MEDICAL INFORMATICS IV
DETECTION OF BIOMARKERS AFFECTING MORTALITY: AN EMERGENCY DEPARTMENT STUDY
Yayıncı:
İstanbul Üniversitesi Yayınları
In this study, data mining techniques were employed to evaluate a dataset of 1106 patients between 15-84 ages who were admitted to the emergency department of a hospital, between February 1, 2021 - May 1, 2021. All patients experienced mortality within 24 hours or after 24 hours following the admission of the emergency department. The aim was to identify significant biomarkers affecting mortality. Accordingly, the focus was on various tests providing essential information about the patient’s metabolic functions and general health status, such as Be, Hbg, HCO3, Htc, Inr, Lac, Mean Arterial Pressure (MAP), pCO2, pH, heart rate, and pulse pressure. Furthermore, variables related to the Injury Severity Score (ISS) used for classifying injuries in trauma patients and the Glasgow Coma Scale (GCS) to assess the overall level of consciousness were included into the dataset. Supervised machine learning algorithms, including Support Vector Machines (SVM), Decision Trees (DT), Naive Bayes (NB), Logistic Regression (LR), and Multi-Layer Perceptrons (MLP) were performed to classify cases. As a result, the highest accuracy rate for mortality prediction was achieved with SVM. MAP, Lac, ISS, and GCS were determined to be the significant variables for separating the classes.
