Artificial Intelligence in Cardiovascular Care Evidence Synthesis and Economic Evaluation

Abstract

Abstract Introduction/background/issues

Cardiovascular disease remains the leading global cause of mortality, contributing over 18 million deaths annually. Despite advances in treatment, early detection and personalised intervention remain under research, particularly in resource-constrained settings. This thesis systematically evaluated the clinical performance and economic implications of artificial intelligence applications in cardiovascular disease management through integrated systematic review and economic modelling within the Australian healthcare context.

Methods

A systematic review following Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) standards was conducted across database such as PubMed, Scopus, MEDLINE, ClinicalTrials.gov, and Google Scholar (2010-2025). Studies reporting clinical and economic outcomes for AI interventions in adult cardiovascular care were included. Data extraction and data synthesis were conducted with an emphasis on diagnostic accuracy, clinical outcomes, and economic indicators A decision-analytic model was developed for artificial intelligence-guided coronary computed tomography angiography within Australian healthcare perspective, utilising diagnostic performance data from Kübler et al. (2024) and Australian cost schedules.

Results/discussions

Ten studies demonstrated robust AI diagnostic performance across ECG, echocardiography, and coronary imaging modalities. Deep learning approaches achieved superior clinical effectiveness, with area under curve values exceeding 0.95 for coronary artery disease prediction and 21% higher performance than traditional methods for cardiovascular risk stratification. AI-enhanced coronary imaging showed excellent agreement with expert assessment (ICC 0.82-0.95) while reducing interpretation time from 25 minutes to seconds. However, significant economic evaluation gaps existed, with most studies lacking formal cost and cost-effectiveness analyses. Limited international economic evidence indicated promising cost-effectiveness ratios between £1,371-£3,244 per quality-adjusted life year in European contexts. The decision-analytic model, utilising AI performance parameters from the Kübler et al. (2024) examining artificial intelligence-enhanced coronary computed tomography angiography in asymptomatic (sensitivity 91.2%, specificity 63.3% for stenosis >0%), demonstrated that this specific AI-guided imaging approach was not cost-effective under Australian healthcare conditions, despite incorporating delayed treatment consequences leading to percutaneous coronary intervention and discounted lifetime costs.

Conclusions/implications

AI demonstrates significant clinical potential for cardiovascular management across diagnostic and prognostic modalities. However, economic evidence remains incomplete. The economic evaluation demonstrated that AI-guided coronary imaging was not cost-effective under Australian healthcare perspective, despite improved diagnostic performance. This reflects the critical need for standardised cost-effectiveness frameworks before widespread adoption