Abstract 191: Identification of Adults With Congenital Heart Disease From Administrative Data

Abstract

Introduction: With improved management, there are now more adults with congenital heart disease (ACHD) than children. However, long-term survival with moderate or severe ACHD remains limited, and there is relatively sparse literature on the intensity or quality of end-of-life care for these patients. The goal of this analysis is to examine the accuracy of administrative data for identifying patients who died with ACHD to facilitate study of care provided near the end of life. Methods: We created a list of ICD-9 and ICD-10 codes representing ACHD of moderate or great complexity. We performed a search for these codes in the electronic health record (EHR) of adults who received care 2010-2016 within our healthcare system. We used state death records to identify which of these patients died during the same timeframe. Manual EHR review was completed to evaluate performance of this search strategy. Identified patients were also compared to a list of patients seen in our ACHD clinic and known to have died during 2010-2016. Results: Using ICD data, 121 patients were identified, of which 66 actually had the moderate or greater complexity ACHD conditions by EHR review (positive predictive value, 0.55; 95% confidence interval 0.45, 0.63). EHR review confirmed 12 patients with Eisenmenger Syndrome, for which there is no specific ICD code. “Cyanosis+other” did not identify any of these, “VSD+other” (ventricular septal defect) identified 6, and there were 6 whose only ACHD code was VSD. Of the remaining 55 patients, 24 had ACHD not on the targeted list, largely due to coding error. In addition, despite being coded as having ACHD, 31 patients had no identified ACHD on EHR review. These misidentifications were attributed to coding error for 15 patients. Another 11 patients (35%) had acquired VSD due to myocardial infarction or endocarditis, for which there is no separate ICD code. Codes with the highest degree of error, incorrect more than 50% of the time, were those for congenitally corrected transposition, endocardial cushion defect, and hypoplastic left heart syndrome. The list of known deceased clinic patients included 21 with ACHD of interest. Only 1 of these was not identified by the ICD search, yielding a sensitivity for our list of ICD codes in this small sample of 0.95 (0.77,0.99). Conclusion: Use of administrative data to identify patients with ACHD of moderate or great complexity who have died had good sensitivity but suboptimal positive predictive value. Strategies to improve accuracy can be employed. Excluding patients who have codes for myocardial infarction or endocarditis in addition to VSD and using “VSD+other” as an additional proxy for Eisenmenger Syndrome are two examples. Administrative data is not ideal for identification of patients with ACHD of moderate or great complexity who have died, and manual EHR review is necessary to confirm these diagnoses.