The utility of the nationwide readmissions database in understanding contemporary transcatheter aortic valve replacement outcomes

This commentary refers to ‘Is the National Readmission Database the right database to identify valve-in-valve TAVR patients?’, by S Nazir et al., 2020;41:4357. We thank the authors for their interest in our article, which compared real-world outcomes between valve-in-valve transcatheter aortic valve replacement (VIV-TAVR) and re-operative surgical aortic valve replacement (Re-SAVR).2 In an effort to understand the optimal treatment options for patients with failed bioprotheses, this study used the Nationwide Readmissions Database (NRD), a robust administrative claims database that enables national assessment of hospital inpatient stays/readmissions among patients of all ages with all types of insurance payer. Overseen by the Agency for Healthcare Research and Quality, the NRD represents the largest inclusive source of longitudinal information about US hospital stays. As such, it was chosen as the best choice to study the entire patient population in question and follow them forward over time. It is true that, like most data sources based on billing codes including clinical registries, the NRD is subject to the limitations of International Classification of Diseases (ICD-9/10-CM) diagnostic descriptions and events that are able to be captured during hospital stays. We appreciate Nazir et al.1 drawing attention to these issues. To address them, we utilized a carefully chosen combination of ICD-9/10-CM diagnosis codes to identify ‘bioprosthesis failure’ and ‘aortic valve disease’.2 Consistent with expectations for the US population, between 2012 and 2016, a nationally weighted total of 2.91 million hospital admissions for aortic valve disease were identified as were 478 522 instances of a failed bioprosthesis. Utilization of the codes in combination identified 46 061 admissions with documented reporting of both. Of these, 6815 admissions met additional inclusion/exclusion criteria. We captured 3443 VIV-TAVR patients, but since we excluded concomitant percutaneous coronary intervention and VIV procedures during same admission, endocarditis patients, and given the uncertainty and possibly under-capturing of patients based on national-weighting, we expect our numbers to be in line albeit lower than those reported in the Society of Thoracic Surgeons (STS)/American College of Cardiology Transcatheter Valve Therapy Registry (n = 4358).3 Similarly, we reported 3372 Re-SAVRs, which are keeping with estimates (n = 3688) extrapolated from our STS Adult Cardiac Surgery Database analysis of 2213 isolated repeat aortic valve replacement with bioprosthetic valve from 2011 to 2013.4 While it is possible that some cases could represent complications of other bioprotheses concomitant with aortic valve disease, they are unlikely to be the majority of patients or to have their inclusion meaningfully alter the results. The 30-day mortality definition measured from the day of a patient’s index hospital admission is one commonly employed in health services research.5 Akin to many such studies involving clinical registries, it is not possible to track outside hospital deaths e.g. deaths which occur following discharge alive and prior to readmission. But due to the inclusion of index hospital deaths, the results are not ‘30-day readmission mortality’. It is, however, worth noting that incomplete mortality reporting is a potential limitation of the paper as is the reality that such numbers represent consistent findings with the majority of single-centre and multicentre investigations on this topic to date6—many of which suffer from the same limitation when data are not able to be death-certificate linked. Conflict of interest: T.K. is a speaker for Edwards Life Sciences, Medtronic, Abbott and Baylis Medical and is a consultant for 4C Medical. There are no other potential conflicts that exist.