Machine learning models in predicting acute kidney injury following acute aortic dissection surgery: Meta-analysis of diagnostic accuracy
Main Article Content
Keywords
acute kidney injury, aortic dissection, artificial intelligence, machine learning, surgery
Abstract
Introduction: Acute kidney injury (AKI) continues to represent a major complication following surgical repair of acute aortic dissection (AAD). Recent advancements in artificial intelligence (AI) demonstrate significant potential by leveraging extensive electronic health record data to forecast AKI. This meta-analysis seeks to elucidate the diagnostic accuracy and comparative performance of various machine learning (ML) models in predicting AKI occurrence in patients with AAD repair.
Methods: A comprehensive systematic literature search was conducted across PubMed, EuropePMC, and EBSCO for studies evaluating ML algorithms (including Neural Network [NN], Gradient Boosting Machine [GBM], K-Nearest Neighbors [KNN], LightGBM [LGB], Logistic Regression [LR], Random Forest [RF], Support Vector Machine [SVM], and eXtreme Gradient Boosting [XGB]) in predicting post-AAD repair AKI. Study quality and risk of bias were systematically evaluated using the Quality Assessment of Prognostic Accuracy Studies (QUAPAS) tool. Pooled sensitivity, specificity, diagnostic odds ratios (DOR), and summary receiver operating characteristic (SROC) curves with corresponding area under the curve (AUC) were calculated using a bivariate random-effects model. Heterogeneity was evaluated using I2 statistics.
Results: Six studies comprising 1,713 participants and evaluating 22 distinct ML models met the inclusion criteria. Predictors evaluated to predict AKI comprised demographic parameters; pre-existing diseases; baseline, intraoperative, and postoperative hemodynamic parameters; laboratory findings; intraoperative or postoperative events; medication usage; and postoperative management. Meta-analysis revealed acceptable discrimination power for all models (sensitivity=0.655, specificity=0.685, AUC=0.709). Subgroup analysis revealed that NN showed the best discrimination power (sensitivity=0.771, specificity=0.882, AUC=0.907). Decent discrimination power was demonstrated by KNN (sensitivity=0.563, specificity=0.766, AUC=0.748), LGB (sensitivity=0.667, specificity=0.788, AUC=0.793), LR (sensitivity=0.755, specificity=0.660, AUC=0.723), and RF (sensitivity=0.750, specificity=0.499, AUC=0.737).
Conclusion: ML models exhibit acceptable overall diagnostic capability for predicting post-AAD repair AKI, with NN demonstrating superior discrimination over other architectures. External validation in broader international cohorts is warranted.
References
2. Chen Z, Lu X, Liu M, Yu Y, Yu L, Cheng S, et al. Artificial Intelligence–Driven Prediction of Acute Kidney Injury Following Acute Type A Aortic Dissection Surgery in a Chinese Population. J Cardiothorac Vasc Anesth. 2025;39(10):2729–38. Available from: http://dx.doi.org/10.1053/j.jvca.2025.05.047
3. Du R, Wang L, Wang Y, Zhao Z, Zhang D, Zuo S. AKI prediction model in acute aortic dissection surgery: nomogram development and validation. Front Med. 2025;May 15(12):1562956. Available from: http://dx.doi.org/10.3389/fmed.2025.1562956
4. Li W, Yu W, Chen Y, Tan W, Zhang F, Zhang Y. Development and validation of a nomogram for predicting acute kidney injury risks in patients undergoing acute Stanford type A aortic dissection repair surgery. BMC Nephrol. 2025;26(1):257. Available from: http://dx.doi.org/10.1186/s12882-025-04150-y
5. Liu X, Fang M, Wang K, Zhu J, Chen Z, He L, et al. Machine learning-based model to predict severe acute kidney injury after total aortic arch replacement for acute type A aortic dissection. Heliyon. 2024;10(13):e34171. Available from: http://dx.doi.org/10.1016/j.heliyon.2024.e34171
6. Dai A, Zhou Z, Jiang F, Guo Y, Asante DO, Feng Y, et al. Incorporating intraoperative blood pressure time-series variables to assist in prediction of acute kidney injury after type A acute aortic dissection repair: an interpretable machine learning model. Ann Med. 2023;55(2):2266458. Available from: http://dx.doi.org/10.1080/07853890.2023.2266458
7. Song X, Liu X, Liu F, Wang C. Comparison of machine learning and logistic regression models in predicting acute kidney injury: A systematic review and meta-analysis. Int J Med Inform. 2021;151:104484. Available from: http://dx.doi.org/10.1016/j.ijmedinf.2021.104484
8. Goyal A, Daneshpajouhnejad P, Bashir K. Acute Kidney Injury. In Treasure Island (FL); 2026. Available from: https://pubmed.ncbi.nlm.nih.gov/28722925/
9. Wang Z, Xu J, Zhang Y, Chen C, Kong C, Tang L, et al. Prediction of acute kidney injury incidence following acute type A aortic dissection surgery with novel biomarkers: a prospective observational study. BMC Med. 2023;21(1):503. Available from: http://dx.doi.org/10.1186/s12916-023-03215-9
10. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. Updating guidance for reporting systematic reviews: development of the PRISMA 2020 statement. J Clin Epidemiol. 2021;134:103–12. Available from: http://dx.doi.org/10.1016/j.jclinepi.2021.02.003
11. Lameire N, Kellum JA, Aspelin P. Kidney disease: improving global outcomes (KDIGO) acute kidney injury work group. KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl. 2012;2:1–138. Available from: http://dx.doi.org/10.1038/kisup.2012.1
12. Lee J, Mulder F, Leeflang M, Wolff R, Whiting P, Bossuyt PM. QUAPAS: An Adaptation of the QUADAS-2 Tool to Assess Prognostic Accuracy Studies. Ann Intern Med. 2022;175(7):1010–8. Available from: http://dx.doi.org/10.7326/m22-0276
13. Zhou Y, Dendukuri N. Statistics for quantifying heterogeneity in univariate and bivariate meta-analyses of binary data: The case of meta-analyses of diagnostic accuracy. Stat Med. 2014;33(16):2701–17. Available from: http://dx.doi.org/10.1002/sim.6115
