Safety and efficiency of modified HEART scores without troponin or electrocardiography: an analysis based on two European out-of-hours primary care cohorts

Abstract

Abstract Background Differentiating between acute chest pain of cardiac versus non-cardiac origin poses a diagnostic challenge. Risk stratification tools, such as the HEART (History, ECG, Age, Risk and Troponin) score, have been developed to aid in decision-making. These scores typically rely on troponin or electrocardiography (ECG), often unavailable in (low-resource) community-based healthcare settings. Purpose We evaluated the diagnostic performance of the HEART score in a low-risk setting when one of its five metrics is missing Methods We relied on data from two cohorts, OUT-ACS (n=1,711; for analyses with and without troponin) and TRACE (n=664; for analyses without troponin) of consecutive patients evaluated at primary care emergency clinics in respectively Norway (2016-2018; 47.7% female, median age: 56, 25th-75th percentiles:45-68) and the Netherlands (2017, 56.9% female, median age: 48, 25th-75th percentiles:32-67). We evaluated the HEAR score (all elements except troponin) and the HART score (all elements except ECG). The primary outcome was the diagnostic performance for ruling out myocardial infarction (MI) at index presentation, with a secondary outcome consisting of the composite of all-cause death or MI at 90-day follow-up. Results Among study participants, 61 (3.6%) were diagnosed with an MI in the OUT-ACS and 20 (3.0%) in the TRACE cohort. The percentages of patients deemed at low risk by HEAR (<=3) were 925 (54.1%) and 527 (79.4%) in OUT-ACS and TRACE, respectively. Sensitivity and specificity were 70.5% (95 % confidence interval 57.4-81.5) and 55.0% (52.5-57.4), and 75.0% (50.9-91.3) and 81.1% (77.8-84.0) for HEAR in the two respective cohorts. For the HART score, low risk was found in 696 (40.7%) OUT-ACS participants, with a respective sensitivity and specificity of 96.7% (88.7-99.6) and 42.1% (39.7-44.5), Positive predictive values ranged from 5.5-15.3%, whilst negative predictive values were 98.1-99.1% for HEAR and 99.7% for HART. By comparison, the HEART score determined 639 (37.3%) patients as low risk, had a sensitivity and specificity of 98.4% (91.2-100) and 38.7% (36.3-41.1), and positive and negative predictive values of 5.6% (5.3-5.9) and 99.8% (98.9-100). Similar findings were found for analyses using the secondary outcome. Conclusion HEART score variants in which troponin or ECG are omitted have higher efficiency but at the cost of decreased safety, particularly when leaving out troponin. Therefore, these biomarkers are paramount for adequate safety, despite the overall high negative predictive values in this low-prevalence primary care setting.