1. VITA-An Interactive 3-D Visualization System to Enhance Student Understanding of Mathematical Concepts in Medical Decision-making M Sriram Iyengar, PhD Asst. Professor, School of Health Information Sciences, University of Texas Health Science Center at Houston Informatics Research Scientist, Medical Informatics and Health Care Systems, NASA Johnson Space Center, Houston, TX John R Svirbely, MD TriHealth, Cincinnati, Ohio Mirabela Rusu, MS Graduate Student,School of Health Information Sciences, University of Texas Health Science Center at Houston Jack W Smith, MD, PhD Professor and Dean, School of Health Information Sciences, University of Texas Health Science Center at Houston
2. Mathematics of Diagnostic Testing A diagnostic test may be a biochemical assay (BNP, PSA, other) or other analysis. If the assay result is greater (or lesser) than a cutoff value then the presence (or absence) of disease is concluded. To analyze the test performance we use: Sensitivity = P(Test positive | Disease present ) Specificity = P(Test negative | Disease not present ) False Positive rate = 1 - specificity ROC curve: Plot of Sensitivity vs. False Positive rate Area under ROC: the closer to 1 the better
3. Mathematics of Diagnostic Testing 2 However, the Post-test Predictive Values are more useful for diagnostic purposes. Positive Predictive Value = P(Disease exists| Test positive) Negative Predictive Value = P(Disease not present | Test negative) Both incorporate disease prevalence and are computed using Bayes formula.
5. VITA Interactive software for 3-D visualization 3-D and 4-D views Provides rotation, zooming and similar functions for graph manipulation Can use a table of cut-off values and generate direct relationships between PPV, NPV and cutoff values for various prevalences.
9. VITA Benefits Teaching: Important and complex mathematical concepts in medical decision-making Understand non-linearity in predictive values Understand how a test will perform in a different population Research: Determine optimal cutoff-vales for diagnostic tests Compare performance of diagnostic tests on the basis of predictive vales that are more meaningful to diagnosticians