Enhancing the clinical decision making of sepsis detection: a machine learning approach to early diagnosis

Abstract

Sepsis is a perilous, life-threatening illness and is a leading cause of death in the world; with an annual death toll of 6 million people worldwide (Reyna et al., 2019). The syndromic nature of the condition is convoluted and almost indistinguishable from uncomplicated infections. Sepsis requires immediate admission to an intensive care unit, and every hour of delay was associated with an absolute mortality rate of 0.3% for sepsis and 1.8% for sepsis shock. To further current endeavours in early sepsis detection, this study offers a machine learning model that predicts sepsis 12 hours before the onset time of sepsis-3 clinical criteria; the current norm for diagnosing sepsis.