Use of Artificial Intelligence in Stress Echocardiography in NHS Coronary Artery Disease Risk Prediction: A Cost Effectiveness Analysis Study

OBJECTIVES: This economic evaluation was conducted alongside a randomized controlled trial (RCT) to assess cost-effectiveness of implementing AI-augmented decision-making (EchoGo Pro, Ultromics Ltd) in stress echocardiography for coronary artery disease (CAD) pathway within NHS.

METHODS: Data were collected from 2,213 patients across 20 NHS hospitals, who were randomized to receive either standard care or standard care with AI-augmented decision-making. Appropriateness of clinical management decisions was assessed based on confirmation of severe CAD or related cardiac events, helping to derive probabilities for patients falling into different diagnostic categories. Data on consequences were collected from Seattle Angina Questionnaire (SAQ-7) and EQ-5D-5L, completed at baseline, 3, and 6 months. EQ-5D-5L data were converted into QALYs. Hospital costs were obtained from a costing study in a similar setting. Cost-consequence (CCA) and cost-effectiveness (CEA) analyses were conducted. CCA evaluated costs and consequences separately, using ANOVA and Friedman tests for within-group analyses, and t-tests and Mann-Whitney tests for between-group comparisons. CEA utilized a decision tree model to compare standard care and AI-based stress echocardiography. Probabilistic sensitivity analysis (PSA) using Monte Carlo simulation was conducted to assess uncertainty regarding CEA outcomes.

RESULTS: In CCA, SAQ-7 dimensions—physical limitation, angina frequency, and quality of life—showed significant improvements in both groups from baseline to 6 months (all p<.001), with no statistically significant differences in change patterns between groups (p=0.99, 0.324, 0.181). CEA indicated that AI-based stress echocardiography has slightly higher effectiveness compared to standard care, albeit at a higher cost.

CONCLUSIONS: AI-based stress echocardiography proved to be a more optimal strategy than standard care, offering slightly higher effectiveness but increased costs whilst remaining within NICE-recommended willingness-to-pay thresholds in specific scenarios. Overall, whether AI-based stress echocardiography remains within NICE-recommended willingness-to-pay thresholds at observed accuracy and effectiveness will depend on actual implementation costs and potential savings in clinician time, inter-alia.