Prediction of Elevated Troponin T Levels from Prehospital Electrocardiograms.

BACKGROUND: Early and reliable predictions of elevated cardiac troponin levels from electrocardiograms (ECGs) in the prehospital setting could serve as a valuable risk stratification tool, guiding triage and early intervention in patients with suspected acute coronary syndrome, and especially in patients presenting with electrocardiographic non-ST elevation (NSTE). Therefore, the primary objective of this study was to investigate whether machine learning applied to the prehospital ECG can enable early identification of patients at high risk of myocardial infarction. METHODS AND RESULTS: A total of 100,334 patients with a prehospital ECG and in-hospital troponin measurement available were included in this study. A random forest model was developed to predict elevated cardiac troponin T (>14 ng/L) from the prehospital ECG. Mean age was 64.72 (17.05) and 55.13% of the cohort were male. Five-fold cross-validation showed an area under the receiver operating characteristics curve of 0.88 and area under the precision-recall curve of 0.89. Positive predictive value was 0.80 and negative predictive value was 0.79. Results on the internal independent test cohort and achieved similar performance. Supplementary analyses showed that the model was able to identify NSTE patients with elevated troponin T as well as identified a gap in time to definitive treatment for NSTE patients compared to those with ST elevation (STE). CONCLUSION: Machine learning applied to the prehospital ECG can identify patients at high risk of myocardial injury before biomarker results are available, with potential to streamline patient flow and reduce time to definitive treatment in patients with suspected acute coronary syndrome.