Multicenter validation of AI-enabled ECG for pediatric biological sex prediction

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Pediatric patients aged 0-18 years undergoing electrocardiogram (ECG) testing with known biological sex (n=67578). Intervention: AI-enabled electrocardiogram (AI-ECG) biological sex prediction model vs. None (model performance evaluated against actual biological sex). Primary outcome: Accuracy of AI-enabled ECG model in predicting biological sex across pubertal stages measured by AUROC (AUROC 0.80 overall).

Donnchadh O’Sullivan; Joshua Mayourian; Scott Anjewierden; Kan Liu; Zachi Itzhak Attia; Francisco Lopez-Jimenez; Paul A. Friedman; Tam T. Doan; Lance Patterson; Jennifer Dugan; Jonathan N. Johnson; Santiago O. Valdés; Daniel J. Penny; John K. Triedman; Jeffrey J. Kim; Talha Niaz; Stephen Morris; Malini Madhavan

doi:10.1038/s41746-026-02416-9

February 26, 2026npj Digital MedicineOpen Access

Multicenter validation of AI-enabled ECG for pediatric biological sex prediction

View Full Paper

Ask AI

Bookmark

Share

Key result

AI-enabled ECG predicts pediatric biological sex with an overall AUROC of 0.80, improving post-puberty.

AUROC 0.80 overall
n=67,578

Figures

Why the study?

Biological sex is closely linked to patterns embedded within the ECG, but verification of an AI-enabled ECG model across pediatric development was needed.

Does an AI-enabled ECG model accurately predict biological sex across different stages of pediatric development?

Key Points

To provide multicenter verification of an artificial intelligence-enabled electrocardiogram model to predict biological sex across pediatric development stages.
Externally validated Mayo Clinic AI-ECG models at Texas Children’s Hospital across 118,003 ECGs from 67,578 patients (median age 9.0 years).
Independently trained and tested an AI-ECG model at Boston Children’s Hospital using 437,350 training ECGs (151,215 patients) and 145,784 testing ECGs (50,405 patients).
Conducted saliency mapping to identify key electrophysiologic features driving sex classification across pubertal stages and common cardiac abnormalities.
External validation at Texas Children’s Hospital confirmed a puberty-linked AUROC gradient: pre-puberty AUROC 0.64, peri-puberty AUROC 0.84, and post-puberty AUROC 0.94.
Replication at Boston Children’s Hospital demonstrated the same pubertal gradient across normal tracings and distinct cardiac conditions, retaining AUROC > 0.9 in post-pubertal ECGs even with prolonged QTc or left ventricular hypertrophy.
Saliency maps identified the QRS complex and T waves as primary discriminatory features, linking male prediction to wider QRS complexes and taller T waves, and female prediction to longer QTc intervals.

Study Design

Type

Observational (n=67,578)

Multicenter

Yes

Structured PICO

Does an AI-enabled ECG model accurately predict biological sex across different stages of pediatric development?

P

Population

Pediatric patients ≤18 years old across multiple institutions. Texas Children's Hospital (TCH) cohort: n=67,578, 52.7% male, median age 9.0 years. Boston Children's Hospital (BCH) training cohort: n=151,215, 49% male, median age 11.6 years. BCH test cohort: n=50,405, 49% male, median age 12.0 years.

I

Intervention

AI-enabled 12-lead ECG model (one-dimensional convolutional neural network) for biological sex prediction

O

Outcome

Prediction of biological sex measured by Area Under the Receiver Operating Characteristic curve (AUROC) across pubertal strata (pre-puberty, peri-puberty, post-puberty)surrogate

Main Result

Effect estimate: AUROC 0.80 overall

Limitations

Absence of clinical pubertal staging or hormone assays to corroborate pubertal status
Reliance on binary biological sex labels from electronic health records
Incomplete race and ethnicity data across cohorts
Variability in ECG recording sample rates between institutions
No assessment of impact of ECG recording duration on model performance
Absence of clinical pubertal staging or hormone assays
Reliance on binary sex labels in the electronic health record
Race/ethnicity data were not uniformly available

Implication

AI-ECG sex prediction in children requires pubertal status consideration; confirms hormone-linked electrophysiologic profile emerges during puberty.

Abstract

Biological sex is closely linked to patterns embedded within the electrocardiogram (ECG) with essential health and disease implications. We report multicenter verification of an AI-enabled ECG model to predict biological sex across pediatric development. A previously published Mayo Clinic model confirmed puberty-linked AUROC gradient during external validation at Texas Children’s Hospital (pre-puberty AUROC 0.64, peri-puberty AUROC 0.84, post-puberty AUROC 0.94). This phenomenon was replicated at Boston Children’s Hospital. Saliency mapping revealed established sex-related electrophysiologic patterns.

Discussion

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

Cite This Study

O’Sullivan et al. (2026) conducted an observational in Pediatric patients aged 0-18 years undergoing electrocardiogram (ECG) testing with known biological sex (n=67,578). AI-enabled electrocardiogram (AI-ECG) biological sex prediction model vs. None (model performance evaluated against actual biological sex) was evaluated on Accuracy of AI-enabled ECG model in predicting biological sex across pubertal stages measured by AUROC (AUROC 0.80 overall).

synapsesocial.com/papers/69a286b80a974eb0d3c01d9a https://doi.org/10.1038/s41746-026-02416-9

Topics

Artificial intelligence in cardiology

View Full Paper

Ask AI

Bookmark

Share

Multicenter validation of AI-enabled ECG for pediatric biological sex prediction

Figures

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Topics

Multicenter validation of AI-enabled ECG for pediatric biological sex prediction

Figures

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Topics

Keep reading