A Randomized, Controlled Trial of a C-reactive Protein (CRP)-Guided Management Algorithm for Adults With Acute Cough Illness in the Emergency Department

We have developed a CRP-guided management algorithm based on recent studies in which the accuracy of CRP for identifying patients with pneumonia was confirmed.(10;11) Importantly, these 2 studies each confirmed that CRP levels provide additional diagnostic support beyond that gained from application of clinical prediction rules. In the proposed study, we will recommend that physicians first assess the probability of pneumonia based on clinical findings, in accordance with current practice recommendations as follows: 1) Low Clinical Suspicion (<5% probability of pneumonia): Absence of fever (T>38C), tachycardia (HR>100), or tachypnea (RR>24), and normal chest examination. These patients should receive symptomatic therapy, and do not require CXR or antibiotic treatment; 2) High Clinical Suspicion (>30% probability of pneumonia): At least 1 abnormal vital sign and abnormal chest examination. These patients receive CXR and antibiotics based on CXR findings or high clinical suspicion; 3) Intermediate Clinical Suspicion (5-30%): All others. Current practice guidelines recommend considering CXR in these patients, however no further guidance is provided for the management of these patients.

After assessing the clinical probability of pneumonia, CRP levels will be provided for a random subset of patients. A CRP level <10 mg/L corresponds with a very low probability of bacterial pneumonia. Thus, for patients with low and intermediate-probability of pneumonia based on clinical grounds, a CRP <10 mg/L should confirm that no CXR or antibiotics are necessary in routine cases (See attached algorithm). A CRP level > 100 mg/L corresponds with a substantially elevated probability of pneumonia. Thus, for patients with low and intermediate-probability of pneumonia based on clinical grounds, a CRP >100 mg/L would support CXR ordering, and for patients with high clinical suspicion, antibiotic therapy should be considered regardless of the CXR result (since studies show that CXR misses about 10% of community-acquired pneumonia confirmed with follow-up CXR or chest computed tomography). Overall, the addition of CRP test results would reduce the need for CXR ordering and antibiotic therapy in cases with an intermediate clinical probability of pneumonia and a low CRP level; and increase the need for CXR ordering and possibly antibiotic therapy in cases with intermediate or high clinical probability of pneumonia and a high CRP level. In this way, we hypothesize that the CRP-guided management algorithm will improve the proportion of patients with acute cough illness who receive appropriate diagnostic and therapeutic interventions in the ED.

We suspect that most patients with clinically apparent pneumonia will have high CRP levels. However, pneumonia can be a difficult diagnosis to make when patients present with atypical clinical manifestations (e.g. the elderly or COPD patients), or in practice settings where radiography is not readily available. In these settings, the addition of CRP testing may help identify a small number of pneumonia cases that might otherwise go undetected and result in delays in appropriate care. In the large number of patients with intermediate probability of pneumonia, such as adults with acute cough illness who have abnormal vital signs, but normal examinations, particularly in RSV and influenza season, a CRP level < 10 mg/L could provide clinically significant reassurance to refrain from further testing or empiric antibiotic treatment. A normal CRP level may also provide support to clinicians who are receiving pressure from patients to prescribe antibiotics when they are not clinically indicated. Indirect or future benefits to society are possible if research results show that CRP-guided management strategies can decrease the costs of health care.

Randomization using a random number generator will be conducted by investigators at the data coordinating center (Philadelphia, PA), and results placed in sequentially number study packets. The study personnel that enroll patients will be blinded to the randomization result until after the patient has provided informed consent to participate in the study.

Interventions