Developing and evaluating machine learning models in the prediction of recurrent sudden cardiac arrest

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Sudden cardiac arrest (n=7062). Intervention: Gradient boosting machine learning model vs. Other machine learning models. Primary outcome: Prediction of recurrent sudden cardiac arrest (AUROC 0.90, p=0.003).

Q: What does this research mean for the field?

A gradient boosting machine learning model accurately predicts recurrent sudden cardiac arrest with high discrimination (AUROC 0.90 in external validation) based on explainable clinical and electrocardiographic features. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Paulomi Gohel; Michael Daidone; Tanmay Gokhale; Nathan T. Riek; Floyd W. Thoma; Suresh R Mulukutla; Seung J. Moon; Chinmay Patel; Sandeep Kumar Jain; Murat Akçakaya; Salah Al-Zaiti; Samir Saba

doi:10.3389/fcvm.2026.1883279

October 14, 2026Frontiers in Cardiovascular MedicineOpen Access

Developing and evaluating machine learning models in the prediction of recurrent sudden cardiac arrest

View Full Paper

Ask AI

Bookmark

Share

Key result

Machine learning model predicts recurrent sudden cardiac arrest with an AUROC of ~0.90.

AUROC 0.90
P=0.003
n=7,062

Why the study?

Sudden cardiac arrest has a significant recurrence rate, and predicting recurrent SCA risk has important implications for patient counseling and management.

Key Points

To develop and evaluate explainable machine learning models for predicting recurrent sudden cardiac arrest in survivors of an initial arrest.
Selected 5,286 patients with an index arrest as a primary diagnosis and split them into training (n = 3,382, 3.8% event rate), internal validation (n = 846, 3.8% event rate), and holdout test (1,058, 3.8% event rate) sets.
Tested multiple machine learning discriminators and externally validated the top-performing model in an independent cohort of 1,776 patients (2.3% event rate) using AUROC, sensitivity, specificity, PPV, NPV, and F1 scores.
The gradient boosting model achieved an AUROC of 0.93 in internal validation, 0.92 in the holdout set, and 0.90 in external validation (sensitivity = 0.92, specificity = 0.83, F1 score = 0.19).
Shapley additive explanations showed that shockable rhythm (ventricular tachycardia), structural heart disease, conduction abnormalities, and initial blood work at the index arrest were the main predictive contributors.
Stratification into low (NPV ≥ 99%), intermediate, and high (PPV≥25%) risk groups strongly predicted time to recurrent arrest across the validation cohorts (p = 0.003).

Study Design

Type

Cohort (n=7,062)

Multicenter

Yes

Structured PICO

P

Population

7,062 survivors of an initial sudden cardiac arrest were evaluated retrospectively to train and externally validate a machine learning model predicting recurrent events over a median 197-day follow-up.

O

Outcome

Recurrent sudden cardiac arrest (defined as a new hospital admission after discharge following the index admission for SCA, with discharge diagnostic code for VT, VF, or unspecified SCA)hard clinical

Main Result

Effect estimate: AUROC 0.90

p-value: p=0.003

A machine learning gradient boosting model based on explainable clinical factors demonstrated excellent discrimination for predicting recurrent sudden cardiac arrest, potentially supporting risk stratification for secondary prevention.

Limitations

Small number of recurrent sudden cardiac arrest events limiting statistical power
Retrospective observational data subject to coding inaccuracies, missing data, and ascertainment bias
Model hyperparameters were selected manually rather than through systematic optimization
Lack of advanced cardiac imaging, electrophysiologic testing, or genetic data in the prediction model
Relatively small overall number of recurrent SCA events
Retrospective, observational data subject to inaccuracies in coding, missing data, unmeasured confounding, and ascertainment bias
Further validation in additional populations and across different healthcare systems is necessary to establish generalizability
Model hyperparameters were selected manually predominantly using default values rather than systematic optimization
Model was limited to variables available within the dataset and did not incorporate advanced cardiac imaging, electrophysiologic testing, or genetic data

Implication

May aid post-arrest risk stratification; leaves open whether ML models improve outcomes in prospective trials.

Abstract

Background Sudden cardiac arrest (SCA) is a major killer of adults in the United States and has a significant recurrence rate. Predicting the risk of recurrent SCA has implications to patients’ counseling and management. Objective The goal of this study is to develop and evaluate machine learning (ML) models to predict recurrent SCA in this high-risk population. Methods Patients ( n = 5,286) whose index arrest was coded as a primary diagnosis were selected and followed to the endpoint of second SCA. The cohort was divided into training ( n = 3,382, 3.8% event rate), internal validation ( n = 846, 3.8% event rate), and holdout test (1,058, 3.8% event rate) sets. External validation was performed using a cohort of 1,776 patients (2.3% event rate). Multiple ML discriminators were tested using model area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and F1 scores. The model with highest AUROC was evaluated on a holdout test set and was externally validated. Results The gradient boosting (GB) model outperformed other models with AUROC of 0.93 in the internal validation set. It also achieved high discrimination in the holdout (AUROC=0.92), and external validation (AUROC = 0.90, sensitivity = 0.92, specificity = 0.83, F1 score = 0.19) sets. Shapley additive explanations (SHAP) demonstrated that a diagnostic code of shockable rhythm (ventricular tachycardia), evidence of structural heart disease, conduction abnormalities, and initial blood work results at the time of the index arrest were main contributors to model prediction. Low (NPV ≥ 99%), intermediate, and high (PPV≥25%) risk groups defined from the internal validation set were applied to the holdout and external validation sets with strong prediction of time to recurrent SCA ( p = 0.003). Conclusion Our ML prediction model demonstrates very good and stable discrimination in the prediction of recurrent SCA based on explainable clinical factors. External validation results confirm the generalizability of our findings. This model could support risk stratification and counseling for survivors of SCA.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Gohel et al. (2026) conducted a cohort in Sudden cardiac arrest (n=7,062). Gradient boosting machine learning model vs. Other machine learning models was evaluated on Prediction of recurrent sudden cardiac arrest (AUROC 0.90, p=0.003). A gradient boosting machine learning model demonstrated high discrimination for predicting recurrent sudden cardiac arrest, achieving an AUROC of 0.90 in an external validation cohort.

synapsesocial.com/papers/6ad34de0727baf805bb60e75 https://doi.org/10.3389/fcvm.2026.1883279

Topics

Artificial intelligence in cardiology

View Full Paper

Ask AI

Bookmark

Share

Developing and evaluating machine learning models in the prediction of recurrent sudden cardiac arrest

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Topics

Developing and evaluating machine learning models in the prediction of recurrent sudden cardiac arrest

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Topics

Keep reading