Reply to Tung and Chang and to Suntharalingam et al.

From the Authors:

We thank Tung and Chang and Suntharalingam and colleagues for their interest in our work (1). Tung and Chang highlight three potential limitations of using claims data for population-based analyses: misdiagnosis, proxy variable bias, and underdiagnosis or underreferral. Suntharalingam and colleagues propose a centrally organized, networked approach as one solution to standardize access to specialist pulmonary arterial hypertension (PAH) care in the United States, referencing the United Kingdom model. We appreciate these excellent discussion points and welcome the opportunity to respond.

We share Tung and Chang’s first concern about the potential for disease misclassification within claims data. For this reason, we specifically chose not to study the “appropriateness” of therapy, which our dataset was not well suited to address. Rather, the question we sought to answer was where therapy intended to treat PAH is being prescribed, irrespective of diagnostic accuracy. Questions about the appropriateness of diagnosis or treatment across locations are indeed important and merit future research using alternative approaches. However, regardless of the diagnostic accuracy of PAH, claims data provide insight into where therapeutic decisions are being made. Regarding their second point, we agree that there is some potential for misclassification if patients continue to receive PAH-focused prescriptions from other physicians while receiving care at a center. However, given the stringent insurance approval process for PAH medications in the United States, we anticipate that the fraction of misclassified patients and data would be minimal, especially for patients followed at centers who are on more than monotherapy. It would be uncommon for all of a patient’s prescriptions to be issued in the community in this case. We agree with and share the final concern regarding underdiagnosis or underreferral of PAH in general, which, although not a topic we could address in our study, remains an important consideration for improving access to PAH care centers.

Suntharalingam and colleagues provide a promising example of how intentional structuring of care delivery can improve access to specialized PAH care. This is exactly the kind of discussion and sharing of experience needed to understand and improve PAH care delivery in the United States and beyond. A key challenge to translating this approach across the Atlantic is that both the “halo effect” they mention, as well as our proximity analysis, suggest that optimizing the number of centers by geographic area may be critical. They describe 7 core centers and 33 shared care centers serving England and Scotland. To make this work in the United States, with approximately 50 times the geographic area of England and Scotland combined, we would need 350 core centers and 1,650 shared care centers (2, 3). Currently, the Pulmonary Hypertension Association has almost 100 accredited centers as a mix of comprehensive and regional care centers. Increasing the geographic density to that described in England and Scotland may improve access to care, but the logistics of expanding this network and ensuring high-quality care throughout would be expensive and daunting and may diffuse expertise. Nevertheless, the United Kingdom example shows that clustering care within centers is feasible and may improve access. It is possible that telemedicine and other technologies may help overcome geographic barriers to care in the United States, potentially reducing the number of centers needed.

In summary, we thank the authors of both replies for their thoughtful contributions to important discussions around how to optimize care for patients with PAH.