Michael A. Kohn, MD, MPP - Associate Professor, Epidemiology and Biostatistics, UCSF, Attending Emergency Physician Mills-Peninsula Medical Center Burlingame, CA
Title: Predictive Models and Continuous Diagnostic Tests: Wells Score and D-Dimer for Pulmonary Embolism
Summay: In medicine, we often use risk models and quantitative test results to guide decisions. With risk models, we generally assume that the patient comes from a known population with a common prevalence of disease and use the risk model to place the patient into one of several risk categories. With diagnostic tests, we assume that the pretest probability varies from patient to patient, but that we can use the same likelihood ratio for a test result to update these varying pretest probabilities. Although the likelihood ratio of a test result is not necessarily independent of pretest probability, it is often reasonable to assume it is. This discussion will elaborate on these points using the workup of emergency department patients with dyspnea for pulmonary embolism as an example.