A machine learning based predictive model for the diagnosis of sepsis

The early recognition and treatment of sepsis is essential to increase the probability of survival of the patient. Sepsis is a complex and heterogeneous syndrome influenced by the site of infection, causative microorganisms, acute organ dysfunctions and co-morbidities. So, early diagnosis is a challenge in which complex physiologic, metabolic, biochemical markers and clinical signs must be evaluated simultaneously for a reliable and early identification of sepsis. In this paper, a list of relevant variables involved in sepsis diagnosis is provided. Furthermore, to help answering the question of whether a patient is suffering from sepsis when entering the emergency room, a model based on the machine learning gradient boosting algorithm is proposed. Using three histones H2B, H3, H4 and the activated protein C together with other variables a gradient boosting classifier was trained and evaluated with cross validation. Results show that the model can achieve up to a 97% mean per-class accuracy.