In this study, we sought to determine whether clinical, laboratory and imaging features of patients who are hospitalized for CAP and test positive for a respiratory virus are sufficiently distinct to safely avoid empiric antibiotic therapy at admission. Our study differs from previous ones: (1) we only included patients who provided a high-quality purulent sputum, defined as showing ≥ 25WBC/epithelial cell on microscopic examination, at or shortly after admission; (2) quantitative bacteriologic testing was done on all sputum samples; and (3) patients with pneumonia due to recognized bacterial pathogens or due to commensal bacteria were included. Admittedly, reliance on a valid sputum sample introduces a selection bias, but there is no other way to establish the diagnosis in most cases, and, with this methodology, we were able to achieve an etiologic diagnosis in every case and could, with high likelihood, establish or exclude a bacterial infection.

Our results show that, in patients hospitalized for pneumonia, the clinical presentation, laboratory and radiologic findings do not differ sufficiently among those whose PCR is positive for a respiratory virus to determine whether a bacterial coinfection is present. An exception might be the absence of bacteria on microscopic examination of a high-quality sputum sample from a patient who has not received an antibiotic. Otherwise, there were no differences that would support a decision to withhold antibiotic treatment in patients admitted for CAP who have a positive viral PCR. These findings are consistent with current ATS/IDSA guidelines that recommend antibiotic therapy in all patients who are sick enough to be hospitalized for CAP even if they test positive for the presence of a respiratory virus [6]. Patients with bacterial pneumonia were more likely to have been smokers and to have chronic pulmonary disease. Upper respiratory symptoms were more common in patients with viral pneumonia, and peripheral blood WBC counts were higher in patients with bacterial pneumonia, but there was substantial overlap. Patients with viral pneumonia may have been more likely to be immunocompromised, as has been reported previously [15]. The similarity in radiologic findings, with the finding of consolidation in viral pneumonia, is especially worth noting, since it is at odds with earlier reports [26, 27] and general opinion that viral pneumonias are ‘patchy’ whereas bacterial pneumonias are consolidative.

A unique feature of this study is the inclusion of patients who were infected with commensal bacteria, a finding that was made possible by selection only of patients who provided high-quality sputum, the use of quantitative bacteriology and MALDI-TOF idenfication of all organisms. This is also the first study to compare clinical features of pneumonia due to recognized bacterial pathogens and commensal bacteria; interestingly, no differences were observed.

Interestingly, although this study was confined to patients whose sputum contained large numbers of WBCs and was, in many instances, frankly purulent, 13% of our patients had only viral infection, emphasizing that patients with purely viral pneumonia clearly may produce purulent sputum. Because previous studies did not have adequate sputum samples on all their patients and certainly did not include quantitative bacteriology to identify commensal bacteria, they may have diagnosed viral pneumonia when bacterial coinfection was actually present.

Existing literature regarding the utility of peripheral WBC counts in differentiating bacterial from viral pneumonia presents inconsistent results. Some studies, including our own [1], found higher median peripheral WBC counts in bacterial than in viral infection as was shown again in the present study, while others have found no significant difference [28]. A very high serum procalcitonin level was observed in 3 patients with bacterial infection but there was great variability in results, and procalcitonin was normal in 29.3% patients with bacterial infection, indicating that a decision to treat with, or to withhold antibiotics can not be based on this test [19, 21].

The primary objective of the present study was to determine, in patients with CAP and a documented respiratory virus by PCR, the possibility of excluding a bacterial etiology in order to avoid prescribing empiric antibiotic therapy. Current guidelines recommend empiric antibiotics for patients who are hospitalized with CAP. However, debate persists regarding the appropriateness of antibiotics in patients with a positive viral PCR test and negative or inconclusive bacterial microbiological testing. A receiver-operator curve suggested that a patient with a positive viral PCR, known sick contact, normal WBC, and normal procalcitonin might not have bacterial infection and, therefore, might not require immediate antibiotics. However, the negative predictive value of this tool was only 0.38 and therefore could not be used to justify withholding antibiotics. Thus, our study further supports consensus guidelines for initiating empiric antibiotic therapy in all patients who are deemed sufficiently ill to require hospitalization for CAP even if a respiratory virus is shown to be present.

This study has several limitations. The total number of patients is small and it was done at a single center, but this kind of intense study is unlikely to be done in large groups of patients, and a single center assures uniform quality of laboratory work. The population consisted largely of older men, many of whom had comorbidities, and all of whom were hospitalized. Patients were only included if they provided high-quality purulent sputum. Although this requirement introduces an important inclusion bias, it was felt to be necessary because, without a high-quality sputum specimen, the diagnosis of bacterial pneumonia can not be established or excluded in about one-half of cases. The standard teaching that viral infection does not cause purulent sputum was not supported by our results. Finally, the number of patients with a pure viral pneumonia was too small to allow for meaningful comparisons in some categories, potentially limiting the generalizability of our results, although this degree of overlap in small numbers of cases probably means that a physician caring for an individual patient can not make therapeutic decisions based on any of the criteria studied.

In conclusion, the present study shows modest differences in clinical presentation of patients with bacterial and viral pneumonia or bacterial/viral coinfection, with substantial overlap in symptoms, laboratory, and imaging findings, precluding the ability to identify patients who may not require antibiotic therapy. If empiric antibiotics are to be withheld in patients hospitalized for CAP, further studies are needed to identify potential biomarkers or other clinical signs that can more clearly exclude a bacterial etiology.