Aernoud Fiolet

321 Text-mining in electronic healthcare records can be used for screening and data collection disease [found as a coronary angiography; CT coronary angiography or Coronary Artery Calcium Score]: n = 38; no known renal function: n = 41; date of previous Coronary Artery Bypass unknown: n = 41). Characteristics of missed participants did not differ substantially from identified participants (median difference of all variables 1.6%, IQR 3.1%); values were therefore assumed to be missing at random. Table 1. Number of patients visiting, eligible, and enrolled per participating medical center Medical center Enrollment period Total no. of patients visiting No. of trial participants (%) No. of patients automatically identified as potentially eligible (% of all visiting patients) No. of trial participants identified as potentially eligible (%; % of participants) Hospital A February 1, 2017– October 1, 2018 51,943 169 (0.3) 10,705 (20.6) 151 (1.4; 89.3) Hospital B July 1, 2017– October 1, 2018 14,206 69 (0.5) 2,966 (20.9) 65 (2.2; 94.2) Hospital C October 1, 2016– December 1, 2018 26,317 330 (1.3) 4,932 (18.7) 252 (5.1; 76.4) Total 92,466 568 (0.7) 18,603 (20.1) 468 (2.5; 82.4) Data collection accuracy Of the 568 trial participants, 540 (95.1%) enrolled trial participants were automatically retrieved on their trial identification number or unique on-site identifier with the data mining tool. On an aggregate level, availability of baseline characteristics for participants using automated EHR text-mining differed by 2.8% (median; IQR across all variables 0.4– 8.5%) with manually collected trial data (Table 2; center-specific distributions are presented in Supplement 2a). Notably larger differences between automated EHR text-mining data and manually collected trial data were found for hypertension (26.2%), antiplatelet therapy (29.1%), and beta-blocker use (24.4%).