964) and EDNN with the features excluding ICD-10 (AUROC of 0.9383). CONCLUSIONS Our proposed EDNN with all features outperforms other state-of-the-art models, including the traditional diagnostic code-based prediction model and triage scale.