Abstract
Rationale
Despite clear guideline recommendations, few patients with pulmonary hypertension (PH) are referred to expert care, including high-risk patients with pulmonary arterial hypertension (PAH) and chronic thromboembolic pulmonary hypertension (CTEPH). Characterizing population-level care networks for patients with PH may inform understanding of referral patterns and help improve PH care quality.
Objectives
We leveraged social network analysis to characterize existing provider networks for patients with PH and to identify provider-level factors associated with connections to PH experts.
Methods
We linked patient-level data from the Massachusetts All-Payer Claims Database with provider-level data to identify all adults diagnosed with PH in 2014-2018 and all relevant providers who saw those patients for PH. We constructed provider networks among all patients with PH (“All-PH network”) and subsets of patients with risk factors for PAH or CTEPH (“PAH network” and “CTEPH network,” respectively). Our provider-level outcome was connection to PH experts, defined as sharing at least 1 patient with an expert. Within each network, we conducted multivariable regression models to determine the association between provider-level variables (specialty, practice location, PH panel volume) and our outcome.
Results
We identified 4766 providers and 8970 patients with PH, of whom 1768 (19.7%) had PAH risk factors and 2164 (24.1%) had CTEPH risk factors. Few providers shared patients with PH experts (31% All-PH network, 35% PAH network, 19% CTEPH network). Within the All-PH network, primary care providers had 59% decreased odds of PH expert connection compared to nonexpert pulmonologists (adjusted odds ratio, 0.41 [95% confidence interval, 0.32-0.51]). Providers practicing outside the greater Boston area and those with smaller PH panel volumes were also less likely to be connected to a PH expert. Findings were similar among the PAH and CTEPH networks.
Conclusions
We found significant gaps in connections to expert care, even among providers caring for patients at risk for PAH or CTEPH, which may be driven in part by limited provider experience, geographic barriers, and disconnected providers and care networks. Multifaceted strategies may be needed to improve referral rates for patients with PH.
Keywords: pulmonary hypertension, health services accessibility, community networks, social network analysis
Introduction
Pulmonary hypertension (PH) diagnosis and management is increasingly challenging, with evolving hemodynamic definitions of the disease, development of new drug classes, and expanding indications for PH therapies.1,2 The need for specialist or even PH expert involvement in PH care is increasingly vital, as primary care providers and nonexpert specialists often do not have the requisite knowledge, expertise, or resources to care for this complex patient population.3,4
Guidelines recommend that, at a minimum, patients with certain PH subtypes including those with suspected or confirmed pulmonary arterial hypertension (PAH) or chronic thromboembolic pulmonary hypertension (CTEPH), those with high-risk features or warning signs (eg, signs of right heart failure), and those with lung disease with severe PH be promptly referred to dedicated PH care centers, where multidisciplinary teams with specific PH expertise can offer the full array of diagnostic and therapeutic modalities (Table S1).2,5,6 Indeed, prior work has shown that outcomes for patients with PAH are improved within PH centers compared to non-PH center care.7 Yet, few patients with PH are seen by PH experts or within PH centers, with significant patient-level geographic, economic, and social barriers to expert referral.8,9
Whether additional provider-specific characteristics influence referral to PH experts for patients with PH is not well known. To understand care patterns for PH care and to evaluate the association between provider characteristics and PH expert connections, we applied social network analysis10 to identify existing provider networks for patients with PH within Massachusetts, 1 of few US states that has multiple PH centers and maintains an All-Payer Claims Database (APCD), allowing construction of population-level care networks. We hypothesized that provider specialty and practice location would be associated with PH expert connections.
Methods
Overview
A social network is defined as a set of “nodes” (representing subjects of interest) and “edges” (representing connections between nodes). Using patient-level claims data linked to provider-level data, we constructed networks of providers caring for patients with PH in Massachusetts, with providers representing the nodes in the network and edges representing shared patients between respective providers (Figure 1).
Figure 1.
Schematic representation of provider networks. Provider–patient encounters are collapsed onto a single network, in which providers are nodes in the network, and connections or “edges” between nodes are shared patients. In this schematic of 4 providers (labeled A through D) and 5 patients (numbered 1 through 5), edge colors represent the actual patient shared between providers (eg, the purple edge is patient 1). Edges are weighted based on number of shared patients. For example, the edge weight between providers A and B is 1, while the edge weight between providers B and D is 3.
Data sources to identify study cohorts
The Massachusetts APCD is the most comprehensive source of health claims in Massachusetts that includes claims from all commercial payers, self-insured employers, Medicaid, and Medicare and captures more than 90% of the state’s population.11 The APCD provides patient demographics and claims data for Massachusetts residents and employees and includes encounters that occurred both within and outside Massachusetts. The APCD also includes provider National Provider Identification (NPI) numbers for each claim, allowing linkages to provider databases. Release 8.0 (2014-2018) was the most current version of the APCD at study inception.
The National Plan and Provider Enumeration System (NPPES) is a publicly available database that includes provider-level information based on NPI numbers, including provider specialty (based on taxonomy codes) and practice location.
Patient cohort and patient-level variables
We used a previously validated approach (sensitivity = 28%; specificity = 100%)12 to identify all adults (age ≥18 years) with PH between January 1, 2014 and December 31, 2018, defined as at least 2 visits (either inpatient or outpatient) linked to an International Classification of Diseases, Ninth Revision (416.x) or Tenth Revision (I27.x) PH diagnosis code. We required that all patients have continuous enrollment in a health plan for 12 months prior to and after the first PH diagnosis code (index date) and have at least 1 PH-related encounter with a provider in the network during the study period. We identified patients with a risk factor for PAH (ie, those with a preceding PAH-associated diagnosis such as connective tissue disease in the 12 months prior to the index date) and CTEPH (ie, those with a preceding diagnosis of acute or chronic thromboembolism in the 12 months prior to the index date) (Figure 2A; Table S2). We captured patient demographics, comorbidities, and PH visit details.
Figure 2.
Derivation of patient (A) and provider (B) study cohorts. Abbreviations: APCD, All-Payer Claims Database; CTEPH, chronic thromboembolic pulmonary hypertension; PAH, pulmonary arterial hypertension; PH, pulmonary hypertension.
Provider cohort and provider-level variables
We identified all healthcare providers who saw at least 1 patient from the patient sample above for a PH visit. To exclude spurious contacts with providers, we limited the sample to specialties (based on taxonomy codes) who commonly care for patients with PH, including primary care providers, cardiologists, pulmonologists, and rheumatologists. We included all providers regardless of practice location as we wanted to capture PH care occurring both in and out of the state of Massachusetts. We then identified subsets of providers who cared for patients with risk factors for PAH and CTEPH (Figure 2B; Table S2). We captured provider specialty, practice location, details of PH patient panel, and whether the provider was a PH expert. Our approach to identifying PH experts has been previously described.9 In brief, for PH experts within Massachusetts and bordering states (Connecticut, New Hampshire, New York, Rhode Island, Vermont), we (1) gathered names of self-reported PH experts through the publicly available Pulmonary Hypertension Association (PHA) website,13 and (2) emailed leaders at health centers in these states known locally to have PH expertise and asked them to identify PH experts who practiced at their center during the study period. For the minority of providers in our cohort practicing outside Massachusetts or bordering states (n = 530 [11.1%]), we identified PH experts from the PHA website.
Network construction and analyses
We constructed the PH care network in Massachusetts for all patients with PH and providers who saw those patients (termed “All-PH network”) based on patient–provider encounters as above (Figure 1). As patients with suspected or confirmed PAH or CTEPH have a stronger indication for PH expert referral, we created 2 subsets of the network limited to patients with risk factors for PAH and CTEPH (Table S2) and providers who saw those specific patients (termed “PAH network” and “CTEPH network,” respectively).
Our primary provider-level outcome was connection to a PH expert, defined as sharing at least 1 patient with PH with an expert (ie, presence of an “edge” between the provider and a PH expert in the network). Within each of the 3 networks (All-PH network, PAH network, and CTEPH network), we determined the association between provider characteristics (including specialty, practice location, and PH panel volume) and our outcome using multivariable logistic regression. We mapped practice locations based on zip codes to Massachusetts regions (Figure 3). For PH panel volume, we assessed for the assumption of linearity using the Box-Tidwell test14 and found that the variable violated this assumption. For this variable, we took 2 approaches: (1) we created categories of PH panel volume and (2) we used restricted cubic splines with knots placed at quintiles.15 To depict the results based on cubic splines, we report effect estimates for the association between PH panel volume and PH expert connection across a range of panel volumes.
Figure 3.
Massachusetts regions. Map courtesy of Commonwealth of Massachusetts.
We used UCINET and Netdraw software16 to create and visualize the networks and SAS statistical software, version 9.4 (SAS Institute, Cary, NC, USA) to conduct statistical analyses. In the figure representation of the networks, we excluded edge weights of 1 (ie, 2 providers who share only 1 patient with PH) to visualize networks more clearly. The Boston University Institutional Review Board approved this study (#H-41419).
Results
Patient and provider study cohorts
We identified 8970 patients with PH in the All-PH network (median age, 73 [IQR, 62-82] years; 53 788 [60.0%] women), of whom 1768 (19.7%) had a preceding diagnosis of a PAH-associated condition and 2164 (24.1%) had a preceding diagnosis of acute or chronic thromboembolism (Table 1). Among the All-PH network, nearly a quarter of patients (22.1%) saw only primary care providers for PH, with a minority (7.8%) seeing a PH expert. In the subnetwork of patients at risk for PAH, 17.0% saw only primary care providers for PH and 13.6% saw a PH expert. In the subnetwork of patients at risk for CTEPH, even more patients (39.1%) saw only a primary care provider for PH, with few patients (8.4%) seeing a PH expert.
Table 1.
Patient and provider characteristics among network of all patients with PH (All-PH network) and networks of patients at risk for PAH and CTEPH.
| Characteristic | All-PH network | PAH network | CTEPH network |
|---|---|---|---|
| Patient characteristics | N = 8970 | N = 1768 | N = 2164 |
| Age at PH diagnosis, y, median (IQR) | 73.0 (62.0-82.0) | 68.0 (57.0-79.0) | 68.0 (55.0-78.0) |
| Female sex | 5378 (60.0) | 1182 (66.9) | 1252 (57.9) |
| Number of outpatient PH visits over study period, median (IQR) | 3.0 (2.0-5.0) | 3.0 (2.0-6.0) | 3.0 (2.0-6.0) |
| Number of unique providers seen for PH over study period, median (IQR) | 4.0 (3.0-6.0) | 4.0 (3.0-7.0) | 4.0 (3.0-6.0) |
| PH provider specialties | |||
| Saw a PH expert | 701 (7.8) | 241 (13.6) | 182 (8.4) |
| Saw a nonexpert specialist | 6286 (70.1) | 1226 (69.3) | 1137 (47.4) |
| Saw primary care provider only | 1983 (22.1) | 301 (17.0) | 845 (39.1) |
| Comorbidities | |||
| Cardiac disease | 6275 (70.0) | 1261 (71.3) | 1119 (51.7) |
| Obstructive lung disease | 3401 (37.9) | 688 (38.9) | 785 (36.3) |
| Interstitial lung disease | 3145 (35.1) | 661 (37.4) | 683 (31.6) |
| Provider characteristics | N = 4766 | N = 2097 | N = 2400 |
| Specialty | |||
| Cardiologists | 973 (20.4) | 619 (29.5) | 532 (22.2) |
| Pulmonologists | 545 (11.4) | 343 (16.4) | 322 (13.4) |
| Rheumatologists | 71 (1.5) | 54 (2.6) | 18 (0.8) |
| Primary care providers | 3177 (66.7) | 1081 (51.6) | 1528 (63.7) |
| PH experta | 41 (0.9) | 34 (1.6) | 35 (1.5) |
| Practice location | |||
| Greater Boston area | 1126 (23.6) | 598 (28.5) | 661 (27.5) |
| Northeast Massachusetts | 296 (6.2) | 139 (6.6) | 150 (6.3) |
| Southeast Massachusetts | 571 (12.0) | 227 (10.8) | 284 (11.8) |
| MetroWest Massachusetts | 938 (19.7) | 418 (19.9) | 474 (19.8) |
| Central Massachusetts | 501 (10.5) | 230 (11.0) | 272 (11.3) |
| Western Massachusetts | 451 (9.5) | 192 (9.2) | 246 (10.3) |
| Neighboring state (CT, NY, NH, RI, VT) | 312 (6.6) | 97 (4.6) | 124 (5.2) |
| Other state | 571 (12.0) | 196 (9.4) | 189 (7.9) |
| Number of unique patients with PH seen during study period, median (IQR) | 2.0 (1.0-4.0) | 1.0 (1.0-2.0) | 1.0 (1.0-2.0) |
Data are presented as No. (%) unless otherwise specified.
Abbreviations: CT, Connecticut; CTEPH, chronic thromboembolic pulmonary hypertension; IQR, interquartile range; NH, New Hampshire; NY, New York; PAH, pulmonary arterial hypertension; PH, pulmonary hypertension; RI, Rhode Island; VT, Vermont.
Additional PH experts identified in the All-PH network but not identified in the PAH or CTEPH networks (n=7) were predominantly providers practicing outside Massachusetts.
We identified 4766 providers in the All-PH network: 3177 (66.7%) primary care providers, 973 (20.4%) cardiologists, 545 (11.4%) pulmonologists, and 71 (1.5%) rheumatologists. There were 41 PH experts, representing 1 PH expert for every 220 patients with PH (Table 1). The plurality of providers practiced within the greater Boston area. In the PAH and CTEPH networks, there was a slightly higher proportion of specialists compared to the All-PH network, with a greater proportion practicing in the greater Boston area.
Association between provider characteristics and connection to PH experts
Among the All-PH network, few (31.3%) providers shared patients with a PH expert. Provider specialty, practice location, and PH panel volume were all associated with connection to a PH expert. Both primary care providers (adjusted odds ratio [aOR], 0.41 [95% CI, 0.32-0.51]) and nonexpert cardiologists (aOR, 0.60 [95% CI, 0.47-0.77]) had decreased odds of connection to PH experts compared to nonexpert pulmonologists. Likewise, providers with practice locations outside the greater Boston area all had decreased odds of PH expert connection compared to those practicing within Boston, with the greatest effect seen for practice locations in Central Massachusetts, where there are no PH experts (aOR, 0.13 [95% CI, 0.10-0.17]) (Table 2). As expected, larger patient panels were associated with higher odds of PH expert connection (aOR, 1.93 [95% CI, 1.65-2.26] for providers with 2-4 patients vs 1 patient on their panel; aOR, 4.65 [95% CI, 3.86-5.61] for providers with ≥5 patients vs 1 patient on their panel). Based on cubic splines, the association between provider PH panel volume and expert connection was greatest at low panel volumes, was attenuated as the panel volume increased, and disappeared above a threshold of 16 patients (Figure S1). For example, an additional patient on a provider’s panel at a volume of 1 patient was associated with 80% increased odds of PH expert connection (aOR, 1.80 [95% CI, 1.48-2.17]), while an additional patient on a provider’s panel at a volume of 16 patients was only associated with 4% increased odds of PH expert connection (aOR, 1.04 [95% CI, 1.01-1.06]).
Table 2.
Association between provider specialty and practice location and connection to a PH expert among entire provider network and networks of providers caring for patients at risk for PAH and CTEPH.
| Provider characteristics | Odds ratio (95% CI) |
||
|---|---|---|---|
| All-PH network | PAH network | CTEPH network | |
| Specialty | |||
| Cardiologistsa | 0.60 (0.47-0.77) | 0.68 (0.50-0.93) | 0.83 (0.60-1.15) |
| Pulmonologistsa | Ref | Ref | Ref |
| Rheumatologists | 1.09 (0.62-1.90) | 0.88 (0.45-1.71) | 1.08 (0.37-3.12) |
| Primary care providers | 0.41 (0.32-0.51) | 0.64 (0.47-0.85) | 0.33 (0.24-0.45) |
| Practice location | |||
| Greater Boston area | Ref | Ref | Ref |
| Northeast Massachusetts | 0.23 (0.17-0.32) | 0.22 (0.14-0.34) | 0.22 (0.13-0.40) |
| Southeast Massachusetts | 0.23 (0.18-0.30) | 0.14 (0.10-0.21) | 0.28 (0.19-0.42) |
| MetroWest Massachusetts | 0.40 (0.33-0.49) | 0.30 (0.23-0.40) | 0.54 (0.40-0.73) |
| Central Massachusetts | 0.13 (0.10-0.17) | 0.10 (0.06-0.15) | 0.12 (0.07-0.20) |
| Western Massachusetts | 0.28 (0.22-0.37) | 0.33 (0.23-0.48) | 0.26 (0.16-0.40) |
| Neighboring state (CT, NY, NH, RI, VT) | 0.53 (0.40-0.71) | 0.72 (0.46-1.14) | 0.88 (0.56-1.40) |
| Other state | 0.22 (0.17-0.29) | 0.22 (0.15-0.33) | 0.34 (0.21-0.54) |
| PH panel volume | |||
| 1 patient | Ref | Ref | Ref |
| 2-4 patients | 1.93 (1.65-2.26) | 2.10 (1.69-2.61) | 1.66 (1.32-2.08) |
| ≥5 patients | 4.65 (3.86-5.61) | 4.38 (2.98-6.44) | 2.77 (1.85-4.14) |
Abbreviations: CI, confidence interval; CT, Connecticut; CTEPH, chronic thromboembolic pulmonary hypertension; NH, New Hampshire; NY, New York; PAH, pulmonary arterial hypertension; PH, pulmonary hypertension; RI, Rhode Island; VT, Vermont.
Non-PH expert cardiologists and pulmonologists.
Among the PAH and CTEPH networks, 735 (35.0%) and 461 (19.2%) providers shared patients with a PH expert, respectively. The patterns of associations between provider characteristics and PH expert connection were similar to the All-PH network for provider specialty and practice location. Primary care providers were less likely to share patients with a PH expert compared to nonexpert pulmonologists (aOR, 0.64 [95% CI, 0.47-0.85] for PAH network; aOR, 0.33 [95% CI, 0.24-0.45] for CTEPH network; Table 2). Nonexpert cardiologists were less likely to share patients with a PH expert compared to nonexpert pulmonologists in the PAH network but not the CTEPH network. Providers with practice locations outside the greater Boston area (except for providers with practice locations in neighboring states) all had decreased odds of PH expert connection compared to those within Boston. Larger panel volumes were associated with increased odds of PH expert connection, though with smaller effect sizes seen for the CTEPH network versus the PAH network. Like the All-PH network, the association between panel volume and connection to an expert was attenuated with increasing panel size, though the threshold was lower compared to the All-PH network, with the association disappearing above a volume of 1 and 2 patients within the PAH and CTEPH networks, respectively (Figure S1).
Network visualization
Visually, the All-PH network with edge weights ≥2 consisted of a core component of providers inclusive of most of the PH experts, a second component disconnected from the main component with fewer PH experts, and many isolated providers (providers not connected to any other providers; shown aligned at the left of the graph) (Figure 4). Patient sharing patterns tended to follow geographic boundaries, with less sharing between regions of the state or across state lines. The PAH and CTEPH networks were more disconnected than the All-PH network, with many small clusters of providers, some focused around PH experts and others not. As with the All-PH network, patient sharing tended to follow geographic boundaries in these subnetworks as well.
Figure 4.
Graphical depiction of provider networks in Massachusetts for patients with pulmonary hypertension (PH). Nodes represent providers and connections between providers represent shared patients. Nodes arranged at the left of the graph are isolated providers not connected to any other provider. Nodes are colored by geographic region and shaped by provider specialty (square nodes = pulmonologists; circle nodes = cardiologists; diamonds = primary care providers; triangle nodes = rheumatologists). Large nodes represent PH experts and small nodes represent nonexperts. Edge weights of 1 were excluded to more clearly visualize the networks. (A) Network for all patients with PH. (B) Network for patients at risk for pulmonary arterial hypertension (PAH) (diagnosis of a PAH-associated condition prior to PH diagnosis). (C) Network for patients at risk for chronic thromboembolic pulmonary hypertension (patients with a diagnosis of acute or chronic thromboembolism prior to PH diagnosis)
Discussion
In this population-level study, we found that provider specialty influences connections to PH experts, with primary care providers being far less likely to be connected to PH experts, including primary care providers caring for patients at risk for PAH or CTEPH, patients for whom referral to expert care is strongly advised.2 Furthermore, we found that many patients, including those at risk for PAH or CTEPH, are cared for solely by primary care providers with no contact with PH-relevant specialists. In a disease like PH that requires a nuanced understanding of the pathophysiology and navigation of complex diagnostic and treatment algorithms, this represents a significant gap in care and a missed opportunity to bolster referral networks and connect patients with specialists. Indeed prior studies have identified significant PH-specific knowledge gaps among board-certified cardiologists and pulmonologists, gaps that are likely even more pronounced among primary care providers.3
Relatedly, we found that provider experience with PH may also influence referral rates regardless of specialty, with higher PH panel volumes being associated with increased likelihood of PH expert connections. This finding is in line with prior literature showing improved quality and patient outcomes with increasing provider caseloads across a range of specialties and contexts,17,18 and is also consistent with our prior qualitative work,19 in which patients and providers alike identified lack of PH knowledge among nonexperts (including primary care providers and nonexpert cardiologists and pulmonologists) as a key driver of diagnostic and referral delays.
We also found that there may be geographic barriers to referring patients with PH to expert care, with providers practicing outside the greater Boston area being less likely to share patients with PH experts and with patient sharing occurring generally within the same geographic area. PH care in rural and other underserved areas (such as Cape Cod in Massachusetts) remains an area of significant concern as the vast majority of PH centers are located within urban counties.20 Indeed, 14 US states currently do not have a PH center and there are immense areas of the United States, particularly in the Midwest and West, with no reasonable access to expert PH care.21,22 It is likely that geographic barriers to PH expert referral are insurmountable in these underserved areas.
Importantly, we found that the rate of PH expert connection among providers and the proportion of patients cared for solely by primary care providers did not increase substantially in the PAH network compared to the All-PH network and actually decreased in the CTEPH network, suggesting that current referral patterns may not align with patient risk profiles or guideline recommendations. While guidelines are clear on indications for referral for certain subgroups of patients with PH (Table S1), the “correct” referral rate for other patients, including those with moderate PH secondary to left heart or chronic lung disease and those with World Health Organization Group 5 PH, remains unclear. Further guidance for primary care providers and nonexpert specialists is needed. This is particularly salient as PH experts are a limited resource. While we found 1 PH expert for every 220 patients with PH in Massachusetts, this ratio is likely much lower in states with few PH experts and the capacity to manage PH referrals in these areas is likely much more limited. Although efforts to improve the pipeline of providers with PH expertise are underway,23 with the rising incidence and identification of patients with PH and expanding indications for PH-specific treatment, shortages of PH experts will likely worsen.24
Improving referral rates to expert PH care by addressing geographic barriers to referrals and bolstering nonexpert PH knowledge will likely require a multifaceted, carefully designed approach that does not overburden PH experts and referral centers. Expanding use of technology such as provider-to-patient telehealth visits within expert centers may help reach rural patients.25 Likewise, establishing provider-to-provider electronic consults may increase accessibility to PH experts, streamline referrals, and help identify patients that can be appropriately managed outside of expert centers, thus decreasing strain on PH centers.26 Importantly, the use of telehealth for PH is currently understudied and more work is needed to identify the appropriate application and evaluate outcomes of telehealth in PH.27
Building telementoring models such as Project ECHO (Extension for Community Healthcare Outcomes) could also help disseminate expert knowledge and best practices to community providers, improving primary care knowledge about PH and quality of PH care.28 Such mentoring models would help facilitate appropriate and timely referrals by disseminating clear referral guidelines and ensuring that adequate screening tests such as echocardiograms are performed prior to referral. Indeed, Project ECHO has been successfully applied in other complex diseases such as cancer and interstitial lung disease, resulting in higher-quality care.29,30 Finally, the use of artificial intelligence in PH is a promising area to improve referral rates. For example, AI could be used to detect patients at risk for PH based on readily available health record data and prompt providers to consider referral.31
The care patterns seen in this study are likely not unique to PH. Indeed, challenges with delayed referrals and barriers to accessing expert care have been well documented in other complex pulmonary diseases such as interstitial lung disease and sarcoidosis.32–34 Identifying provider networks using social network analysis for other diseases may likewise lend clarity to the drivers of delayed care in other contexts.
Our study has limitations. First, we restricted our study to Massachusetts in order to leverage the APCD to conduct an inclusive population-level analysis. However, the care patterns seen in this study may not be applicable to all states, and repeating this work in other states (particularly more rural states and states without PH centers) with all-payer claims databases would add further insight into our findings. Second, we chose a PH definition to maximize specificity at a cost to sensitivity. Thus, it is likely that we did not capture all patients with PH in Massachusetts and that provider networks for the entire PH population are even sparser and connections to PH experts are even lower. Third, the APCD provides only 5-year cuts of data, limiting the longitudinal assessment of provider connections. We required that all patients have 12 months of continuous enrollment after PH diagnosis to identify follow-up care; however, it is possible that patient sharing with PH experts occurred after the end of our study period that we were unable to identify. Fourth, the granularity of our data, including lack of hemodynamic data or prescription data, did not allow us to assess for severity of PH or PH phenotypes to confirm if patients met guideline criteria for expert referral. For example, it is likely that many patients in the All-PH network have PH secondary to left heart disease or chronic lung disease given the high prevalence of comorbid heart and lung disease, for whom expert referral recommendations are less clear. Fifth, we did not have additional provider information such as which healthcare system providers practiced within, years of practice, or prescribing practices, which may further influence decision making around referrals. Finally, while it is likely that a connection between a nonexpert provider and PH expert represents a referral, our social network analysis did not allow us to confirm this.
In summary, we found significant gaps in connections to expert PH care among providers in Massachusetts, which may be driven by limited provider knowledge of the disease, geographic barriers to care, and limited access to PH experts. Comprehensive changes on the provider, healthcare center, and health policy level may be needed to address these care gaps and improve quality of care and outcomes for patients living with PH.
Supplementary Material
Acknowledgments
We would like to thank Emily Sisson and Rabindra Kadel for their assistance in data cleaning.
Contributor Information
Kari R Gillmeyer, Email: kari.gillmeyer@va.gov, Center for Health Optimization and Implementation Research, VA Boston Healthcare System, Boston, MA, United States; The Pulmonary Center, Boston University Chobanian and Avedisian School of Medicine, Boston, MA, United States.
Seppo T Rinne, Center for Health Optimization and Implementation Research, VA Bedford Healthcare System, Bedford, MA, United States; Department of Medicine, Geisel School of Medicine at Dartmouth, Lebanon, NH, United States.
Elizabeth S Klings, The Pulmonary Center, Boston University Chobanian and Avedisian School of Medicine, Boston, MA, United States.
A Rani Elwy, Center for Health Optimization and Implementation Research, VA Bedford Healthcare System, Bedford, MA, United States; Department of Psychiatry and Human Behavior, Warren Alpert Medical School of Brown University, Providence, RI, United States.
Renda Soylemez Wiener, Center for Health Optimization and Implementation Research, VA Boston Healthcare System, Boston, MA, United States; The Pulmonary Center, Boston University Chobanian and Avedisian School of Medicine, Boston, MA, United States.
Author contributions
Study concept and design: All authors. Acquisition of data: K.R.G. Analysis and interpretation of data: All authors. Drafting of the manuscript: K.R.G. Critical revision of the manuscript for important intellectual content: All authors. K.R.G. had full access to all of the data and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Supplementary material
Supplementary material is available at Annals of the American Thoracic Society online.
Conflicts of interest
Please see the ICMJE disclosure forms, which have been provided as supplementary material. E.S.K. receives research support from Novartis, Pfizer, Novo Nordisk, and United Therapeutics. She has served as a consultant/advisory board member for Novo Nordisk, Pfizer, and CSL Behring for sickle cell disease–related clinical trials, unrelated to the present work.
Funding
This work was supported by the Parker B. Francis Fellowship Program; a Department of Veterans Affairs Career Development Award (HSR CDA 22-140); the Doris Duke Foundation; the National Center for Advancing Translational Sciences (NCATS), National Institutes of Health (NIH), through BU-CTSI (1UL1TR001430); and in part with resources from the VA Boston and VA Bedford Healthcare Systems. E.S.K. is supported by the NIH/National Heart, Lung, and Blood Institute 1UG3 HL143192-01A1, NCATS 2UL1TR001430-05A1, and Health Resources and Services Administration U1EMC27864-08-00. A.R.E. is supported by a Department of Veterans Affairs Research Career Scientist Award (RCS 23-081).
Disclaimer
The views expressed in this article do not necessarily represent the views of the Department of Veterans Affairs or the United States government.
Artificial intelligence disclaimer
No artificial intelligence tools were used in writing this manuscript.
Data availability
The data underlying this article were provided by The Center for Health Information and Analysis (CHIA) by permission. Data will be shared on request to the corresponding author with permission of CHIA.
References
- 1. Johnson S, Sommer N, Cox-Flaherty K, Weissmann N, Ventetuolo CE, Maron BA. Pulmonary hypertension: a contemporary review. Am J Respir Crit Care Med. 2023;208:528-548. 10.1164/rccm.202302-0327SO [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Humbert M, Kovacs G, Hoeper MM, et al. ESC/ERS Scientific Document Group. 2022 ESC/ERS guidelines for the diagnosis and treatment of pulmonary hypertension: developed by the Task Force for the Diagnosis and Treatment of Pulmonary Hypertension of the European Society of Cardiology (ESC) and the European Respiratory Society (ERS). endorsed by the International Society for Heart and Lung Transplantation (ISHLT) and the European Reference Network on Rare Respiratory Diseases (ERN-LUNG.). Eur Heart J. 2022;43:3618-3731. 10.1093/eurheartj/ehac237 [DOI] [PubMed] [Google Scholar]
- 3. de Belen E, McConnell JW, Elwing JM, et al. Gaps in the care of pulmonary hypertension: a cross-sectional patient simulation study among practicing cardiologists and pulmonologists. J Am Heart Assoc. 2023;12:e026413. 10.1161/JAHA.122.026413 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Sahay S, Melendres-Groves L, Pawar L, Cajigas HR; Pulmonary Vascular Diseases Steering Committee of the American College of Chest Physicians. Pulmonary Hypertension Care Center Network: improving care and outcomes in pulmonary hypertension. Chest. 2017;151:749-754. 10.1016/j.chest.2016.10.043 [DOI] [PubMed] [Google Scholar]
- 5. McLaughlin VV, Shah SJ, Souza R, Humbert M. Management of pulmonary arterial hypertension. J Am Coll Cardiol. 2015;65:1976-1997. 10.1016/j.jacc.2015.03.540 [DOI] [PubMed] [Google Scholar]
- 6. Frost A, Badesch D, Gibbs JSR, et al. Diagnosis of pulmonary hypertension. Eur Respir J. 2019;53:1801904. 10.1183/13993003.01904-2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Pi H, Kosanovich CM, Handen A, et al. Outcomes of pulmonary arterial hypertension are improved in a specialty care center. Chest. 2020;158:330-340. 10.1016/j.chest.2020.01.046 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Leary PJ, Le RJ, Panjabi S, et al. Location of care delivery for pulmonary arterial hypertension in the United States. Am J Respir Crit Care Med. 2025;211:619-627. 10.1164/rccm.202404-0816OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Gillmeyer KR, Rinne ST, Rucci JM, Klings ES, Elwy AR, Wiener RS. Factors associated with referral to expert providers among patients with pulmonary hypertension. Ann Am Thorac Soc. 2025;22:838-845. 10.1513/AnnalsATS.202408-901OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. DuGoff EH, Fernandes-Taylor S, Weissman GE, Huntley JH, Pollack CE. A scoping review of patient-sharing network studies using administrative data. Transl Behav Med. 2018;8:598-625. 10.1093/tbm/ibx015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Center for Health Informatics and Analysis. Massachusetts All-Payer Claims Database. Accessed May 9, 2025. https://www.chiamass.gov/ma-apcd/
- 12. Wijeratne DT, Housin A, Lajkosz K, et al. Using health administrative data to identify patients with pulmonary hypertension: a single center, proof of concept validation study in Ontario, Canada. Pulm Circ. 2022;12:e12040. 10.1002/pul2.12040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Pulmonary Hypertension Association. Doctors who treat pulmonary hypertension. Accessed June 10, 2025. https://phassociation.org/patients/doctorswhotreatph/
- 14. Box GEP, Tidwell PW. Transformation of the independent variables. Technometrics. 1962;4:531-550. 10.1080/00401706.1962.10490038 [DOI] [Google Scholar]
- 15. Schuster NA, Rijnhart JJM, Twisk JWR, Heymans MW. Modeling non-linear relationships in epidemiological data: the application and interpretation of spline models. Front Epidemiol. 2022;2:975380. 10.3389/fepid.2022.975380 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Borgatti S, Everett M, Freeman L. UCINET 6 for Windows: Software for Social Network Analysis. Analytic Technologies; 2002. [Google Scholar]
- 17. Ashy N, Nguyen T-N, Denhaerynck K, et al. Hierarchical modeling of patient and physician determinants of blood pressure outcomes in hypertensive patients with and without diabetes: pooled analysis of six observational valsartan studies with 15,282 evaluable patients. Int J Chronic Dis. 2017;2017:9842450. 10.1155/2017/9842450 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Khatana SAM, Fiorilli PN, Nathan AS, et al. Association between 30-day mortality after percutaneous coronary intervention and education and certification variables for New York state interventional cardiologists. Circ Cardiovasc Interv. 2018;11:e006094. 10.1161/circinterventions.117.006094 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Gillmeyer KR, Shusterman S, Rinne ST, Elwy AR, Wiener RS. Gaps in access to pulmonary hypertension care and opportunities for improvement: a multi-site qualitative study. BMC Pulm Med. 2025;25:355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Issa R, Minhas AMK, Issa R, et al. Urban-rural disparities in pulmonary hypertension–related mortality between 2004 and 2019: a call to improve access to specialty care centers for rural residents in the United States. Curr Probl Cardiol. 2023;48:101623. 10.1016/j.cpcardiol.2023.101623 [DOI] [PubMed] [Google Scholar]
- 21. Gillmeyer KR, Johnson SW. Pulmonary hypertension care delivery in the United States: is it time to get with the guidelines? Am J Respir Crit Care Med. 2025;211:552-554. 10.1164/rccm.202501-0107ED [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Pulmonary Hypertension Association. PH care centers—accredited centers. Accessed June 10, 2025. https://phassociation.org/phcarecenters/accredited-centers/
- 23. Kudelko K, Zamanian RT, De Jesus Perez VA. Pulmonary vascular fellowship training to promote excellence in PH clinical care: the Stanford perspective. 2020. Accessed June 10, 2025. https://www.acc.org/latest-in-cardiology/articles/2020/09/21/19/13/pulmonary-vascular-fellowship-training-to-promote-excellence-in-ph-clinical-care
- 24. Wijeratne DT, Lajkosz K, Brogly SB, et al. Increasing incidence and prevalence of World Health Organization groups 1 to 4 pulmonary hypertension: a population-based cohort study in Ontario, Canada. Circ Cardiovasc Qual Outcomes. 2018;11:e003973. 10.1161/circoutcomes.117.003973 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Butzner M, Cuffee Y. Telehealth interventions and outcomes across rural communities in the United States: narrative review. J Med Internet Res. 2021;23:e29575. 10.2196/29575 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Vimalananda VG, Orlander JD, Afable MK, et al. Electronic consultations (E-consults) and their outcomes: a systematic review. J Am Med Inform Assoc. 2020;27:471-479. 10.1093/jamia/ocz185 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Li B, Gillmeyer KR, Molloy-Paolillo B, et al. Scoping review of pulmonary telemedicine consults: current knowledge and research gaps. Ann Am Thorac Soc. 2023;20:456-465. 10.1513/AnnalsATS.202205-404OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Osei-Twum J-A, Wiles B, Killackey T, Mahood Q, Lalloo C, Stinson JN. Impact of Project ECHO on patient and community health outcomes: a scoping review. Acad Med J Assoc Am Med Coll. 2022;97:1393-1402. 10.1097/ACM.0000000000004749 [DOI] [PubMed] [Google Scholar]
- 29. ILD Collaborative. Project ECHO for interstitial lung disease. Accessed May 9, 2025. https://www.ildcollaborative.org/echo
- 30. Arora S, Brakey HR, Jones JL, Hood N, Fuentes JE, Cirolia L. Project ECHO for cancer care: a scoping review of provider outcome evaluations. J Cancer Educ. 2023;38:1509-1521. 10.1007/s13187-023-02292-x [DOI] [PubMed] [Google Scholar]
- 31. Young JA, Chang C-W, Scales CW, Menon SV, Holy CE, Blackie CA. Machine learning methods using artificial intelligence deployed on electronic health record data for identification and referral of at-risk patients from primary care physicians to eye care specialists: retrospective, case-controlled study. JMIR AI. 2024;3:e48295. 10.2196/48295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Pritchard D, Adegunsoye A, Lafond E, et al. Diagnostic test interpretation and referral delay in patients with interstitial lung disease. Respir Res. 2019;20:253. 10.1186/s12931-019-1228-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Johannson KA, Lethebe BC, Assayag D, et al. Travel distance to subspecialty clinic and outcomes in patients with fibrotic interstitial lung disease. Ann Am Thorac Soc. 2022;19:20-27. 10.1513/AnnalsATS.202102-216OC [DOI] [PubMed] [Google Scholar]
- 34. Gerke AK, Judson MA, Cozier YC, Culver DA, Koth LL. Disease burden and variability in sarcoidosis. Ann Am Thorac Soc. 2017;14:S421-S428. 10.1513/AnnalsATS.201707-564OT [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article were provided by The Center for Health Information and Analysis (CHIA) by permission. Data will be shared on request to the corresponding author with permission of CHIA.