Abstract
OBJECTIVES: This study aims to improve early detection of cardiac surgery-associated acute kidney injury (CSA-AKI) compared to classical clinical scores. METHODS: Data from 7633 patients who underwent cardiac surgery between 2008 and 2018 in our institution were analysed. CSA-AKI was defined according to the Kidney Disease Improving Global Outcomes (KDIGO) criteria. Cleveland Clinical Score served as the reference with an area under the curve (AUC) 0.65 in our cohort. Based on that, stepwise logistic regression modelling was performed on the training data set including creatinine (Cr), estimated glomerular filtration rate (eGFR) levels and deltas (ΔCr, ΔeGFR) at different time points and clinical parameters as preoperative haemoglobin, intraoperative packed red blood cells (units) and cardiopulmonary bypass time (min) to predict CSA-AKI in the early postoperative course. The AUC was determined on the validation data set for each model respectively. RESULTS: Incidence of CSA-AKI in the early postoperative course was 22.4% (n = 1712). The 30-day mortality was 12.5% in the CSA-AKI group (n = 214) and in the no-CSA-AKI group 0.9% (n = 53) (P < 0.001). Logistic regression models based on Cr and its delta gained an AUC of 0.69; 'Model eGFRCKD-EPI' an AUC of 0.73. Finally, 'Model DynaLab' including dynamic laboratory parameters and clinical parameters as haemoglobin, packed red blood cells and cardiopulmonary bypass time improved AUC to 0.84. CONCLUSIONS: Model DynaLab' improves early detection of CSA-AKI within 12 h after surgery. This simple Cr-based framework poses a fundament for further endeavours towards reduction of CSA-AKI incidence and severity.