Document resource
Introduction Artificial intelligence-assisted electrocardiogram interpretation (AI-ECG) has shown promise in detecting occlusive myocardial infarction (OMI), with previous studies focusing mainly on electrocardiograms (ECG) in primary care and emergency settings. We investigated the potential impact of AI-ECG on cardiac catheterisation laboratory activation in a tertiary centre delivering primary percutaneous coronary intervention (PPCI).Methods We retrospectively reviewed all acute coronary syndrome (ACS) referrals to the Golden Jubilee National Hospital (GJNH) over a 4-week period in February 2020. ECGs were anonymised and analysed using the Queen of Hearts (QoH) AI programme (PMcardio v3.3.0).Patients were categorised according to the cardiac team’s initial decision following ECG review:Emergency angiography (PPCI) at GJNHAdmission to GJNH for urgent angiography (within 24 hours)Transfer to a local hospital for further assessment with subsequent angiography(3) without angiographyWe defined OMI as the following:Confirmed OMI: Acute occlusive culprit lesion with Thrombolysis in Myocardial Infarction (TIMI) 0–1 on angiographyPresumed OMI:With angiography – acute non-occlusive culprit lesion (TIMI 2–3) with significant infarct size as demonstrated by (i) troponin >100 times upper limit of normal (TnT >1300 ng/L, TnI >3400 ng/L), or rise >20 ng/L/hour, or (ii) new regional wall motion abnormality on echocardiographyWithout angiography – criteria (i) and (ii) as aboveDescriptive statistics were calculated, and diagnostic performance of QoH versus clinicians was compared using Area Under the Receiver Operating Characteristic Curve (AUROC) with De Long’s test.Results Among 294 ACS referrals, 57 (19%) underwent emergency angiography and 18 (6%) were admitted to GJNH for urgent angiography. The remaining 219 patients (74%) were transferred to their local hospital for further assessment; of these, 45 (21%) later underwent invasive angiography. Following review of electronic health care records after discharge, a total of 96 patients (33%) met OMI criteria, of whom 91 underwent revascularisation (87 PCI; 4 coronary artery bypass graft).QoH classified 87 patients as OMI (table 1). 32 of 87 were not taken for emergency angiography; of these, 24 met OMI criteria and 22 underwent invasive angiography, all of whom received revascularisation. QoH demonstrated superior discriminatory performance for OMI detection compared with clinicians (AUROC: 0.89 vs 0.78; p<0.0001; figure 1), with significantly greater sensitivity and negative predictive value (table 2).Theoretical adoption of QoH would have led to 32 additional emergency cardiac catheterisation activations - a 56% increase.Conclusion Adoption of AI-ECG could significantly inform activation of emergency coronary angiography for suspected OMI. Larger prospective studies are required to determine improvements in patient outcomes and cost effectiveness.Abstract 52 Table 1DemographicsAbstract 52 Table 2Diagnostic performance of QoH versus the clinicianAbstract 52 Figure 1Receiver operating characteristic curves for OMI detection comparing QoH with clinician