Clinical prediction rules in Staphylococcus aureus bacteremia demonstrate the usefulness of reporting likelihood ratios in infectious diseases.

ABSTRACT

Infectious diseases specialists often use diagnostic tests to assess the probability of a disease based on knowledge of the diagnostic properties. It has become standard for published studies on diagnostic tests to report sensitivity, specificity and predictive values. Likelihood ratios are often omitted. We compared published clinical prediction rules in Staphylococcus aureus bacteremia to illustrate the importance of likelihood ratios. We performed a narrative review comparing published clinical prediction rules used for excluding endocarditis in S. aureus bacteremia. Of nine published clinical prediction rules, only three studies reported likelihood ratios. Many studies concluded that the clinical prediction rule could safely exclude endocarditis based on high sensitivity and high negative predictive value. Of the studies with similar high sensitivity and high negative predictive value, calculated negative likelihood ratios were able to differentiate and identify the best clinical prediction rule for excluding endocarditis. Compared to sensitivity, specificity and predictive values, likelihood ratios can be more directly used to interpret diagnostic test results to assist in ruling in or ruling out a disease. Therefore, a new standard should be set to include likelihood ratios in reporting of diagnostic tests in infectious diseases research.