Predicting the triage level and clinical outcome of emergency department patients using machine learning approaches

Emergency departments are critical units where fast and accurate decisions must be made within the healthcare system. In emergency departments, where the density is usually high, incoming patients are evaluated and examined according to their urgency levels by passing through the triage area for prioritization purposes. In recent years, has been conducted for predicting patient triage levels to help triage professionals. Predicting patients' clinical outcomes is another common topic in the literature when looking at the prediction studies done with triage data. In this context, an application was made within the scope of the study aiming to estimate both the triage levels and the clinical outcomes of patients applying to the emergency department. In the study, the MIMIC-IV-ED dataset used in the literature was utilized. In the data pre-processing stage, text-based expressions were converted into numerical form with natural language processing methods. Then prediction models were created using random forest, XGBoost, and MLP algorithms. Finally, the performances of these algorithms were presented comparatively.