EARLY PREDICTION OF SEPSIS BASED ON PATIENT VITAL SIGNS AND LABORATORY VALUES USING XGBOOST

Sepsis is a fatal disease with a high mortality rate, especially in intensive care patients. Early and accurate diagnosis of sepsis is important because delay in treatment increases mortality. Infectious Disease Prevention, Clinical Abnormalities and Early Warning Signs are the usual diagnostic criteria for diagnosing sepsis in practice. However, the score cannot provide an early prediction for sepsis, in which the mortality rate will decrease if patients receive emergency treatment. The person applying for isolation can predict the fact of sepsis 6 hours before the diagnosis of the disease. To achieve this, a patient's electronic medical record, demographic information, and vital signs are used. This study uses a data preprocessing strategy adapted to the dataset. This plan introduces a new outlier-based mean data evaluation method, increasing the value of existing data and thus improving the overall accuracy of the prediction. It is made easier for physicians to understand the model by providing an explanation of the main points that affect the distributor's estimate.