Development with external validation of a prediction model for postoperative acute kidney injury following noncardiac surgery in elderly patients

Study objectiveTo develop and externally validate a risk prediction model for postoperative acute kidney injury (PO-AKI) in elderly patients undergoing noncardiac surgery, addressing the current gap in predictive tools for this vulnerable population.DesignA multicenter retrospective cohort study presented according to TRIPOD + AI statement.SettingConducted in 21 tertiary hospitals across 11 provinces in China from January 2009 to April 2022.PatientsElderly patients (>= 65 years) undergoing noncardiac procedures.Interventions and measurementsThe endpoint was PO-AKI within seven days post-surgery, diagnosed using the KDIGO criteria. Data were extracted from electronic medical records for model derivation and validation.Main resultsThe study included 163,131 elderly patients, with 52,494 for model discovery, 7,899 and 80,641 for external validation. The model incorporated nine variables: age, heart disease history, preoperative hyponatremia, renal surgery (yes/no), surgery type, surgery duration, intraoperative diuretics usage, first-aid vasopressors usage, and blood transfusion. The model demonstrated acceptable discriminative ability with AUROC values of 0.803, 0.793, 0.770, and 0.774 across the training, internal validation, and two external validation datasets, respectively. The calibration plots and decision curve analyses yielded commendable results in both training and validation sets. To streamline usability, we employed risk scores and categorized the population into low-, medium-, and high-risk subgroups.ConclusionsClinicians could implement this externally validated risk prediction model to stratify PO-AKI risks in elderly patients during the early postoperative phases of noncardiac surgery.