Characterizing the Accuracy of ICD-10 Administrative Claims for Aortic Valve Disease
Background: Administrative claims for aortic stenosis (AS) regurgitation (AR) may be useful but their accuracy and ability to identify individuals at risk for valve-related outcomes has not been well characterized.
Methods: Using echocardiographic (TTE) reports linked to US Medicare claims, 2017-2018, the performance of candidate International Classification of Diseases, 10th Revision (ICD-10) claims to ascertain AS/AR was evaluated. The optimal performing algorithm was tested against outcomes at 1-year after TTE in a separate 100% sample of US Medicare claims, 2017-2019.
Results: Of those included in the derivation (N = 5497, mean age 74.4 ± 11.0 years, 49.7% female), any AS or AR was present in 24% and 38.8%, respectively. The sensitivity and specificity of ICD-10 code I35.0 for identification of any AS was 53.1% and 94.8%, respectively. Amongst those with an I35.0 code, 40.3% had severe AS. Claims were unable to distinguish disease severity (i.e. severe vs. non-severe) or subtype (e.g. bicuspid or rheumatic AS), and were insensitive and nonspecific for AR of any severity. Among all beneficiaries who received a TTE (N = 4,033,844), adjusting for age, sex, and 27 comorbidities, those with an I35.0 code had a higher adjusted risk of all-cause mortality (adjusted hazard ratio [HR] 1.33, 95% CI 1.31-1.34), heart failure hospitalization (adjusted HR 1.37, 95% CI 1.34-1.41), and aortic valve replacement (adjusted HR 34.96, 95% CI 33.74-36.22).
Conclusions: Among US Medicare beneficiaries receiving a TTE, ICD-10 claims, though identifying a population at significant greater risk of valve-related outcomes, failed to identify nearly half of individuals with AS and were unable to distinguish disease severity or subtype. These results argue against the widespread use of ICD-10 claims to screen for patients with AS and suggest the need for improved coding algorithms and alternative systems to extract TTE data for quality improvement and hospital benchmarking.