June 27, 2026EP EuropaceOpen Access

AI tool for SCD in EHR improves phenotyping precision to ~86% over existing methods.

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Sudden cardiac death (n=7000000). Intervention: AI-based model for ascertainment of SCD (aSCD) vs. ESCAPE-NET phenotype. Primary outcome: Precision (positive predictive value) of sudden cardiac death phenotyping (95% CI 0.83-0.89).

View Full Paper

Ask AI

Bookmark

Why the study?

Does an AI-based model improve the ascertainment of sudden cardiac death in electronic health records compared to existing phenotypes?

Population

>7 million individuals in the Clinical Practice Research Datalink, fine-tuned using 1,084 definite cases of…

Comparison

Transformer-based artificial intelligence model… vs Existing ESCAPE-NET Consortium phenotype for SCD…

Design

Other

Key result

An AI-based tool for ascertaining sudden cardiac death in electronic health records improved phenotyping precision, achieving a positive predictive value of 86% compared to 55% for existing methods.

Authors

BPB Petrazzini NAN Ahmed ZFZ Fan

Discussion

Member takes

Overview

Supports SCD ascertainment in research databases; leaves open prospective validation before clinical adoption.

Key Points

To develop and validate an AI-based model for the ascertainment of sudden cardiac death in electronic health records and demonstrate its utility across national datasets.
Trained an AI model using diagnostic and procedural codes from over 7 million individuals in Clinical Practice Research Datalink (CPRD).
Tested the AI model with a focus on minimizing label leakage, using 1,084 definite cases of sudden cardiac death and 6,142 curated controls.
Applied the AI model to ascertain sudden cardiac death in multiple national datasets including CPRD, UK Biobank, and Mount Sinai Data Warehouse.
Achieved an area under the curve (AUC) of 0.95 with a positive predictive value (PPV) of 0.86 in the test set.
The model demonstrated significant differentiation between sudden and expected death types (P<0.0001).
Identified 138,387 cases of sudden cardiac death in the UK, revealing mismatches in automated external defibrillator (AED) density and sudden cardiac death incidence.

Study Design

Type

Observational (n=7,000,000)

Multicenter

Yes

Structured PICO

Does an AI-based model improve the ascertainment of sudden cardiac death in electronic health records compared to existing phenotypes?

Population

Over 7 million individuals from the UK and US were used to develop and validate an AI-based model for ascertaining sudden cardiac death in electronic health records.

Exposure

Transformer-based artificial intelligence (AI) model for ascertainment of sudden cardiac death (aSCD) using diagnostic and procedural codes

Comparator

Existing ESCAPE-NET Consortium phenotype for SCD in EHR

Outcome

Area under the curve (AUC) and positive predictive value (PPV) of the aSCD model in a test set

Main Result

Absolute Event Rate: 86% vs 55%

An AI-based tool significantly improves the precision of sudden cardiac death phenotyping in electronic health records, enabling large-scale research, pharmacovigilance, and health policy applications.

Cite This Study

Petrazzini et al. (2026) conducted an observational in Sudden cardiac death (n=7,000,000). AI-based model for ascertainment of SCD (aSCD) vs. ESCAPE-NET phenotype was evaluated on Precision (positive predictive value) of sudden cardiac death phenotyping (95% CI 0.83-0.89). An AI-based tool for ascertaining sudden cardiac death in electronic health records improved phenotyping precision, achieving a positive predictive value of 86% compared to 55% for existing methods.

synapsesocial.com/papers/6a3f97ce125782b61d865a57 https://doi.org/10.1093/europace/euag105.1232

View Full Paper

Ask AI

Bookmark

June 27, 2026EP EuropaceOpen Access

AI tool for SCD in EHR improves phenotyping precision to ~86% over existing methods.

View Full Paper

Ask AI

Bookmark

Why the study?

Does an AI-based model improve the ascertainment of sudden cardiac death in electronic health records compared to existing phenotypes?

Population

>7 million individuals in the Clinical Practice Research Datalink, fine-tuned using 1,084 definite cases of…

Comparison

Transformer-based artificial intelligence model… vs Existing ESCAPE-NET Consortium phenotype for SCD…

Design

Other

Key result

An AI-based tool for ascertaining sudden cardiac death in electronic health records improved phenotyping precision, achieving a positive predictive value of 86% compared to 55% for existing methods.

Authors

BPB Petrazzini NAN Ahmed ZFZ Fan

Discussion

Member takes

Overview

Supports SCD ascertainment in research databases; leaves open prospective validation before clinical adoption.

Key Points

To develop and validate an AI-based model for the ascertainment of sudden cardiac death in electronic health records and demonstrate its utility across national datasets.
Trained an AI model using diagnostic and procedural codes from over 7 million individuals in Clinical Practice Research Datalink (CPRD).
Tested the AI model with a focus on minimizing label leakage, using 1,084 definite cases of sudden cardiac death and 6,142 curated controls.
Applied the AI model to ascertain sudden cardiac death in multiple national datasets including CPRD, UK Biobank, and Mount Sinai Data Warehouse.
Achieved an area under the curve (AUC) of 0.95 with a positive predictive value (PPV) of 0.86 in the test set.
The model demonstrated significant differentiation between sudden and expected death types (P<0.0001).
Identified 138,387 cases of sudden cardiac death in the UK, revealing mismatches in automated external defibrillator (AED) density and sudden cardiac death incidence.

Study Design

Type

Observational (n=7,000,000)

Multicenter

Yes

Structured PICO

Does an AI-based model improve the ascertainment of sudden cardiac death in electronic health records compared to existing phenotypes?

Population

Over 7 million individuals from the UK and US were used to develop and validate an AI-based model for ascertaining sudden cardiac death in electronic health records.

Exposure

Transformer-based artificial intelligence (AI) model for ascertainment of sudden cardiac death (aSCD) using diagnostic and procedural codes

Comparator

Existing ESCAPE-NET Consortium phenotype for SCD in EHR

Outcome

Area under the curve (AUC) and positive predictive value (PPV) of the aSCD model in a test set

Main Result

Absolute Event Rate: 86% vs 55%

Cite This Study

synapsesocial.com/papers/6a3f97ce125782b61d865a57 https://doi.org/10.1093/europace/euag105.1232

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark