Abstract 16090: Human Chart Abstraction Combined With Electronic Medical Record Programmed Queries Improves Identification of Patients With Cardiac Implanted Electronic Devices

Abstract

Background The Society of Thoracic Surgery cardiac surgery database (STSD) is used for quality improvement and research. The STSD relies on human chart abstraction (STS-HA) to accurately identify cardiac patients with existing Cardiac Implanted Electronic Devices (CIED’s). Perioperative CIED management may impact surgical outcomes, yet the sensitivity of the STS-HA for CIED patients is unknown. Methods All 7 major STS cardiac surgeries performed at Johns Hopkins Hospital (JHH) and Temple University Hospital (TUH), between 7/2017 and 6/2022, were assessed for the presence of a CIED. CIED patients were identified using both STS-HA and a programmed EMR (Epic) reports (EMR-pro) relying on CPT codes, problem lists, structured data elements, and CIED interrogation reports. The presence of existing CIED in patients was confirmed by same-day chest imaging, which is not uniformly searchable using EMR-pro. McNemar’s chi square test was used to compare the sensitivity of STS-HA with EMR-pro to identify preoperative patients with existing CIED’s. Results Using the STSD, we identified 4770 surgical patients across both institutions during the study period (JHH n= 2627, TUH n=2143), of whom 167 patients had CIED’s (3.5%; JHH 77 (2.9%), TUH 90 (4.2%). Of these, STS-HA identified 144 (86.2% pooled sensitivity) as having existing CIED’s (JHH 69, 89.6%; TUH 75, 83.3%). EMR-pro identified 164 patients (98.2% pooled sensitivity; JHH 77, 100%; TUH 87, 96.7%). Each of the 3 TUH patients identified by STS-HA but not by the EMR-pro had CIED documentation in paper chart scans. The pooled sensitivity of EMR-pro was better than that of STS-HA in identification of CIED’s at the time of surgery (p<0.0001). Discussion The programmed EMR algorithm’s superior sensitivity to human abstraction in identifying cardiac surgical patients with existing CIED’s is an important enhancement in identifying this population. This improvement in sensitivity illustrates the power of EMR programmed reports to address the unreliability of human chart abstraction. Understanding the inherent limitations of the STS database and improving it are essential to the interpretation of past STSD based research and improving the reliability of STSD based studies performed in the future.