A machine learning algorithm to predict a culprit lesion after out of hospital cardiac arrest.

ABSTRACT

BACKGROUND: We aimed to develop a machine learning algorithm to predict the presence of a culprit lesion in patients with out-of-hospital cardiac arrest (OHCA). METHODS: We used the King's Out-of-Hospital Cardiac Arrest Registry, a retrospective cohort of 398 patients admitted to King's College Hospital between May 2012 and December 2017. The primary outcome was the presence of a culprit coronary artery lesion, for which a gradient boosting model was optimized to predict. The algorithm was then validated in two independent European cohorts comprising 568 patients. RESULTS: A culprit lesion was observed in 209/309 (67.4%) patients receiving early coronary angiography in the development, and 199/293 (67.9%) in the Ljubljana and 102/132 (61.1%) in the Bristol validation cohorts, respectively. The algorithm, which is presented as a web application, incorporates nine variables including age, a localizing feature on electrocardiogram (ECG) (≥2 mm of ST change in contiguous leads), regional wall motion abnormality, history of vascular disease and initial shockable rhythm. This model had an area under the curve (AUC) of 0.89 in the development and 0.83/0.81 in the validation cohorts with good calibration and outperforms the current gold standard-ECG alone (AUC 0.69/0.67/0/67). CONCLUSIONS: A novel simple machine learning-derived algorithm can be applied to patients with OHCA, to predict a culprit coronary artery disease lesion with high accuracy.