Prolonged ICU Stay After Coronary Revascularization: Development and Validation of a Predictive Model

Cardiac artery bypass grafting (CABG) is a common cardiac procedure with standardized postoperative intensive care unit (ICU) management. However, prolonged ICU stays increase complications, costs, and reduce service efficiency. We developed a prediction model for prolonged ICU stays in CABG patients based on a retrospective cohort. Prolonged ICU stay was defined as a stay of ≥ 7 days. Demographic data, laboratory results, and complications during the first 48 hours of ICU admission were analyzed. For the prediction model, variables with p  < 0.25 in the univariate analysis were selected, followed by backward stepwise selection (significance level = 0.1). Interaction terms and fractional polynomials were evaluated, and bootstrapping was applied for external validation. Model performance was assessed using logistic regression diagnostics, ROC curve analysis (AUC), and calibration. The adjusted model equation was g(x) = age (0.0357) + left ventricular ejection fraction (LVEF) (-0.040) + perfusion time (0.0082) + hypothyroidism (0.754) + postoperative renal failure (1.653) + reoperation due to bleeding (1.885) + postoperative cerebrovascular event (1.287) + postoperative acute myocardial infarction (1.163) and the probability of in-hospital death was p (x) = e ^{g (x)} /(1 + e ^{g (x)} ). The AUC of the ROC curve was 0.7671 (95% CI 0.7161 and 0.8066).