ADVANCED EVIDENCE-BASED DIAGNOSIS (ROC CURVES, INTERVAL LIKELIHOOD RATIOS, AND BIAS IN TEST ACCURACY STUDIES)

Format Requirements: The workshop will consist of brief didactic presentations followed by small groups discussing and solving illustrative problems based on studies of diagnostic tests. Among other tests, the problems will cover: rapid antigen testing for influenza; CT and LP for subarachnoid hemorrhage; BNP for acute heart failure; D-Dimer for pulmonary embolism; serum lactate for severe sepsis: the ABCD2 Score for stratifying risk of stroke after TIA, and abdominal CT for diagnosing appendicitis in children. Participants should be comfortable with the basics of diagnostic testing such as the definitions of sensitivity, specificity and predictive value, and should also have some initial exposure to receiver operating characteristic (ROC) curves and likelihood ratios. As should be clear from the topics covered, we created this workshop for physicians or medical students who have completed some clinical training. However, we welcome non-physicians interested in test evaluation, utilization, and interpretation.

Diagnosis and diagnostic testing are fundamental to medical decision making.  When and how should we use a rapid antigen test for influenza?  How should we interpret BNP for acute heart failure, D-Dimer for pulmonary embolism, or serum lactate for severe sepsis?  How can we determine the risk of stroke after a TIA?  In evaluating children for possible appendicitis, how do we balance the radiation exposure of CT scanning against its greater accuracy?   In this full-day, interactive workshop, we will discuss using the evidence to evaluate and interpret diagnostic tests.  We will cover dichotomous tests, multilevel and continuous tests, ROC curves, interval likelihood ratios, and bias in studies of test accuracy -- and believe it or not, we will have fun!  Instead of giving and passively listening to a series of lectures, we will discuss and solve real-world diagnostic testing problems in small groups.

 The course will cover the following topics (examples):

1)       Review of dichotomous tests, sensitivity, specificity, LR(+), LR(-), case-control vs. cross-sectional sampling, and the false negative rate confusion (rapid antigen testing for influenza, pertssis, and Group A strep)

2)       Multilevel and continuous tests, interval likelihood ratios and the perils of making multi-level tests dichotomous (BNP for acute heart failure, D-Dimer for PE, ABCD2 Score to estimate stroke risk)

3)       Going beyond the area under the ROC curve and getting the most out of published ROC curves (peripheral WBC count for meningitis in infants; serum lactate for severe sepsis)

4)       Studies of diagnostic test accuracy -- beyond the check list: incorporation and spectrum biases (CT scans for appendicitis and intracranial bleeding, echocardiogram for wall motion abnormalities in chest pain patients)

5)       Studies of diagnostic test accuracy, partial and differential verification bias (history and exam findings for appendicitis, hemorrhagic CVA, and elbow fracture)