Abstract
BACKGROUND: Acute kidney injury (AKI) is frequent among intensive care unit (ICU) patients and is linked with high morbidity and mortality. In the absence of specific pharmacological treatments for AKI, continuous renal replacement therapy (CRRT) is a primary treatment option. This study aimed to develop and validate a predictive model for 90-day mortality in critically ill patients with AKI undergoing CRRT. METHODS: Clinical data from DATADRYAD were used. We randomly divided 1121 adult patients receiving CRRT for AKI into training (80%, n = 897) and validation (20%, n = 224) cohorts. A nomogram prediction model was developed using Cox proportional hazards regression with the training set, and was validated internally. Model performance was evaluated based on calibration, discrimination, and clinical utility. RESULTS: The model, incorporating seven predictors-SOFA score, serum creatinine, blood urea nitrogen, albumin levels, Charlson comorbidity index, mean arterial pressure at CRRT initiation, and phosphate levels 24 h after CRRT initiation-demonstrated robust performance. It achieved a C-index of 0.810 in the training set and 0.794 in the validation set. CONCLUSIONS: We developed and validated a predictive model based on seven key clinical predictors, showing excellent performance in identifying high-risk patients for 90-day mortality in AKI patients undergoing CRRT.