The Recognition of STEMI by Paramedics and the Effect of Computer inTerpretation (RESPECT): a randomised crossover feasibility study.

Background: The appropriate management of patients with ST-segment elevation myocardial infarction (STEMI) depends on accurate interpretation of the 12-lead ECG by paramedics. Computer interpretation messages on ECGs are often provided, but the effect they exert on paramedics' decision-making is not known. The objective of this study was to assess the feasibility of using an online assessment tool, and collect pilot data, for a definitive trial to determine the effect of computer interpretation messages on paramedics' diagnosis of STEMI.

Methods: The Recognition of STEMI by Paramedics and the Effect of Computer inTerpretation (RESPECT) feasibility study was a randomised crossover trial using a bespoke, web-based assessment tool. Participants were randomly allocated 12 of 48 ECGs, with an equal mix of correct and incorrect computer interpretation messages, and STEMI and STEMI-mimics. The nature of the responses required a cross-classified multi-level model.

Results: 254 paramedics consented into the study, 205 completing the first phase and 150 completing phase two. The adjusted OR for a correct paramedic interpretation, when the computer interpretation was correct (true positive for STEMI or true negative for STEMI-mimic), was 1.80 (95% CI 0.84 to 4.91) and 0.58 (95% CI 0.41 to 0.81) when the computer interpretation was incorrect (false positive for STEMI or false negative for STEMI-mimic). The intraclass correlation coefficient for correct computer interpretations was 0.33 for participants and 0.17 for ECGs, and for incorrect computer interpretations, 0.06 for participants and 0.01 for ECGs.

Conclusions: Determining the effect of computer interpretation messages using a web-based assessment tool is feasible, but the design needs to take clustered data into account. Pilot data suggest that computer messages influence paramedic interpretation, improving accuracy when correct and worsening accuracy when incorrect.