Mortality after an Emergency Department Visit for Exacerbation of Chronic Obstructive Pulmonary Disease

Abstract

Objective. To describe the mortality after emergency department (ED) visits for chronic obstructive pulmonary disease (COPD) exacerbation. Design. Retrospective cohort study of ED patients with COPD exacerbation. Setting. Administrative data analysis. Participants. Patients age 55 and over who visited the ED during a 2-year period with primary ICD-9 codes of 491, 492, or 496. Measurements. Demographic characteristics, comorbid conditions, hospital utilization for COPD, and vital status. Results. During the study period, there were 482 index visits with a median follow-up of 1,1 28 days (3.1 years). Demographic characteristics of the cohort were as follows: mean age 72 years, 56% female, 93% White, and 37% currently married. Mortality increased over time: 5% at 30 days, 9% at 60 days, 11% at 90 days, 16% at 180 days, 23% at 1 year, 32% at 2 years, and 39% at 3 years. At the end of follow-up, 220 (46%) patients had died. On multivariate analysis, independent predictors of mortality were increasing age (hazard ratio [HR] 1.3 per 5-year increase, 95% CI 1.2–1.4), having congestive heart failure (HR 1.6, 95% CI 1.2–2.1), having metastatic solid tumor (HR 3.3, 95% CI 2.0–5.5), and hospital utilization for COPD exacerbation during past year (HR 1.9, 95% CI 1.4–2.6). Conclusion. The mortality rate after an ED visit for COPD exacerbation is quite high. Mortality is related to older age, specific comorbid conditions, and history of prior COPD exacerbations.

Key words: :

• Chronic obstructive pulmonary disease
• Exacerbation
• Mortality
• Emergency department

ABBREVIATIONS
CHF	=	Congestive Heart Failure;
CI	=	Confidence Interval;
COPD	=	Chronic Obstructive Pulmonary Disease;
ED	=	Emergency Department;
FEV1	=	Forced Expiratory Volume in One Second;
FVC	=	Forced Vital Capacity;
HR	=	Hazard Ratio;
IQR	=	Interquartile Range;
IRB	=	Institutional Review Board;
SSADMF	=	Social Security Administration Death Master File.

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is characterized by progressive irreversible airway obstruction ([1]). This progression is punctuated by exacerbations, which are defined as a sustained worsening of the COPD patient's condition, from the stable state, and beyond normal day-to-day variations, that is acute in onset and necessitates a change in regular medication ([2], [3]). While many of the mild COPD exacerbations are never brought to a physician's attention ([4]), moderate-to-severe exacerbations, which by definition require increased medication or hospitalization ([2]), involve formal medical evaluation and treatment. In the United States, the ED is an important venue for this care, with more than 1 million ED visits for COPD exacerbations every year ([5]).

In the United States, COPD is the fourth cause of death (following heart disease, cancer, and cerebrovascular disease) and also is one of the fastest growing causes of mortality and morbidity ([6]). Prior studies have examined long-term outcomes among patients with COPD ([7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21]) and they have identified several risk factors for mortality: older age, marital status, low forced expiratory volume in 1 second (FEV₁), hypercapnia, hypoxia, pulmonary hypertension, co-morbid conditions, and baseline functional status. While some of these studies followed COPD cohorts ([7], [8], [9], [10], [15], [16], [21]), several of these studies focused on long-term outcomes after an exacerbation ([11], [12], [13], [14], [17], [18], [19], [20]). Though these studies provided insight regarding patient outcomes after a COPD exacerbation, we are not aware of any studies assessing mortality among patients who visited the ED for a COPD exacerbation. Since the ED is an important venue for the treatment of COPD exacerbations, and only about 60% of COPD patients who are treated in the ED are admitted ([22]), the mortality of patients who are treated in the ED for a COPD exacerbation is of interest.

This study used a large administrative database, with data on demographic factors, clinical tests, hospital utilization records, and comorbid conditions. Mortality data were imported from Social Security Administration Death Master File (SSADMF). We aim to describe the mortality experience of patients treated in the ED for a COPD exacerbation, and to identify risk factors of mortality in this high-risk group.

MATERIALS AND METHODS

Study design

This was a retrospective cohort study.

Methods of measurements/data collection

This study used data obtained from the Research Patient Data Repository at Partners HealthCare (Boston, MA). The Research Patient Data Repository contains data about patients and their visits, including demographic and contact information, diagnoses and procedures, medication dispensed by the inpatient pharmacy, laboratory results, and provider information. The Research Patient Data Repository provides a query tool used by researchers to obtain aggregate totals for requested data in addition to specific patient databases. Researchers with institutional review board approval may then proceed to obtain more detailed data about their sets of patients from the Research Patient Data Repository ([23]).

We obtained aggregate numbers of patients who visited the Massachusetts General Hospital ED between January 1, 1999 and December 31, 2001 with primary ICD-9 codes 491, 492 or 496 using the online query tool. We then requested more detailed information about the queried patients, including demographic and encounter information, pulmonary function test results, and vital status. The Research Patient Data Repository determines vital status using SSADMF. If a patient is missing his/her Social Security number, the Research Patient Data Repository search engine uses the patient's name, date of birth, and zip code to find the patient's vital status. If the patient died in the hospital, the system obtains the information from the hospital medical records system. The study protocol was reviewed and approved by the institutional review board.

The compiled dataset was carefully reviewed for errors and inconsistencies. Encounter records were examined for all co-morbid conditions. Any encounter, including hospitalization, ED visit, or outpatient visit within 2 years from the index visit was included for co-morbid condition assessment. All of the ICD-9 codes and procedure codes associated with the encounter were examined, and if any ICD-9 code in the Deyo et al.'s ICD-9-CM definitions of Charlson Index calculation ([24]) was found, the patient was assigned that co-morbid condition. All patients were assigned “chronic pulmonary diseases” since diagnosis of COPD at the index visit was required for inclusion in the study. The Charlson Index was calculated by summing the weighted comorbidity scores ([25]).

Median family income was estimated using patients' home ZIP code ([26]) and is presented as U.S. dollars. Body mass index was calculated by weight (in kilograms) divided by square of height (in meters) using height and weight values obtained during pulmonary function test. Hospital utilization for COPD exacerbation during past year (from the index visit) was defined as any ED visit and hospitalization with primary diagnosis of COPD (ICD-9 codes 491, 492 or 496) or ED visits/ hospitalizations with a secondary diagnosis of COPD and a primary diagnosis related to COPD exacerbation (e.g., acute bronchitis, pneumonia). COPD stage was determined for patients who had pulmonary function test results within 2 years from the index visit. According to the GOLD criteria ([1]), patients were assigned stages 0 to 4. Reversibility was defined as an FEV₁ increase of > 12% and > 200 ml after bronchodilator administration ([1]).

Duration of follow-up was calculated as the duration between the date of the index visit and the date of death for deceased patients. For living patients, it was defined as the duration between the date of the index visit and the last date when the SSADMF data were imported (October 4, 2004). For patients who were missing Social Security number, and did not have any mortality information in hospital records, the last date of any Partners Healthcare System visit or laboratory test was defined as last date on which patient was known to be alive, and the duration between the date of index visit and this last date was defined as the follow-up duration. For this study, “early mortality” was defined as mortality within 60 days from the index visit.

Statistical analysis

All analyses were performed using STATA 7.0 (StataCorp, College Station, TX). Data are presented as proportions (with 95% confidence intervals [CI]), means (with standard deviation [SD]), and medians (with interquartile range [IQR]). The association between survival time and other factors was examined initially using an unadjusted Cox proportional hazard analysis and Kaplan–Meier survival curves. Age and sex were included in the multivariate Cox model. Other variables associated with mortality at p < 0.10 in univariate analysis were evaluated for inclusion in multivariate Cox models. All hazard ratios (HR) are presented with 95% CI. A logistic regression model was used to examine predictive factors for early mortality. The area under the receiver operator curve was calculated for the final multivariate logistic regression models. The final logistic regression model was further evaluated using the Hosmer–Lemeshow statistic ([27], [28]). All p-values are two-tailed, with p < 0.05 considered statistically significant.

RESULTS

There were 482 index ED visits for COPD between January 1, 1999 and December 31, 2001. Only 16 (3%) of the 482 patients were missing Social Security number. Among the 16 patients, 9 were identified as deceased by the SSADMF using name and address, and hospital records when the patient died in the hospital. There was no evidence of vital status for 7 patients of the 16 patients, so data were censored at the last time point when the patients had any information in the administrative database (e.g., laboratory test or outpatient visit).

Table 1 shows the baseline characteristics of the cohort. Of 482 patients, 220 patients (46%) died by the end of the study. Median follow-up duration was 1,128 days (IQR, 436-1,575 days). Figure 1 shows the Kaplan–Meier estimate of mortality by time, with a more rapid decline in survival in the initial months: survival at 30 days, 60 days, 90 days, 6 month and 1 year were 95%, 91%, 89%, 84% and 77%, respectively.

Figure 1. Estimated survival function of 482 emergency department patients with COPD exacerbation

Table 1. Baseline characteristics of 482 emergency department patients with COPD exacerbation

Age (years), mean ± SD	72 ± 9
Female (%)	56
Race/ethnicity (%)
White	94
Black	3
Hispanic	2
Asian	1
Other	< 1
Married (%)	38
Median household income, U.S. dollars, median (IQR)	47,119 (39,670–55,291)
Body mass index, kg/m^2, mean ± SD	27.7 ± 7.8
Co-morbid conditions (%) Myocardial infarction	23
Congestive heart disease	40
Peripheral vascular disease	15
Dementia	2
Chronic pulmonary disease	100
Rheumatologic disease	3
Peptic ulcer disease	5
Mild liver disease	2
Diabetes (mild to moderate)	23
Diabetes with chronic complication	4
Hemiplegia or paraplegia	< 1
Renal disease	6
Any malignancy, including lymphoma and leukemia	18
Moderate or severe liver disease	< 1
Metastatic solid tumor	5
Acquired Immune Deficiency Syndrome	0
Charlson index, median (IQR)	2 ([1], [2], [3], [4])
COPD-related ED visit or hospitalization during past year (%)	23
Hospitalization at the index visit (%)	62

SD denotes standard deviation; IQR, interquartile range; COPD, chronic obstructive pulmonary disease; GOLD, Global initiative for chronic obstructive lung disease.

*Baseline body mass index available for 103 patients with measured height and weight within 2 years before their index visit.

Risk factors of mortality were evaluated on univariate analysis (Table 2). Predictive factors were: older age, non-married status, BMI less than 25 kg/m², having congestive heart failure (CHF), renal disease, any malignancy, or metastatic solid tumor, higher Charlson index, hospital utilization for COPD exacerbation during past year, and hospitalization after the index visit. Among 117 patients with an available pulmonary function test within 2 years of the index ED visit, COPD stage and airway reversibility were not related to mortality. The association between COPD stage and airway reversibility with mortality are shown in Table 3.

Table 2. Unadjusted risk factors for mortality in 482 emergency department patients with COPD exacerbation

	Hazard Ratio	95% CI
Age (per 5-year increase)	1.3	1.2–1.4
Female	1.0	0.8–1.4
Race/ethnicity
White	1.0	Reference
Black	0.6	0.2–1.6
Hispanic	0.3	0.1–1.4
Asian	0.3	0.04–1.9
Other	1.8	0.2–12.6
Marital status
Married	1.0	Reference
Single, divorced, widowed, separated	1.4	1.04–1.84
Median household income
< $44,999	1.0	Reference
$45,000–$54,999	0.9	0.7–1.3
$55,000 +	1.1	0.8–1.5
Body mass index (kg/m^2)
< 25.0	2.1	1.2–3.6
25.0–29.9	1.0	Reference
30.0 +	1.4	0.8–2.6
Co-morbid conditions
Myocardial infarction	1.2	0.9- 1.7
Congestive heart disease	1.8	1.4–2.4
Peripheral vascular disease	1.2	0.8–1.7
Dementia	1.9	0.9–4.3
Chronic pulmonary disease	N/A	N/A
Rheumatologic disease	0.4	0.1–1.3
Peptic ulcer disease	1.0	0.6–1.9
Mild liver disease	1.4	0.6–3.5
Diabetes (mild-to-moderate)	0.9	0.7–1.3
Diabetes with chronic complication	1.3	0.7–2.4
Hemiplegia or paraplegia	0.7	0.1–5.1
Renal disease	2.0	1.3–3.2
Any malignancy, including lymphoma and leukemia	1.5	1.1–2.0
Moderate or severe liver disease	1.2	0.2–8.4
Metastatic solid tumor	2.4	1.5–4.0
Charlson index	1.11	1.05–1.16
Hospital utilization for COPD exacerbation during past year
None	1.0	Reference
1+	2.0	1.5–2.6
Hospitalization at the index visit	1.33	1.00–1.77

CI denotes confidence interval; and COPD, chronic obstructive pulmonary disease.

Table 3. Pulmonary function test results and univariate analysis of mortality by pulmonary function

	Percentage	HR	95% CI
COPD stage per GOLD
0	27	1.0	Reference
1	1	0.9	0.5–1.8
2	15
3	30	1.3	0.8–2.0
4	27
Airway reversibility	12	0.6	0.3–1.3

HR denotes hazard ratio; CI, confidence interval; COPD, chronic obstructive pulmonary disease; and GOLD, Global initiative for chronic obstructive lung disease.

*COPD stage available for 117 patients in whom spirometry data were available within 2 years before their index visit. COPD stages 1 and 2 and stages 3 and 4 were each combined for the purposes of assessing the association with COPD stage and mortality.

^†Reversibility available for 102 patients in whom spirometry data were available within 2 years before their index visit.

Table 4 shows the multivariate mortality models. BMI and pulmonary function tests were not included in the model because these data only were available for a small subset of patients (< 30%). Significant predictors of mortality in a Cox model, over the approximately 3 years of follow-up, were: older age, having CHF, having metastatic solid tumor, and hospital utilization for a COPD exacerbation during past year. To explore the risk factors for “early mortality,” we used a multivariate logistic regression model. Sex and marital status did not have independent association with mortality in univariate analysis and were excluded in multivariate model. Older age, having metastatic solid tumor, and hospital utilization for COPD exacerbation during the past year were independent risk factors of early mortality. The main results did not change when patients with stage 0 COPD were excluded (data not shown).

Table 4. Multivariate models of overall and early mortality in 482 emergency department patients with COPD exacerbation

	Overall Mortality		Early Mortality &%
	HR	95% CI	OR	95% CI
Age (per 5-year increase)	1.3	1.2–1.4	1.3	1.1–1.6
Female	0.9	0.7–1.2	—	—
Single, divorced, widowed, separated	1.2	0.9–1.6	—	—
Congestive heart failure	1.6	1.2–2.1	1.4	0.7–2.8
Metastatic solid tumor	3.3	2.0–5.5	7.0	2.6–19.1
Hospital utilization for COPD exacerbation during past year	1.9	1.4–2.6	2.2	1.1–4.6
Hospitalization at the index visit	1.2	0.9–1.6	1.2	0.6–2.4

COPD denotes chronic obstructive pulmonary disease; HR, hazard ratio; CI, confidence interval; OR, odds ratio.

*The Cox regression model the proportional hazards assumption is valid. Univariate results are shown in Table 2.

^†Early mortality denotes death within 60 days of the index visit. The logistic regression model gave a good fit to the data (Hosmer-Lemeshow test, p = 0.51). The area under the receiver operator curve was 0.73.

DISCUSSION

We examined the mortality experience of patients who visited the ED with a COPD exacerbation and found that mortality increased over time: 5% at 30 days, 9% at 60 days, 11% at 90 days, 16% at 180 days, 23% at 1 year, 32% at 2 years, and 39% at 3 years. Our study extends previous studies on long-term outcomes of patients who were hospitalized with a COPD exacerbation ([17], [20]) to the large patient population who visit the ED for treatment of their COPD exacerbations and demonstrated a comparable mortality rate.

Our results are comparable with those of prospective studies assessing the mortality of patients hospitalized for COPD ([13], [17], [20], [29]). However, our study showed a considerably lower mortality rate at 1 year compared to the SUPPORT study. The SUPPORT study included patients who were admitted to the hospital and had a PCO₂ of 50 mmHg or greater ([12]). Hypercapnia has been reported as a risk factor of mortality among COPD patients ([7]). Our cohort and those inpatient cohorts with 1-year mortality results similar to ours may have more heterogeneous patients compared to the SUPPORT cohort. Patients in the SUPPORT study were admitted and had severe disease as determined by hypercapnia. Less severe disease in our cohort may be an explanation for the observed differences in mortality rates. Age is a universal risk factor for mortality in many studies assessing outcomes after a COPD exacerbation ([12], [13], [17], [20]). Our study confirms this finding, not only for long-term outcomes, but also for early mortality (i.e., first 60 days after the index ED visit). Old age is known to be associated with a decline in lung function ([30]). The increased mortality of older patients might be due, in part, to the decreased lung function reported as a risk factor of mortality in COPD patients ([8], [9], [10], [15]) or to the many other chronic diseases associated with cigarette smoking ([31]).

A notable example is CHF, which has been reported as a risk factor for mortality among patients with a COPD exacerbation ([12]). We found the prevalence of CHF was 40% among our study population. Another study reported the prevalence of COPD was 66% among patients who were admitted with acute myocardial infarction and had CHF ([32]). Despite this high co-morbidity, and our finding of increased mortality of COPD patients who also have CHF, the interaction between the two conditions is not well understood. One possible explanation is deterioration of lung function in individuals with CHF. Lung function abnormalities that accompany CHF include restrictive and obstructive defects in acute heart failure and reduced carbon monoxide diffusion capacity in chronic congestive cardiac failure with proposed mechanisms of respiratory muscle weakness, reduced lung volume caused by cardiomegaly and alvelolar and interstitial fluid, changes in lung compliance, chronic edema, and interstitial fibrosis ([33], [34]). It is possible that COPD patients who have decreased lung function would be further affected by these added deteriorations in lung function caused by CHF, leading to increased mortality.

Another explanation is the potential limitation in available medications among patients with both COPD and CHF. Two important medications known to decrease mortality in CHF patients are angiotensin converting enzyme inhibitor and β blockers ([35], [36], [37], [38]). The American College of Cardiology/American Heart Association guidelines recommend using angiotensin converting enzyme inhibitors and β blockers in treating CHF ([39]). However, angiotensin converting enzyme inhibitors can cause persistent cough, and in COPD patients, it can be confused with respiratory infection, leading to inappropriate discontinuation of the medication. Similarly, β blockers can provoke bronchoconstriction in COPD with a bronchospastic component. Even β -1 selective blockers (e.g., metoprolol) lose their selectivity in doses that are proven to decrease mortality in CHF patients. For this reason, the American College of Cardiology/American Heart Association guidelines do not recommend β blockers for individuals with COPD ([39]). Patients with both COPD and CHF could have higher mortality because they cannot take medications that are known to decrease mortality from CHF. Future studies might focus on the mechanism causing increased mortality in patients with both conditions, and the tolerability of CHF medications in these patient populations.

We also found that having metastatic solid tumor is one of the strongest risk factors, resulting in a 3-fold increase in mortality. This seemingly rare condition was observed more frequently than expected (5%) in our study population of COPD patients, age 55 years or older. The effect of this condition had an even stronger effect on early mortality (OR, 7.0). Though it cannot be stressed enough to obtain a thorough medical history in all patients, even in busy ED settings, it might be desirable to focus on more simple but significant aspects of the medical history–such as known tumors.

Hospital utilization for COPD exacerbations during the past year also was an independent risk factor for mortality. To our knowledge, this is the first study that reports hospital utilization for a COPD exacerbation during the past year as a risk factor of mortality. While 77% of patients did not have record of hospital utilization for a COPD exacerbation, the rest of the patients had 1 to 9 ED visits or hospitalizations during the year preceding the index visit. Considering only about half of COPD exacerbations are brought to a physician's attention ([4]), the frequency of exacerbation during the past year is assumed to be twice as high as suggested by the records in our database.

In a previous study, we found that the frequency of COPD exacerbations during the past year was a risk factor for relapse shortly after an ED visit for COPD exacerbation ([40]). With studies that reported different inflammatory status between patients with frequent and non-frequent exacerbations ([40], [41], [42], [43]), and a report that found that exacerbation frequency was a consistent feature within a patient ([44]), we concluded that there are various degrees of inflammation in COPD patients either induced by imbalance of inflammatory/anti-inflammatory factors or by viral and/or bacterial colonization. Additionally, the degree of inflammation can determine the frequency of exacerbations and the outcome of these exacerbations.

With the findings from the current study, we suspect that different inflammation status also has an effect on disease status and, ultimately, mortality. Additionally, patients who have very progressive disease might experience more frequent exacerbations, ultimately leading to mortality. Though further study on the underlying pathophysiology of frequent exacerbations and their effect on outcomes is necessary, the number of ED visits or hospitalizations can be used as a simple assessment of COPD exacerbations and contribute to mortality prediction.

While studies ([12], [13]) have reported pulmonary function as an important predictor of mortality after a COPD exacerbation, we did not find an association between COPD stage, determined by predicted FEV₁ and FEV₁/FVC, and mortality. However, only 117 (24%) out of 482 patients had pulmonary function test results within 2 years of the index visit, so our data on pulmonary function are incomplete. The lack of pulmonary function testing was reported before ([41]). Damarla et al. studied patients with COPD and found that 30% of 553 patients had a pulmonary function test performed within 7 years, and it was significantly lower than performance of 2-D echocardiography in CHF patients. This underperformance of pulmonary function testing might arise from the fact that COPD may exhibit gradual progression, resulting in physicians waiting long periods of time between pulmonary function testing. Spirometry, however, is the best way to track COPD patients' decline in lung function ([1]), and pulmonary function is an important predictive factor of mortality in some studies in the general COPD population ([42], [43]). GOLD recommends periodic pulmonary function test, not more frequent than once a year ([1]).

For physicians who treat COPD patients in EDs, short-term outcomes are of perhaps greater interest. The mortality rates within 30, 60, and 90 days of the index visit were 5%, 9%, and 11%, respectively. Given that the mortality rates after acute myocardial infarction are approximately 19%, 22% and 24% over this same time frame ([44]), the short-term mortality of ED patients with COPD exacerbation merits more attention.

This study has a few potential limitations. First, case ascertainment depended on ICD-9 codes in the hospital records. Others have reported that ICD-9 codes correctly identified 87% of patients with a COPD exacerbation ([12]). Thus, it is possible that we misclassified some patients with other conditions as cases with COPD exacerbation. Given 27% of 117 patients in whom pulmonary function test results were available had COPD stage 0 by GOLD criteria ([1]), our database might include patients who did not have COPD. However, we limited our database to patients age 55 and over, and patients whose admitting or primary diagnosis was one of three COPD diagnoses, so we were unlikely to include patients in our database with conditions other than COPD. Also, when patients with stage 0 were excluded, the main results did not change.

Second, because this cohort was obtained from an administrative database, we could not include potential risk factors such as medication ([20]), quality-of-life ([17]), respiratory physiologic variables ([11]) or depression ([17]) that have been suggested as risk factors of mortality after COPD exacerbation. However, our study found risk factors that can be easily obtained by administrative database, or by simple questions at patient interview. Accordingly, the information can be used in future outcomes studies or in the clinical setting.

Finally, we used data from the SSADMF to determine vital status. It was reported that the sensitivity of detecting death by using the SSADMF was 82%, with increasing sensitivity in older population ([45]). Our database was restricted to patient age 55+, and we did not only depend on SSADMF but also used hospital records in cases where the patient died in the hospital. With the characteristics of our cohort, and data input from hospital medical records, we believe that mortality information is sound.

In summary, the results of our study suggest that the mortality rate of ED patients with a COPD exacerbation is quite significant, both in the short-term and long-term. Independent, significant risk factors of mortality in these patients were older age, having CHF, having metastatic solid tumor, and hospital utilization for COPD exacerbation during past year. Further study might focus on how CHF affects, survival of COPD patients, and on the pathophysiology of frequent COPD exacerbations and their contribution to mortality and other major clinical outcomes.

Drs. Kim and Camargo were supported, in part, by an Emergency Medicine Foundation Center of Excellence Award (Dallas, TX), and Ms. Clark by grant T32 ES07069 from the National Institutes of Environmental Health Sciences (Research Triangle Park, NC).

We thank Henry Chueh, Shawn N. Murphy, and the entire Research Patient Data Repository team for having created and maintained the Research Patient Data Repository, and for their assistance in the conduct of the current study.