Abstract
OBJECTIVES: Early prediction is crucial for cardiac surgery-associated acute kidney injury (CSA-AKI). We aimed to develop and validate a simple, clinical- and laboratory-based risk score system for better CSA-AKI prediction. METHODS: We developed a new pre-operative risk score system for moderate-to-severe CSA-AKI in a 10-year cohort of patients undergoing coronary artery bypass grafting at one tertiary centre. Most predictive laboratory and clinical variables were identified and constituted a simple and a full model. External testing was performed in patients at another centre. The risk score system was compared with 2 established clinical models. RESULTS: The overall cohort comprised 27 534, 6403, and 1733 patients with moderate-to-severe CSA-AKI rates of 3.3%, 2.8%, and 8.4% for training, validation, and external testing, respectively. A simple 6-variable AB2C-S2 score (Age, Biomarkers of N-terminal pro-B-type natriuretic peptide and haemoglobin, Clinical history of preoperative critical state, Surgical factors of isolated surgery and on-pump surgery) and a full 9-variable AB2C2-S4 score (AB2C-S2 score plus hypertension, urgent surgery, and previous surgery) were developed. The simple model achieved similar performance as the full model in validation (area under the receiver-operating characteristic curve [AUC] 0.78 vs 0.79, P = .37) and external testing (AUC 0.74 vs 0.75, P = .17), and both significantly outperformed than 2 established clinical models: Cleveland Clinic model (validation: AUC 0.71, external testing: AUC 0.65, all P < .001) and Ng model (validation: AUC 0.64, external testing: AUC 0.65, all P < .001). CONCLUSIONS: A simple preoperative risk score system for moderate-to-severe CSA-AKI was developed and outperformed established complex clinical models.