Abstract
BACKGROUND: Patients in intensive care units (ICUs) who receive enteral nutrition (EN) treatment frequently experience feeding intolerance (FI), which, if not promptly managed, can adversely affect treatment outcomes and overall prognosis. This study aims to identify the risk factors associated with enteral nutrition feeding intolerance (ENFI) in critically ill ICU patients and to develop a predictive model to assess the risk of ENFI. METHODS: This study enrolled 144 patients, who were categorized into an ENFI group and a non-ENFI group. Variable selection for model development was conducted through univariate analysis, multicollinearity testing, and binary logistic regression. Based on the logistic regression results, a visual predictive model for ENFI risk was constructed using a nomogram. The model's discriminative performance was evaluated using the area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Internal validation was performed using the bootstrap method with 1,000 resamples of the original dataset. A calibration curve was generated, and the Hosmer-Lemeshow goodness-of-fit test was applied to assess the model's calibration accuracy. RESULTS: Based on the results of the binary logistic regression analysis, a nomogram model was developed to predict enteral nutrition feeding intolerance (ENFI) in critically ill ICU patients. The model incorporated five variables: Acute Physiology and Chronic Health Evaluation II (APACHE II) score, mechanical ventilation (MV), albumin (ALB), intra-abdominal pressure (IAP), and EN start time. AUC was 0.800 (95% confidence interval, 0.725-0.875), with a cutoff value of 0.306. The model demonstrated a sensitivity of 82.5%, specificity of 72.4%, positive predictive value (PPV) of 67.2%, and negative predictive value (NPV) of 86.3%. Following internal validation using the bootstrap method, the Hosmer-Lemeshow goodness-of-fit test produced a χ(2) value of 2.9954 (p = 0.9346). The lack of statistically significant deviation between the predicted and observed risk values indicates that the model demonstrates good calibration and accurately reflects the actual risk of ENFI. CONCLUSION: The model demonstrated good predictive performance and can effectively assess the risk of ENFI in critically ill ICU patients.