Abstract
Sepsis, a syndrome with disturbed host response to severe infection, is a critical health problem worldwide. It is urged to develop and update novel therapeutic strategies for improving the outcome of sepsis. In this study, we demonstrated that different bacteria clustering in sepsis patients may generate differences of prognosis results. We extracted all the sepsis patients from Medical Information Mart for Intensive Care IV 2.0 (MIMIC-IV 2.0) critical care data set according to certain standards and clinical score, a total of 2,339 patients were included in our study. Then we used multiple data analytics and machine learning methods to make all data deeply analyzed and elucidated. The results showed that the types of bacteria infected by patients with different ages, sex and race are different, the types of bacteria infected by patients with different SIRS values and GCS scores of the first day are different, and the severity of patients with different clusters is different, and most importantly, the survival rate of patients with different clusters also has this significant difference. We concluded prognostic assessment predicated by bacteria clustering might be a relatively potentially novel strategies and perspectives on prevention and management for sepsis in the future.