Accuracy of signs and symptoms for the diagnosis of acute rhinosinusitis and acute bacterial rhinosinusitis

Image credit: Flickr

Abstract

PURPOSE To evaluate the accuracy of signs and symptoms for the diagnosis of acute rhinosinusitis (ARS). METHODS We searched Medline to identify studies of outpatients with clinically suspected ARS and sufficient data reported to calculate the sensitivity and specificity. Of 1,649 studies initially identified, 17 met our inclusion criteria. Acute rhinosinusitis was diagnosed by any valid reference standard, whereas acute bacterial rhinosinusitis (ABRS) was diagnosed by purulence on antral puncture or positive bacterial culture. We used bivariate meta-analysis to calculate summary estimates of test accuracy. RESULTS Among patients with clinically suspected ARS, the prevalence of imaging confirmed ARS is 51% and ABRS is 31%. Clinical findings that best rule in ARS are purulent secretions in the middle meatus (positive likelihood ratio [LR+] 3.2) and the overall clinical impression (LR+ 3.0). The findings that best rule out ARS are the overall clinical impression (negative likelihood ratio [LR−] 0.37), normal transillumination (LR− 0.55), the absence of preceding respiratory tract infection (LR− 0.48), any nasal discharge (LR− 0.49), and purulent nasal discharge (LR− 0.54). Based on limited data, the overall clinical impression (LR+ 3.8, LR− 0.34), cacosmia (fetid odor on the breath) (LR+ 4.3, LR− 0.86) and pain in the teeth (LR+ 2.0, LR− 0.77) are the best predictors of ABRS. While several clinical decision rules have been proposed, none have been prospectively validated. CONCLUSIONS Among patients with clinically suspected ARS, only about one-third have ABRS. The overall clinical impression, cacosmia, and pain in the teeth are the best predictors of ABRS. Clinical decision rules, including those incorporating C-reactive protein, and use of urine dipsticks are promising, but require prospective validation.

Publication
Annals of Family Medicine
Avatar
Brian McKay
Lecturer

My research interests include distributed robotics, mobile computing and programmable matter.

Related