A Quantitative Model for Sepsis Stratification

Abstract

Sepsis is a kind of systemic inflammatory response syndrome caused by infection and it endangers the life of patients seriously due to its rapid development progression and high mortality rate. In clinic it is highly demanded to quantitatively stratify the severity of sepsis for individual management. This work aimed to build a quantitative model for sepsis patients which can stratify the disease severity in three levels. For this purpose, clinical data were collected and preprocessed, i.e. screening, normalization and data replenishing. Afterwards, sepsis sensitive parameters were tested and selected, which were utilized as the input of the stratification model. For the model, the algorithm of Support Vector Machine was applied. Eventually, the model was tested in total of 522 clinical cases and an accuracy of 67.5% in stratification was achieved. The performance of the established model is superior to the conventional APACHE scoring method. Preliminary results exhibited that the established model is potential to help improve the patients’ management by quickly stratifying the sepsis severity.