Heart failure remains a major global cause of morbidity and mortality, and delayed diagnosis continues to limit timely intervention, particularly in settings with restricted access to echocardiography and specialist assessment. This study presents an explainable hybrid artificial intelligence framework designed for early heart failure prediction by integrating structured clinical variables with language model-derived contextual information. A total of 918 de-identified cardiovascular records containing key demographic, symptom, electrocardiographic, and physiological variables were preprocessed using imputation, standardization, categorical encoding, and class balancing with SMOTE and generative adversarial network augmentation. The structured data pathway employed convolutional neural networks, while large language models processed unstructured clinical narratives to generate complementary embeddings. These multimodal representations were fused in a unified predictive architecture and benchmarked against standalone convolutional neural network, recurrent neural network, GAN-augmented convolutional neural network, and language model-only baselines using stratified five-fold cross-validation and a held-out test set. The hybrid model achieved the highest performance, with 95.1% accuracy, 94.8% precision, 95.7% recall, 95.2% F1-score, and an area under the receiver operating characteristic curve of 97.6%. The most influential predictors were chest pain type, maximum heart rate, exercise-induced angina, oldpeak, and age, which align with established cardiovascular risk markers. The integration of explainable artificial intelligence using SHAP further improved interpretability and clinical relevance. These findings suggest that hybrid AI can provide accurate, transparent, and scalable decision support for earlier heart failure detection and triage, with particular value for resource-constrained healthcare environments.

Abedalrahman Alshraideh is an Internal Medicine Trainee and Medical Registrar with a strong clinical and academic interest in cardiology and cardiovascular research. He obtained his MD from the University of Jordan in 2021. His work focuses on the application of artificial intelligence and machine learning in clinical medicine, with particular interest in cardiovascular risk prediction, early heart failure detection, and decision-support tools that can improve patient outcomes. He is committed to advancing clinically relevant, interpretable, and scalable digital solutions for modern healthcare practice.