Development and validation of a nomogram for predicting acute kidney injury risks in patients undergoing acute stanford type A aortic dissection repair surgery

Abstract

Background: This study aims to construct and internally validate a comprehensive nomogram designed for accurately predicting the incidence of acute kidney injury (AKI) in patients undergoing repair surgery for acute Stanford Type A aortic dissection (ATAAD), thereby enhancing postoperative risk management and patient care strategies.

Methods: A retrospective analysis of 1471 consecutive patients diagnosed with ATAAD through computed tomography angiography (CTA) and confirmed by surgery at four tertiary medical centers from February 2010 to July 2023 was conducted. The study involved a comprehensive evaluation of 36 variables, categorizing patients into non-AKI and AKI groups. Advanced statistical techniques, including LASSO regression and Logistic regression, were employed. A sophisticated nomogram prediction model was developed using R language, and its efficacy was assessed using the concordance index (C-index), area under the receiver operating characteristic curve (AUC-ROC), and decision curve analysis.

Results: Seven key factors independently predicting AKI were identified, including heart failure (a condition where the heart can't pump blood as well), hyperlipidemia (high levels of fats in the blood), arterial dissection (a serious condition where there is a tear in the wall of a blood vessel), renal insufficiency, blood urea nitrogen (BUN), abnormal electrocardiogram (ECG), and total cholesterol (TC). The AUC-ROC, a measure of the model's ability to distinguish between classes, was 0.850 (95% CI: 0.823-0.877) for the training set, with high sensitivity (76%) and specificity (99%). For the validation set, the AUC-ROC was 0.840 (95% CI: 0.798-0.833), with sensitivity and specificity of 78% and 94%, respectively. The nomogram demonstrated a recalibrated C-index of 0.854 for the training set and 0.752 for the validation set. Decision curve analysis revealed the nomogram's significant net benefit across various clinical threshold probabilities.

Conclusion: The AKI nomogram exhibits robust predictive capabilities, establishing itself as a crucial clinical tool for the early identification of patients at risk for AKI following ATAAD repair surgery. By delivering personalized risk assessments, this nomogram not only optimizes postoperative management strategies but also plays a vital role in enhancing patient outcomes through timely and proactive interventions.

Keywords: Acute kidney injury; Acute stanford type A aortic dissection; Nomogram; Predictive model; Retrospective analysis.

Fig. 1

Flow diagram showing the method for identifying patients undergoing ATAAD repair surgery

Fig. 2

Propensity score distributions before and after matching. Density curves illustrate the overlap of propensity scores between the AKI (red) and non-AKI (blue) groups. Dashed lines represent post-matching distributions, demonstrating improved balance after propensity score matching

Fig. 3

Love plot visualizing covariate balance improvement through propensity score matching. Standardized mean differences (SMDs) for all covariates are plotted before (green) and after (purple) matching. The red dashed line marks the SMD threshold of 0.1, indicating adequate balance after propensity score matching

Fig. 4

This figure delineates the LASSO coefficient trajectories for 36 features in predicting postoperative AKI, showcasing the shrinkage effect as lambda varies (A). The accompanying graph illustrates LASSO regression’s cross-validation outcomes, highlighting optimal model performance at λmin and a sparser, yet effective model at λ1SE, with variable counts annotated (B)

Fig. 5

This nomogram, designed for early detection of postoperative AKI, is based on multivariable logistic regression and key predictors identified via LASSO. It assigns scores to each predictor, summing them to estimate AKI risk, with a probability scale for clinical application in high-risk patient assessment

Fig. 6

ROC Curves of the AKI Prediction Model in Training and Validation Cohorts”. The receiver operating characteristic (ROC) curves demonstrate the discriminative performance of the model, with an area under the curve (AUC) of 0.850 (95% CI: 0.823–0.877) for the training set and 0.840 (95% CI: 0.798–0.883) for the validation set

Fig. 7

Calibration plots for the postoperative AKI model using training (A) and testing (B) sets compare nomogram predictions with actual frequencies, where alignment with the 45-degree line indicates accurate predictions and model consistency

Fig. 8

Decision Curve Analysis for Clinical Utility of the Hypoglycemia Prediction Model”. Standardized net benefit curves compare the model’s clinical utility across risk thresholds in both training and validation cohorts. The dashed lines represent strategies of “treat all” and “treat none,” while the colored curves reflect net benefits under varying cost-benefit ratios