Development and Validation of a Web-Based Prediction Model for AKI after Surgery

Area of Science:

• Nephrology
• Surgical Outcomes
• Health Informatics

Background:

• Postoperative acute kidney injury (AKI) is a significant complication associated with increased mortality and morbidity.
• Identifying patients at risk for AKI requiring renal replacement therapy (AKI-dialysis) is crucial for timely intervention.
• Existing prediction tools may not adequately capture the complexity of AKI development in surgical populations.

Purpose of the Study:

• To develop and validate a predictive model for postoperative acute kidney injury requiring renal replacement therapy (AKI-dialysis).
• To create a clinically applicable bedside tool for risk stratification of surgical patients.
• To improve the management and outcomes of patients susceptible to AKI post-surgery.

Main Methods:

• Retrospective cohort study of 2,299,502 surgical patients from the ACS NSQIP Database (2015-2017).
• Multivariable logistic regression used to develop a risk prediction model based on eleven key predictors.
• Model trained on 2015-2016 data (n=1,487,724) and validated on 2017 data (n=811,778).

Main Results:

• AKI-dialysis occurred in 0.3% of patients (n=6853), with a 30-day postoperative mortality rate of 37.5%.
• The developed risk prediction model demonstrated high predictive accuracy (AUC: 0.89 training, 0.90 test cohort).
• Key predictors included age, comorbidities (CHF, diabetes, ascites), emergency surgery, hypertension, and preoperative lab values.

Conclusions:

• A robust and validated risk prediction model for postoperative AKI-dialysis has been developed.
• The model offers a clinically useful bedside tool for identifying surgical patients at high risk of AKI requiring dialysis.
• This tool can aid in proactive patient management and potentially reduce AKI-related morbidity and mortality.