Performance of current claims-based approaches to identify aortic dissection hospitalizations

To describe index visits for acute aortic dissection (AD) to an academic center and validate the prevailing claims-based methodology to identify and stratify them.

Inpatient hospitalizations at a single center assigned an International Classification of Diseases, Ninth Revision (ICD-9) diagnosis code for AD from January 2005 to September 2015 were identified. Diagnoses were verified by review of medical records and imaging studies. All visits were secondarily stratified with the algorithm based on ICD-9 codes. Sensitivity and specificity analyses were conducted to evaluate the ability of the algorithm to correctly identify acute AD by Stanford class and treatment modality (type A open repair [TAOR], type B open repair [TBOR], thoracic endovascular repair [TEVAR], medical management [MM]).

In the study interval, there were 1245 visits coded for AD attributed to 968 unique patients. Chart review verification demonstrated that the majority of visits were for AD (79%; n = 981), of which 32% (n = 310) were for an index acute AD event. The true distribution of acute AD visit classifications was TAOR (46.1%; n = 143), TBOR (5.2%; n = 16), TEVAR (7.7%; n = 24), and MM (39.4%; n = 122). The algorithm, which used ICD-9 codes, identified 631 acute visits and stratified them as TAOR (27.1%; n = 171), TBOR (4.1%; n = 26), TEVAR (4.9%; n = 31), and MM (63.9%; n = 403). Analyses demonstrated high specificities, but generally low sensitivities of the algorithm (TAOR: sensitivity, 58%, specificity, 92%; TBOR: sensitivity, 13%, specificity, 98%; TEVAR: sensitivity, 17%, specificity, 98%; MM: sensitivity, 73%, specificity, 72%).

The prevalent claims-based strategy to identify hospitalizations with acute AD is specific, but lacks sensitivity. Caution should be exercised when studying AD with ICD-9 codes and improvements to existing claims-based methodologies are necessary to support future study of acute AD.