Abstract 16120: Efficacy of an Electronic Medical Record Alert System in Identifying Cardiac Implantable Electronic Device Infection

Abstract

Introduction: Cardiac Implantable Electronic Device infection (CIED-I) is common, morbid, and frequently under recognized. Electronic Medical Record (EMR) based alerts have been proposed as a solution to improve detection of CIED-I. Hypothesis: Development of a real-time EMR algorithm to identify CIED-I can improve recognition of patients with CIED-I. Methods: An EMR (Epic, Verona, WI) alert system to identify inpatients with CIED-I was developed based on the presence of CIED and positive blood cultures. The presence of CIED was determined using patient problem list, structured data elements describing device implantation, and prior device interrogation. Alerts were generated each time a patient with CIED had a positive blood culture. We examined patients admitted between January 2020 - March 2023. Identified patients were adjudicated manually to confirm: (1) the presence of CIED and (2) true bacteremia versus contaminant based on patient and microbiological characteristics. Results: The algorithm generated 255 alerts, corresponding to 100 unique patients over 142 hospitalizations. Of these alerts, 213 appropriately identified 88 unique patients with positive blood cultures and CIEDs, resulting in an alert specificity of 84% and patient specificity of 88%. The remaining 42 (16.6%) alerts occurred in 10 patients who no longer had CIED due to prior CIED extraction (6 at the time of heart transplant and 4 with extraction without reimplantation) as well as 2 patients with erroneous blood culture identification (one positive syphilis antibody test and one lab documentation error). In patients with appropriate alerts, 27 patients had 4/4 positive blood culture bottles, 5 patients had 3/4, 10 patients had 2/4, 41 patients had 1/4, 2 patients had 2/2, and 1 patient had 1/2. The positive predictive value of the algorithm to identify true bacteremia was 0.65 (138/213). Conclusions: The algorithm has high specificity to identify patients with possible CIED infection. Improving the accuracy of device detection, by incorporating radiology imaging or history of extraction, could further improve specificity. Positive predictive value and alert fatigue could be improved by stratifying alerts by blood culture species and total number of positive blood cultures.