EUVIMEDDécouvrir EUVIMED
Open access

Early Prediction of Cardiac Arrest in the Intensive Care Unit Using Explainable Machine Learning: Retrospective Study

Frühe Vorhersage von Herzstillstand auf der Intensivstation mit erklärbarem Machine Learning: retrospektive Studie

Yun Kwan Kim iD, Won-Doo Seo iD, Sun Jung Lee iD, Ja Hyung Koo iD, Gyung Chul Kim iD, Hee Seok Song iD, Minji Lee iD

Journal of Medical Internet Research · 2024-09-17 · 26 · e62890

DOI: 10.2196/62890 ↗

Auf einen Blick

Eine retrospektive Analyse mit MIMIC-IV und eICU-CRD prüfte die Vorhersage von Herzstillstand innerhalb von 24 Stunden. Der vorgeschlagene Ansatz schnitt in verschiedenen Kohorten und ICU-Subtypen besser ab als Vergleichsmodelle.

Fragestellung

Kann ein klinisch interpretierbarer Ensemble-Ansatz Herzstillstand auf der Intensivstation innerhalb von 24 Stunden zuverlässig vorhersagen – auch bei unterschiedlicher Patient:innenheterogenität?

Methode

Retrospektive Analyse mit Daten aus MIMIC-IV und eICU-CRD; Feature-Extraktion aus Vitalzeichen, multiresolution statistischer Analyse und Gini-Index; TabNet mit cost-sensitive learning; 10-fold leave-one-patient-out CV und Cross-Dataset-Validierung.

Ergebnisse

Der Ansatz erzielte in beiden Datensätzen und über verschiedene ICU-Subtypen hinweg eine bessere Vorhersageleistung und höhere Genauigkeit als Basis- und Vergleichsmodelle. Die interpretierbaren Ergebnisse zeigten statistische Unterschiede zwischen CA- und Nicht-CA-Gruppen.

Schlussfolgerung der Autor:innen

Das neue Framework zeigte eine stabile Vorhersageleistung in unterschiedlichen ICU-Umgebungen. Die interpretierbaren Informationen spiegeln statistische Unterschiede zwischen CA- und Nicht-CA-Gruppen wider und sind für klinische Entscheidungen nutzbar.

Kernaussagen

  • Vorhersage von CA innerhalb von 24 Stunden
  • Bessere Leistung als konventionelle Modelle in MIMIC-IV und eICU-CRD
  • Interpretierbare Ergebnisse zeigten Unterschiede zwischen CA- und Nicht-CA-Gruppen

Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.

At a glance

Journal article in Journal of Medical Internet Research, published online on 2024-09-17. Written by Yun Kwan Kim and 6 co-authors. Institution: Technology Development, Seers Technology Co. Ltd., Pyeongtaek-si, Gyeonggi-do, Republic of Korea. Major MEDLINE topics: Machine Learning, Intensive Care Units, Heart Arrest. Cited 28 times so far (Crossref); 54 references. The full text is freely available in PubMed Central.

Auf einen BlickRésuméSubjectsRéférences bibliographiquesCiteRelated studies

Abstract (original language)

Background

Cardiac arrest (CA) is one of the leading causes of death among patients in the intensive care unit (ICU). Although many CA prediction models with high sensitivity have been developed to anticipate CA, their practical application has been challenging due to a lack of generalization and validation. Additionally, the heterogeneity among patients in different ICU subtypes has not been adequately addressed.

Objective

This study aims to propose a clinically interpretable ensemble approach for the timely and accurate prediction of CA within 24 hours, regardless of patient heterogeneity, including variations across different populations and ICU subtypes. Additionally, we conducted patient-independent evaluations to emphasize the model’s generalization performance and analyzed interpretable results that can be readily adopted by clinicians in real-time.

Methods

Patients were retrospectively analyzed using data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) and the eICU-Collaborative Research Database (eICU-CRD). To address the problem of underperformance, we constructed our framework using feature sets based on vital signs, multiresolution statistical analysis, and the Gini index, with a 12-hour window to capture the unique characteristics of CA. We extracted 3 types of features from each database to compare the performance of CA prediction between high-risk patient groups from MIMIC-IV and patients without CA from eICU-CRD. After feature extraction, we developed a tabular network (TabNet) model using feature screening with cost-sensitive learning. To assess real-time CA prediction performance, we used 10-fold leave-one-patient-out cross-validation and a cross–data set method. We evaluated MIMIC-IV and eICU-CRD across different cohort populations and subtypes of ICU within each database. Finally, external validation using the eICU-CRD and MIMIC-IV databases was conducted to assess the model’s generalization ability. The decision mask of the proposed method was used to capture the interpretability of the model.

Results

The proposed method outperformed conventional approaches across different cohort populations in both MIMIC-IV and eICU-CRD. Additionally, it achieved higher accuracy than baseline models for various ICU subtypes within both databases. The interpretable prediction results can enhance clinicians’ understanding of CA prediction by serving as a statistical comparison between non-CA and CA groups. Next, we tested the eICU-CRD and MIMIC-IV data sets using models trained on MIMIC-IV and eICU-CRD, respectively, to evaluate generalization ability. The results demonstrated superior performance compared with baseline models.

Conclusions

Our novel framework for learning unique features provides stable predictive power across different ICU environments. Most of the interpretable global information reveals statistical differences between CA and non-CA groups, demonstrating its utility as an indicator for clinical decisions. Consequently, the proposed CA prediction system is a clinically validated algorithm that enables clinicians to intervene early based on CA prediction information and can be applied to clinical trials in digital health.

Subjects

MeSH terms (MEDLINE)

Machine LearningIntensive Care UnitsHeart ArrestHumansRetrospective StudiesMaleFemaleMiddle AgedAged

Author keywords

cost-sensitive learningearly cardiac arrest warning systemelectric medical recordensemble learningexplainable clinical decision support systempseudo-real-time evaluation

Authors and institutions

  • Technology Development, Seers Technology Co. Ltd., Pyeongtaek-si, Gyeonggi-do, Republic of Korea.
  • Department of Brain and Cognitive Engineering, Korea University, Seoul, Republic of Korea.
  • Department of Biomedical Software Engineering, The Catholic University of Korea, Bucheon-si, Gyeonggi-do, Republic of Korea.

ORCID profiles: Yun Kwan Kim, Won-Doo Seo, Sun Jung Lee, Ja Hyung Koo, Gyung Chul Kim, Hee Seok Song, Minji Lee

Cite

APA

Kim, Y. K., Seo, W. D., Lee, S. J., Koo, J. H., Kim, G. C., Song, H. S., & Lee, M. (2024). Early Prediction of Cardiac Arrest in the Intensive Care Unit Using Explainable Machine Learning: Retrospective Study. Journal of Medical Internet Research, 26, e62890. https://doi.org/10.2196/62890

Vancouver

Kim YK, Seo WD, Lee SJ, Koo JH, Kim GC, Song HS, et al. Early Prediction of Cardiac Arrest in the Intensive Care Unit Using Explainable Machine Learning: Retrospective Study. Journal of Medical Internet Research. 2024;26:e62890. doi:10.2196/62890

BibTeX
@article{Kim2024,
  author = {Kim, Yun Kwan and Seo, Won-Doo and Lee, Sun Jung and Koo, Ja Hyung and Kim, Gyung Chul and Song, Hee Seok and Lee, Minji},
  title = {Early Prediction of Cardiac Arrest in the Intensive Care Unit Using Explainable Machine Learning: Retrospective Study},
  journal = {Journal of Medical Internet Research},
  year = {2024},
  volume = {26},
  pages = {e62890},
  doi = {10.2196/62890}
}
Start for free →