Objective  To construct a predictive model for early death of multiple organ dysfunction syndrome in the elderly (MODSE) based on integrated learning model XGBoost, so as to provide evidence for clinical decision-making and treatment.

  Methods  Using the openly available electronic medical record-based database Medical Information Mart for Intensive Care (MIMIC)-Ⅲ, 14 329 patients with MODSE were included, of which 2 341 (16.3%) died in the hospital. Eighty percent of the sample were randomly selected as the training set, and the remaining 20% were the testing set. According to the prognosis, the patients were divided into death group (n=1 864) and survival group (n=9 599). Demographic characteristics, vital signs on the first day admitted to ICU, clinical intervention measures as model parameters were collected, and differences between the two groups were compared. XGBoost model algorithm was used to carry out model training and investigation of the distribution and ranking of death-related features, and the predictive value of the model for the risk of death in patients with MODSE was also evaluated using receiver operating characteristic (ROC) curve.

  Results  Compared with the survival group, patients in the death group had significantly higher mean Glasgow coma score (GCS), mean age, maximum heart rate, but lower BMI and systolic blood pressure (all P < 0.01). The ten most important variables to predict death risk in the XGBoost model were respiratory rate, APTT, age, body temperature, BMI, systolic blood pressure, platelet, blood glucose, shock index, and white blood cell count. In the prediction of death in MODSE patients, the XGBoost model had a sensitivity of 0.824, specificity of 0.725, and accuracy of 0.854 with AUC of 0.853.

  Conclusion  XGBoost model is a better predictive model with high accuracy, which can provide assistance in clinical decision-making and treatment of MODSE and thus improving the prognosis of elderly patients in the ICU.