What does this research mean for the field?

An AI-estimated electrocardiographic age exceeding chronological age by more than eight years independently predicts increased all-cause mortality, even among individuals with normal ECGs. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.NEUTRAL.

February 23, 2021Open Access

Deep neural network estimated electrocardiographic-age as a mortality predictor

Q: What are the key findings of this study?

An AI-estimated electrocardiographic age more than 8 years greater than chronological age was associated with a significantly higher risk of mortality (HR 1.79) compared to an ECG-age closer to chronological age.

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: General population undergoing ECG screening (n=1574282). Intervention: AI-estimated ECG-age more than 8 years greater than chronological age vs. ECG-age within 8 years of chronological age. Primary outcome: Overall mortality (HR 1.79, 95% CI 1.69-1.90, p=<0.001).

View Full Paper

Ask AI

Bookmark

Key result

AI-estimated ECG age >8 years older than chronological age is linked to ~79% higher mortality.

HR 1.79
95% CI 1.69-1.90
P<0.001
n=1,574,282

Why the study?

The ECG is widely used for cardiovascular screening, but the potential of AI-predicted ECG-age as a prognostic marker for mortality was not established.

Does deep neural network estimated ECG-age predict mortality in patients undergoing 12-lead ECG?

Population

1,558,415 patients from the CODE study cohort with external validation in ELSA-Brasil (14,236) and SaMi-Trop (1,631) cohorts

Comparison

ECG-age estimated by deep neural network vs chronological age

Design

Cohort study using deep convolutional neural network analysis of 12-lead ECGs

Follow-up

Mean 3.67 years

Authors

ELEmilly M. Lima ARAntônio H. Ribeiro ARAntônio H. Ribeiro

Discussion

Member takes

Overview

Does not support routine clinical use of ECG-age discrepancies; leaves open incremental value in validated risk models.

Study Design

Type

Cohort (n=1,574,282)

Multicenter

Yes

Structured PICO

Does deep neural network estimated ECG-age predict mortality in patients undergoing 12-lead ECG?

Population

Over 1.5 million patients who underwent a 12-lead ECG in Brazil, followed for a mean of 3.67 years to evaluate the prognostic value of AI-estimated electrocardiographic age.

Exposure

Deep convolutional neural network estimation of electrocardiographic-age (ECG-age) from raw 12-lead ECG tracings.

Comparator

Chronological age (specifically comparing patients with ECG-age >8 years greater vs. >8 years less than chronological age).

Outcome

Mortality rate at a mean follow-up of 3.67 years.hard clinical

Main Result

Hazard Ratio: 1.79 (95% CI 1.69–1.9)

p-value: p=<0.001

AI-estimated ECG-age that is significantly higher than chronological age is an independent predictor of increased mortality, even in patients with apparently normal ECGs.

Limitations

Retrospective observational design.
Lack of interpretability of the deep neural network features, as cardiologists could not visually distinguish the ECG traces associated with higher predicted ages.
Potential residual confounding despite adjustment for cardiovascular risk factors.

Cite This Study

Lima et al. (2021) conducted a cohort in General population undergoing ECG screening (n=1,574,282). AI-estimated ECG-age more than 8 years greater than chronological age vs. ECG-age within 8 years of chronological age was evaluated on Overall mortality (HR 1.79, 95% CI 1.69-1.90, p=<0.001). An AI-estimated electrocardiographic age more than 8 years greater than chronological age was associated with a significantly higher risk of mortality (HR 1.79) compared to an ECG-age closer to chronological age.

synapsesocial.com/papers/6ab0b5c19398050a857f79e4 https://doi.org/10.1101/2021.02.19.21251232

View Full Paper

Ask AI

Bookmark

February 23, 2021Open Access

Deep neural network estimated electrocardiographic-age as a mortality predictor

View Full Paper

Ask AI

Bookmark

Key result

AI-estimated ECG age >8 years older than chronological age is linked to ~79% higher mortality.

HR 1.79
95% CI 1.69-1.90
P<0.001
n=1,574,282

Why the study?

The ECG is widely used for cardiovascular screening, but the potential of AI-predicted ECG-age as a prognostic marker for mortality was not established.

Does deep neural network estimated ECG-age predict mortality in patients undergoing 12-lead ECG?

Population

1,558,415 patients from the CODE study cohort with external validation in ELSA-Brasil (14,236) and SaMi-Trop (1,631) cohorts

Comparison

ECG-age estimated by deep neural network vs chronological age

Design

Cohort study using deep convolutional neural network analysis of 12-lead ECGs

Follow-up

Mean 3.67 years

Authors

ELEmilly M. Lima ARAntônio H. Ribeiro ARAntônio H. Ribeiro

Discussion

Member takes

Overview

Does not support routine clinical use of ECG-age discrepancies; leaves open incremental value in validated risk models.

Study Design

Type

Cohort (n=1,574,282)

Multicenter

Yes

Structured PICO

Does deep neural network estimated ECG-age predict mortality in patients undergoing 12-lead ECG?

Population

Over 1.5 million patients who underwent a 12-lead ECG in Brazil, followed for a mean of 3.67 years to evaluate the prognostic value of AI-estimated electrocardiographic age.

Exposure

Deep convolutional neural network estimation of electrocardiographic-age (ECG-age) from raw 12-lead ECG tracings.

Comparator

Chronological age (specifically comparing patients with ECG-age >8 years greater vs. >8 years less than chronological age).

Outcome

Mortality rate at a mean follow-up of 3.67 years.hard clinical

Main Result

Hazard Ratio: 1.79 (95% CI 1.69–1.9)

p-value: p=<0.001

AI-estimated ECG-age that is significantly higher than chronological age is an independent predictor of increased mortality, even in patients with apparently normal ECGs.

Limitations

Retrospective observational design.
Lack of interpretability of the deep neural network features, as cardiologists could not visually distinguish the ECG traces associated with higher predicted ages.
Potential residual confounding despite adjustment for cardiovascular risk factors.

Cite This Study

synapsesocial.com/papers/6ab0b5c19398050a857f79e4 https://doi.org/10.1101/2021.02.19.21251232

Also Consider

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

View Full Paper

Ask AI

Bookmark