Daily Cardiology Research Analysis
Three high-impact cardiology studies stood out. A multinational analysis showed an AI model applied to ECG images robustly predicts incident heart failure and improves risk stratification beyond established equations. A biomarker-driven subphenotyping framework for cardiogenic shock improved prognostication over SCAI shock staging and suggested treatment heterogeneity. A contemporary AMI in-hospital mortality model from 313,825 U.S. hospitalizations achieved excellent discrimination and bedside
Summary
Three high-impact cardiology studies stood out. A multinational analysis showed an AI model applied to ECG images robustly predicts incident heart failure and improves risk stratification beyond established equations. A biomarker-driven subphenotyping framework for cardiogenic shock improved prognostication over SCAI shock staging and suggested treatment heterogeneity. A contemporary AMI in-hospital mortality model from 313,825 U.S. hospitalizations achieved excellent discrimination and bedside usability.
Research Themes
- AI-enabled cardiovascular risk prediction
- Biomarker-driven shock phenotyping and precision cardiology
- Contemporary outcome modeling and quality benchmarking in AMI
Selected Articles
1. Identifying biomarker-driven subphenotypes of cardiogenic shock: analysis of prospective cohorts and randomized controlled trials.
Across two prospective cohorts, unsupervised clustering of plasma biomarkers revealed four reproducible cardiogenic shock subphenotypes. The inflammatory and cardiopathic classes had the highest 28-day mortality and improved risk stratification beyond SCAI stages. Applying a simplified classifier to three RCTs suggested potential heterogeneity of treatment effect by subphenotype.
Impact: This work operationalizes precision cardiology in shock by linking molecular phenotypes to outcomes and potential treatment heterogeneity, advancing beyond traditional hemodynamic staging.
Clinical Implications: Subphenotype assignment could refine prognostication, guide trial stratification/enrichment, and eventually inform phenotype-targeted therapies for cardiogenic shock.
Key Findings
- Unsupervised clustering in two cohorts identified four biomarker-defined CS subphenotypes (adaptive, non-inflammatory, cardiopathic, inflammatory).
- Inflammatory and cardiopathic subphenotypes had significantly higher 28-day mortality; adding subphenotype improved Harrell’s C-index beyond SCAI staging.
- A simplified classifier assigned subphenotypes in three RCTs, enabling exploration of heterogeneity of treatment effect on 28-day mortality.
Methodological Strengths
- Prospective cohorts with independent replication of clusters
- Application and validation across multiple RCT datasets with a simplified classifier
Limitations
- Biomarker panels and sampling times may differ between cohorts and trials
- Observational nature of associations; not a randomized phenotype-guided intervention
Future Directions: Prospective, phenotype-guided trials to test tailored therapies; standardization of biomarker panels; integration with hemodynamics and imaging.
BACKGROUND: Cardiogenic shock (CS) is a heterogeneous clinical syndrome, making it challenging to predict patient trajectory and response to treatment. This study aims to identify biological/molecular CS subphenotypes, evaluate their association with outcome, and explore their impact on heterogeneity of treatment effect (ShockCO-OP, NCT06376318). METHODS: We used unsupervised clustering to integrate plasma biomarker data from two prospective cohorts of CS patients: CardShock (N = 205 [2010-2012, NCT01374867]) and the French and European Outcome reGistry in Intensive Care Units (FROG-ICU) (N = 228 [2011-2013, NCT01367093]) to determine the optimal number of classes. Thereafter, a simplified classifier (Euclidean distances) was used to assign the identified CS subphenotypes in three completed randomized controlled trials (RCTs) (OptimaCC, N = 57 [2011-2016, NCT01367743]; DOREMI, N = 192 [2017-2020, NCT03207165]; and CULPRIT-SHOCK, N = 434 [2013-2017, NCT01927549]) and explore heterogeneity of treatment effect with respect to 28-day mortality (primary outcome). FINDINGS: Four biomarker-driven CS subphenotypes ('adaptive', 'non-inflammatory', 'cardiopathic', and 'inflammatory') were identified separately in the two cohorts. Patients in the inflammatory and cardiopathic subphenotypes had the highest 28-day mortality (p (log-rank test) = 0.0099 and 0.0055 in the CardShock and FROG-ICU cohorts, respectively). Subphenotype membership significantly improved risk stratification when added to traditional risk factors including the Society for Cardiovascular Angiography and Interventions (SCAI) shock stages (increase in Harrell's C-index by 4% ( INTERPRETATION: Subphenotypes with the highest concentration of biomarkers of endothelial dysfunction and inflammation (inflammatory) or myocardial injury/fibrosis (cardiopathic) were associated with mortality independently from the SCAI shock stages. FUNDING: Dr Sabri Soussi was awarded the Canadian Institutes of Health Research (CIHR) Doctoral Foreign Study Award (DFSA) and the Merit Awards Program (Department of Anesthesiology and Pain Medicine, University of Toronto, Canada) for the current study.
2. Heart failure risk stratification using artificial intelligence applied to electrocardiogram images: a multinational study.
An AI model applied to 12-lead ECG images predicted incident heart failure across three cohorts, with hazard ratios 3.9–23.5 and C-statistics up to 0.81. Adding AI-ECG to PCP-HF improved discrimination, supporting ECG-based digital biomarkers for HF risk stratification.
Impact: Demonstrates scalable, low-cost risk stratification using existing ECGs with AI, potentially enabling earlier HF prevention and more efficient population screening.
Clinical Implications: AI-ECG probability could triage patients for echocardiography, intensify risk modification, and augment PCP-HF in primary care and health systems.
Key Findings
- AI-ECG positivity was associated with markedly higher incident HF risk across YNHHS, UKB, and ELSA-Brasil (HRs 3.88–23.50).
- Discrimination ranged from 0.718 to 0.810; integrating AI-ECG with PCP-HF significantly improved risk prediction.
- Risk increased monotonically with higher AI-ECG probabilities, robust to comorbidities and competing risk of death.
Methodological Strengths
- Large, multinational external validation across three diverse cohorts
- Comparison to guideline-based PCP-HF and demonstration of incremental predictive value
Limitations
- Observational design limits causal inference about interventions triggered by AI-ECG
- Model transportability to other healthcare systems and ECG acquisition workflows requires further testing
Future Directions: Prospective impact studies to test AI-ECG-triggered pathways, calibration in underrepresented groups, and integration with imaging and biomarkers.
BACKGROUND AND AIMS: Current heart failure (HF) risk stratification strategies require comprehensive clinical evaluation. In this study, artificial intelligence (AI) applied to electrocardiogram (ECG) images was examined as a strategy to predict HF risk. METHODS: Across multinational cohorts in the Yale New Haven Health System (YNHHS), UK Biobank (UKB), and Brazilian Longitudinal Study of Adult Health (ELSA-Brasil), individuals without baseline HF were followed for the first HF hospitalization. An AI-ECG model that defines cross-sectional left ventricular systolic dysfunction from 12-lead ECG images was used, and its association with incident HF was evaluated. Discrimination was assessed using Harrell's C-statistic. Pooled cohort equations to prevent HF (PCP-HF) were used as a comparator. RESULTS: Among 231 285 YNHHS patients, 4472 had primary HF hospitalizations over 4.5 years (inter-quartile range 2.5-6.6). In UKB and ELSA-Brasil, among 42 141 and 13 454 people, 46 and 31 developed HF over 3.1 (2.1-4.5) and 4.2 (3.7-4.5) years. A positive AI-ECG screen portended a 4- to 24-fold higher risk of new-onset HF [age-, sex-adjusted hazard ratio: YNHHS, 3.88 (95% confidence interval 3.63-4.14); UKB, 12.85 (6.87-24.02); ELSA-Brasil, 23.50 (11.09-49.81)]. The association was consistent after accounting for comorbidities and the competing risk of death. Higher probabilities were associated with progressively higher HF risk. Model discrimination was 0.718 in YNHHS, 0.769 in UKB, and 0.810 in ELSA-Brasil. In YNHHS and ELSA-Brasil, incorporating AI-ECG with PCP-HF yielded a significant improvement in discrimination over PCP-HF alone. CONCLUSIONS: An AI model applied to a single ECG image defined the risk of future HF, representing a digital biomarker for stratifying HF risk.
3. Predicting Mortality in Patients Hospitalized With Acute Myocardial Infarction: From the National Cardiovascular Data Registry.
Using 313,825 AMI hospitalizations across 784 U.S. sites, the authors developed and validated a 14-variable in-hospital mortality model (C-statistic 0.868) and a 0–25 bedside score. Out-of-hospital cardiac arrest, cardiogenic shock, and STEMI were the strongest predictors, and performance was consistent across subgroups and pandemic periods.
Impact: Provides an updated, validated benchmark for AMI care quality and bedside prognostication, enabling risk-standardized outcomes and informed clinical decisions.
Clinical Implications: Supports hospital benchmarking, triage, and shared decision-making; the simplified score facilitates quick risk assessment and resource allocation.
Key Findings
- A 14-variable model achieved excellent discrimination (C-statistic 0.868) for AMI in-hospital mortality with good calibration.
- Strongest predictors were out-of-hospital cardiac arrest, cardiogenic shock, and STEMI; a 0–25 point bedside score mapped to 0.3%–49.4% mortality risk.
- Model performance was stable across MI type, hospital volume, and pre-/during COVID-19 periods.
Methodological Strengths
- Very large, contemporary national registry with split-sample validation and bootstrapping
- Parsimonious variable selection and creation of a user-friendly bedside score
Limitations
- Registry-based observational data; potential residual confounding and coding variability
- In-hospital outcome only; does not capture post-discharge events
Future Directions: Integration into EHRs for real-time risk dashboards; evaluation of impact on care pathways and outcomes; extension to post-discharge risk.
BACKGROUND: In-hospital mortality risk prediction is an important tool for benchmarking quality and patient prognostication. Given changes in patient characteristics and treatments over time, a contemporary risk model for patients with acute myocardial infarction (MI) is needed. METHODS: Data from 313 825 acute MI hospitalizations between January 2019 and December 2020 for adults aged ≥18 years at 784 sites in the National Cardiovascular Data Registry Chest Pain-MI Registry were used to develop a risk-standardized model to predict in-hospital mortality. The sample was randomly divided into 70% development (n=220 014) and 30% validation (n=93 811) samples, and 23 separate registry-based patient characteristics at presentation were considered for model inclusion using stepwise logistic regression with 1000 bootstrapped samples. A simplified risk score was also developed for individual risk stratification. RESULTS: The mean age of the study cohort was 65.3 (SD 13.1) years, and 33.6% were women. The overall in-hospital mortality rate was 5.0% (n=15 822 deaths). The final model included 14 variables, with out-of-hospital cardiac arrest, cardiogenic shock, and ST-segment elevation MI as the strongest independent predictors of mortality. The model also included age, comorbidities (dyslipidemia, diabetes, prior percutaneous coronary intervention, cerebrovascular disease, and peripheral artery disease), heart failure on admission, heart rate, systolic blood pressure, glomerular filtration rate, and hemoglobin. The model demonstrated excellent discrimination (C-statistic, 0.868 [95% CI 0.865-0.871]) and good calibration, with similar performance across subgroups based on MI type, periods before and during the COVID-19 pandemic, and hospital volume. The simplified risk score included values from 0 to 25, with mortality risk ranging from 0.3% with a score of 0 to 1 up to 49.4% with a score >11. CONCLUSIONS: This contemporary risk model accurately predicts in-hospital mortality for patients with acute MI and can be used for risk standardization across hospitals and at the bedside for patient prognostication.