|Home | About | Journals | Submit | Contact Us | Français|
Aspiration pneumonia is a common syndrome, although less well characterized than other pneumonia syndromes. We describe a large population of patients with aspiration pneumonia.
In this retrospective population study, we queried the electronic medical record at a tertiary-care, university-affiliated hospital from 1996–2006. Patients were initially identified by ICD-9 code 507.x; subsequent physician chart review excluded patients with aspiration pneumonitis and those without a confirmatory radiograph. Patients with community-acquired aspiration pneumonia were compared to a contemporaneous population of community-acquired pneumonia (CAP) patients. We compared CURB-65 predicted mortality with actual 30-day mortality.
We identified 628 patients with aspiration pneumonia, of which 510 were community-acquired. Median age was 77, with 30-day mortality of 21%. Compared to CAP patients, patients with community-acquired aspiration pneumonia had more frequent inpatient admission (99% vs. 58%) and ICU admission (38% vs. 14%), higher Charlson comorbidity index (3 vs. 1), and higher prevalence of “do not resuscitate/intubate” orders (24% vs. 11%). CURB-65 predicted mortality poorly in aspiration pneumonia patients (AUC 0.66).
Patients with community-acquired aspiration pneumonia are older, have more comorbidities, and demonstrate higher mortality than CAP patients, even after adjustment for age and comorbidities. CURB-65 poorly predicts mortality in this population.
Pneumonia is a common clinical syndrome with well-described epidemiology and microbiology. Aspiration pneumonia comprises 5–15% of patients with pneumonia1, but is less well-characterized despite being a major syndrome of pneumonia in the elderly2,3. Difficulties in studying aspiration pneumonia include the lack of a sensitive and specific marker for aspiration, the overlap between aspiration pneumonia and other forms of pneumonia, and the lack of differentiation between aspiration pneumonia and aspiration pneumonitis by many clinicians4–6. Aspiration pneumonia, which develops after the aspiration of oropharyngeal contents, differs from aspiration pneumonitis, wherein inhalation of gastric contents causes inflammation without the subsequent development of bacterial infection7,8.
A number of validated mortality prediction models exist for community-acquired pneumonia (CAP), using a variety of clinical predictors. We favor eCURB, a version of the CURB-65 model which uses continuously weighted variables to more accurately predict mortality, validated in CAP populations9. Most studies validating pneumonia severity scoring systems excluded aspiration pneumonia from their study population10–12. Severity scoring systems for CAP may not accurately predict disease severity patients with aspiration pneumonia.
The aims of our study were to: (1) identify a population of patients with aspiration pneumonia; (2) compare characteristics and outcomes in patients with community-acquired aspiration pneumonia to those with CAP; and (3) study the performance of eCURB and CURB-65 in predicting mortality for patients with community-acquired aspiration pneumonia.
The study was performed at LDS Hospital, a university-affiliated community teaching hospital in Salt Lake City, Utah, USA, with 520 beds. In retrospective analysis of data from the electronic medical record, we identified all patients older than 18 evaluated in the emergency department at LDS hospital or admitted patients from other sources (direct admission, transfer from other hospital) from 1996 to 2006 with ICD-9 (International Statistical Classification of Disease and Health Related Problems – version 9) codes specific for aspiration pneumonia and pneumonitis (507.x). The treating physicians were mostly hospitalists and intensivists. Two physicians (ML and ND) manually reviewed the electronic medical record, including the emergency room physician's note, the admission history and physical, the discharge summary, and radiographic reports of the patients identified in the query. Consensus regarding the diagnosis of aspiration pneumonia was achieved in all patients reviewed using criteria listed in Table 1. This study was approved by the LDS Hospital Institutional Review Board, and permission was granted to use the Utah Population Database for determining mortality (#1008505), with waiver of informed consent. For the contemporaneous group of CAP patients, we used a previously described population identified using ICD-9 codes 481.x-487.x, captured from the same hospital during the same period13.
Inclusion and exclusion criteria are listed in Table 1. In order to exclude patients with recurrent pneumonia, we included only the first episode of pneumonia in a given 12-month period. LDS hospital frequently receives patients transferred from surrounding emergency departments and intensive care units. We excluded patients who were transferred >48 hours from their initial emergency department admission and therefore were late in their disease course. Exclusion criteria 8–10 were used to exclude patients with clinical presentations more consistent with aspiration pneumonitis. We also excluded immunocompromised patients (criteria 4–7).
Healthcare-associated aspiration pneumonia was defined as receipt of chronic hemodialysis, residence in a nursing facility, or hospitalization within any Intermountain Healthcare affiliated hospital within the past 90 days14. The remaining patients were defined as community-acquired aspiration pneumonia.
The first vital signs, orientation status, and first 12 hours of routine laboratory results were extracted from the electronic medical record and used to calculate predicted mortality by eCURB and CURB-65. We determined 30-day mortality from the merger of the electronic medical record (EMR) with vital status information from the Utah Population Database15. The first measured SpO2 and FiO2 were used to estimate the PaO2/FiO2 ratio, using the Severinghaus calculation16 if no arterial blood gas was available. Presence of ATS/IDSA 2007 minor criteria for severe community-acquired pneumonia (SCAP)17 were obtained from baseline patient characteristics (Table 2). A Charlson comorbidity index was calculated from ICD-9 codes, using published methodology18,19. Presence of an abnormal swallow was defined as dysphagia or aspiration on modified barium swallow study, fiberoptic endoscopic evaluation, or clinical determination by a speech language pathologist during the index hospitalization. We also looked for causative pathogens, defined by a positive pneumococcus or legionella urinary antigen, or a positive culture from blood, bronchoalveolar lavage, pleural fluid, or tracheal aspirate, collected within 24 hours of admission. Antibiotics administered within the first 24 hours of admission were classified into four broad groups, based on local physician prescribing patterns. Clindamycin and metronidazole were considered anaerobic-specific antibiotics. Vancomycin or linezolid were considered methicillin-resistant S. aureus (MRSA) antibiotics. Broad spectrum antibiotics included any of the following: carbapenems, aztreonam, piperacillin/tazobactam, ticarcillin/clavulanate, cefepime, ceftazidime. Macrolides, respiratory fluoroquinalones, and third generation cephalosphorins were considered standard care antibiotics.
We compared baseline patient characteristics and clinical outcomes using Fisher's exact test to compare proportions of categorical variables, and used Mann-Whitney U-tests or Student's T-test to compare central tendencies of continuous variables, as dictated by the normality of the data. Receiver-operating characteristic curves were calculated the ability of eCURB and CURB-65 to predict 30-day mortality prediction in patients with community-acquired aspiration pneumonia and CAP, as well as the ability of IDSA/ATS minor criteria for SCAP to predict admission to the ICU. We performed multivariate logistic regression to predict 30 day mortality in patients with community-acquired aspiration pneumonia and CAP, using stepwise backwards elimination. Confounders were included if they were significant at a 0.05 level or they altered the coefficient of the main variable by more than 10 percent. For logistic models, we evaluated goodness of fit with the Hosmer-Lemeshow technique; comparisons of area under the curve (AUC) among models were made using the technique of DeLong20. Two-tailed p-values of ≤ 0.05 were considered statistically significant. Stata v12 (StataCorp, College Station, TX) was used for all analyses.
Our initial query identified 1165 patients. Physician review of the medical record resulted in 628 patients, 118 of which were classified as healthcare-associated aspiration pneumonia (Figure 1,Table 3). Of All aspiration pneumonia patients, 80% were seen in the emergency department, 12.5% were directly admitted from the community, and 7.5% were transferred from another health care facility. Almost all patients seen in the emergency department (99.0%) were admitted to the hospital, with median length of hospitalization 6.7 days among survivors.
Observed mortality was 21.0%. eCURB significantly underestimated mortality in this group, predicting a mortality rate of 10.6%. When classifying patients by the 2007 IDSA/ATS guidelines, 24.7% of the patients had three or more minor criteria for SCAP17. The PaO2/FiO2 ratio was obtained in 99.7% of patients. The median PaO2/FiO2 ratio observed in this population was 221 mm Hg (equivalent to 260 mm Hg at sea level barometric pressure, adjusted for our altitude of 1400 meters), near the threshold sea level definition (250 mm Hg) for SCAP13,17. Admission to the ICU was common, as were admission orders for “Do Not Resuscitate” (DNR) or “Do Not Intubate” (DNI). Patients with healthcare-associated aspiration pneumonia had a higher comorbidity index and had a higher mortality rate than patients with community-acquired aspiration pneumonia, although we found no significant difference in the rate of hospital or ICU admission or the receipt of critical care therapies. Inpatient assessment of dysphagia and aspiration was conducted in 177 patients. Abnormal swallow was noted in 96% of those tested.
We found several differences between patients with community-acquired aspiration pneumonia and 2,584 patients with CAP identified during the same time period13(Table 4). Patients with community-acquired aspiration pneumonia were older, more likely to have multilobar disease or effusion on imaging, and had greater disease severity. They also had a higher frequency of ICU and hospital admission, IDSA/ATS minor criteria for SCAP, and higher Charlson comorbidity indices. Patients with community-acquired aspiration pneumonia were more likely to receive mechanical ventilation than CAP patients, although there was no difference in 30-day mortality among intubated patients, nor a difference in ventilator-free days.
Thirty-day mortality for patients with community-acquired aspiration pneumonia was significantly higher than in CAP patients. Patients with community-acquired aspiration pneumonia also had higher eCURB and CURB-65 scores. However, eCURB was a poor predictor of 30-day mortality, with an AUC of 0.71, compared to 0.86 calculated for the CAP population (Figure 1). CURB-65 performed similarly: AUC was 0.66 versus 0.81. The presence of a DNR/DNI order was twice as prevalent in the community-acquired aspiration pneumonia population vs. the CAP population; those patients with a DNR/DNI order were 3 times as likely to die. Excluding patients with a DNR/DNI order did not improve performance of eCURB or CURB-65 (Table 4). The presence of IDSA/ATS minor criteria for SCAP was not predictive of triage to the ICU in the group of patients with community-acquired aspiration pneumonia (AUC 0.51), compared with CAP patients (AUC 0.88, p < 0.01 for comparison, Figure 2). This finding persisted in the subset of patients without a DNR/DNI order (AUC 0.52 in community-acquired aspiration pneumonia vs. 0.88 in CAP, p < 0.01).
Our regression model of mortality incorporated gender, presence of effusion or multilobar pneumonia, presence of a DNR/DNI order, and all the components of the CURB-65, IDSA/ATS minor criteria for SCAP. and Charlson comorbidity index. The regression model demonstrated that even after adjustment for age, comorbidities, disease severity, and presence of a DNR/DNI order, the presence of aspiration pneumonia was associated with higher mortality than CAP (OR 3.46, p <0.001, Table 5). In this model, systolic blood pressure did not predict mortality, and diabetes with complications was associated with decreased mortality.
Blood cultures were performed at admission in 67.4% of aspiration pneumonia patients, and a tracheal aspirate in half (50.7%) of intubated patients with aspiration pneumonia. Organisms were recovered in 90 patients (14.3%), although 41 of those patients had tracheal aspirates of organisms commonly thought to be non-pathogenic (non-pneumococcal alpha-hemolytic streptococcus, non-hemolytic streptococcus, diphtheroids, micrococci, coagulase negative staphylococccus). Tracheal aspirate was the most common method of recovering an organism (7.8% of patients), followed by blood culture (4.3%). Bronchoalveolar lavage, urinary antigen, and pleural fluid culture were less common (1.3%, 1.1%, 0.3%, respectively). The microbiologic results were grouped into: S. aureus, S. pneumoniae, enteric bacilli, Hemophilus species, Neisseria species, M. catarrhalis, and P. aeruginosa (Figure 4). Comparing healthcare-associated with community-acquired aspiration pneumonia, healthcare-associated patients were more likely to have a confirmed infection with MRSA (4.2% vs. 1.4%, p= 0.06), and enteric bacteria (5.1% vs. 1.6%, p = 0.03). There were no other statistically significant differences in microbiologic recovery between the two groups. Antibiotics targeting anaerobic pathogens were administered in 28.7% of patients with aspiration pneumonia, with no correlation to the presence of healthcare-associated risks. Healthcare-associated patients were more likely to receive broad-spectrum antibiotics (47.5% vs. 32.5%, p < 0.01) and MRSA coverage (15.3% vs 5.7%, p < 0.01) than patients with community-acquired aspiration pneumonia.
Our study identifies a larger cohort of patients with aspiration pneumonia than previous studies21–25. Patients with community-acquired aspiration pneumonia are older and more likely to die than CAP patients. They are more likely to be admitted to the hospital or intensive care unit. Thirty-day mortality in this patient population was significantly underestimated by CURB-65 and eCURB, models developed and validated in CAP populations9,26. This finding supports a prior study27. It appears that a traditional prognostic model assessing mortality risk in the CAP patient does not apply to the aspiration pneumonia patient. One reason for eCURB and CURB-65's poor utility in community-acquired aspiration pneumonia may be their reliance on objective clinical features rather than comorbidities, which may influence mortality to a greater degree in aspiration pneumonia.
This study has several limitations. There is no gold standard for the definition of aspiration pneumonia, it is difficult to distinguish aspiration pneumonia from typical pneumonia. It is plausible that older patients with greater comorbidities are being designated as aspiration pneumonia. If this is the case, then aspiration pneumonia merely represents the end of the pneumonia spectrum with highest mortality risk, and it is no surprise that these patients fare poorly.
It appears that the hospitalist or emergency department physician implicitly appreciates that aspiration pneumonia has a higher mortality risk than predicted by traditional severity assessment. With such high mortality and morbidity, a patient presenting to the emergency room with aspiration pneumonia is almost always admitted to the hospital. Further work in this area should investigate other factors to improve prognostic modeling in patients with aspiration pneumonia, although the utility of such a model may be limited to determining ICU admission. Our data indicate that IDSA/ATS minor criteria for SCAP are not useful in predicting admission to the ICU in patients with aspiration pneumonia.
In this study, a DNR/DNI order was twice as common in the community-acquired aspiration pneumonia population than the CAP population. However, patients with community-acquired aspiration pneumonia and a DNR/DNI order were more than three times more likely to die than patients with CAP and a DNR/DNI order. Our regression model suggested that the presence of a DNR/DNI order was an independent predictor of mortality (OR 1.75, p <0.001). While a DNR/DNI order may correlate with the withholding or withdrawal of medical therapy, it is also a surrogate for increased age or comorbidities28. In our study, however, the increased prevalence of DNR/DNI orders did not explain the poor mortality prediction of the eCURB or CURB-65, as exclusion of those patients did not significantly alter the AUCs in either the aspiration group or the CAP group.
Controversy exists regarding treatment of aspiration pneumonia. Historically, some have advocated for treatment of aspiration pneumonia with a regimen designed to cover anaerobic bacteria29. This recommendation was based on early microbiologic studies which obtained the samples late in the course of illness, or other studies where the sample was obtained transtracheally, where oropharyngeal flora may contaminate the sample30–32. Our clinically obtained microbiologic recovery of organisms was similar to the flora recovered in more recent CAP studies, in respect to both the incidence of pathogen recovery and the relative frequencies of recovered organisms33,34. Our data do not support inferences regarding the prevalence of anaerobic infections, as the recovery of anaerobic organisms was limited to blood and pleural fluid cultures in this study, rather than techniques used in research settings that might have greater yield. As expected, patients with healthcare associated risk factors trended toward increased incidence of MRSA. Given the similarity of the organisms recovered to those recovered in CAP35, this study supports IDSA/ATS recommendations that antibiotic therapy in aspiration pneumonia be similar to that of higher-risk CAP, with the addition of vancomycin or linezolid for MRSA coverage in patients with risk factors for healthcare-associated pneumonia17.
Our study is limited by its single-center, retrospective design. However, beginning in 1995, the LDS Hospital ED initiated a standardized pneumonia therapy protocol and deployed an electronic medical record, which prospectively recorded a wide array of clinical, therapeutic, and biometric data. Most data elements used in this analysis were routinely charted for clinical purposes in real time. While the eCURB, CURB-65, and some comorbidities could be extracted electronically for each patient, it was not possible to calculate the pneumonia severity index score due to our inability to rigorously identify the necessary comorbid illness elements. Other comorbidities, not present in our model, may have been identified by the physician who makes a diagnosis of “aspiration pneumonia”. Our identification of swallow impairment is also methodologically limited. The decision to obtain a swallow study was clinical, usually occurring upon convalescence. Therefore, it is not possible to distinguish between antecedent oropharyngeal dysfunction and post-critical illness dysfunction.
Our definition of aspiration pneumonia required the treating physician to diagnose and code the patient as having aspiration pneumonia, followed by excluding patients more likely to have aspiration pneumonitis. Although we relied on ICD-9 codes to initially identify aspiration pneumonia, all patients in our database were confirmed by physician chart review. Our incidence of community-acquired aspiration pneumonia is congruent with other studies using different methodologies1,36,37. Unfortunately, there is no standard and widely accepted definition for separating aspiration pneumonia from usual CAP. A younger and healthier patient who has developed pneumonia subsequent to aspiration may be more likely to be diagnosed with CAP, resulting in selection bias for older patients with greater comorbidities.
Patients diagnosed with aspiration pneumonia are older, have more comorbid conditions, and demonstrate greater disease severity and higher 30-day mortality than CAP patients. Mortality prediction using CURB-65 and eCURB in this population was poor, possibly due to a greater effect of comorbidities on mortality. The pneumonia severity index, which incorporates patient comorbidities, might perform better than the eCURB or CURB-65, and should be studied in aspiration pneumonia populations where comorbid illness information is prospectively collected. Further areas of study include creating an improved mortality prediction model for aspiration pneumonia that incorporates comorbid conditions, DNR/DNI status, and disease severity.
The authors would like to acknowledge Al Jephson for database support, Yao Li for statistical analysis, and Anita Austin for help reviewing the medical records. Dr. Lanspa had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Support: This study was supported by grants from the Intermountain Research and Medical Foundation.
Dr. Brown is supported by a career development award from National Institute of General Medical Sciences (K23GM094465).
Dr. Dean serves on an advisory board for Merck, has been a paid consultant for Cerexa, and has received an investigator-initiated competitive grant from Pfizer for development of an electronic pneumonia decision support tool.
Preliminary versions of this work were presented as posters at the American Thoracic Society on May 17th, 2011, in Denver, Colorado.
Potential conflicts of interest: All other authors report no relevant financial disclosures.