RAPIDxAI: Can AI-augmented chest pain assessment improve cardiovascular outcomes?

An AI tool that distinguished between type 1 myocardial infarction (MI) and non-type 1 MI in the setting of assessing cardiac chest pain at the emergency department did not reduce major cardiovascular adverse events compared with usual care in a hospital setting. However, exploratory analyses suggest there may be value in this tool with some finetuning in diagnosing and managing patients with elevated high-sensitivity cardiac troponin (hs-cTn).

Dr Derek Chew (Monash University, Australia) and colleagues designed an AI tool to guide cardiac chest pain assessment in the emergency department [1]. The clinical decision support tool mostly aimed to differentiate type 1 MI from other forms of myocardial injury. The investigators compared the use of this AI tool to usual care in a multicentre, cluster-randomised trial.

The participants (n=3,029) from 12 Australian hospitals were randomised 1:1 to the intervention arm, in which patients stratified by phenotype by the AI tool received treatment recommendations that fitted this phenotype, or to the control arm, in which patients were not stratified by this tool. The primary endpoint was a composite of cardiovascular death, MI, and cardiac readmission at 6 months.

No effect was observed on the primary endpoint (HR 0.99; 95% CI 0.86–1.14; P_cluster=0.87). In the type 1 MI cohort of participants (n=578), the authors noted an increase in statin use, P2Y12 inhibitors, and mineralocorticoid receptor antagonists in the intervention arm, but not in invasive management or revascularisation. In the non-type 1 MI cohort (n=2,441), the authors saw a slight decrease in angiography, beta-blocker use, and revascularisation in the intervention arm compared with the control arm. An exploratory analysis, excluding 165 participants with ST-elevation MI, a population in which the use of an algorithm may be less useful, did suggest that the AI tool helps to reduce the rate of cardiovascular death or MI (HR 0.81; 95% CI 0.66–0.99; P-cluster=0.048).

Although the new AI tool did not improve cardiovascular outcomes in this study, the trial produced some hypothesis-generating data that may help to assess the value of the tool in a more specific group of patients.

1. 
   
   1. Lambrakis K, et al. RAPIDxAI: Re-engineering the clinical approach to suspected cardiac chest pain assessment in the emergency department using artificial intelligence. HOTLINE 12, ESC Congress 2024, 30 Aug–02 Sept, London, UK.

Copyright ©2024 Medicom Medical Publishers



Posted on 7 November 2024