Xinyan Cai, MSPH; Mark H. Ebell, MD, MS; Garth Russo, MD
Corresponding Author: Mark H. Ebell, MD, MS; Department of Epidemiology and Biostatistics, College of Public Health - University of Georgia.
Contact Email: ebell@uga.edu
Section: Original Research
Publication Date: TBD
BACKGROUND: Ordering a serological test for infectious mononucleosis (IM) in all young patients with sore throat is costly and impractical. The test threshold to determine when to order a diagnostic test for IM based on the patient’s symptoms has not been previously studied.
OBJECTIVE: To determine the test threshold for IM in the management of patients with sore throat.
DESIGN AND SETTING: Online surveys were sent to a convenience sample of US primary care clinicians regarding their decision-making about whether or not to order a test for IM in a patient with sore throat.
METHOD: Seven clinical vignettes were created, each with a different combinations of symptoms and signs. The probability of IM for each vignette was estimated by the investigator based on the number of symptoms present to generate a plausible range of disease probabilities. Clinicians were then asked to decide whether to test or not test for IM, and mixed-effect logistic regression was used to determine the test threshold for IM where half of physicians chose to test and half chose not to test.
RESULTS: A total of 117 clinicians provided responses for a total of 819 clinical vignettes. The overall test threshold for IM as estimated using the logistic regression was 9.5% (95% CI: 8.2% to 10.9%). The test threshold for clinicians practicing greater than 10 years was significantly higher than for those practicing less or equal to 10 years (10.5% vs. 7.3%, p=0.02). No significant differences between specialties and practice sites were found with respect to the test threshold.
CONCLUSION: This study identified a test threshold for IM of approximately 10% based on realistic clinical vignettes. This threshold was stable regarding the clinician’s specialty and practice sites and could be used in the development of a clinical prediction rule to determine the cutoff for low versus high risk groups.