28-day all-cause mortality in patients with alcoholic cirrhosis: a machine learning prediction model based on the MIMIC-IV

To develop and validate a machine learning prediction model for 28-day all-cause mortality in patients with alcoholic cirrhosis using data from the MIMIC-IV database. The data of 2134 patients diagnosed with alcoholic cirrhosis (AC) were obtained from Medical Information Mart for Intensive Care IV database. Machine learning algorithms, including decision trees, random forests, extreme gradient boosting, Logistic Regression and support vector machines were employed to develop the prediction model. The model was trained on 70% of the data and validated on the remaining 30% randomly. Performance was assessed using the area under the receiver operating characteristic curve, calibration curves and decision curve analysis (DCA). SHAP analysis was used to assess the marginal effects of each independent variable. The mean age was 56.2 years, and 69.5% were male. The primary factors associated with 28-day mortality included Age, SOFA score, ASPIII score, OASIS score, LODS score, Temperature, Chloride, Lactate, Total bilirubin (Tbil), international normalized ratio (INR), Activated partial thromboplastin time (Aptt), Stroke, Malignancy, Congenital coagulation defect (Ccd). The machine learning model demonstrated good predictive performance in the training and validation group, higher than traditional MELD score. Our machine learning prediction model effectively identifies patients with alcoholic cirrhosis at high risk of 28-day mortality. This model could assist clinicians in early risk stratification and guide clinical decision-making. Further validation in external cohorts is warranted to confirm its generalizability.