Development and external validation of an acute kidney injury risk model in older ICU patients with aspiration pneumonia

Abstract Acute kidney injury (AKI) is a major complication in older patients with aspiration pneumonia. This retrospective multicohort study developed and externally validated an interpretable AKI risk assessment model for patients aged ≥ 65 years admitted to intensive care units (ICUs). The study included 3,125 patients from MIMIC-IV for model development and internal validation, and 1,928 from MIMIC-III, 1,551 from eICU, and 500 from Guangxi for external validation. AKI was defined using serum creatinine-based Kidney Disease: Improving Global Outcomes criteria. Predictors were collected within the first 24 hours of ICU admission. Three feature-selection methods identified 13 predictors for seven models. Discrimination, calibration, potential clinical utility, and interpretability were evaluated. Logistic regression achieved an area under the receiver operating characteristic curve of 0.817 (95% confidence interval, 0.790–0.844) internally, and 0.807 (0.784–0.830), 0.793 (0.772–0.813), and 0.772 (0.728–0.815) in eICU, MIMIC-III, and Guangxi, respectively. Shapley additive explanations identified blood urea nitrogen, phosphate, chronic kidney disease, lactate, and bicarbonate as the principal contributors to model predictions. The model demonstrated consistent discrimination across cohorts and may support early AKI risk assessment in older ICU patients with aspiration pneumonia, although prospective validation is required before clinical implementation.