COPD is associated with increased mortality in patients with community-acquired pneumonia

Study sites and inclusion and exclusion criteria

All patients admitted to the study hospitals between January 1, 1999 and December 1, 2002 with a primary discharge diagnosis of pneumonia (International Classification of Diseases, ninth revision (ICD-9) codes 480.0–483.99 or 485–487.0) or secondary discharge diagnosis of pneumonia with a primary diagnosis of respiratory failure (518.81) or sepsis (038.xx) were identified. Patients were included in the present study if they were: 1) aged ≥18 yrs; 2) had an admission diagnosis of CAP; and 3) underwent chest radiography or chest computerised tomography within 24 h of admission with findings that were consistent with CAP.

Exclusion criteria included: 1) discharge from an acute care facility within 14 days of admission; 2) transfer after being admitted to another acute care hospital; 3) HIV/AIDS; and 4) having “comfort measures only” status during the admission. If a subject was admitted more than once during the study period, only the first hospitalisation was abstracted.

Definition of COPD

The definition of COPD was based on baseline clinical data obtained by chart review. Owing to the absence of pulmonary function test results, terms including “chronic obstructive pulmonary disease”, “emphysema” and “chronic bronchitis” were used as proxy measures for COPD. Missing values or data were assumed to be normal and not show COPD. This strategy is widely used in the application of prognostic prediction rules and reflects the methods used in the original pneumonia severity index (PSI) score studies 10, 12.

The primary outcomes were 30- and 90-day mortality, and secondary outcomes were length of hospital stay, need for intensive care unit (ICU) admission and need for mechanical ventilation. Both 30- and 90-day mortality were included to differentiate between the proportion of deaths attributable to pneumonia and other causes, respectively 13.

Mortality was assessed using information from the Texas Department of Health and Department of Veterans Affairs clinical database (Austin, TX, USA). Mortality status was assessed during December 2002.

Data abstraction

Chart review data included: demographics, comorbid conditions, physical examination findings, laboratory data, and chest radiographic reports. Comorbid conditions were identified from either the admission note or the chart problem list. In addition, data on important processes of care measures for patients hospitalised with CAP were also abstracted: 1) timing of first dose of antibiotics; 2) collection of blood cultures prior to antibiotic administration; and 3) obtention of blood cultures and oxygen saturation measurements within 24 h of presentation 14, 15. Antimicrobial therapy was considered guideline-concordant if it agreed with either the 2003 Infectious Diseases Society of America (IDSA) or 2001 American Thoracic Society (ATS) guidelines 14, 15. Information on outpatient corticosteroid use was recorded from either the admission note or the electronic medical record.

Diagnostic criteria

Microbiological data were reviewed, and a microbiological cause assigned independently by two of the principal investigators (M.I. Restrepo and E.M. Mortensen). The cause of pneumonia was stratified as definitive or presumptive. A definitive diagnosis was considered to be present if one of the following conditions were met: 1) positive blood cultures for bacterial or fungal pathogens (in the absence of an extrapulmonary source of infection); 2) pleural fluid cultures yielding a bacterial pathogen; 3) endotracheal aspirates with moderate or heavy growth of bacterial pathogens; and 4) significant quantitative culture growth from bronchoscopic respiratory samples (≥1×10³ colony-forming units (cfu) per mL in protected specimen brush cultures, and ≥1×10⁴ cfu·mL^-1 in bronchoalveolar lavage fluid). In addition, a presumptive diagnosis was made if qualitative valid sputum samples yielded one or more predominant bacterial pathogen. Definitive and presumptive causes were combined for reporting purposes. When two or more microbiological causes were present, the cause was classified as polymicrobial pathogens. A patient was considered to have CAP of unknown cause if no diagnostic tests were performed, or tests were performed but test results did not meet criteria for assigning a microbiological cause (including a contaminant pathogen).

Risk adjustment

The PSI was used to assess severity of illness on presentation. The PSI is a validated prediction rule for 30-day mortality in patients with CAP 10. This rule was based on three demographic characteristics, five comorbid illnesses, five physical examination findings, and seven laboratory and radiographic findings from the time of presentation. Patients were classified into five risk classes, with 30-day mortality ranging from 0.1% for Class I to 27% for Class V for patients enrolled in the PORT cohort study 10.

Statistical analyses

Univariate statistics were used to test the association of demographic and clinical characteristics with all-cause 30- and 90-day mortality. Categorical variables were analysed using the Chi-squared test and continuous variables using an unpaired t-test.

Two Cox’s proportional-hazards models were used to estimate, and plot, baseline survivor functions with either 30- or 90-day mortality as the dependent variable 16. Variables were included in the survival analysis if they had either been previously demonstrated to be associated with CAP-related outcomes (e.g. PSI and processes of care) 10 or a p-value of <0.10 in the univariate analyses. History of COPD was entered into the model as an independent dichotomised variable, and PSI score was used as the risk adjustment tool 10. In addition, processes of care measures (initial antibiotics within 4 h, obtention of blood cultures prior to initial dose of antibiotics, and whether antimicrobial therapy was guideline-concordant) were used as potential confounding variables.

Patient characteristics

The cohort consisted of 582 (78%) males and 162 (22%) females. Their mean±sd age was 61±16 yrs (range 18–105 yrs). Of the patients, 83% were admitted via the emergency department from their own home and 7% from a nursing home; 128 (17%) had received outpatient antibiotic therapy prior to admission (table 1⇓).

One or more concomitant comorbid medical conditions were present in 635 (85%) patients. The most frequently associated conditions were congestive heart failure (123 patients) and a prior history of stroke (105 patients). No known prior significant medical disorders existed in 109 (15%) patients. CAP patients with COPD were significantly more likely to be older males, admitted from a nursing home and placed in the ICU within 24 h of admission (table 1⇑). In addition, COPD patients with CAP showed higher rates of congestive heart failure and a history of neoplastic disease. Physical examination and laboratory/radiological data demonstrated that CAP patients with COPD were significantly more likely to be tachypnoeic, acidotic and hypoxaemic, but less likely to exhibit hyperglycaemia than non-COPD patients. COPD patients used inhaled corticosteroids more frequently; however, they used similar amounts of systemic steroids to patients without COPD (table 1⇑).

Administration of antibiotics within 4 h occurred more commonly in COPD patients (35 versus 26%; p = 0.02). However, COPD patients received other processes of care less commonly, including collection of appropriate blood cultures prior to antibiotics and within the first 24 h (70 versus 77%; p = 0.04), and tended to have received antimicrobial therapy not concordant with the recommendation from national guidelines (74 versus 80%; p = 0.05). There were no significant differences in the rate of oxygenation status assessment.

Pneumonia aetiology

An aetiological diagnosis could not be obtained in 77% of the cohort. A microbiological diagnosis was assigned in 172 (23%) patients with microorganisms identified from cultures of blood and/or sputum. Diagnosis was established by a positive blood culture in 63 (8.5%) patients. Bacteraemia was present in 53 (10%) hospitalised CAP patients without COPD and 10 (4.6%) patients with COPD. The organisms most frequently involved were Streptococcus pneumoniae (56 cases) and Staphylococcus aureus (39 cases). Hospitalised CAP patients with COPD showed more infections attributable to Pseudomonas aeruginosa, a trend of higher rates of Haemophilus influenzae, but less S. aureus than patients without COPD (table 2⇓). However, there were no other significant differences between other pathogens in either group.

Clinical outcomes

For this study cohort, the median length of stay was longer by 2 days in COPD versus non-COPD patients (7±8 versus 9±25 days; p = 0.05). The overall 30- and 90-day mortality were 10 and 14%, respectively. The unadjusted mortality was lower for non-COPD patients than COPD patients: 30 day, 8.7 versus 10.6% (p = 0.4); 90 day, 11.7 versus 18.6% (p = 0.013). The mean PSI score was significantly higher for COPD patients than for CAP patients without COPD (105±32 versus 87±34; p = 0.05). Of the patients, 148 (20%) were admitted to the ICU and 83 (14%) required mechanical ventilation. When comparing COPD and non-COPD patients, the proportion of patients that required hospitalisation in the ICU was higher among COPD patients, but the percentage needing mechanical ventilation was not (table 3⇓). There were no differences in mortality within 30 or 90 days for CAP patients with COPD who needed ICU admission, received mechanical ventilation or were bacteraemic (table 3⇓). In addition, CAP patients with COPD receiving any form of corticosteroids, whether inhaled or systemic, did not show any significant differences in 30- or 90-day mortality compared with non-COPD patients (table 3⇓).

In the Cox’s proportional-hazards model, after adjusting for potential confounders, including processes of care and severity of illness, patients with a history of COPD exhibited significantly increased 30- (hazard ratio (HR) 1.32; 95% confidence interval (CI) 1.01–1.74) and 90-day mortality (HR 1.34; 95% CI 1.02–1.76). The Cox survival curve shows the variation in mortality from the initiation of the hospitalisation to 30 and 90 days for patients with and without COPD (fig. 1⇓).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Cox survival curve showing proportion of surviving hospitalised community-acquired pneumonia patients by the presence (-----; n = 215) or absence (––––; n = 529) of chronic obstructive pulmonary disease after adjusting for other potential confounders.

Discussion

In the present study, it was found that COPD patients hospitalised with CAP, compared to patients without COPD, show significantly higher 30- and 90-day mortality. In addition, hospitalised CAP patients with COPD exhibit significantly higher rates of ICU admission and a longer length of hospital stay compared with those without COPD. These data confirm that COPD should be considered for inclusion as a comorbid condition for pneumonia severity of illness measures.

COPD is considered a risk factor for the development of CAP, and previous studies of CAP including outpatient, inpatient and ICU cohorts have shown that COPD is a frequently reported comorbid condition 3, 4, 9, 17–22. However, COPD has not been previously identified as being a risk factor for mortality in CAP patients 6–9. The present study showed that hospitalised CAP patients with COPD show higher mortality at 30- and 90-days compared to patients without CAP. The present authors believe that this difference was found by examining only patients with COPD, and excluding other pulmonary conditions, including asthma, bronchiectasis and interstitial lung disease. Therefore, it was possible to examine the impact of COPD without dealing with other potential confounding pulmonary conditions. Torres et al. 23 showed, in a large Spanish multicentric study, an in-hospital mortality rate of 8% in a cohort of 124 CAP patients with COPD. However, the main difference, compared with the present study, was the lack of a comparison group of CAP patients without COPD, which limited their ability to compare clinical outcomes.

There were important differences in COPD versus non-COPD patients, which included being older males, more admissions from nursing homes and requiring ICU treatment within 24 h of admission. In addition, COPD patients with CAP were more tachypnoeic, acidotic and hypoxaemic. All of these variables are also included in the PSI score, the severity of illness predictor used in the present study 10. The PSI score assesses five comorbid conditions (cardiovascular, history of malignancy, cerebrovascular, renal and liver diseases), but does not include COPD as one of them 10.

It was expected that COPD patients hospitalised with CAP, who had higher PSI scores, rates of ICU admission and a longer length of stay in the hospital, would also show a higher mortality. One of the possible explanations for not finding a higher mortality in these specific groups is that the PSI score does not completely adjust for all of the abnormalities that are common in COPD patients. In addition, the ability to accurately predict medical outcomes in CAP influences patient management decisions made by physicians. These include the decision to hospitalise the patient, the length of inpatient care if the patient is hospitalised, and the choice of antimicrobial or other types of immediate therapy. The choice of appropriate empirical antibiotic regimens depends on several factors, including the aetiology of CAP.

The present results show that hospitalised CAP patients with COPD had more infections attributable to P. aeruginosa. Other studies have found that P. aeruginosa is an important pathogen in patients with pulmonary comorbid conditions, especially those with bronchiectasis 3, 23, 24. The present data show that P. aeruginosa was the second-most-common organism in patients with COPD; therefore, appropriate anti-pseudomonal coverage should be considered in patients with COPD, whether or not bronchiectasis is present. The present data support the IDSA 14 and ATS 15 clinical practice treatment guidelines, which recommend stratifying patients based on the presence of coexisting cardiopulmonary disease (COPD and congestive heart failure) in order to select an appropriate antimicrobial agent. Therefore, it is important to recognise COPD in patients with CAP so that they may receive appropriate antimicrobial therapy.

The present study has several limitations that are important to acknowledge. First, it was a retrospective cohort study, and inherent problems related to this design include ascertainment and selection bias. However, the present authors do not feel that this study has significant problems with either of these biases, due to the methods involving admission and discharge diagnosis ICD-9 codes to identify patients and the fact that only a small amount of missing data was encountered. Furthermore, it was possible to verify that all of the patients had a radiological diagnosis of CAP. Secondly, the present sample was predominantly male since one of the sites was a Veterans Administration hospital and so it was not possible to examine whether or not females with COPD and CAP may exhibit a different clinical course, or outcomes, compared with males. Thirdly, unfortunately, no serological information was available, including Legionella urinary antigen. Fourthly, patients were identified as having COPD by medical history, which has been demonstrated in past studies to be the same method as that used to identify other comorbid conditions included to create the PSI score 6, 10, 23, 25, 26. However, it was not possible to collect data regarding pulmonary function tests or COPD disease severity. Pulmonary function data could be helpful in predicting which patients with COPD might show the highest morbidity and mortality when they develop CAP.

In conclusion, the present study demonstrates significantly higher 30- and 90-day mortality, and increased length of stay and intensive care unit admission in chronic obstructive pulmonary disease patients hospitalised for community-acquired pneumonia compared with patients without chronic obstructive pulmonary disease. These findings have implications regarding how to evaluate patients with community-acquired pneumonia and chronic obstructive pulmonary disease, and how to decide which antimicrobial agents should be used for initial empirical therapy. Further prospective cohort studies are warranted to determine the impact of chronic obstructive pulmonary disease severity (confirmed by pulmonary function tests) and the use of anti-pseudomonal antibiotics on clinical outcomes for chronic obstructive pulmonary disease patients with community-acquired pneumonia.