With emergency departments (EDs) and urgent care centers brimming at this time of year with patients experiencing influenza-like illness (ILI), clinicians would welcome a more effective, faster means of differentiating viral versus bacterial ILI. Both present with similar symptoms and can have indeterminate radiology and lab test results. Researchers recently investigated the use of C-reactive protein (CRP) for this purpose, and their findings are the subject of this issue of Strategies.
During influenza season, many patients present to EDs with respiratory symptoms, but perhaps only one-third actually have influenza. The others typically have viral or bacterial illnesses that share symptoms with influenza, like fever, cough, and chills. Diagnostic testing in patients with ILI often includes chest x-ray and white blood cell count, both of which can be equivocal in this patient population. As a consequence, the final diagnosis of influenza, bacterial infection, or other viral illness remains a clinical one, based on the physician’s assessment of the patient’s physical examination and history, and radiology and lab test results. Yet being able to more accurately and promptly distinguish bacterial from viral illnesses has implications for improved patient care and throughput in busy EDs. This led researchers at Brown University’s Alpert Medical School and the University of Massachusetts Medical School at Worcester (UMass) to investigate the utility of CRP for this purpose (Am J Emer Med 2013;31:137–44).
“During flu season a wide range of people come to the ED with respiratory complaints, and the question of whether or not to give antibiotics is left up to the physician’s discretion. From my experience and the literature, it’s a clinical diagnosis, and one of the things that is frustrating is the testing behind it,” said lead author, John P. Haran, MD, an assistant professor of emergency medicine at UMass. “My interest in this is from a clinical standpoint, recognizing that white blood cell count is a poor marker of whether or not someone has a bacterial infection and x-ray seems to be poor depending on your definition of bacterial pneumonia. As medicine advances, we need more markers to help us with our clinical decision-making. Anything that speeds up the decision process is helpful. But we don’t want to just move people through faster. We want to make the right decisions, so that we avoid either having patients come back sicker or giving them antibiotics unnecessarily.”
He went on to explain why he and his colleagues explored the utility of CRP, an acute-phase protein produced by the liver in response to inflammation. “CRP is a readily available marker with results typically turned around within a half-hour from the lab. I wanted to investigate a marker that would be readily accessible. There are other markers out there under development but not in a rapid system for physicians to have those markers readily available within a short period of time,” he said.
Haran and his colleagues enrolled subjects from October to March in three successive influenza seasons: 2009, 2010, and 2011. All subjects were adults, and presented to the ED with symptoms of ILI. They completed surveys about their past medical history, current illness, symptom severity, and presenting vital signs. Those who had had symptoms longer than 3 days or reported having a bacterial infection within 4 weeks of presenting to the ED were excluded. Nasal wash and blood samples were collected from all patients and either an enzyme-linked immunosorbent assay or viral culture for influenza were performed, along with CRP and other blood tests, including white blood cell count. After treatment, the researchers followed patients via either chart review or telephone survey to determine the duration of their symptoms and whether they experienced complications like hospitalization or subsequent diagnosis of pneumonia.
Patients were classified as having bacterial infection, influenza, or other viral process. The primary outcome of interest was presence of bacterial versus non-bacterial infection and baseline CRP result. The secondary outcome was total white blood cell count and percentages of band and segmented neutrophils in patients with bacterial versus viral infections.
The researchers found that patients in the bacterial infection group had significantly higher CRP levels at presentation, with a mean of 135.96 mg/L, compared with 25.65 mg/L for influenza patients and 13.73 mg/L for those with other viral infections. They performed a fitted receiver operating characteristic curve (ROC) analysis for CRP for bacterial infection and found an area under the curve of 0.978. A CRP <20 mg/L equated to a sensitivity of 100%, while CRP >80 mg/L equated to a specificity of 100%.
“We found that CRP had a range. In the ED, the D-dimer test is used to differentiate among patients with chest pain, and CRP functions in a similar way,” said Haran. “If it is above a certain threshold that raises your suspicion of bacterial infection, and if it’s below a certain threshold it helps you eliminate suspicion for it.”
Haran went on to explain that these findings had not been adopted as part of any protocols at UMass, but that he and his colleagues were using it on a case-by-case basis. “Obviously this was not a randomized, controlled trial, but that would be the way to go if we wanted to use a certain marker to help determine whether or not a patient has a bacterial illness. In the trial, the decision to give antibiotic therapy would be based on a specific range of the biomarker and then assessing medical outcome,” he said. “Even without that level of evidence, my group has found it helpful to use as far as adding more information in making the diagnosis, but it’s not part of a protocol.”
Roger L. Bertholf, PhD, DABCC, FACB, concurred that more research about CRP would be needed to demonstrate its value in this population. He also wondered whether aspects of the authors’ data analysis might have ascribed more utility to CRP than warranted. “The authors acknowledged the limitation to this study that there was a significant age difference between the groups, with the bacterial group being significantly older than either of the non-bacterial groups. They point out correctly that there may be some bias in the way clinicians classify these patients based on their age, which would introduce a selection bias into their study,” he explained. “The other possibility here, which they pointed out, is that older people are more prone to bacterial infections.” Bertholf, who was not involved in the study, is a professor of pathology at the University of Florida College of Medicine and director of clinical chemistry, toxicology, and point of care testing at Shands Jacksonville Hospital. The mean age of patients with bacterial infections was 48 years, compared with 30 for both the influenza and other viral groups.
Bertholf also questioned whether the data did not show white blood cell count results to be as good as CRP results in discerning bacterial infections. “When you compare the reported CRP levels with white blood cell count results looking strictly at the means and confidence intervals it appears as though white blood cell count performs just as good as CRP in predicting bacterial infection. Yet the ROC curves for CRP, white blood cell count, and subsets of white blood cell counts show a dramatic difference between the predictive values of CRP compared to white blood cell count,” he explained. “The data presented showing means and confidence intervals do not appear to agree with presentation of the ROC curves.”
Haran responded that “ROC analysis is related in a direct and natural way to cost/benefit analysis of diagnostic decision-making. The confidence intervals do overlap when it comes to comparing the bacterial and viral groups. However, when the question is asked how good is white blood cell count at diagnosing pneumonia—that is the ROC curve—we see that it is in fact a poorer test. We aren’t stating that white blood cell count has no added value in diagnosing pneumonia, just that it’s much poorer in comparison to CRP.”
While they disagreed about the presentation of the data, Haran and Bertholf concurred that the premise of the study was good. “Emergency physicians want to make a diagnosis as quickly as possible so they can either initiate treatment or refer that patient to a service that is appropriate for their illness,” said Bertholf. “A laboratory test to help them make that decision is what we’re all about in the clinical lab.”