Event
Nandini Dendukuri, PhD, Ï㽶ÊÓƵ
Friday, January 20, 2017 15:30to16:30
Purvis Hall
1020 avenue des Pins Ouest, Montreal, QC, H3A 1A2, CA
Latent class analysis: An indispensable tool for diagnostic research
The lack of an adequate diagnostic test is a fairly common problem in many disease areas including childhood tuberculosis, pneumonia and Alzheimer’s disease. Latent class analysis has been applied to estimate disease prevalence or sensitivity and specificity of a new diagnostic test in these settings. It can also be used to develop prediction models, estimate overdiagnosis and support cost-effectiveness analyses. Despite its utility, the number of applications of latent class analysis in diagnostic research remains limited. This presentation will discuss some of the reasons for this and some solutions to make latent class models more accessible. One particular impediment has been the promotion of simple but biased alternative approaches, particularly composite reference standards. Other challenges include the handling of non-identifiability and conditional dependence, which are common in these problems, and model validation in the absence of a perfect reference test.