A Low Dimensional Algorithm for Detection of Sepsis from Electronic Medical Record Data

Aruna Deogire
A C Patil College of Engineering, Navi Mumbai


Aim: Aim of this work is an early detection of sepsis according to sepsis-3 criteria. Sepsis is one of the life threatening condition caused by organ dysfunction. Early detection of sepsis can prevent severe damage of body functioning. This work is result of participation in Physionet CinC 2019 challenge. Method: After studying the training dataset provided by the organizers of the challenge a simple algorithm is developed to detect the sepsis. Out of 40 parameters the available parameters as per Sepsis-3 guidelines, Mean Atrial Pressure (MAP), Creatinine, Bilirubin and Platelets are examined and scored. The other two parameters which are observed are Systolic Blood Pressure (SBP) and Respiration rate. Score for every parameter is calculated according to sepsis-3 guidelines. The scores are then converted to probability. Applying the threshold criteria patient is labelled as either positive (1) or negative (0). Result : currently the training database is tested and entry is submitted for scoring. The score is not yet declared. According to the tested records approximately 72% of records are correctly detected. The algorithm is still in fine tuning process and expected