Background

Pulmonary hypertension (PH) is a condition characterized by elevation in pulmonary artery pressures of varying etiologies, which may result in substantial morbidity. PH is clinically classified into five groups, defined by the World Symposium on PH, according to differing pathological findings between groups, including underlying cause of disease, clinical presentation, and hemodynamic characteristics [1]. PH clinical classifications are also used by healthcare providers for medical management and by the Food and Drug Administration (FDA) for labeling of new drugs approved for the treatment of PH [1, 2].

Group 3 PH describes a subpopulation of patients with PH due to chronic lung disease and/or hypoxia. In particular, chronic obstructive pulmonary disease (COPD) and interstitial lung disease (ILD) are two large subgroups of chronic lung disease patients who often develop PH [2, 3]. In COPD, the prevalence rate ranges from 30 to 70% [4]. Because ILD is composed of multiple lung diseases, its prevalence is difficult to estimate. In idiopathic pulmonary fibrosis (IPF), the most common type of ILD, prevalence rates of PH range from 8 to 15% at initial diagnosis to 46% at evaluation for lung transplant and 86% at the time of transplant [5,6,7,8,9,10]. Wide ranges in prevalence can be due to heterogeneity in the definitions for PH, diagnostic modalities for PH, differences in patient populations, physiologic characteristics, and severity of underlying lung disease [4, 11].

Administrative claims are generated following healthcare utilization for the purposes of payment. Medical documentation is converted to standardized codes using uniform coding systems. The International Classification of Diseases (ICD), a medical classification system, is the international standard for reporting diseases and health conditions. In the US, the ICD, 9th Revision, Clinical Modification (ICD-9-CM) and ICD, 10th Revision, Clinical Modification (ICD-10-CM) provide a system of diagnostic codes assigned for each encounter.

ICD-9-CM diagnostic codes provide the level of detail to indicate PH but are not specific to groups of PH defined by World Symposium on PH. All groups of PH due to lung disease are generally coded under the same four-digit codes. The newer ICD-10-CM, implemented in October 2017, provides five-digit codes for PH with greater differentiation but was not required for billing and reimbursement until October 2018 [12]. The designated code for Group 3 PH is I27.23, and it remains to be seen how frequently and accurately the utilization of this code will be.

While collected for billing purposes, claims data can provide real-world evidence outside the setting of a clinical trial about the treatment patterns, risk factors, patient outcomes, healthcare resource utilization, and costs for patients covered by health insurance plans. However, using diagnostic codes in claims databases for the identification of patients with Group 3 PH associated with lung disease conditions like COPD and ILD is not sufficient. Additional considerations must be applied to patient identification to reduce misclassification and improve the usefulness of claims data to understand the medical management and health-related outcomes of patients with these conditions.

Recently, a focused review provided recommended algorithms for the identification of patients with pulmonary arterial hypertension (Group 1 PAH) [13]. Others have explored the use of machine-learning approaches to reduce selection bias in patient identification [14]. Our goal is to provide readers with information to determine the most appropriate methodology for claims-based patient selection under different types of research questions about Group 3 PH associated with COPD or ILD.

Methods

A panel of US-based healthcare providers and researchers with expertise in PH was convened to discuss methodologies to identify patients with Group 3 PH associated with COPD or ILD in retrospective claims databases. Panel members included US-based practicing pulmonologists (n = 2), a nurse (n = 1), pharmacists (n = 6) with expertise in PH and backgrounds in public health and/or claims-based analysis, and researchers (n = 2) with expertise in pharmacoeconomics, epidemiology, and claims-based analysis.

We conducted a literature review to identify studies and explore considerations when using claims-based data to identify patients with Group 3 PH associated with COPD or ILD. The recommendations in this article are those of the authors convened for the discussion and are based on group consensus.

The literature review utilized EMBASE and MEDLINE (via EMBASE) to identify English language articles published from 2010 through June 2020, on adult patient population, including both US and international studies. We looked for claims-based studies, retrospective studies, or healthcare management-related studies, focused on both PH and lung diseases causing Group 3 PH, or that specifically mentioned Group 3 PH. The search terms for lung diseases and PH were derived from several sources [3, 15] and decided upon by the panel. Search terms were required in the title or abstract (Supplemental Table 1). We also referred to a similar effort around conceptualization of Group 1 PAH in order to provide further support [13].

Additionally, we reviewed abstracts from 2018 through 2020 from the American Thoracic Society, the American College of Chest Physicians, and the Pulmonary Vascular Research Institute to identify relevant studies.

Results

The broad literature search strategy resulted in 2,646 potential observational studies in patients with Group 3 PH associated with COPD or ILD (Supplemental Table 1). The targeted title and abstract review identified 11 studies and 2 conference abstracts for claims-based studies focusing on Group 3 PH associated with COPD or ILD; 8 studies based in the US and 5 non-US (Table 1) (Supplemental Fig. 1).

Table 1 Administrative claims-based studies of group 3 PH associated with COPD or ILD in the literature

The focus of these studies varied. Four studies evaluated measures across multiple PH groups or Group 3 PH in general [16,17,18,19]. There were 3 studies on patients with COPD [20,21,22]: 2 studies each in ILD [23, 24] and IPF [25, 26] and 1 study in systemic sclerosis classified as Group 3 PH [27]. One study looked at PH in ILD, COPD, and combined ILD and COPD [28].

Most studies (n = 9) claimed to be identifying Group 3 PH and included steps in their methodology to filter for these patients (such as ensuring that PH occurred after COPD/ILD or excluded non-Group 3 PH) [16,17,18,19,20,21,22, 27, 28]. Other studies took additional steps in their methodology but did not specifically claim to identify Group 3 PH [26] or reported PH as a comorbidity to the lung disease [23,24,25].

In the following discussion section, we outline key components to consider in the development of an algorithm for Group 3 PH associated with COPD or ILD. For each, we outline the findings from the literature search that relate to the component followed by our summary and recommendations.

Discussion

There are several components to consider when choosing an algorithm. The methods used in the published literature for identifying adult patients with Group 3 PH associated with COPD or ILD relied on the following considerations: (a) identification of COPD or ILD, (b) identification of PH, (c) defining the sequence between COPD/ILD and PH diagnoses, and (d) other PH groups or other Group 3 PH exclusions. We provide a summary of the published literature findings followed by considerations as to the impact of how restricting or relaxing the criteria for each of these components can impact the diagnostic performance of the algorithm. Components of the recommended algorithm are provided in Fig. 1.

Fig. 1
figure 1

Recommended algorithm components – components of the recommended algorithm. COPD chronic obstructive pulmonary hypertension, ILD interstitial lung disease, LD lung disease, PH pulmonary hypertension

Identification of COPD or ILD

Findings in the Literature

The criteria used for the identification of COPD and ILD within the examined studies included (a) diagnosis codes, (b) relevant procedures, and (c) claims for medication. All 13 studies used diagnosis codes to identify patients with COPD or ILD, with 11 studies reporting ICD-9-CM codes [16,17,18,19,20,21,22, 24,25,26, 28] and 5 reporting ICD-10-CM codes [19, 20, 23, 27, 28].

Except for one COPD study conducted in the US that required the COPD diagnosis to be in the primary position [21], the remaining studies allowed the COPD or ILD diagnosis codes to be in either the primary or secondary position. Twelve studies identified patients from either inpatient or outpatient claims [16,17,18,19,20,21,22,23, 25,26,27,28] and the one remaining study used inpatient claims only [24]. In addition, 6 of the 12 studies using outpatient claims included a requirement for multiple claims (i.e., ≥ 2 claims) with a specified diagnosis [17, 20,21,22, 26, 28].

Only two studies utilized procedures specific to COPD or ILD to identify patients (i.e., pulmonary function tests in patients evaluated for ILD). Lautsch et al. excluded patients with prior lung transplant, while Frank et al. required patients to have at least one diagnostic procedure including bronchoscopy, lung computerized tomography (CT), pulmonary function testing, or assessment of autoantibodies [20, 23]. Two additional studies utilized procedures specific to COPD or ILD in sensitivity or subset analyses, including lung biopsy and CT [25, 26]. Requiring a procedure for COPD or ILD identification resulted in a significant drop in case count but did not change the general outcome of the studies.

Summary and Recommendations

After reviewing codes across studies, recommended diagnosis codes for COPD and ILD are provided in Table 2. While the US and non-US studies utilized similar ICD-9-CM codes for COPD and ILD, they differed more frequently when it came to ICD-10-CM. The US studies used codes down to the fifth digit, whereas non-US studies tended to use the higher, more general code down to the fourth digit. Thus, corresponding codes should be verified if international versions of ICD are used.

Table 2 Lung disease diagnosis codes for COPD and ILD

In addition, we recommend searching for COPD or ILD diagnoses in the primary or secondary position. When PH symptom exacerbations are due to underlying COPD or ILD, PH may be the primary reason for utilization diagnosis. We also recommend using both inpatient and outpatient claims unless doing so would affect study objectives (e.g., a study looking at hospital readmissions). When outpatient codes are used, we recommend using ≥ 2 to reduce the likelihood that a single diagnostic claim is used for patient identification.

We also note that PH is a complication of connective tissue disease and can be due to mechanisms other than ILD. Thus, caution should be considered with including connective tissue disease-related codes in the identification of ILD. We specifically recommend limiting diagnostic codes to connective tissue disease with respiratory or lung involvement. We also suggest alternative methods for the assessment of patients whose underlying conditions overlap across PH classification, such as analyzing patients meeting criteria for multiple PH groups separately.

Based on the findings from the evaluated studies, we do not recommend using any COPD/ILD-related procedure codes or medication claims for patient identification.

Identification of PH

Findings in the Literature

The criteria used for the identification of PH within examined studies can be classified into the following: (a) diagnosis codes, (b) relevant procedures, (c) claims for medication, and (d) exclusions. All 13 studies used ICD-9-CM and ICD-10-CM diagnosis codes to identify patients with Group 3 PH associated with COPD or ILD. The majority (n = 11) [16,17,18,19,20,21,22, 24,25,26, 28] contained ICD-9-CM codes for patient identification, while six contained ICD-10-CM codes [19, 20, 22, 23, 27, 28]. No studies required the PH diagnosis to be in the primary position. Twelve studies identified patients from either inpatient or outpatient claims [16,17,18,19,20,21,22,23, 25,26,27,28] and the remaining one study used inpatient claims only [24]. The use of additional criteria such as a requirement for multiple claims was reported in five studies [16, 18, 20, 22, 28].

Only one study utilized PH-related procedures (i.e., right heart catheterization [RHC] or echocardiogram) in identification of patients with PH [16], and no studies required claims for any PH-related medications except Kim et al., which focused on appropriate use of phosphodiesterase-5 inhibitors (PDE5is) [17].

Summary and Recommendations

As all studies included PH-related diagnosis codes and there was general consensus on these codes across studies, we recommend utilizing a limited list of consensus codes (Table 3). Corresponding codes should be verified if international versions of the ICD are used. We recommend that PH diagnosis be in the primary or secondary position, as COPD or ILD associated with PH symptom exacerbation may be reported in the primary position. We also recommend using both inpatient and outpatient claims unless the study objectives are limited to one or the other. When outpatient codes are used, require at least two to reduce the likelihood that a single diagnostic claim is used for patient identification and specify a minimum time frame of at least 30 days for how far apart the outpatient claims need to be in order to qualify.

Table 3 PH diagnosis codes

We generally do not suggest utilizing PH-related procedures. While the guideline-driven practice is to confirm cases by means of RHC, studies suggest that less than two-fifths of patients with PH have an RHC prior to diagnosis [16, 19] within 3 months before or after medication [29] or within 12 months following diagnosis [30]. We also note that claims data indicate if a patient received a diagnostic test but do not include test results, thus utilizing echocardiography, which is non-specific to Group 3 PH, would not improve the sensitivity of patient identification. Thus, we recommend limiting the use of procedure codes for sensitivity analysis in patient identification or to limit the study population when a purer cohort is required. Lastly, we do not suggest utilizing PH-related medications for patient identification, as therapies used in Group 1 PAH are often used on- and off-label for Group 3 PH associated with both COPD and ILD.

The Sequence of the COPD/ILD and PH Diagnoses

Findings in the Literature

In order to confirm that COPD or ILD was a contributing factor to the development of PH, patients must have developed lung disease prior to PH. Eight of the 13 evaluated studies required the COPD or ILD diagnosis code to be prior to the PH diagnosis code [16, 19,20,21,22, 26,27,28], and one study required an underlying cause of PH (not limited to lung disease) to be documented prior to PDE5i prescription [17]. In fact, of the 8 studies that specifically claimed to be studying Group 3 PH, seven of them included this criterion [16, 19,20,21,22, 27, 28].

Summary and Recommendations

We recommend that PH diagnosis occur after the COPD or ILD diagnosis to align with the natural progression of this disease. Patients should be identified and indexed on their first claim with a PH diagnosis code and require a COPD or ILD diagnosis code in the baseline period at least six to 12 months prior to the PH diagnosis.

Identification of Other PH Group and Other Group 3 PH Conditions

Findings in the Literature

When identifying patients with Group 3 PH, five studies used diagnosis codes to identify and exclude non-Group 3 PH patients [16, 17, 20, 22, 28] and two studies used diagnosis codes to separate patients who met criteria for multiple PH groups from those who met only Group 3 PH [18, 19]. In addition to diagnosis codes, two studies utilized claims for PAH-indicated medications [20] and two studies used procedure codes [16, 28] to exclude patients from Group 3 PH. One study used PDE5i guidelines to assign patients with multiple diagnoses to an “appropriate use” group [17]. Algorithm assignment across PH groups was compared to chart abstraction, resulting in a positive predictive value of 86% for possible inappropriate use across groups of PH.

Summary and Recommendations

We recommend methodology to exclude patients with other PH groups prior to their Group 3 PH using diagnosis codes provided in Table 4. We do not recommend excluding patients with other PH conditions that develop after their Group 3 PH diagnosis, as these are relatively uncommon occurrences and may arise from diagnostic workup rather than diagnostic confirmation, as claims data reflect clinical care provided rather than results. We acknowledge that approximately 34% of patients with PH have overlapping diagnoses [19] and that excluding these patients may bias results for some research objectives, and the combination of multiple comorbidities may render patients vulnerable to developing PH and contribute to poor prognosis [19]. Thus, an alternative method would be to analyze patients meeting criteria for multiple PH groups separately or assign patients to a group based on the objectives of the study [17].

Table 4 Other exclusionary PH diagnosis codes

Using diagnosis codes in the identification of Group 3 PH does have some limitations. ICD-9-CM codes do not have specificity of classification for secondary PH, and it was not until ICD-10-CM that diagnosis codes provided for greater clinical classification of secondary PH. In addition, diagnosis codes do not reflect severity of the disease, particularly as they relate to distinguishing between multiple underlying causes of disease. So, for a patient with mild heart disease but severe lung disease that contributes significantly to PH, using real-world claims data may inadvertently misclassify this person as Group 2 PH.

In addition, there are some notable considerations. Given the overlap of secondary PH-related conditions in Group 1 PAH and Group 3 PH, we do not recommend using I27.x codes when developing exclusion criteria. Further, when conducting a study to identify Group 3 PH with COPD only, we recommend all patients with diagnosis codes for connective tissue disease be excluded. Lastly, given the high prevalence of sleep disorder breathing in COPD and ILD, we do not recommend excluding patients with this diagnosis.

None of the examined studies utilized PH-related medications to directly identify Group 3 PH, but one study did exclude patients with evidence of PAH-indicated medications [20].

As inhaled treprostinil is now FDA approved in both Group 1 PAH and Group 3 PH, and it is likely that medications approved for use in Group 1 PAH are being used off-label in Group 3 PH to improve exercise capacity, we do not recommend relying on medication use to identify or exclude patients.

Conclusion

Correctly identifying adult patients with Group 3 PH associated with COPD or ILD in claims-based studies can improve the value of research findings for application in clinical care and population health, the utility of real-world evidence in support of FDA regulatory approvals, and more accurately inform formulary decision-making. When using the recommendations provided herein, care should be taken to consider policy and regulatory changes, such as FDA approvals and updated guidelines and their impact on how patients are identified. While some studies reported that they used validated codes or algorithms for either COPD/ILD or PH, it is important to note that only one reviewed study provided results on the validation of their patient identification algorithm. Future research should be conducted to validate patient identification algorithms, especially the combination of criteria required for the identification of Group 3 PH associated with COPD or ILD.