Development and validation of a risk prediction model for perioperative acute kidney injury in non-cardiac and non-urological surgery patients: a retrospective cohort study.

BACKGROUND

This study presents a predictive model designed to fill the gap in tools for predicting perioperative acute kidney injury (AKI) in patients undergoing non-cardiac, non-urological surgeries, with the goal of improving clinical decision-making and patient outcomes.

METHODS

A retrospective cohort of 40,520 patients aged 65 and older who underwent non-cardiac, non-urological surgeries was analyzed. Key risk factors were identified using univariable logistic regression and LASSO, while multivariate logistic regression was applied to develop and validate the model.

RESULTS

The prediction model, based on 18 key variables including demographic data, comorbidities, and intraoperative factors, demonstrated strong discriminatory power for predicting perioperative AKI (AUC = 0.803; 95% CI, 0.783-0.823). It also showed a good fit in the validation cohort (Hosmer-Lemeshow test, χ = 5.895, = 0.750). Decision curve analysis further confirmed the model's significant clinical utility.

CONCLUSION

This model effectively predicts perioperative AKI, providing a valuable tool for personalized risk assessment and prevention strategies in non-cardiac, non-urological surgeries. Further validation in diverse populations is recommended to optimize its clinical application.