Samir Saba

Background: Sudden cardiac arrest (SCA) is a major killer of adults in the United States and has a significant recurrence rate. Predicting the risk of recurrent SCA has implications to patients’ counseling and management. Objective: The goal of this study is to develop and evaluate machine learning (ML) models to predict recurrent SCA in this high-risk population. Methods: Patients ( n = 5,286) whose index arrest was coded as a primary diagnosis were selected and followed to the endpoint of second SCA. The cohort was divided into training ( n = 3,382, 3.8% event rate), internal validation ( n = 846, 3.8% event rate), and holdout test (1,058, 3.8% event rate) sets. External validation was performed using a cohort of 1,776 patients (2.3% event rate). Multiple ML discriminators were tested using model area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and F1 scores. The model with highest AUROC was evaluated on a holdout test set and was externally validated. Results: The gradient boosting (GB) model outperformed other models with AUROC of 0.93 in the internal validation set. It also achieved high discrimination in the holdout (AUROC=0.92), and external validation (AUROC = 0.90, sensitivity = 0.92, specificity = 0.83, F1 score = 0.19) sets. Shapley additive explanations (SHAP) demonstrated that a diagnostic code of shockable rhythm (ventricular tachycardia), evidence of structural heart disease, conduction abnormalities, and initial blood work results at the time of the index arrest were main contributors to model prediction. Low (NPV ≥ 99%), intermediate, and high (PPV≥25%) risk groups defined from the internal validation set were applied to the holdout and external validation sets with strong prediction of time to recurrent SCA ( p = 0.003). Conclusion: Our ML prediction model demonstrates very good and stable discrimination in the prediction of recurrent SCA based on explainable clinical factors. External validation results confirm the generalizability of our findings. This model could support risk stratification and counseling for survivors of SCA.

Background: Sudden cardiac arrest (SCA), a leading cause of mortality, is administratively coded as ventricular tachycardia (VT), ventricular fibrillation (VF), or unspecified SCA. Diagnostic codes are widely used in research, yet their accuracy in reflecting true SCA events is uncertain. The objective of this study is to evaluate the relationship between administrative diagnostic code rank (listing order) and the likelihood of true SCA in patients presenting to the emergency department (ED). Methods: From a database of 22 369 patients with VT/VF/SCA, we randomly selected 380 patients for detailed chart review. Diagnostic accuracy was confirmed when patients (1) had sustained a true VT/VF/SCA event during the index hospitalization and (2) when the event was not precipitated by non‐cardiac causes. Results: Manual chart review for the 380 patients (age 71 ± 12 years, 53% women) confirmed that 65% experienced true VT/VF/SCA and 55% had an arrest unrelated to non‐cardiac causes. Diagnostic accuracy was 100% at priority rank #1 and declined significantly at lower ranks. In the overall larger cohort of 22 369 patients, recurrence of VT/VF/SCA trended 3% lower ( p = 0.086) and all‐cause mortality was 5% higher ( p < 0.001) with each diagnostic priority rank decrease, over a mean follow‐up of 1.5 years. Conclusion: Diagnostic codes show variable predictive accuracy for true VT/VF/SCA events, with the highest accuracy at priority rank #1. Higher‐priority codes identify patients most likely to have true SCA, while lower‐priority codes capture sick patients with other in‐hospital events. These results have important implications for the interpretation of administrative health datasets.

Aims: AI in electrocardiography (ECG) has diverged into two paths: traditional signal processing with machine learning, and deep learning of raw waveforms. The former preserves physiological interpretability but may miss novel patterns, while the latter overlooks a century of ECG science. No framework yet integrates both. This study introduces a fusion model combining handcrafted feature-based random forest and waveform-based convolutional neural network to improve the diagnosis of acute coronary events. Methods and results: This prospective, multicentre cohort study included pre-hospital ECGs recorded on emergency medical services from patients with a chief complaint of chest pain. Using this dataset, we developed four fusion frameworks that integrate ECG features with waveform information: (1) decision-level fusion combining predictions from random forest and convolutional neural network (CNN); (2) decision-level fusion with retraining; (3) feature-level fusion by concatenating CNN embeddings with ECG features; and (4) feature-level fusion combining CNN embeddings with tree-level outputs from the random forest. This study comprised of 10 393 ECGs from 7397 patients. On both outcomes, acute coronary syndrome and occlusion myocardial infarction, the proposed decision fusion with retraining (Approach 2) achieved an AUROC of 0.878 (95% CI, 0.837–0.915) and 0.961 (95% CI, 0.934–0.983) on the test set and an average precision score of 0.749 (95% CI, 0.681–0.810) and 0.850 (95% CI, 0.763–0.918), respectively, outperforming all other proposed fusion and individual models. Conclusion: This study demonstrates the significance of fusion models that integrate complementary ECG representations to improve diagnostic accuracy in acute coronary events, highlighting a promising approach for real-time AI-augmented cardiovascular disease detection across diverse use cases.

Background: Rhythm control with class 1C antiarrhythmic drugs (AADs) is a widely used treatment among patients with atrial fibrillation (AF), but cardiovascular benefits of class 1C pharmacotherapy over rate control strategy are still unclear. Methods: Adult patients with AF (n=100 748; 45% women) evaluated at a multihospital health system between 2010 and 2023 were assessed on clinical characteristics and class 1C AAD use and long‐term outcomes ascertained. Covariate balance across 36 factors was achieved using propensity score matching. Adjusted Cox modeling with competing risk analysis was used to compare patients using rhythm control (class 1C) with those offered a rate control strategy. Results: Patients receiving a class 1C AAD (n=5239 [5%]) (70% flecainide) were significantly younger with lower cardiovascular risk factors (hypertension, diabetes, coronary artery disease, heart failure) and less comorbidity burden. Over a mean follow‐up of 5±3 years, matched patients receiving a class 1C AAD had significantly less all‐cause death (14% versus 19%; hazard ratio [HR], 0.65 [95% CI, 0.57–0.74]; P P =0.017), CHA ₂ DS ₂ ‐VASc score ( P =0.0003), and comorbidity burden ( P =0.032) modify this relationship, while prior revascularization ( P =0.83) or presence of coronary artery disease had no significant ( P =0.54) impact. Class 1C recipients also had lower stroke hospitalization (3.6% versus 4.8%; HR, 0.82 [95% CI, 0.8–0.95]; P =0.01), while those aged 65 to 80 years experienced less heart failure (HR, 0.84 [95% CI, 0.71–0.98]; P =0.03) and major bleeding (HR, 0.8 [95% CI, 0.65–0.98]; P =0.039) as well. Conclusions: In a real‐world contemporary registry of patients with AF from a North American health care system, use of class 1C AADs for rhythm control was associated with improved survival and hospitalization for stroke compared with rate control.

Background: Patients with mitral valves prolapse have a heterogeneous course, and long‐term outcomes among those P =0.003), mitral annular calcification (HR, 2.15 [95% CI, 1.4–3.2]; P P P =0.022) were independent predictors of mortality. Cardiovascular hospitalization was associated significantly with mitral regurgitation severity (HR increase, 1.1–6.7) and presence of left atrial enlargement (HR, 1.4 [95% CI, 1.03–1.9]; P =0.03). Conclusions: In a large contemporary real‐world cohort of patients <65 years of age with MVP, the clinical and echocardiographic profile varies by age at presentation and MVP subtype with appreciable morbidity and mortality.

Objective: Acute coronary syndrome (ACS) is a life-threatening condition requiring accurate diagnosis for better outcomes. However, variability in signs and symptoms among racial subgroups could cause disparities in diagnostic accuracy. In this study, we use machine learning models to diagnose ACS, focusing on mitigating disparities and ensuring fairness between Black and non-Black populations. Methods: We built on a state-of-the-art random forest classifier to compare three mitigation strategies. The first two approaches involved resampling or partitioning the data prior to training, while the third approach proposed an innovative framework called adversarial debiasing. To evaluate our model performance, we used the receiver operating characteristic (ROC) curve and an operating point at 80% specificity for clinical importance. Results: After mitigation with adversarial debiasing, the difference in sensitivities between the two subgroups decreased from 9.8% to 1.3%. Specifically, this approach achieved areas under the ROC of 0.810 and 0.817, and sensitivities of 70.1% and 71.4%, respectively for Black and non-Black subgroups. Conclusion: The proposed adversarial debiasing model outperformed the other two methods in both diagnostic accuracy and effectiveness in minimizing disparities. Significance: We expect this framework to achieve fair diagnostic models across diverse demographic populations globally and be generalizable to other outcomes.

Aims: A typical left bundle branch block (LBBB) contraction pattern prior to cardiac resynchronization therapy (CRT) has been demonstrated to be highly associated with response beyond QRS morphology and duration. This study investigates whether septal deformation type and lateral wall strain amplitude may be of particular importance for long-term outcome after CRT implantation. Methods: From two centres, 208 CRT candidates with LBBB, New York Heart Association functional class II–IV, left ventricular ejection fraction ≤35%, and QRS duration ≥ 120 ms underwent echocardiography before CRT-defibrillator (CRT-D) implantation. Results: Four septal contraction patterns were identified: 36 patients (17%) had double-peaked systolic pattern (LBBB-1), 51 (25%) had early pre-ejection peak shortening followed by systolic stretch (LBBB-2), 43 (21%) had shortening with one systolic peak inside 70% of the ejection phase (LBBB-3), and 78 (38%) had normal septal peak timing outside early 70% of ejection phase (LBBB-4). The primary outcome (freedom from death, implantation of a left ventricular assist device, and heart transplantation) occurred in 48 patients during a median follow-up of 4 years (interquartile range 3.25–4). The most favourable outcome was associated with LBBB-1 and LBBB-2, which had one-seventh the risk of the primary outcome compared to LBBB-4 (HR: 0.14 [95% confidence interval: 0.06–0.34]; P < 0.001). Patients in the lowest tertile of lateral wall strain (>−7.4%) were at high risk of an event compared to the other patients (HR: 2.57 [1.46–4.55]; P = 0.001). Conclusion: The long-term clinical benefit of CRT-D in patients with LBBB pre-implantation varies. Subtypes of septal contraction patterns and, to a lesser degree, lateral function are important determinants of outcome in patients with CRT-D.