Assessing diagnostic and screening tests: Part 2. How to use the research literature on diagnosis

Case scenario One night when on call for inpatient admissions, you receive the third call in a week to admit a child who looks well but has fever and a petechial rash. Few of these children actually seem to have bacterial (pa rticularly meningococcal) sepsis, but you are not sure how good the "diagno stic test" of the child's clinical appearance (well vs not well) performs i n this circumstance, You estimate that it would be 20 times worse to miss a child who has bacteremia than to admit a child unnecessarily (that is, the action threshold is around 5%) and wonder whether the child's appearance c an be incorporated into the decision to admit children who have fever and p etechial rash. You frame the question, "ln children with fever and petechia e (population), does looking ill (diagnostic test) increase the risk of bac teremia (outcome)?" and search for literature on this topic using the searc h strategy "petechiae AND bacteremia" in the "Clinical Queries" section of PubMed's web site (highlighting diagnosis and sensitivity). This strategy y ields 24 studies.