The course covers the epidemiological aspects of diagnostic test evaluation (DTE) studies – from the technical aspects of planning, conducting and subsequently analyzing data from such studies, to the underlying assumptions regarding disease definitions, test characteristics and the implications of using a perfect reference tests vs. a latent class approach.
The first two days of the course concentrates on the classic diagnostic test evaluation against a so called ‘golden standard’ or perfect reference test. Here the concepts of sensitivity, specificity and predictive values of a test will be put into context of planning the diagnostic test evaluation, including the concept of fitness for purpose, and the subsequent analysis and presentation of the DTE using the STARD guidelines. The concept of conditional independence given disease status will also be presented and discussed in detail. This concept is a fundamental issue in the development of the latent class models used in the remaining part of the course.
The design and analysis of latent class models for test evaluation without a perfect reference test (i.e. a ‘gold’ standard) will be the main focus of the course and will be given attention on day 3 and 4 of the course. The participants will be introduced to Latent Class Analysis (LCA) in a Bayesian framework using OpenBUGS and R. Through a mixture of lectures/discussion of the theory and biological implications and exercises/tutorial based on published examples, the course participants will be exposed to the necessary concepts and ideas of LCA as well as gain working skills in using OpenBUGS and the presented examples to explore the possibilities of diagnostic test evaluation using LCA methods. The course will present models at population and individual based levels, including models that include covariates affecting the test characteristics. The steps involved in planning a DTE based on latent class analysis will be the topic of an exercise and discussion during the course. It is of particular importance to understand the differences and implications of the choice of candidate test(s) and possible imperfect reference tests. The concept of conditional independence and the possible lack of thereof is treated in detail.
The last day of the course is dedicated to models where the primary purpose is not test evaluation, but more a means of coping with mis-classification in the analyses.
Participants are encouraged to bring their own data so that they have an opportunity to analyze these using the models presented in the course.
A post-course assignment shall be submitted. It will be evaluated by course teachers.
- At the end of the course, the participant will be able to plan and carry out a diagnostic test evaluation (DTE) of a (set of) candidate test(s) against a reference test. This involves analysis of the data, including ROC analyses for continuous test data, assessment of conditional dependence, and reporting according to the STARD guidelines.
- The participant understands the concept of fitness for purpose of a test and can use this for a given test (and associated DTE).
- The participant is capable of analyzing a given dataset using a latent class approach with the programs provided in the course.
- The participant can discuss the implications of the chosen set of tests included in a latent class DTE and the conditional independence given disease status.
Evaluation will be through a Multiple Choice test based on the pre-course e-learning and a post-course assignment.
The learning process will be based on three parts:
- Students will read some introduction texts before the course.
- The course itself will be based upon lectures and supervised assignments with considerable time on specific software.
- After the course each student will work on an assignment based upon own data. This assignment will be discussed with another student before submitted for the final evaluation form course teachers.
TOTAL: 90 hours
- Before course: Reading, e-learning + MC test, discussions (20 hours)
- At the 5 days course: Lectures (15 hours), group discussions (12 hours), independent work (13 hours)
- After course: Individual assignment (30 hours)
Basic training in epidemiology/ veterinary epidemiology including some knowledge on the use of diagnostic tests in populations.
Admission for NOVA courses is handled by the course organiser/ the NOVA member institution organising the course. Please see the links in the margin for more information.
For non-NOVA and non-BOVA students, there is a course fee of EUR 600.