Diagnostic and prognostic test assessment in emergency medicine: likelihood and diagnostic odds ratios

Emergency physicians use diagnostic and prognostic tests on a daily basis to assess for life-threatening illness and to inform clinical decisions. Current and new tests must be scientifically evaluated for their diagnostic utility. We discuss the evaluation of diagnostic and prognostic tests using the Bayesian likelihood ratio (LR) and logistic regression diagnostic odds ratio (OR) frameworks. These approaches can be applied to a single test in isolation using univariate techniques, or to a group of tests as commonly applied in clinical practice using multivariate methods. We compare and contrast the relative benefits and challenges of the LR and OR approaches, and assess their interchangeability. The concepts of diagnostic multivariate testing also underlie the framework of clinical decision rules which have gained acceptance in emergency medicine. Clinical decision rules can be viewed as a subanalysis within the joint LR framework. Ultimately, a variety of approaches may be acceptable and even complementary to assess a diagnostic test, each with its own merits and limitations.