IP017. Performance of Current Claims-Based Approaches to Identify Aortic Dissection Hospitalizations

Abstract

Methods that use claims data to identify patients with aortic dissection (AD) for study are not well validated. We aimed to evaluate the accuracy of previously reported diagnosis and procedural coding mechanisms to identify AD admissions and treatment types for patients with acute AD at our institution. All inpatient hospitalizations at a single center between 2000 and 2015 were retrospectively reviewed for an AD diagnosis code (International Classification of Diseases, Ninth Revision and Tenth Revision) through institutional billing databases. Diagnoses were verified using medical records and imaging studies. Characteristics of the aortic disease, treatment modalities, and procedural codes were abstracted. Sensitivity and specificity analyses were performed to test the ability of diagnosis and procedural codes to correctly identify the anatomic involvement of the aorta and treatment type for acute AD (TASR, type A open surgery repair; TBSR, type B open surgery repair; TEVAR, thoracic endovascular repair; MM, medical management). Diagnosis codes identified 1697 hospitalizations among 1312 patients. AD was present in 1295 (76%) visits. In 281 (17%) visits, no aortic disease was present, and 121 (7%) were for intramural hematoma or penetrating aortic ulcer without AD. Among visits with AD, 438 (34%) were first-time visits for acute AD, 11 (<1%) were readmissions for acute AD, 112 (9%) were for subacute AD, 620 (48%) were for chronic AD, and 114 (9%) were for AD of unknown age. There were 223 (51%) acute AD visits that ranked AD among the top three discharge diagnoses, and 202 (46%) did not rank AD at all. Stanford type A AD was present for 260 (59%) acute visits and Stanford type B for 178 (41%). Acute AD management included 229 (52%) open surgery, 28 (6%) endovascular surgery, 10 (2%) open and endovascular surgery, and 171 (39%) medical. Diagnosis coding demonstrated sensitivities and specificities for aortic involvement as follows: thoracic (sensitivity, 46%; specificity, 75%), thoracoabdominal (sensitivity, 22%; specificity, 98%), and abdominal (sensitivity, 65%; specificity, 98%). Procedural coding had low sensitivities: TASR (sensitivity, 29%; specificity, 98%), TBSR (sensitivity, 0%; specificity, 100%), TEVAR (sensitivity, 9%; specificity, 100%), MM (sensitivity, 46%; specificity, 97%). After exclusion of cases with diagnosis of aneurysm, sensitivities and specificities were minimally changed: TASR (sensitivity, 25%; specificity, 99%), TBSR (sensitivity, 0%; specificity, 100%), TEVAR (sensitivity, 9%; specificity, 100%), and MM (sensitivity, 45%; specificity, 98%). At our institution, one in four inpatient hospitalizations coded for AD demonstrated no evidence of dissection. Current billing stratification methods to determine type of acute AD and treatment modality have poor sensitivities but are highly specific. To more effectively use claims data to study AD, more sensitive methods are necessary.