Journal of Cardiology and Heart Failure

ABSTRACT

Heart failure is a major cardiovascular condition in which clinical status may change over time and may require timely reassessment. Artificial intelligence (AI) has been investigated for diagnosis, prognosis, risk stratification, and prediction of outcomes in heart failure, but predictive performance alone does not establish clinical usefulness. This theoretical article proposes IMPETUS, an AI-based early-warning framework designed to detect possible heart-failure worsening from longitudinal, multimodal patient-health information. The framework integrates patient-reported symptoms, vital signs, body weight, oxygen saturation, electrocardiographic information, medication-related information, previous clinical records, laboratory and clinical information, and telemonitoring data. Its theoretical novelty is not the individual use of these components, which are already present in the literature, but their organization into a temporal, multimodal, explainable, human-reviewed warning pathway in which changing patient trajectories are treated as the central unit of early-warning reasoning. For this article, heart-failure worsening is defined prospectively as a predefined transition toward clinically meaningful deterioration, such as urgent clinical assessment, emergency presentation, heart-failure hospitalization, or another validated decompensation endpoint, within a prespecified prediction window. Six testable propositions are formulated with explicit comparators and outcomes. The article also specifies candidate temporal AI methodologies, alert-generation logic, explainability requirements, leakage prevention, external validation, clinical usefulness assessment, ethical safeguards, and the relationship of IMPETUS to existing heart-failure guidance. No clinical or experimental results are claimed. The framework remains a theoretical proposal requiring retrospective, prospective, independent, external, and real-world evaluation.

Keywords: IMPETUS; Artificial Intelligence; Heart Failure; Early Warning; Temporal Prediction; Multimodal Data; Telemonitoring; Clinical Decision Support; Explainable AI