A cohort study of developing a death risk prediction model in patients with septic shock based on MIMIC-Ⅲ database / 中华危重病急救医学

ABSTRACT

Objective:To develop and validate a model for predicting death risk in septic shock patients using LASSO-Logistic methods.Methods:A retrospective cohort study was conducted. Based on the open-source database Medical Information Mart for Intensive Care-Ⅲ v1.4 (MIMIC-Ⅲ v1.4), the septic shock patients meeting the Sepsis-3 criteria were included, and the data on demographic characteristics, major signs, laboratory examinations, hospitalization, and outcomes were extracted. Predictive variables were selected by LASSO regression and predictive models were derived using Logistic regression. The calibration of the model was evaluated using the Hosmer-Lemeshow test and discrimination was evaluated using the receiver operator characteristic curve (ROC curve).Results:A total of 693 patients with septic shock were enrolled, in which 445 patients survived and 248 patients dead within 30 days and the mortality was 35.8%. Logistic regression model was constructed according to nine predictive variables and outcome variables screened by LASSO regression method, which showed that advanced age, Elixhauser index, blood lactic acid (Lac), K + level and mechanical ventilation were associated with increased 30-day mortality [odds ratio ( OR) and 95% confidence interval (95% CI) was 1.023 (1.010-1.037), 1.047 (1.022-1.074), 1.213 (1.133-1.305), 2.241 (1.664-3.057), 2.165 (1.433-3.301), respectively, all P < 0.01], and reduced systolic blood pressure (SBP), diastolic blood pressure (DBP), body temperature, and pulse oxygen saturation (SpO 2) were also associated with increased 30-day mortality [ OR (95% CI) was 0.974 (0.957-0.990), 0.972 (0.950-0.994), 0.693 (0.556-0.857), 0.971 (0.949-0.992), respectively, all P < 0.05]. The calibration curve showed that the predicted risk of septic shock death risk prediction model had good agreement with the real situation. ROC curve analysis showed that the area under the ROC curve (AUC) of the prediction model was 0.839 (95% CI was 0.803-0.876), which could distinguish patients at risk of death from those at risk of survival. Conclusions:The septic shock death risk prediction model has a good ability to identify the 30-day mortality risk of septic shock patients, including nine hospital readily variables (age, Elixhauser index, mechanical ventilation, Lac, K +, SBP, DBP, body temperature and SpO 2). The model could be used by clinicians to calculate the risk of death in septic shock individuals.