Construction of a machine learning-based risk prediction model for acute stroke in maintenance hemodialysis patients
Entwicklung eines machine-learningbasierten Risikovorhersagemodells für akuten Schlaganfall bei Patient:innen unter Erhaltungshämodialyse
PeerJ · 11.11.2026 · 14 · e21894
DOI: 10.7717/peerj.21894 ↗Auf einen Blick
Bei 1.100 Patient:innen unter Erhaltungshämodialyse sagte ein XGBoost-Modell das Risiko für akuten Schlaganfall im Testdatensatz am besten voraus. Die Güte lag bei einem AUC von 0,847.
Fragestellung
Lässt sich das Risiko für akuten Schlaganfall bei Patient:innen unter Erhaltungshämodialyse mit Machine-Learning-Modellen vorhersagen?
Methode
1.100 MHD-Patient:innen aus einem Krankenhaus; Aufteilung in Trainings- und Testset 7:3; Entwicklung von 7 Machine-Learning-Modellen nach LASSO-Selektion; Bewertung mit AUC, Sensitivität, Spezifität und Brier-Score.
Ergebnisse
XGBoost war im Testset am besten: AUC 0,847 (95 %-KI 0,791–0,902), Sensitivität 0,73, Spezifität 0,83, F1-Score 0,61, Brier-Score 0,115. Wichtigste Prädiktoren: prädialytischer SBP, Albumin, Harnsäure, Hämoglobin, CRP, KHK-Anamnese.
Schlussfolgerung der Autor:innen
Die Autor:innen berichten, dass ein interpretierbares XGBoost-basiertes Modell zur Vorhersage des akuten Schlaganfallrisikos bei MHD-Patient:innen erfolgreich entwickelt wurde.
Kernaussagen
- XGBoost zeigte im Testset die beste Vorhersagegüte
- AUC 0,847 (95 %-KI 0,791–0,902)
- Sechs wichtigste Prädiktoren wurden per SHAP identifiziert
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift PeerJ, online erschienen am 11.11.2026. Verfasst von Jinping Ying und 6 weiteren Autor:innen. Beteiligte Einrichtung: Department of Nursing, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China. Hauptthemen laut MEDLINE: Stroke, Renal Dialysis, Machine Learning. Die Arbeit nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Zhejiang Medical and Health Science and Technology Plan Project, Discipline Construction Project, Zhejiang Provincial Medical Association.
Abstract (Originalsprache)
Background
Acute stroke represents a significant complication in patients undergoing maintenance hemodialysis (MHD) and is associated with increased mortality rates in this population. This study aimed to construct a prediction model for acute stroke in these patients by using seven machine learning algorithms.
Methods
This study included 1,100 patients receiving MHD at the First Affiliated Hospital, Zhejiang University School of Medicine, between January 1, 2015, and December 31, 2024. The patients were divided into training ( n = 770) and test ( n = 330) sets at a 7:3 ratio. Following feature selection using least absolute shrinkage and selection operator (LASSO) regression, predictive models were developed using seven machine learning algorithms. Model performance was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, and the Brier score. SHapley Additive exPlanations (SHAP) were used to quantify the contribution of individual features to model predictions. A web-based prediction tool was subsequently developed using the Shiny framework in R.
Results
The results showed that the extreme gradient boosting (XGBoost) model performed the best in the test set, with an AUC of 0.847 (95% confidence interval (CI) [0.791–0.902]), sensitivity = 0.73, specificity = 0.83, F1-score = 0.61 and a Brier score of 0.115. The SHAP analysis identified six most important predictors, which were ranked in the descending order as follows: pre-dialysis systolic blood pressure (SBP), serum albumin level, uric acid level, hemoglobin level, C-reactive protein level, and a history of coronary heart disease.
Conclusions
We successfully developed an interpretable XGBoost-based model that effectively predicts acute stroke risk in MHD patients.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Nursing, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China.
- Kidney Disease Center, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China.
Förderung
- Zhejiang Medical and Health Science and Technology Plan Project
- Discipline Construction Project
- Zhejiang Provincial Medical Association
Zitieren
APA
Ying, J., Cai, G., Chen, L., Zhang, Y., Zhang, R., Han, F., & Zhang, P. (2026). Construction of a machine learning-based risk prediction model for acute stroke in maintenance hemodialysis patients. PeerJ, 14, e21894. https://doi.org/10.7717/peerj.21894
Vancouver
Ying J, Cai G, Chen L, Zhang Y, Zhang R, Han F, et al. Construction of a machine learning-based risk prediction model for acute stroke in maintenance hemodialysis patients. PeerJ. 2026;14:e21894. doi:10.7717/peerj.21894
BibTeX
@article{Ying2026,
author = {Ying, Jinping and Cai, Genlian and Chen, Linglin and Zhang, Yujiao and Zhang, Rui and Han, Fei and Zhang, Ping},
title = {Construction of a machine learning-based risk prediction model for acute stroke in maintenance hemodialysis patients},
journal = {PeerJ},
year = {2026},
volume = {14},
pages = {e21894},
doi = {10.7717/peerj.21894}
}
EUVIMED