Introduction This study aimed to develop and validate an interpretable machine-learning model based on clinical features for early predicting in-hospital mortality in critically ill patients with sepsis. Methods We enrolled all patients with sepsis in the Medical Information Mart for Intensive Care IV (MIMIC-IV, v.1.0) database from 2008 to 2019. Lasso regression was used for feature selection. Seven machine-learning methods were applied to develop the models. The best model was selected based on its accuracy and area under curve (AUC) in the validation cohort. Furthermore, we employed the SHapley Additive exPlanations (SHAP) method to illustrate the effects of the features attributed to the model, and to analyze how the individual features affect the output of the model, and to visualize the Shapley value for a single individual. Results In total, 8,817 patients with sepsis were eligible for participation, the median age was 66.8 years (IQR, 55.9–77.1 years), and 3361 of 8817 participants (38.1%) were women. After selection, 25 of a total 57 clinical parameters collected on day 1 after ICU admission remained associated with prognosis and were used for developing the machine-learning models. Among seven constructed models, the eXtreme Gradient Boosting (XGBoost) model achieved the best performance with an AUC of 0.884 and an accuracy of 89.5% in the validation cohort. Feature importance analysis showed that Glasgow Coma Scale (GCS) score, blood urea nitrogen, respiratory rate, urine output, and age were the top 5 features of the XGBoost model with the greatest impact. Furthermore, SHAP force analysis illustrated how the constructed model visualized the individualized prediction of death. Conclusions We have demonstrated the potential of machine-learning approaches for predicting outcome early in patients with sepsis. The SHAP method could improve the interpretability of machine-learning models and help clinicians better understand the reasoning behind the outcome. Graphical abstract

Introduction The association between thiamine use and clinical outcomes among patients with sepsis and alcohol use disorder (AUD) is unclear. Methods In this retrospective cohort study of patients from Medical Information Mart for Intensive Care III (MIMIC-III, version 1.4), we evaluated the association of thiamine use with clinical outcomes in patients with AUD and sepsis. The primary outcome was 28-day survival, and secondary outcomes included ICU, in-hospital, and 90-day mortality, ICU and hospital length of stay, duration of vasopressor use, need and duration of continuous renal replacement therapy (CRRT), and dynamic changes for variables up to day 7 after ICU admission. Results A total of 944 patients with sepsis and AUD were included in this cohort [median age, 53.1 years; women, 26.0% (245 of 944)]. Among all patients, 24.6% (233 of 944) received thiamine with a dose of 200 mg (IQR 100–345 mg). The 28-day mortality was 11.2% (26 of 233) in the thiamine use group compared with 18.6% (132 of 711) in the no thiamine use group (P = 0.009). After adjustment for a series of confounders, the mixed-effects Cox proportional hazards models showed that administration of thiamine was associated with a lower risk of 28-day mortality compared with no administration of thiamine. Conclusions In critically ill patients with alcohol use disorder admitted for sepsis, treatment with thiamine may be associated with a decreased risk of death. However, the present results should be interpreted with caution due to the limitations of retrospective design. Additional larger, multicenter randomized controlled trials are needed to confirm our findings.