Abstract 11645: Use of an Electronic Health Record to Identify Implantable Cardiac Rhythm Device Lead Failure

• Published in: Circulation (New York, N.Y.) Vol. 144; no. Suppl_1; p. A11645
• Main Authors: Brock, Jonathan, Bundy, Richa, Rosenthal, Gary, Rodgers, Larisa, Bhave, Prashant
• Format: Journal Article
• Language: English
• Published: Lippincott Williams & Wilkins 16-11-2021
• ISSN: 0009-7322; 1524-4539
• DOI: 10.1161/circ.144.suppl_1.11645

IntroductionElectronic health records (EHR) offer the potential to facilitate research examining the real-world effectiveness and safety of medical interventions. However, few studies have used EHR data to evaluate medical devices. This study sought to devise a method using current procedural terminology (CPT) and international classification of disease (ICD) codes captured in EHR data to identify patients with implantable cardiac rhythm device lead failures. HypothesisCPT and ICD diagnosis codes within an EHR can be used to correctly identify implantable cardiac rhythm device lead failures. MethodsStudy data were extracted from the EHR of a large North Carolina health system. Patients who had implantable cardiac rhythm devices implanted during January 2013 to December 2018 were identified using predetermined CPT codes. Patients were followed longitudinally to identify those undergoing subsequent lead insertions or removals through December 2019. Medical records were reviewed to determine reasons for the re-insertion or removal and identify true lead failures. Random forest modeling was used to develop an algorithm for predicting true lead failure using CPT and ICD codes for mechanical breakdown or cardiac device complication. Patients with potential lead failure were split into two cohorts60% for model training and 40% for testing. ResultsA total of 4,148 encounters with billed CPT codes for implantable cardiac rhythm devices were initially identified. After applying study exclusion criteria, 2,390 patients met study inclusion criteria. Of those, 175 patients had a subsequent insertion indicating a potential lead failure. A total of 31 patients were found to have true lead failures. A random forest algorithm predicted true lead failures with good discrimination, achieving an AUROC (area under receiver operating curve) of 0.908. The model accurately detected 66% of true lead failure cases in the testing dataset. ConclusionsApplying a specific combination of CPT and diagnosis codes to a cohort of patients undergoing subsequent implantable cardiac rhythm device procedures following an index insertion can correctly identify patients with lead failure with strong accuracy and moderate sensitivity.