ROC Curves: Teaching Prototype
2018
In this video, ROC curves, Professor Myriam Hunink introduces students to tests with continuous or categorical results. In contrast to a test with dichotomous results (e.g., positive versus negative), she poses a scenario in which a test has multiple possible results.
Access the video. ROC Curves: Teaching Prototype (~20 min)
Professor Hunink asks students to consider a CT coronary artery calcium for the diagnosis of cardiovascular disease. She explains that a high score is more abnormal than a low score - but provides data from a study group of patients with results separated into three categories. Using the data from this study group, students learn how to calculate a likelihood ratio for a specific “test result” when there are multiple results.
An ROC curve is defined and plotted, and the conceptual basis for a “lenient” and “strict” zone of the ROC curve explained. Professor Hunink walks through the implications of a ‘cutoff point’ for a test with continuous results, or in this case, the implications of calling a particular test category result, “positive” or “negative”, for which “positive” would indicate “treat” and “negative” would indicate “do not treat”. Students acquire an introduction to the factors that influence an optimal positivity criterion of the test variable - recognizing these include the prior probability of disease, the test performance, and the harm to benefit ratio of treatment.
This video is one of a series developed by Professor Myriam Hunink during an immersion residency at the Center for Health Decision Science (CHDS) Media Hub. The video series reflect experiments to augment brick and mortar teaching with multimedia materials that emphasize visualization of basic concepts.
ROC Curves: Teaching Prototype
Source:
ROC Curves: Teaching Prototype. Teaching Pack: Teaching Prototypes for Decision Analysis. Center for Health Decision Science, Harvard T.H. Chan School of Public Health 2018. https://vimeo.com/236607964/040d28bbbf