Beyond the end of surgery: early postoperative laboratory trends improve prediction of acute kidney injury after noncardiac surgery

Acute kidney injury (AKI) is a common and consequential complication of noncardiac surgery, but most perioperative risk models are static, combining preoperative variables into a single estimate that does not update as intraoperative and postoperative data accumulate. We examined the association between perioperative hemodynamic and laboratory variables and postoperative AKI, and tested whether adding early postoperative laboratory trends improves risk discrimination beyond a preoperative/intraoperative model alone. We performed a retrospective cohort analysis of the publicly available VitalDB dataset, comprising perioperative clinical, intraoperative, and time-stamped laboratory data from 6,388 patients undergoing noncardiac surgery at a single tertiary center. AKI was defined by KDIGO serum creatinine criteria. A multivariable logistic regression model examined preoperative, intraoperative, and early postoperative (day-1) predictors of AKI. Discrimination of a preoperative–intraoperative model (variables available by the end of surgery) was compared with a perioperative model additionally incorporating day-1 maximum lactate, day-1 hemoglobin, and day-1 maximum potassium, using 5-fold cross-validated area under the receiver operating characteristic curve (AUC) and bootstrap resampling. Of 3,834 analyzable cases, 256 (6.7%) developed AKI. Independent predictors included ASA class ≥ III (OR 1.91), male sex (OR 1.42), longer surgery duration (OR 1.22 per hour), day-1 maximum lactate (OR 1.13 per mmol/L), and day-1 maximum potassium (OR 2.83 per mmol/L), while older age, higher preoperative hemoglobin, and higher preoperative albumin were associated with lower odds. The preoperative–intraoperative model discriminated AKI with an AUC of 0.75 (95% CI 0.72–0.78); adding early postoperative laboratory trends (perioperative model) improved this to 0.77 (95% CI 0.74–0.80; bootstrap difference 0.023, 95% CI 0.008–0.041, p = 0.002). Postoperative AKI after noncardiac surgery was associated with recognized preoperative and intraoperative risk factors, and incorporating early postoperative laboratory trends produced a modest but statistically significant improvement in discrimination. These findings support further development of dynamic, updatable AKI risk models that evolve across the perioperative period rather than relying on a single preoperative estimate.