Aernoud Fiolet

345 Accuracy of using routinely collected electronic healthcare data to identify cardiovascular endpoints events per 100 person-years using investigator-reported endpoints; Figure 3). Similar agreement was observed for MACE+ (2.9 events per 100 person-years using EHR data retrieval and 3.3 events per 100 person-years using investigator- reported endpoints). Figure 2: Sensitivity and specificity of endpoint identification using EHR data retrieval. The primary endpoint (MACE) was the composite of cardiovascular death, myocardial infarction, or ischemic stroke. The secondary endpoint (MACE+) was the composite of cardiovascular death, myocardial infarction, ischemic stroke, or ischemia-driven coronary revascularization. The formula for sensitivity is: true positive / (false negative + true positive). The formula for specificity is: true negative / (false positive + true negative). The whiskers of the points depict the 95% confidence interval. Abbreviations: FN: false negative; FP: false positive; TN: true negative; TP: true positive; MACE: major adverse cardiovascular event Cumulative incidence rates for all-cause mortality did not differ between both methods of collection (0.48 events per 100 person-years). Similar agreement was found for the cumulative incidence rates for myocardial infarction (1.4 events per 100 person-years using EHR data retrieval and 1.5 events per 100 person-years using investigator-reported endpoints) and ischemic stroke (0.24 events per 100 person-years using EHR data retrieval and 0.29 events per 100 person-years using investigator-reported endpoints). Cumulative incidence rates for ischemia-driven coronary revascularization were estimated lower using EHR retrieval (1.8 events per 100 person-years) as compared to the investigator-reported endpoints (2.3 events per 100 person-years).