CoChat

Explainable AI associates ECG aging effects with increased cardiovascular risk in a longitudinal population study

Open access

The DOI matches its Crossref record; 44 of the 44 references it cites were found in a registry or an index.

Abstract

Aging affects the 12-lead electrocardiogram (ECG) and correlates with cardiovascular disease (CVD). AI-ECG models estimate aging effects as a novel biomarker but have only been evaluated on single ECGs—without utilizing longitudinal data. We validated an AI-ECG model, originally trained on Brazilian data, using a German cohort with over 20 years of follow-up, demonstrating similar performance (r 2 = 0.70) to the original study (0.71). Incorporating longitudinal ECGs revealed a stronger association with cardiovascular risk, increasing the hazard ratio for mortality from 1.43 to 1.65. Moreover, aging effects were associated with higher odds ratios for atrial fibrillation, heart failure, and mortality. Using explainable AI methods revealed that the model aligns with clinical knowledge by focusing on ECG features known to reflect aging. Our study suggests that aging effects in longitudinal ECGs can be applied on population level as a novel biomarker to identify patients at risk early.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41746-024-01428-7

Cites

44 of 44 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Scatter plot illustrating the age prediction of the AI-ECG. Data stems from the SHIP-START cohort study, which includes serial ECGs taken 5–6 years apart for the same individuals. The chronological age of subjects is plotted against the predicted ECG-age to assess the accuracy of the AI-ECG predictions. The color code represents the prediction accuracy: “Correct Prediction” (blue) is defined as a predicted ECG-age within ±8 years of the chronological age. “Overestimation” (orange) indicates a predicted ECG-age more than 8 years higher, while “Underestimation” (green) refers to a predicted ECG-age more than 8 years lower. Histograms display the frequency distributions of chronological and predicted ECG-ages. The correlation coefficient r and the coefficient of determination r2 show a decreasing trend over time.
Fig. 1 Scatter plot illustrating the age prediction of the AI-ECG. Data stems from the SHIP-START cohort study, which includes serial ECGs taken 5–6 years apart for the same individuals. The chronological age of subjects is plotted against the predicted ECG-age to assess the accuracy of the AI-ECG predictions. The color code represents the prediction accuracy: “Correct Prediction” (blue) is defined as a predicted ECG-age within ±8 years of the chronological age. “Overestimation” (orange) indicates a predicted ECG-age more than 8 years higher, while “Underestimation” (green) refers to a predicted ECG-age more than 8 years lower. Histograms display the frequency distributions of chronological and predicted ECG-ages. The correlation coefficient r and the coefficient of determination r2 show a decreasing trend over time.CC BY · PubMed Central
Fig. 2 Kaplan–Meier survival estimates based on the gap between predicted ECG-age and chronological age. Results are adjusted for age and sex. Subjects with pronounced aging effects (“Overestimation”) exhibit the highest mortality rates, whereas those with reduced aging effects (“Underestimation”) exhibit the lowest.
Fig. 2 Kaplan–Meier survival estimates based on the gap between predicted ECG-age and chronological age. Results are adjusted for age and sex. Subjects with pronounced aging effects (“Overestimation”) exhibit the highest mortality rates, whereas those with reduced aging effects (“Underestimation”) exhibit the lowest.CC BY · PubMed Central
Fig. 3 HRs for overall mortality based on serial ECG recordings. The first row presents data exclusively from a single ECG at baseline examination, illustrating initial risk association. The second row focuses on single ECG from the first follow-up examinations conducted after 5–6 years. The third row includes “Serial ECGs—Consistent Predictions” that showed “Overestimation” in both, baseline and follow-up ECGs. To analyze the influence of serial ECGs, “Follow-up” is compared to “serial ECG,” with both having a follow-up period of up to 16 years w.r.t. mortality data. “Baseline” shows the influence of a longer follow-up period w.r.t. mortality data of up to 21 years in the same individuals.
Fig. 3 HRs for overall mortality based on serial ECG recordings. The first row presents data exclusively from a single ECG at baseline examination, illustrating initial risk association. The second row focuses on single ECG from the first follow-up examinations conducted after 5–6 years. The third row includes “Serial ECGs—Consistent Predictions” that showed “Overestimation” in both, baseline and follow-up ECGs. To analyze the influence of serial ECGs, “Follow-up” is compared to “serial ECG,” with both having a follow-up period of up to 16 years w.r.t. mortality data. “Baseline” shows the influence of a longer follow-up period w.r.t. mortality data of up to 21 years in the same individuals.CC BY · PubMed Central
Fig. 4 Example of two subjects that have a similar chronological age but different ECG-age. The left side depicts the precordial leads of a 51-year-old woman who is predicted as 72 years old. Relevances are highlighted in orange. The right side shows the precordial leads of a 52-year-old woman who is predicted as 41 years old. Relevances are highlighted in green. Both ECG show regular sinus rhythm whereas the first shows early R progression, which is caused by ventricular hypertrophy and partial atrial fibrosis; all findings are typical for “cor pulmonale.” The depicted grid is identical to standard ECG showing 10 mm/mV and 25 mm/s.
Fig. 4 Example of two subjects that have a similar chronological age but different ECG-age. The left side depicts the precordial leads of a 51-year-old woman who is predicted as 72 years old. Relevances are highlighted in orange. The right side shows the precordial leads of a 52-year-old woman who is predicted as 41 years old. Relevances are highlighted in green. Both ECG show regular sinus rhythm whereas the first shows early R progression, which is caused by ventricular hypertrophy and partial atrial fibrosis; all findings are typical for “cor pulmonale.” The depicted grid is identical to standard ECG showing 10 mm/mV and 25 mm/s.CC BY · PubMed Central
Fig. 5 Boxplots illustrating the importance of precordial leads in the AI-ECG’s decision-making process. Results are categorized by predicted aging effects. Each boxplot shows how relevant each precordial lead was in determining whether an ECG was classified as “Overestimation,” “Correct Prediction,” or “Underestimation” based on the AI's aging predictions. This visualization helps to understand which specific leads contributed most to the AI's assessment of aging effects in the ECG recordings.
Fig. 5 Boxplots illustrating the importance of precordial leads in the AI-ECG’s decision-making process. Results are categorized by predicted aging effects. Each boxplot shows how relevant each precordial lead was in determining whether an ECG was classified as “Overestimation,” “Correct Prediction,” or “Underestimation” based on the AI's aging predictions. This visualization helps to understand which specific leads contributed most to the AI's assessment of aging effects in the ECG recordings.CC BY · PubMed Central
Fig. 6 XAI relevances for leads V1 and V4 indicating the decision-making of the AI-ECG. The relevance values are normalized between −1 and 1 and aligned by the detected R peaks (dashed line) to obtain an average relevance over all subjects and heartbeats. Standard ECG parameters such as P-wave or QRS-complex are defined via physiological boundaries. Positive relevances indicating more pronounced aging effects resulting in a higher age prediction whereas negative relevances result in the prediction of younger age.
Fig. 6 XAI relevances for leads V1 and V4 indicating the decision-making of the AI-ECG. The relevance values are normalized between −1 and 1 and aligned by the detected R peaks (dashed line) to obtain an average relevance over all subjects and heartbeats. Standard ECG parameters such as P-wave or QRS-complex are defined via physiological boundaries. Positive relevances indicating more pronounced aging effects resulting in a higher age prediction whereas negative relevances result in the prediction of younger age.CC BY · PubMed Central
Fig. 7 Boxplots illustrating clinical features known for aging effects. Shown are P duration, PR interval, QRS duration, and QTc interval22. Each feature is positively correlated with higher age groups. Boxplots display significant differences across age groups, as indicated by the asterisks (***), denoting a highly significant statistical difference (p < 0.001) calculated with a two-sided t-test.
Fig. 7 Boxplots illustrating clinical features known for aging effects. Shown are P duration, PR interval, QRS duration, and QTc interval22. Each feature is positively correlated with higher age groups. Boxplots display significant differences across age groups, as indicated by the asterisks (***), denoting a highly significant statistical difference (p < 0.001) calculated with a two-sided t-test.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 44 references

Research works printed but not linked: 1

  1. Xie, C. et al. Waveform database software package (WFDB) for Python (version 3.4. 1) Available at https://doi.org/10.13026/gjvw-1m31 (2021). · DOI

Cited by

See all 46

Metadata

Authors
Hempel, P., Ribeiro, A. H., Vollmer, M., Bender, T., Dörr, M., Krefting, D., & Spicher, N.
Affiliations
  • Universitätsmedizin Göttingen (DE)
  • Uppsala University (SE)
  • Universitätsmedizin Greifswald (DE)
Corresponding author
Hempel, Philip
Venue
npj Digital Medicine
Published
2025-01-13
Volume
8
Issue
1
Pages
25-25
ISSN
2398-6352 (ISSN-L)
OpenAlex
W4406332045
Cited by
31
References
44
Citations by year
2026: 15 · 2025: 16
FWCI
21.53
Type
journal article
Language
en
Open access
Licence
Locations
  • npj Digital Medicine (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • GoeScholar The Publication Server of the Georg-August-Universität Göttingen (Georg-August-Universität Göttingen) (publishedVersion, cc-by)
  • KTH Publication Database DiVA (KTH Royal Institute of Technology) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Lower Saxony “Vorab” of the Volkswagen Foundation and the Ministry for Science and Culture of Lower Saxony
  • German Federal Ministry for Education and Research, the Ministry for Education, Research and Cultural Affairs, and the Ministry for Social Affairs of the State Mecklenburg-Western Pomerania
Topics
  • Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › ECG Monitoring and Analysis
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
  • Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › Cardiac electrophysiology and arrhythmias
Keywords
cardiovascular risk, explainable AI, model validation, mortality, atrial fibrillation, heart failure, aging biomarkers, 12-lead ECG

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Explainable AI associates ECG aging effects with increased cardiovascular risk in a longitudinal population study. https://app.cochat.ai/papers/10.1038/s41746-024-01428-7
The paper
Hempel, P., Ribeiro, A. H., Vollmer, M., Bender, T., Dörr, M., Krefting, D., & Spicher, N. (2025). Explainable AI associates ECG aging effects with increased cardiovascular risk in a longitudinal population study. npj Digital Medicine. https://doi.org/10.1038/s41746-024-01428-7

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.