Development and validation of machine learning models for predicting short- and long-term mortality in gastroparesis patients: a retrospective cohort study using the MIMIC-IV database

Background: Gastroparesis is a debilitating disorder characterized by delayed gastric emptying without mechanical obstruction, carrying significant mortality risks. Accurate prediction of mortality remains challenging due to the condition's heterogeneous nature. This study aimed to develop and validate machine learning models for predicting both short-term and long-term mortality in patients with gastroparesis. Methods: This retrospective cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Adult patients diagnosed with gastroparesis between 2003 and 2019 were included. Comprehensive clinical data encompassing demographics, laboratory parameters, comorbidities, and severity scores were extracted. Seven machine learning algorithms were employed to predict 30-day and 365-day mortality: logistic regression, support vector machine, random forest, k-nearest neighbors, decision tree, artificial neural network, and gradient boosting machine. Model performance was evaluated using area under the receiver operating characteristic curve (AUC-ROC), accuracy, precision, recall, and F1-score. Results: Among 6,125 eligible patients with gastroparesis, the random forest model demonstrated superior performance for both 30-day (AUC: 0.39) and 365-day mortality prediction (AUC: 0.905). Key predictors included age, red cell distribution width (RDW), Charlson comorbidity index, white blood cell count, and renal function markers. Significant mortality associations were identified for ICU admission, sepsis, acute kidney injury, and liver diseases. Conclusions: Machine learning algorithms, particularly random forest, effectively predict mortality risk in gastroparesis patients using routinely available clinical data. These models provide valuable tools for risk stratification and informed clinical decision-making, potentially improving patient outcomes through early identification of high-risk individuals.