Abstract
Background: Bordetella pertussis (BP) is a common cause of prolonged cough. Our objective was to perform an updated systematic review of the clinical diagnosis of BP without restriction by patient age.
Methods: We identified prospective cohort studies of patients with cough or suspected pertussis and assessed study quality using QUADAS-2. We performed bivariate meta-analysis to calculate summary estimates of accuracy and created summary receiver operating characteristic curves to explore heterogeneity by vaccination status and age.
Results: Of 381 studies initially identified, 22 met our inclusion criteria, of which 14 had a low risk of bias. The overall clinical impression was the most accurate predictor of BP (positive likelihood ratio [LR+], 3.3; negative likelihood ratio [LR−], 0.63). The presence of whooping cough (LR+, 2.1) and posttussive vomiting (LR+, 1.7) somewhat increased the likelihood of BP, whereas the absence of paroxysmal cough (LR−, 0.58) and the absence of sputum (LR−, 0.63) decreased it. Whooping cough and posttussive vomiting have lower sensitivity in adults. Clinical criteria defined by the Centers for Disease Control and Prevention were sensitive (0.90) but nonspecific. Typical signs and symptoms of BP may be more sensitive but less specific in vaccinated patients.
Conclusions: The clinician's overall impression was the most accurate way to determine the likelihood of BP infection when a patient initially presented. Clinical decision rules that combine signs, symptoms, and point-of-care tests have not yet been developed or validated.