Early Prediction of Sepsis from Clinical Data Using Artificial Intelligence

Abstract

Sepsis is a major cause of death in the world. World Health Organization estimates 30 million people developing sepsis and 6 million people die from sepsis each year; an estimated 4.2 million newborns and children are affected. The mortality rate is highest in septic shock in poor and developing countries. Early prediction of sepsis is critical for improving sepsis outcomes. The late prediction of sepsis in non-sepsis patients is a challenging problem. The aim of this study is to develop an artificial intelligence-based early warning and therapeutic decision support system which reduces sepsis-associated hospital mortality. We propose two compatible Boolean switchable Partially Observable Markov Decision Processes (POMDP) under a general risk-sensitive optimization criterion with finite time horizon. It is based on Spectral analysis of unevenly sampled (missing) observations with Demographics, Vital Signs, and Laboratory values for the patient. The policy is a common mixture of sepsis and non-sepsis beliefs on own utility functions which favors to achieve Pareto Optimality from this high dimensional belief space.