Prediction of Venous Thromboembolism in Diverse Populations Using Machine Learning and Structured Electronic Health Records
Vorhersage von venöser Thromboembolie in diversen Populationen mit Machine Learning und strukturierten elektronischen Gesundheitsakten
Arteriosclerosis, Thrombosis, and Vascular Biology · 2024-02 · 44 (2) · 491-504
DOI: 10.1161/atvbaha.123.320331 ↗Auf einen Blick
Modelle aus strukturierten EHR-Daten sagten VTE-Diagnose und das 1-Jahres-Risiko in drei Datensätzen mit unterschiedlichen Leistungen voraus und übertrafen den Padua-Score.
Fragestellung
Lassen sich mit Machine-Learning-Modellen aus strukturierten EHR-Daten VTE-Diagnose und 1-Jahres-Risiko in diversen Populationen vorhersagen?
Methode
Entwicklung und Validierung auf 159 001 Teilnehmenden; externe Tests in der UK Biobank und All of Us. Kleine, mittlere und große Modelle mit unterschiedlich vielen Merkmalen wurden trainiert.
Ergebnisse
Im Holdout- und externen Testset lagen die AUCs für die VTE-Diagnose bei 0,80–0,83 bzw. 0,72–0,82 und für das 1-Jahres-Risiko bei 0,76–0,78 bzw. 0,64–0,69. Modelle schnitten besser ab als der Padua-Score.
Schlussfolgerung der Autor:innen
Machine-Learning-Modelle mit strukturierten EHR-Daten können die VTE-Diagnose und die 1-Jahres-Risikovorhersage in diversen Populationen deutlich verbessern.
Kernaussagen
- VTE-Diagnose: AUC 0,80–0,83 im Holdout-Set
- 1-Jahres-Risiko: AUC 0,76–0,78 im Holdout-Set
- Externe Tests zeigten ebenfalls bessere Werte als der Padua-Score
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 Arteriosclerosis, Thrombosis, and Vascular Biology, 2024-02. Verfasst von Robert Chen und 4 weiteren Autor:innen. Beteiligte Einrichtung: Charles Bronfman Institute for Personalized Medicine (R.C., B.O.P., R.D.), Icahn School of Medicine at Mount Sinai, New York. Hauptthemen laut MEDLINE: Venous Thromboembolism, Population Health. Die Arbeit wurde bisher 38-mal zitiert (Crossref) und nennt 55 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: NIGMS NIH HHS.
Themen: Thromboseprophylaxe, Künstliche Intelligenz in der Pflege
Abstract (Originalsprache)
BACKGROUND: Venous thromboembolism (VTE) is a major cause of morbidity and mortality worldwide. Current risk assessment tools, such as the Caprini and Padua scores and Wells criteria, have limitations in their applicability and accuracy. This study aimed to develop machine learning models using structured electronic health record data to predict diagnosis and 1-year risk of VTE. METHODS: We trained and validated models on data from 159 001 participants in the Mount Sinai Data Warehouse. We then externally tested them on 401 723 participants in the UK Biobank and 123 039 participants in All of Us. All data sets contain populations of diverse ancestries and clinical histories. We used these data sets to develop small, medium, and large models with increasing features on a range of optimizing portability to maximizing performance. We make trained models publicly available in click-and-run format at https://doi.org/10.17632/tkwzysr4y6.6 . RESULTS: In the holdout and external test sets, respectively, models achieved areas under the receiver operating characteristic curve of 0.80 to 0.83 and 0.72 to 0.82 for VTE diagnosis prediction and 0.76 to 0.78 and 0.64 to 0.69 for 1-year risk prediction, significantly outperforming the Padua score. Models also demonstrated robust performance across different VTE types and patient subsets, including ethnicity, age, and surgical and hospitalization status. Models identified both established and novel clinical features contributing to VTE risk, offering valuable insights into its underlying pathophysiology. CONCLUSIONS: Machine learning models using structured electronic health record data can significantly improve VTE diagnosis and 1-year risk prediction in diverse populations. Model probability scores exist on a continuum, affecting mortality risk in both healthy individuals and VTE cases. Integrating these models into electronic health record systems to generate real-time predictions may enhance VTE risk assessment, early detection, and preventative measures, ultimately reducing the morbidity and mortality associated with VTE.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Charles Bronfman Institute for Personalized Medicine (R.C., B.O.P., R.D.), Icahn School of Medicine at Mount Sinai, New York.
- Medical Scientist Training Program (R.C.), Icahn School of Medicine at Mount Sinai, New York.
- Department of Genetics and Genomic Sciences (R.C., B.O.P., R.D.), Icahn School of Medicine at Mount Sinai, New York.
- Center for Genomic Data Analytics (B.O.P., R.D.), Icahn School of Medicine at Mount Sinai, New York.
- The Zena and Michael A. Wiener Cardiovascular Institute (W.A.M., R.S.R.), Icahn School of Medicine at Mount Sinai, New York.
ORCID-Profile: Robert Chen, Ben Omega Petrazzini, Robert S Rosenson, Ron Do
Förderung
- NIGMS NIH HHS
Zitieren
APA
Chen, R., Petrazzini, B. O., Malick, W. A., Rosenson, R. S., & Do, R. (2024). Prediction of Venous Thromboembolism in Diverse Populations Using Machine Learning and Structured Electronic Health Records. Arteriosclerosis, Thrombosis, and Vascular Biology, 44(2), 491-504. https://doi.org/10.1161/atvbaha.123.320331
Vancouver
Chen R, Petrazzini BO, Malick WA, Rosenson RS, Do R. Prediction of Venous Thromboembolism in Diverse Populations Using Machine Learning and Structured Electronic Health Records. Arteriosclerosis, Thrombosis, and Vascular Biology. 2024;44(2):491-504. doi:10.1161/atvbaha.123.320331
BibTeX
@article{Chen2024,
author = {Chen, Robert and Petrazzini, Ben Omega and Malick, Waqas A and Rosenson, Robert S and Do, Ron},
title = {Prediction of Venous Thromboembolism in Diverse Populations Using Machine Learning and Structured Electronic Health Records},
journal = {Arteriosclerosis, Thrombosis, and Vascular Biology},
year = {2024},
volume = {44},
number = {2},
pages = {491--504},
doi = {10.1161/atvbaha.123.320331}
}
EUVIMED