Performance of a prehospital HEART score in patients with possible myocardial infarction: a prospective evaluation

Abstract

IntroductionThe History, Electrocardiogram (ECG), Age, Risk Factors and Troponin (HEART) score is commonly used to risk stratify patients with possible myocardial infarction as low risk or high risk in the Emergency Department (ED). Whether the HEART score can be used by paramedics to guide care were high-sensitivity cardiac troponin testing available in a prehospital setting is uncertain.MethodsIn a prespecified secondary analysis of a prospective cohort study where paramedics enrolled patients with suspected myocardial infarction, a paramedic Heart, ECG, Age, Risk Factors (HEAR) score was recorded contemporaneously, and a prehospital blood sample was obtained for subsequent cardiac troponin testing. HEART and modified HEART scores were derived using laboratory contemporary and high-sensitivity cardiac troponin I assays. HEART and modified HEART scores of ≤3 and ≥7 were applied to define low-risk and high-risk patients, and performance was evaluated for an outcome of major adverse cardiac events (MACEs) at 30 days.ResultsBetween November 2014 and April 2018, 1054 patients were recruited, of whom 960 (mean 64 (SD 15) years, 42% women) were eligible for analysis and 255 (26%) experienced a MACE at 30 days. A HEART score of ≤3 identified 279 (29%) as low risk with a negative predictive value of 93.5% (95% CI 90.0% to 95.9%) for the contemporary assay and 91.4% (95% CI 87.5% to 94.2%) for the high-sensitivity assay. A modified HEART score of ≤3 using the limit of detection of the high-sensitivity assay identified 194 (20%) patients as low risk with a negative predictive value of 95.9% (95% CI 92.1% to 97.9%). A HEART score of ≥7 using either assay gave a lower positive predictive value than using the upper reference limit of either cardiac troponin assay alone.ConclusionsA HEART score derived by paramedics in the prehospital setting, even when modified to harness the precision of a high-sensitivity assay, does not allow safe rule-out of myocardial infarction or enhanced rule-in compared with cardiac troponin testing alone.

Selective classification with machine learning uncertainty estimates improves ACS prediction: A retrospective study in the prehospital setting.

“…Modifying the ECG SNR 23 , ECG distribution shift 4,5 ), no guarantee exists these predictions will not be random or overconfident. Accordingly, as mentioned before, we encourage keeping the ECG interpreter in the loop 21 until predictive performance is at desired levels 3 . Whilst we showcase performance improvements, the work is still limited by various factors.…”

“…Accordingly, the community has developed Machine Learning (ML) methods to improve the prediction of ACS with prehospital covariates 1 , 2 . Nevertheless, performance remains insufficient for safe rule-out or rule-in of ACS 3 . Current research in cardiovascular disease detection from ECG has observed a possible trade-off between performance and coverage (i.e.…”