Incorporation of Serial 12-Lead ECG with Machine Learning to Augment the Prehospital Diagnosis of Non-ST Elevation Acute Coronary Syndrome

Background: Ischemic ECG changes are subtle and transient in patients with suspected non-ST-segment elevation acute coronary syndrome (NSTE-ACS), yet the prehospital (PH)-ECG is not routinely used during subsequent evaluation at the emergency department (ED). We sought to compare the diagnostic performance of PH- and ED-ECG and evaluate the incremental gain of artificial intelligence (AI)-augmented ECG analysis. Methods: This prospective observational cohort study recruited patients with prehospital chest pain. We retrieved PH-ECG obtained by paramedics in the field and first ED-ECG obtained by nurses during in-hospital evaluation. Two independent and blinded reviewers interpreted ECG dyads in mixed order as per practice recommendations. Using 179 morphological ECG features, we trained, cross-validated, and tested a random forest classifier to augment NSTE-ACS diagnosis. Results: Our sample included 2,122 patients (age 59 (16); 53% females; 44% Black, 13.5% confirmed ACS). The rate of diagnostic ST elevation and ST depression were 5.9% and 16.2% on PH-ECG and 6.1% and 12.4% on ED-ECG, with ~40% of changes seen on PH-ECG persisting and ~60% resolving. Using PH-ECG alone gave poor baseline performance with AUROC, sensitivity, and negative predictive value of 0.69, 0.50, and 0.92. Using serial ECG changes enhanced this performance (0.80, 0.61, and 0.93). Interestingly, augmenting the PH-ECG alone with AI algorithms boosted its performance (0.83, 0.75, 0.95), yielding NRI of 29.5% against expert ECG interpretation. Conclusion: In this study, sixty percent of diagnostic ST changes resolved prior to hospital arrival, making the ED-ECG suboptimal for in-hospital evaluation of NSTE-ACS. Using serial ECG changes or incorporating AI-augmented analyses would allow correctly reclassifying one in four patients with suspected NSTE-ACS.