Systematic Review and Meta-Analysis of Machine Learning Models for Acute Kidney Injury Risk Classification
Systematische Übersicht und Metaanalyse von Machine-Learning-Modellen zur Risikoklassifikation einer akuten Nierenschädigung
Journal of the American Society of Nephrology · 2025-10 · 36 (10) · 1969-1983
DOI: 10.1681/asn.0000000702 ↗Auf einen Blick
Die gepoolte Diskriminierung war für alle AKI-Risikokategorien in interner und externer Validierung akzeptabel. Die Übertragbarkeit in die klinische Praxis ist wegen starker Heterogenität eingeschränkt.
Fragestellung
Wie gut klassifizieren extern validierte Machine-Learning-Modelle das Risiko für eine akute Nierenschädigung (AKI), die AKI-Schwere und post-AKI-Komplikationen bei hospitalisierten Erwachsenen und Kindern?
Methode
Systematische Übersichtsarbeit mit Metaanalyse; Suche in PubMed, EMBASE, Web of Science und Scopus bis August 2023; gepoolte AUCs mit Random-Effects-Modell.
Ergebnisse
95 eingeschlossene Studien mit 3,8 Millionen Aufnahmen und 302 Modellen. Für AKI-Beginn betrugen die gepoolten AUCs 0,82 (95 %-KI 0,80–0,84) intern und 0,78 (95 %-KI 0,76–0,80) extern. Über alle Settings lagen sie bei 0,78 bis 0,87 intern und 0,73 bis 0,84 extern.
Einschränkungen
Hohe Heterogenität bei Patient:innen, Definitionen, klinischen Prädiktoren und Methoden; 86 % der Studien mit hohem Verzerrungsrisiko.
Schlussfolgerung der Autor:innen
Die meisten extern validierten Modelle sagten AKI-Beginn, AKI-Schwere und post-AKI-Komplikationen bei hospitalisierten Erwachsenen und Kindern gut voraus. Die Heterogenität begrenzt jedoch Generalisierbarkeit und Umsetzung am Bett.
Kernaussagen
- Pooled AUCs >0,70 für alle AKI-Risikokategorien
- AKI-Beginn: 0,82 intern, 0,78 extern
- Hohe Heterogenität und 86 % hohes Bias-Risiko
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Systematische Übersicht in der Zeitschrift Journal of the American Society of Nephrology, online erschienen am 28.03.2025. Verfasst von Augusto Cama-Olivares und 12 weiteren Autor:innen. Beteiligte Einrichtung: Division of Nephrology, Department of Internal Medicine, University of Alabama at Birmingham, Birmingham, Alabama. Die Arbeit wurde bisher 28-mal zitiert (Crossref) und nennt 154 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: NIDDK NIH HHS, NIDDK.
Abstract (Originalsprache)
Key Points Pooled discrimination metrics were acceptable (area under the receiver operating characteristic curve >0.70) for all AKI risk classification categories in both internal and external validation. Better performance was observed in most recently published studies and those with a low or unclear risk of bias. Significant heterogeneity in patient populations, definitions, clinical predictors, and methods limit implementation in real-world clinical scenarios.
Background
Artificial intelligence through machine learning models seems to provide accurate and precise AKI risk classification in some clinical settings, but their performance and implementation in real-world settings has not been established.
Methods
PubMed, Excerpta Medica (EMBASE) database, Web of Science, and Scopus were searched until August 2023. Articles reporting on externally validated models for prediction of AKI onset, AKI severity, and post-AKI complications in hospitalized adult and pediatric patients were searched using text words related to AKI, artificial intelligence, and machine learning. Two independent reviewers screened article titles, abstracts, and full texts. Areas under the receiver operating characteristic curves (AUCs) were used to compare model discrimination and pooled using a random-effects model.
Results
Of the 4816 articles initially identified and screened, 95 were included, representing 3.8 million admissions. The Kidney Disease Improving Global Outcomes (KDIGO)-AKI criteria were most frequently used to define AKI (72%). We identified 302 models, with the most common being logistic regression (37%), neural networks (10%), random forest (9%), and eXtreme gradient boosting (9%). The most frequently reported predictors of hospitalized incident AKI were age, sex, diabetes, serum creatinine, and hemoglobin. The pooled AUCs for AKI onset were 0.82 (95% confidence interval, 0.80 to 0.84) and 0.78 (95% confidence interval, 0.76 to 0.80) for internal and external validation, respectively. Pooled AUCs across multiple clinical settings, AKI severities, and post-AKI complications ranged from 0.78 to 0.87 for internal validation and 0.73 to 0.84 for external validation. Although data were limited, results in the pediatric population aligned with those observed in adults. Between-study heterogeneity was high for all outcomes (I 2 >90%), and most studies presented high risk of bias (86%) according to the Prediction Model Risk of Bias Assessment Tool.
Conclusions
Most externally validated models performed well in predicting AKI onset, AKI severity, and post-AKI complications in hospitalized adult and pediatric populations. However, heterogeneity in clinical settings, study populations, and predictors limits their generalizability and implementation at the bedside.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Division of Nephrology, Department of Internal Medicine, University of Alabama at Birmingham, Birmingham, Alabama.
- Division of Critical Care, Department of Pediatrics, University of Alabama at Birmingham, Birmingham, Alabama.
- Department of Health Policy and Informatics, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University, Tokyo, Japan.
- UAB Libraries University of Alabama at Birmingham, Birmingham, Alabama.
- Department of Health Behavior, School of Public Health, University of Alabama at Birmingham, Birmingham, Alabama.
- Department of Epidemiology, School of Public Health, University of Alabama at Birmingham, Birmingham, Alabama.
ORCID-Profile: Augusto Cama-Olivares, Chloe Braun, Tomonori Takeuchi, Emma C O'Hagan, Kathryn A Kaiser, Lama Ghazi, Jin Chen, Marlies Ostermann, Benjamin Shickel, Jacob Ninan, Javier A Neyra
Förderung
- NIDDK NIH HHS
- NIDDK
Zitieren
APA
Cama-Olivares, A., Braun, C., Takeuchi, T., O'Hagan, E. C., Kaiser, K. A., Ghazi, L., Chen, J., Forni, L. G., Kane-Gill, S. L., Ostermann, M., Shickel, B., Ninan, J., & Neyra, J. A. (2025). Systematic Review and Meta-Analysis of Machine Learning Models for Acute Kidney Injury Risk Classification. Journal of the American Society of Nephrology, 36(10), 1969-1983. https://doi.org/10.1681/asn.0000000702
Vancouver
Cama-Olivares A, Braun C, Takeuchi T, O'Hagan EC, Kaiser KA, Ghazi L, et al. Systematic Review and Meta-Analysis of Machine Learning Models for Acute Kidney Injury Risk Classification. Journal of the American Society of Nephrology. 2025;36(10):1969-1983. doi:10.1681/asn.0000000702
BibTeX
@article{CamaOlivares2025,
author = {Cama-Olivares, Augusto and Braun, Chloe and Takeuchi, Tomonori and O'Hagan, Emma C and Kaiser, Kathryn A and Ghazi, Lama and Chen, Jin and Forni, Lui G and Kane-Gill, Sandra L and Ostermann, Marlies and Shickel, Benjamin and Ninan, Jacob and Neyra, Javier A},
title = {Systematic Review and Meta-Analysis of Machine Learning Models for Acute Kidney Injury Risk Classification},
journal = {Journal of the American Society of Nephrology},
year = {2025},
volume = {36},
number = {10},
pages = {1969--1983},
doi = {10.1681/asn.0000000702}
}
EUVIMED