Latent Class Models for Individual Participant Data Meta-Analyses of Diagnostic Test Accuracy Studies with Imperfect Reference Standards

• Activity Number: 51 - Recent Developments in Modeling High-Dimensional and Complex Data
• Type: Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
• Sponsor: SSC (Statistical Society of Canada)
• Author(s): Zelalem Negeri* and Brooke Levis and Yin Wu and Nandini Dendukuri and Brett Thombs and Andrea Benedetti
• Companies: McGill University and Keele University and Department of Psychiatry, McGill University and McGill University and Jewish General Hospital and McGill University
• Keywords: Meta-analyses; Diagnostic test accuracy studies; Individual participant data; Latent class modelling; Imperfect reference standards; Generalized linear mixed-effects modelling

Abstract

Latent class models have been commonly applied to correct for imperfect reference or gold standards in diagnostic test accuracy studies. Most of these latent class models use the Bayesian analysis approach to estimate the unknown model parameters. To the best of our knowledge, there are no methodological studies that attempted to account for imperfect reference standards in the context of individual participant data meta-analyses (IPDMA) of diagnostic test accuracy studies. Therefore, the objective of this study is to develop and validate latent class models for IPDMA to estimate the diagnostic test accuracy of both an index test and imperfect reference standards by exploring both Frequentist and Bayesian approaches to the problem. We will illustrate the models using our database that consists of more than 100 studies and 46,000 participants on the most commonly used tool for detecting major depression – the Patient Health Questionnaire-9 (PHQ-9), and diagnostic interviews such as the Structured Clinical Interview for DSM (SCID), Composite International Diagnostic Interview (CIDI), and the Mini International Neuropsychiatric Interview (MINI).