Machine Learning–Driven Prediction of Early Enteral Nutrition Intolerance in ICU Patients Using Intra-Abdominal Pressure and Selected Gut Microbial Taxa

Abstract

To develop and validate a machine learning model integrating intra-abdominal pressure (IAP) and gut microbial characteristics for early identification of enteral nutrition intolerance (EENI) in critically ill ICU patients. This cohort study (January 2023–December 2025) included 300 ICU patients receiving early enteral nutrition. Baseline clinical characteristics, intra-abdominal pressure, laboratory indices, and quantitative gut microbial taxa were collected. Candidate predictors were selected using the least absolute shrinkage and selection operator regression, and independent predictors were incorporated into a multivariable logistic regression model, which was presented as a nomogram. Model performance was assessed using discrimination, calibration, and decision curve analyses. The incidence of EENI was 49.00% (147/300). LASSO selected 16 features: age, analgesic use, serum albumin, glucose, IAP, and the absolute counts of Enterococcus, Bacteroides, Escherichia-Shigella, Klebsiella, Bifidobacterium , and Parabacteroides . The absolute counts of Enterococcus, Bacteroides, Escherichia-Shigella, Klebsiella, Bifidobacterium , and Parabacteroides were significant independent predictors (P < 0.05). The model achieved an AUC of 0.900 (95% CI: 0.863–0.936) in the training set and 0.900 (95% CI: 0.859–0.941) in the validation set. Calibration was good (Hosmer-Lemeshow P = 0.425 and P = 0.423, respectively). DCA demonstrated clinical utility across a wide range of risk thresholds. The developed machine learning model, combining IAP and selected genus-level gut microbial markers, demonstrates strong predictive performance and clinical potential for forecasting and managing early enteral nutrition intolerance in ICU patients.