Admission for COPD Exacerbation Is Associated with the Clinical Diagnosis of Pulmonary Hypertension: Results from a Retrospective Longitudinal Study of a Veteran Population

ABSTRACT

Patients with chronic obstructive pulmonary disease and pulmonary hypertension (PH–COPD) have an increased risk of hospitalizations and death compared to COPD alone. Identifying PH in COPD is challenging because performing right heart catheterization, the gold standard for PH diagnosis, is invasive and not routinely performed. Clinical characterization of COPD patients at risk who are progressing toward PH will aid therapeutic development at earlier stages of progressively fatal PH–COPD. We studied the records of 5,45,086 patients in a large Veterans Affairs healthcare network (2000–2012) with a primary discharge diagnosis of COPD based on encounters' ICD-9 codes and further stratified into those who received an additional ICD-9 code for a PH diagnosis. Patients with PH–COPD were assigned to one of the four subgroups: those with (a) no history of exacerbation or hospital admissions, (b) history of exacerbations but no hospital admissions, (c) hospital admissions unrelated to COPD and (d) history of COPD exacerbation-related hospital admissions. We also examined the COPD and COPD-PH cohorts for associated comorbidities such as cardiac disease and the presence of obstructive sleep apnea (OSA). A regression analysis revealed that patients with COPD exacerbation-related hospital admissions had 7 × higher risk of having a concomitant clinical diagnosis of PH compared to non-hospitalized patients. COPD-PH patients had higher rates of cardiac comorbidities (89% vs. 66%) and OSA (34% vs. 16%) compared to COPD alone. We conclude that COPD patients hospitalized for COPD exacerbations are at a higher risk for developing PH, and hospitalized COPD patients with cardiac comorbidities and/or OSA should be screened as at-risk population for developing PH.

KEYWORDS:

• Chronic obstructive pulmonary disease
• pulmonary hypertension

Introduction

Pulmonary hypertension (PH) is a condition in which elevated pulmonary artery pressures lead to exertional dyspnea, progressive right heart failure, and often death (1). It is confirmed as having a mean resting pulmonary artery pressure (mPAP) of greater than or equal to 25 mm Hg when measured by right heart catheterization (RHC) (2). PH confers an increased risk of mortality in the veteran population (3).

Chronic obstructive pulmonary disease (COPD) is marked by persistent airflow limitation and chronic inflammation of the airways (4). It is a common and morbid condition: as of 2013, COPD is the third leading cause of death in the United States (5). Patients with COPD can develop PH (PH–COPD). PH–COPD is classified under group 3 PH according to the World Health Organization (WHO) classification scheme (6). Mortality is high in patients with COPD who develop PH (7–11). Additionally, the presence of PH in patients with COPD is a risk factor for the development of acute exacerbations of COPD (12–14).

Identifying COPD patients who have developed PH is difficult due to insufficient diagnostic modalities, poor understanding of the pathophysiology of the disease, and lack of strong predictive models. Symptoms such as dyspnea on exertion are common to both disease processes. Although trans-thoracic echocardiography (TTE) is often used as a screening test for the presence of PH, the use of this study in patients with structural lung disease is not well validated and sometimes inaccurate (15,16). RHC is the gold standard of diagnosis for PH (2), but in practice, this study is not routinely used in patients with COPD (17).

Another factor complicating the diagnosis of PH in COPD is that the development of elevated pulmonary artery pressures is not necessarily associated with the traditional measures of COPD severity, such as airflow limitations. For some patients, the degree of PH in COPD can be mild to moderate in severity, and develop slowly over time (18). In contrast, a subset of patients appears to have severe PH that is “out-of-proportion” to the expected degree of elevation in pulmonary pressures based on lung parenchymal abnormalities (19). Severe PH has even been found in a subset of COPD patients who have imaging showing emphysematous parenchymal changes, but normal spirometry (20). The reason for the development of PH in COPD is not well understood, but factors including chronic hypoxemia, inflammation, and toxicity from nicotine exposure are thought to contribute to the vascular remodeling, endothelial dysfunction, and destruction of capillary beds that are seen in this disease process (21).

Interventions effective in PH do not improve outcomes in PH–COPD and raise safety concerns. It is important to identify patients with PH–COPD not only because they have worse outcomes, but also because they should be studied to see if targeted PH therapies could be beneficial. Research focusing on therapies for PH in patients with COPD is needed (11,22). Our study was designed to identify factors predicting the presence of a clinical diagnosis of PH in patients with COPD.

Methods

This was a retrospective cohort study which longitudinally followed patients from an inception time point to a set end time point. It was approved by the Institutional Review Board of Baylor College of Medicine (H-30464), the Research and Development Committee of Michael E. DeBakey Veterans Affairs Medical Center, and Veterans Health Administration Corporate Data Warehouse.

Patients

All patients were members of the South Central Veteran's Affairs Healthcare Network (VISN 16) subset of the Veterans Health Administration Corporate Data Warehouse during fiscal years (FY) 2000–2012. Patients who received their care at multiple VA facilities during the study period were excluded. This database was queried using ICD-9 codes (refer to Table 1 in the supplement for the ICD-9 codes used in this study). Patients were included in the study either by having one hospitalization in which COPD was the primary discharge diagnosis or by having two outpatient encounters in which COPD was the primary diagnosis. Inception time was defined as either the first admission with a primary discharge diagnosis of COPD or the second outpatient encounter with a primary diagnosis of COPD.

Table 1. Clinical characteristics of PH–COPD and COPD-alone patients.

	PH-COPD	COPD-alone	p-Value
N	2,176	51,832
Age at COPD diagnosis (mean)	67.2	66.7	0.023
Female % (n)	2.5% (55)	2.5% (1272)	0.828
BMI at COPD diagnosis (mean)	29.3	27.6	<0.001
Race % (n)
White	60% (1,300)	59% (30,532)	0.405
Black	17% (363)	11% (5,931)	<0.001
Others	1% (19)	1% (500)	0.773
Unknown	23% (494)	29% (14,870)	<0.001
Comorbid conditions
OSA diagnosis % (n)	34.8% (757)	16.0% (8,311)	<0.001
Cardiac comorbidities % (n)	88.7% (1,930)	66.1% (34,237)	<0.001
Charlson comorbidity (mean)	4.1	3.2	<0.001
Need for home oxygen % (n)	49% (1,072)	16% (8,045)	<0.001
Health care utilization
COPD exacerbations per year (mean)	1.7	0.5	<0.001
Admissions per year (mean)	5.2	2.3	<0.001
Medication history
Beta blocker % (n)	65.8% (1,431)	52.6% (27,277)	<0.001
Beta blocker days (mean)	1,021	775	<0.001
Statins % (n)	67.6% (1,470)	65.4% (33,877)	0.035
Anti-coagulant/aspirin % (n)	74.3% (1,617)	53.4% (27,615)	<0.001

With recent COPD guidelines emphasizing COPD exacerbation as the main determinant of severity and important clinical outcomes like mortality, we divided COPD cohort into subgroups based on the history of COPD exacerbation and hospital admission (4). The groups included patients with (a) no history of exacerbation or hospital admission, (b) a history of exacerbations but no hospital admission, (c) non-COPD hospital admission but no exacerbations, and (d) a history of exacerbation-related hospital admissions. COPD exacerbation was defined as any hospital admission with COPD as the primary discharge diagnosis, and any ER visit or outpatient clinic visit with COPD as the primary diagnosis that was associated with prescribing a course of antibiotic or systemic corticosteroid. Exacerbations were considered as one episode if they occurred within two weeks of each other.

The main COPD cohort was divided into two subgroups using ICD-9 codes, based on whether the patient had also undergone a diagnosis of PH during the time period between the inception and the endpoint. The two groups were therefore (1) patients having diagnoses of COPD and PH (PH–COPD cohort) and (2) patients having COPD with no PH diagnosis (COPD cohort). Data of the patients were collected from the inception time until death or end of FY 2012.

Data collection

We extracted variables related to the pre-inception period, including demographic data (age, sex, and race), body mass index (BMI), and the presence of co-existing illnesses via assessment of concomitant cardiac diagnoses (Supplement Table 1), presence of obstructive sleep apnea (OSA), and the Charlson comorbidity index (23). Post-inception data from the study period (FY 2000–2012) were also obtained: these included mortality; annual number of all-cause hospitalizations, emergency room visits, and outpatient encounters; and respiratory and cardiac medications.

Statistical analysis

Data analysis was performed using STATA SE version 12.1 (StataCorp, College Station, TX, USA). The primary analysis compared all patients in the two study cohorts. All study variables, including baseline and outcome measures, were analyzed descriptively. Percentages and counts were provided for dichotomous and polychotomous variables. Means and standard deviations were provided for continuous variables. For dichotomous variables, chi-square tests were used to evaluate the statistical significance of differences, and Student's t-test was used for the means of continuous variables. We used a multivariate logistic regression analysis for inpatient and 30-day mortality and Cox regression analysis for overall mortality. p value <0.01 was considered significant in this large population database.

Results

During the study period, a total of 5,45,086 patients from VISN 16 database were reviewed: these patients were sorted by a primary diagnosis of COPD, which identified 54,008 patients with COPD. This represented 10% of the patient population. This cohort was divided into two groups based on whether they had a diagnosis of PH at any time during the study period. Four percent of the patients with COPD carried a diagnosis of PH, encompassing 2,176 patients (PH–COPD cohort); the remaining 51,832 patients comprised the COPD-alone cohort.

Table 1 shows demographic data of the two groups. The PH–COPD group was older and had a higher mean BMI. The PH–COPD carried a much-effective diagnosis of OSA, various cardiac comorbidities and had higher Charlson comorbidity index indicating more chronically ill population compared to COPD alone. A higher percentage of the PH–COPD group received cardiac medications including beta blocker, statins and anticoagulation and aspirin.

Incidence and prevalence of PH

The data showed that the yearly incidence of PH–COPD ranged from 0.39% to 0.57%. Prevalence increased from 1.4% to 4.9% over the study period. The prevalence of PH was then evaluated based on characteristics of COPD. Patients with COPD were stratified into four groups based on the presence of COPD exacerbations, hospitalizations for any cause, and exacerbation-related hospitalizations. Figure 1 shows the prevalence of PH in the four COPD groups for FY 2000–2009. Patients with COPD-related admissions had the highest prevalence of PH diagnosis, while patients with no history of COPD exacerbation and or admission were diagnosed least with PH.

Figure 1. Prevalence of PH–COPD based on COPD admission characteristics.

Table 2 shows the results of regression analysis that was performed on the COPD cohort to evaluate the factors predicting the diagnosis of PH. Patients with a higher number of COPD exacerbations or a higher number of any type of admissions were more likely to have PH diagnosis. Not surprisingly, higher BMI and the presence of OSA predicted higher chance of PH diagnosis. Investigating medication exposures showed that prescription for beta blocker or anticoagulation/aspirin was associated with a higher chance of being diagnosed with PH. In contrast, prescription for statins predicted a lower chance of being diagnosed with PH.

Table 2. Predictors of developing pulmonary hypertension in patients with COPD.

	Odds ratio (confidence interval)	p value
Age at inception	1.027 (1.025–1.028)	<0.05
Female	1.137 (0.771–1.676)	0.517
BMI	1.023(1.021–1.025)	<0.0001
Annual # admissions	1.068 (1.065–1.071)	<0.0001
Annual # COPD exacerbations	1.109 (1.101–1.116)	<0.0001
OSA	2.269 (2.181–2.360)	<0.0001
Days of drug exposure
Beta blocker	1.572 (1.506–1.641)	<0.0001
Statins	0.835 (0.802–0.870)	<0.0001
Anticoagulant/aspirin	2.027 (1.941–2.117)	<0.0001

Mortality

Survival of patients with PH–COPD was reduced over the study period, as illustrated by the Kaplan–Meier curve shown in Figure 2. A diagnosis of PH–COPD is associated with increased odds of overall mortality, as noted in Table 3. Undergoing the diagnosis of PH–COPD also increases the odds of having more health care utilization: this population had a 22% increase in outpatient encounters, a 32% increase in inpatient hospitalizations, and a 27% increase in emergency room visits.

Figure 2. Survival curve for veterans with COPD, with and without PH, over 12 years.

Table 3. Presence of PH in veterans with COPD is associated with increased mortality and health care utilization.

Variable	Odds ratio	p-Value	95% confidence interval
Mortality	1.108	<0.001	1.053–1.166
Outpatient encounters	1.221	<0.001	1.184–1.258
Inpatient admissions	1.318	<0.001	1.267–1.371
Emergency room visits	1.27	<0.001	1.177–1.372

Discussion

This study is the first large-scale, long-term, retrospective assessment of the veteran population with clinical diagnoses of COPD and PH. No similar large-scale studies evaluating the topic of PH–COPD have been done, and our work represents a step toward better characterization of these patients. We found that patients with COPD exacerbation-related hospital admissions had the highest prevalence of PH among this large cohort of COPD patients. Moreover, the annual number of COPD exacerbations was found to be a factor predicting the diagnosis of PH–COPD. Based on these results, we propose that patients with COPD who are hospitalized for an exacerbation should be evaluated for PH and/or the presence of diastolic and/or systolic left heart dysfunction.

Our study has some key limitations. First, the veterans with both COPD and PH were identified using ICD-9 codes. Using this strategy gives false-positive and false-negative diagnoses: it can fail to identify patients with PH, and alternatively could include patients who were incorrectly labeled with PH. Our study is thus dependent on clinical diagnoses based on physician definitions and provider coding practices rather than precise definitions of disease. Given that RHCs are not routinely performed and we do not have accurate numbers on whether patients in this cohort had this invasive procedure, it is a strong possibility that many patient diagnoses of PH were made based on echocardiogram reports, which can generate inaccurate results in patients with lung disease, as discussed previously. However, echocardiograms as a screening test are commonly used in a high-risk population such as veterans, and despite its limitations, our results reflect the importance of identifying those patients that merit a more discerning screening and assessment for PH besides echocardiogram, and perhaps clinical phenotypes offer that option.

Another limitation of using ICD-9 codes is that the definition used to identify COPD exacerbations could inadvertently include patients who were admitted for pneumonia, and we acknowledge that this is a shortcoming. Additionally, our methodology confers an ascertainment bias, as patients with increased healthcare utilization would be more likely to undergo procedures such as echocardiography that would result in them obtaining a diagnosis of PH. Another limitation in the study lies in the fact that although the study is longitudinal, we cannot make assumptions on the time intervals between the diagnosis of COPD, and the occurrence of PH and COPD-related hospital admissions. Thus, no causation can be confirmed. Hypoxemia and long-term oxygen treatment were not assessed in the scope of this study but would be informative in further stratifying a clinical phenotype since these are independent risk factors for the development of COPD-associated PH.

Identifying the PH–COPD population is important because future research could find that PH-specific therapy could be beneficial. Currently, the only therapy that shows clear evidence of reducing progression of PH in COPD is long-term oxygen therapy (24). Results from studies evaluating other PH-specific therapies have been less robust. There have been equivocal results regarding sildenafil (25–28), tadalafil (29), bosentan (30,31), and iloprost (32). Intriguingly, a recent study has evaluated patients with PH that is “out-of-proportion” to their obstructive pulmonary disease: although small and retrospective in nature, treatment with PH-specific medications produced some promising changes in pulmonary hemodynamics (33). The use of PH-specific medications in the PH–COPD population merits further investigation (22). We believe that our data provide important information about how to identify this subgroup of patients so that they can be further studied.

Prior research has suggested that patients with frequent exacerbations of COPD comprise a distinct phenotype of the disease (34). These patients have worsened quality of life indicators compared to the generalized COPD population, and display higher rates of healthcare utilization and hospitalization (35). An analysis of the ECLIPSE database showed that although the rates of exacerbation increased with the higher GOLD stage, there was a subset of patients who were at an increased risk for exacerbations irrespective of disease severity as measured by spirometry (36). We suggest that the presence of PH could be a factor contributing to this independent susceptibility phenotype.

We found that approximately 4% of the patients with a diagnosis of COPD also carried a diagnosis of PH. Comparing this prevalence to results found in other studies is challenging due to differences in how the diagnosis of PH is established. Our study is limited in that it is a database query that relies on ICD-9 diagnostic codes. Presumably, the diagnosis of PH–COPD is made on the basis of TTE, which itself has limitations as noted previously. A subset of patients likely did had RHC that assisted in making this diagnosis, but data on this topic were not obtainable from our database. The prevalence of PH–COPD increased over our study period. It is unclear whether this is due to increased physician recognition of the condition, or perhaps due to increased survival due to initiation of COPD-appropriate therapies.

Our data provide other information regarding PH–COPD. Having a diagnosis of PH–COPD was associated with increased mortality and healthcare utilization. This is in agreement with previous research in the veteran population, which also suggests that the presence of PH in COPD confers mortality risks (13). In our study, patients with PH–COPD were more likely to have cardiac comorbidities and an elevated BMI; additionally, the use of cardiac-related medications predicated the diagnosis of PH–COPD.

One finding in our study was that patients who carried a diagnosis of PH were more likely to have cardiac comorbidities. We acknowledge that this finding represents a potential confounder, since cardiac disease itself is a driver of PH, and is classified as group 2 PH (37). A recent study which assessed all veterans who had undergone RHC nationally between 2007 and 2012 noted that many patients had both cardiac and pulmonary abnormalities (3). Other prevalence studies looking at PH in COPD have noted a large amount of pulmonary venous hypertension secondary to left ventricular disease (10,38,39). Even in the Registry to Evaluate Early and Long-Term Pulmonary Arterial Hypertension Disease Management (REVEAL), which looked at the outcomes of patients with pulmonary arterial hypertension, around 20% of the patients with this clinical diagnosis also have evidence of airway obstruction (40). This information highlights the limitations in the current classification schema, namely that many patients with PH have multiple contributors.

Patients with PH–COPD were more likely to have OSA. In many regards, this is unsurprising because OSA itself can increase pulmonary pressures. The prevalence of PH in OSA ranges from approximately 20% to 40% depending on the method used for the diagnosis and the population being studied (41–44). The treatment of OSA using continuous positive airway pressure has been shown to improve pulmonary hemodynamics (45,46). The presence of sleep apnea in COPD, much like the presence of cardiac comorbidities, serves as an additive and possibly modifiable risk factor that contributes to the development of PH.

Conclusions

Patients who have COPD exacerbations requiring hospitalization have more incidence of being diagnosed with PH. We propose that all patients with COPD exacerbations requiring hospitalization should be screened for PH and/or left heart failure. This clinical rule of thumb could facilitate the identification of this important subset of patients.

Abbreviations

COPD	=	chronic obstructive pulmonary disease
PH	=	pulmonary hypertension
ICD	=	International Classification of Disease
mPAP	=	mean pulmonary artery pressure
RHC	=	right heart catheterization
WHO	=	World Health Organization
TTE	=	trans-thoracic echocardiography
VISN 16	=	South Central Veteran's Affairs Healthcare Network
FY	=	fiscal years
BMI	=	body mass index
OSA	=	obstructive sleep apnea

Declaration of interest

The authors have no conflicts of interest.

Acknowledgments

This work was supported by a Career Development Award #IK20BX002410 (LMP) from the United States Department of Veterans Affairs, Biomedical Laboratory Research and Development.

The authors thank the Center for Translational Research on Inflammatory Diseases (CTRID) at the Veterans Affairs Medical Center in Houston, Texas, for their scholastic support and advice in the preparation of this manuscript.