Trends and Cardiovascular Co-morbidities of COPD Patients in the Veterans Administration Medical System, 1991–1999

Abstract

Persons with Chronic Obstructive Pulmonary Disease (COPD) are at risk for developing other smoking-related complications, including cardiac and vascular diseases. Information about the prevalence and incidence of these conditions is needed to anticipate their occurrence in clinical research. We conducted a cohort study using longitudinal administrative data to describe the prevalence and incidence of cardiovascular (CV) diseases among COPD patients treated by the Veterans Administration Medical System. The COPD cohort included all persons with a diagnosis of COPD admitted to a Veterans Administration Medical System hospital (N = 70,679) or seen in a outpatient clinic (N = 314,209) in fiscal year 1998. Each COPD patient was matched to a Veteran of the same age and gender who did not have a COPD diagnosis, creating a non-COPD cohort for comparison. Among all hospitalized Veterans, the prevalence of COPD steadily increased from 1991 to 1999, although the total number of Veterans hospitalized during this time period decreased by more than one-third. Among COPD patients hospitalized in 1998, the prevalence of coronary artery disease, congestive heart failure, and atrial fibrillation were very high (33.6%, 24.4%, and 14.3%, respectively) and significantly higher than those seen among the matched non-COPD cohort (27.1%, 13.5%, and 10.4%; p < 0.001). Among COPD outpatients, increased complications were found in every CV disease category with rate ratios that were greater than observed among inpatients. We conclude that CV diseases are remarkably prevalent among Veterans with COPD, and their incidence is likely to increase as the Veteran population ages.

• Pulmonary disease
• Chronic obstructive
• Co-morbidity
• Cardiovascular diseases
• United States Department of Veterans Affairs
• Seasons
• Epidemiology

Introduction

Chronic Obstructive Pulmonary Disease (COPD) is currently the fourth most common cause of death in the United States, and the only condition among the 10 leading causes of death projected to have increased mortality over the next decade [1&2]. COPD's impact is growing at an even more rapid pace globally: the World Health Organization projects that it will become the fifth leading cause of disability by the year 2020, an increase from its current rank of 12th [[3]]. The Global Initiative for Chronic Obstructive Pulmonary Disease, an international workgroup created to increase awareness of COPD and decrease morbidity and mortality from the disease, has cited the need for additional information on COPDepidemiology so that countries can plan for increased demand for COPD-related health care services [[4]].

One of the difficulties of conducting clinical outcomes research in COPD is the high prevalence of comorbid conditions, especially those that are also caused by cigarette smoking [[5]]. Persons with chronic airflow obstruction are known to have a substantially increased risk of coronary artery disease and sudden cardiac death [6-10], and the degree of airflow obstruction is highly correlated with systemic inflammatory markers that are associated with coronary artery disease [[11]]. However, remarkably little population-based data is available on the incidence and prevalence of specific cardiovascular (CV) diseases in COPD patients. These data are needed to help anticipate the effects new COPD treatments may have on concomitant CV diseases, and to understand the epidemiology of uncommon but serious adverse events that may be observed during clinical research.

We conducted this study to describe the prevalence and incidence of CV diseases in one of the largest cohorts of COPD patients in the United States, the Veterans Administration (VA) Medical System. We accessed the VA's administrative data to capture all Veterans hospitalized with COPD during the decade of the 1990s to describe this population and important changes in demographic and utilization characteristics. We then described the incidence and prevalence of CV complications among all Veterans treated for COPD as an inpatient or outpatient in 1998 and compared these with an age- and sex-matched cohort without COPD.

Materials and Methods

The Institutional Review Board of the University of New Mexico Health Sciences Center and the Human Research Review Committee of the Albuquerque VAMS granted approval for this study. All data were obtained through the Austin Automation Center (AAC), a centralized computing system that has logged all VAMC hospitalizations since 1970, and all VAMC outpatient encounters since 1980 [[12]]. The AAC currently maintains files on over 28 million Veterans. Inpatient data is maintained in the Patient Treatment File (PTF), and outpatient data is maintained in the Outpatient Care File (OCF). Patient confidentiality is maintained by assigning each Veteran a scrambled Social Security number that cannot be matched to files by non-AAC personnel, and by stripping files of addresses and other information that could be used to identify individuals. A merge with the separate administrative Standard File database was necessary to combine demographic information with the diagnostic coding information and to ensure that all patient visits were captured. Seasonality of admission was categorized by the month of admission.

COPD Case Identification and Control Matching

Hospitalized COPD patients were identified by looking for International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes 491.x (chronic bronchitis), 492.x (emphysema), or 496 (chronic airway obstruction, notelsewhere classified) recorded anywhere in the PTF discharge string. COPD outpatients were identified using the same ICD-9-CM codes; however, comprehensive outpatient diagnostic coding was not available in the OCF until 1997.

Each COPD patient was matched to another Veteran who did not have a diagnosis of COPD by age and gender using the MATCH procedure in SAS (Version 8.0 for Windows; SAS Corporation, Cary, NC). Due to the relatively small proportion of non-white Veterans in the COPD cohort, we did not use race or ethnicity for matching.

Co-morbidity Abstraction and Validation

The presence of comorbid disease by organ system was classified using a modified version of the Charleson-Deyo Index [[13]]. Specific CV and other diseases of interest were identified by their ICD-9-CM codes. Prevalence was defined as the presence of an ICD-9 code for that condition in any discharge string for any hospitalization or outpatient visit at any time during the index year (fiscal year 1992 or 1998). Incidence was defined as identification of a specific ICD-9-CM CV code at any time during fiscal year 1999 among those in each cohort who did not have that code in 1998. Without having personal identifiers, we were not able to merge the PTF with the OCF. While it is not certain what effect this had on our incidence estimates, we feel that it is likely that our incidence figures are slightly underestimated.

In our preliminary survey, we found that if one simply identifies patients with COPD admissions or visits in the PTF or OCF files, then subsequent visits and their co-morbidities may be lost. To validate our methods of identifying patients and controls with co-morbid conditions in the database, we used two different software algorithms to abstract the database. The first was to do an initial survey that captured the patients, then conduct a second run that captured all visits for the cohort. The second strategy was to capture all COPD admissions or visits, then collate the files by patient, which required only one file inquiry. The incidence and prevalence figures did not differ by method.

Statistical Methods

Comparisons of normally distributed measures were made using the Student's t-test, non-normal data was analyzed using Wilcoxon comparisons, and chi-square tests were used for ordinal data. The differences in proportions of various co-morbid conditions between COPD patients and other Veterans were compared by calculating the z statistic. Comparisons of the differences in incidence and prevalence figures for the COPD and non-COPD patients were made by comparing the rate ratios and their 95% Taylor series confidence intervals, with intervals that do not include 1.0 designated as being statistically significant. Length of stay is reported as median due to its log-normal distribution. The authors wrote and submitted SAS programs to extract and analyze all data.

Results

Demographics for all Veterans hospitalized in 1992 and 1998, divided into COPD and non-COPD groups, are presented in Table 1. The decrease in the number of COPD and non-COPD patients in these two samples reflects the steady and dramatic decrease in the number of hospitalized Veterans over the last decade (Table 2). Although the absolute number of Veterans hospitalized with COPD decreased from 1991 to 1999, COPD was found in an increasing proportion of all VAMC admissions. The prevalence of COPD amongfemale and minority Veterans increased slightly during this decade, but the overwhelming majority of Veterans with COPD are non-Hispanic White males.

Table 1.  Demographic Characteristics and Chronic Illnesses Among All Hospitalized Veterans With and Without COPD in 1992 and 1998.

	1992 COPD (n = 87,867)	1992 Non-COPD (n = 481,812)	1998 COPD (n = 70,679)	1998 Non-COPD (n = 308,275)
Age
Mean (median)	66.1 (67)	57.3 (60)	67.8 (70)	59.8 (60)
Interquartile range	61–72	44–69	61–76	49–72
Sex (%)
Male	98.7	98.1	98.1	95.7
Female	1.3	1.9	1.9	4.3
Race/ethnicity (%)
Hispanic	1.8	4.9	2.1	5.5
Non-Hispanic White	85.8	71.4	82.3	67.2
African-American	10.9	21.2	12.0	21.7
Other	1.5	2.5	3.6	5.6
Mean number of admissions per year	1.59	1.62	1.52	1.61
Co-morbidities, n (%)
Cardiovascular Dz	54,587 (62.1)	194,551 (40.4)	49,621 (71.2)	157,549 (51.1)
Cancer	14,718 (16.8)	59,969 (12.4)	10,385 (14.9)	36,467 (11.8)
Cerebrovascular Dz	6614 (7.5)	32,730 (6.8)	5472 (7.9)	23,774 (7.7)
Diabetes	13,244 (15.1)	72,591 (15.1)	14,685 (21.1)	63,799 (20.7)
Hepatic	2601 (3.0)	14,650 (3.0)	2068 (3.0)	9551 (3.1)
Renal	2821 (3.2)	13,468 (2.8)	3327 (4.8)	12,666 (4.1)
Dementia	2160 (2.5)	11,145 (2.3)	1925 (2.8)	8091 (2.6)
Neurologic	3444 (3.9)	23,074 (4.8)	1717 (2.5)	11,320 (3.7)
Rheumatologic	1222 (1.3)	4452 (0.1)	862 (1.2)	2706 (0.1)
Degenerative arthritis	8424 (9.6)	33,025 (6.9)	6607 (9.5)	22,848 (7.4)
Peptic ulcer disease	6878 (7.8)	24,674 (5.1)	3950 (5.7)	13,080 (4.2)

*Indicates significantly higher proportion in COPD, p < 0.01.

Table 2.  Changes in Inpatient Utilization for All Veterans and Veterans with COPD, 1991 to 1999.

Year	All VAMC discharges	All COPD discharges	COPD as % of all discharges	Number of COPD patients	Mean age
1991	944,162	145,495	15.4	91,234	66.8
1992	922,514	139,812	15.2	87,867	66.1
1993	909,896	141,661	15.6	89,032	66.3
1994	901,665	142,500	15.8	89,666	66.6
1995	873,975	139,122	15.9	87,809	66.7
1996	809,717	133,956	16.5	84,680	66.9
1997	671,087	113,897	17.0	73,992	67.4
1998	604,946	106,585	17.6	70,005	67.6
1999	581,226	105,294	18.1	69,711	67.8

*All discharges where COPD was noted anywhere in the diagnosis string.

In our examination of factors affecting hospitalization and incidence rates, we found a large seasonal variation in the rate of COPD admissions (Figure 1). The rate of COPD admissions during the winter months (December through February) was more than 15% greater than the annual monthly average, and more than 40% greater than that seen during the summer months (Figure 2). This fluctuation was directly related to a seasonal variation in admissions for respiratory tract infections.

Figure 1. Mean cumulative monthly admissions for COPD for 1991–1999, U.S. Veterans Affairs Hospitals.

Figure 2. Percent change from mean for COPD admissions by month, U.S. Veterans Affairs Hospitals, 1991–1999.

In an un-adjusted comparison, COPD patients are more likely than non-COPD patients to have a major co-morbid illness, especially diseases that are also associated with smoking including cardiovascular disease and cancer (Table 1). The prevalence of heart disease among both the COPD and non-COPD hospitalized Veterans increased from 1992 to 1998, while the rates of most other major chronic illnesses stayed about the same. The prevalence of heart disease among hospitalized COPD patients was remarkably high, with 71.2% of those admitted with COPD in 1998 having one or more CV conditions.

Most CV diseases increased in prevalence among both the hospitalized COPD patients and their matched controls from 1992 to 1998 (Table 3). COPD patients had a higher prevalence and incidence of almost every cardiovascular disease as compared to their control cohort without COPD, with remarkably higher rates in congestive heart failure, atrial fibrillation and flutter, and coronary artery disease. Obstructive sleep apnea and cor pulmonale, which are both known to overlap clinically with severe chronic bronchitis, were both more than twice as common among COPD patients. Differences in the average number of admissions per year for hospitalized COPD and non-COPD patients were not statistically significant (Table 1). In 1992, COPD patients did have a longer median hospital stay (7 days versus 5 days, p < 0.01), but by 1998, there was no longer any difference (5 days for each).

Table 3.  Prevalence of Cardiac and Vascular Diseases among Veterans Hospitalized for COPD in 1992 and 1998, Disease Incidence in 1999, and Their Respective Rate Ratios as Compared to Age and Gender-Matched Controls.

Condition	COPD 1992 prevalence	Prevalence ratio (95% CI)	COPD 1998 prevalence	Prevalence ratio (95% CI)	COPD 1999 incidence	Rate ratio
Hypertension	35,222	1.01 (1.00–1.03)	46,567	1.04 (1.03–1.06)	4168	0.95 (0.91–0.99)
Hypertensive heart disease	1668	1.59 (1.47–1.72)	1152	1.44 (1.31–1.52)	456	1.49 (1.29–1.72)
Paroxismal supraventricular tachycardia	613	2.06 (1.79–2.36)	464	1.60 (1.34–1.90)	201	2.21 (1.72–2.83)
Paroxismal ventricular tachycardia	1290	1.46 (1.34–1.59)	1576	1.38 (1.28–1.49)	482	1.35 (1.18–1.55)
Atrial fibrillation	10,098	1.50 (1.45–1.54)	14,283	1.37 (1.33–1.41)	2289	1.31 (1.23–1.39)
Atrial flutter	2073	1.67 (1.56–1.79)	2824	1.54 (1.45–1.63)	740	1.55 (1.38–1.74)
Acute myocardial infarction	3844	1.10 (1.05–1.16)	4715	1.06 (1.02–1.11)	1408	1.28 (1.18–1.38)
Angina pectoris	5236	0.95 (0.92–0.99)	6169	0.97 (0.94–1.01)	1430	1.08 (1.00–1.16)
Coronary artery disease	28,514	1.34 (1.31–1.36)	33,601	1.24 (1.22–1.27)	3969	1.25 (1.19–1.31)
Congestive heart failure	20,265	1.97 (1.92–2.02)	24,447	1.81 (1.77–1.85)	3704	1.46 (1.38–1.53)
Peripheral vascular disease	1297	1.25 (1.15–1.36)	2066	1.11 (1.05–1.19)	478	0.96 (0.85–1.09)
Atherosclerosis-aorta	248	1.46 (1.20–1.77)	150	1.16 (0.92–1.47)	38	1.09 (0.69–1.72)
Conduction disorder	4000	1.24 (1.18–1.30)	2875	1.16 (1.10–1.22)	736	1.19 (1.07–1.32)
Aortic stenosis or insufficiency	1436	1.14 (1.06–1.23)	1572	1.05 (0.98–1.13)	351	1.17 (1.00–1.37)
Atherosclerosis of renal artery	228	1.08 (0.90–1.30)	293	0.98 (0.83–1.15)	76	0.89 (0.66–1.22)
Mesenteric ischemia	148	1.37 (1.07–1.76)	144	1.13 (0.89–1.44)	86	1.46 (1.05–2.03)
Hypercholesteremia	4250	0.88 (0.84–0.92)	9805	0.90 (0.87–0.92)	2129	0.98 (0.92–1.04)
Obstructive sleep apnea	1015	2.71 (2.41–3.06)	2850	2.32 (2.16–2.48)	529	2.01 (1.73–2.33)
Cor pulmonale	714	2.88 (2.49–3.33)	1409	2.72 (2.46–3.01)	495	2.50 (2.12–2.95)
Gastroesophageal reflux disease	7769	1.70 (1.61–1.73)	8931	1.59 (1.54–1.65)	2210	1.44 (1.35–1.54)
Death	NA	NA	9047	NA	12,618	NA

*(Cases per 100,000).

The statistics for COPD patients seen in outpatient clinics in 1998 are found in Table 4. The mean age of the VA COPD outpatient cohort was 65.5 years, which was only slightly younger than that of the 1998 inpatient cohort (67.8 years),although the median age was the same (70). Women also comprised a slightly higher proportion of the outpatient COPD cohort (4.4% versus 1.9% in the 1998 inpatient cohort). Race is not catalogued in the outpatient database. Outpatient clinic utilization increased significantly for COPD patients between 1997 and 1999, the only years for which complete outpatient diagnostic coding is available. In 1997, the VAMC incurred 36,631,328 outpatient encounters while in 1999 that number was 42,044,619. The percentage of those visits for COPD was steady at 2.2%, so the absolute number of outpatient COPD encounters increased from 791,466 in 1997 to 905,613 in 1999.

Table 4.  Prevalence and Incidence of Cardiac and Vascular Diseases and Their Respective Rate Ratios as Compared to Age- and Gender-Matched Controls in the 1998 VAMC COPD Outpatient Cohort.

Condition	COPD 1998 prevalence	1998 Prevalence ratio (95% CI)	COPD 1999 incidence	1999 Rate ratio (95% CI)
Hypertension	51,226	4.22 (4.14–4.31)	8815	3.57 (3.41–3.74)
Hypertensive heart disease	1046	5.18 (4.45–6.02)	588	3.89 (3.26–4.66)
Paroxysmal supraventricular tachycardia	246	6.65 (4.71–9.40)	177	4.66 (3.28–6.61)
Paroxysmal ventricular tachycardia	289	5.16 (3.88–6.87)	152	4.47 (3.08–6.49)
Atrial fibrillation	6389	5.64 (5.29–6.01)	2053	4.74 (4.27–5.26)
Atrial flutter	697	6.77 (5.50–8.32)	425	4.89 (3.88–6.15)
Acute myocardial infarction	1208	6.75 (5.77–7.90)	727	5.31 (4.54–6.21)
Angina pectoris	1658	6.63 (5.81–7.58)	1022	5.32 (4.56–6.21)
Coronary artery disease	30,357	5.27 (5.12–5.43)	6465	4.01 (3.80–4.24)
Congestive heart failure	14,116	7.94 (7.56–8.35)	4611	5.94 (5.50–6.42)
Peripheral vascular disease	1848	5.50 (4.90–6.18)	928	5.33 (4.54–6.27)
Atherosclerosis-aorta	91	6.07 (3.51–10.4)	69	6.27 (3.32–11.9)
Conduction disorder	710	4.67 (3.92–5.57)	538	4.24 (3.49–5.14)
Aortic stenosis or insufficiency	1327	4.38 (3.87–4.96)	689	3.83 (3.25–4.51)
Atherosclerosis renal artery	57	4.07 (2.27–7.31)	39	7.80 (3.07–19.8)
Mesenteric ischemia	36	2.77 (1.47–5.22)	27	3.86 (1.68–8.86)
Hypercholesterolemia	22,395	4.01 (3.89–4.14)	7790	3.59 (3.42–3.76)
Obstructive sleep apnea	3285	10.84 (9.64–12.2)	1316	6.93 (5.95–8.07)
Cor pulmonale	349	12.93 (8.74–19.1)	195	12.19 (7.32–20.3)
Gastroesophageal reflux disease	12,940	6.26 (5.97–6.56)	5789	4.70 (4.42–4.99)

*Cases per 100,000.

The differences between COPD and non-COPD cohorts in both prevalence and incidence of CV events were far more pronounced in the outpatient cohort (Table 4). The largest total differences were again observed among patients with coronary artery disease, congestive heart failure, and atrial fibrillation and flutter. However, substantial differences were now seen among the less common complications such as mesenteric ischemia. Also, some of the less serious chronic illnesses such as hypertension, hypercholesterolemia, and gastroesophageal reflux were far more prevalent in the outpatient database, most likely reflecting the differences in inpatient and outpatient coding procedures and the factors that initiated the visits in either area.

Discussion

As expected, the incidence and prevalence of almost all cardiovascular conditions were significantly higher among VA COPD patients as compared to an age- and gender-matched cohort without COPD. The absolute number of hospitalized veterans decreased from 1992 to 1998, most likely due to the changes in hospital utilization throughout the VA system. However, most of the prevalence rate ratios changed very little from 1992 to 1998, indicating that our estimates are relatively robust. The prevalence of some very serious cardiovascular conditions, including congestive heart failure and atrial arrhythmias, were remarkably high in this COPD population. These data illustrate the difficult and common problem of cardiovascular disease complications among COPD patients, and should be kept in mind when examining the incidence of these events in clinical studies of new treatments.

There are several factors that should be considered when comparing our prevalence and incidence data to other populations. First, the large seasonal variation in COPD admissions that we illustrate in Figures 1 and 2 confirm that annual or seasonally adjusted rates should be used when describing incidence and prevalence figures in COPD. Also, the differences in prevalence and incidence figures in our inpatient and outpatient populations illustrate how hospital-based populations can have very different clinical characteristics from clinic-based populations, even when the demographic factors are similar. Hospital billing records also may be less likely to capture chronic illnesses such as hypertension, even though they may contain more diagnoses per visit. The VA treats relatively few women, and because cardiovascular disease risks are affected by gender, it is likely that other COPD populations with a higher proportion of women will have a different experience. The VA patient population is also generally older (36% are 65 years or older) and less economically advantaged (70% have an annual income ofless than $20,000) than most U.S. populations, which could also affect their outcome experience [[14]]. Finally, we were not able to match our COPD population by race. African-American males have a slightly lower incidence of cardiovascular diseases, so the slightly higher proportion of African-Americans in the control population may have introduced a small bias in the estimates.

This database also has several important limitations that must be considered when examining these results. First, COPD is grossly underdiagnosed, so a large proportion of our patient cohort without COPD is likely to have had COPD and its associated risk of CV illness [[2]]. This bias is likely to have resulted in spuriously low risk ratio estimates, so the significantly elevated CV risks we have observed would probably be even greater if we were able to appropriately reclassify each patient. It is also likely that some of the COPD patients have been misdiagnosed and would be more appropriately classified as having asthma or some other chronic respiratory condition. It is difficult to know what effect over-diagnosis had on the risk estimates; however, because most other lung diseases do not have as strong an association with cigarette smoking, we believe that this is also likely to be a bias towards a null effect. Second, we were not allowed to cross-match the inpatient and outpatient files, so the inpatient and outpatient data should be considered as two separate populations, even though there is a very high degree of overlap. Third, complications that often lead to sudden death, such as ventricular tachycardia, are very likely to be missed or underreported. In fact, death, which we were only able to capture in the 1998 inpatient COPD population, was remarkably common: 9.0% died during the index year, and another 12.6% died in 1999.

Population-based studies focused on the prevalence of comorbid conditions among patients with COPD compared to those without COPD are sparse in the published literature. Van Manen and colleagues compared a cohort with evidence of chronic airway obstruction (and COPD or asthma diagnosis) aged 40 and older with a random sample of general practice patients without COPD or asthma in the same age group [[15]]. The prevalence of any patient-reported comorbidity was 10 to 12% higher among the chronic airway obstruction cohort; locomotive disease (odds ratio (OR) = 3.13), insomnia (OR = 1.65), sinusitis (OR = 6.08), migraine (OR = 3.13), depression (OR = 2.10), stomach or duodenal ulcers (OR = 7.33), and cancer (OR = 2.47) were significantly more common among chronic airway obstruction patients than in the matched patient group. Surprisingly, there was not a significant difference for cardiovascular diseases (heart disease, stroke, hypertension, atherosclerosis) between the cohorts even though there were higher levels of current or ex-smokers in the chronic airway obstruction cohort. This finding may be explained by a higher cardiovascular mortality rate in the chronic obstructive airway group causing more deaths and a lower prevalence rate among the survivors in this cross-sectional sample.

Clinical studies in COPD have traditionally focused on airflow obstruction as their primary outcome measure, but our data suggests that clinical investigators need to also carefully consider what effects a new treatment might have on CV disease co-morbidity and mortality. Some agents such as methylxanthines can potentially cause CV complications as direct side effects [16-20]. However, if COPD therapy is able to prevent or reduce the severity of exacerbation episodes, or reduce the systemic inflammation associated with advanced airflow obstruction, it is possible that they could have protective benefits as well [[21]]. Long-term follow-up studies of randomized treated populations are underway that will hopefully provide a better understanding of the interactions between COPD, heart disease risk, and COPD treatment.

Acknowledgments

The authors would like to thank Dr. David B. Coultas and Dr. Eva Lydick for their advice and contributions to this project.