Department seminar by Zelalem Firisa Negeri

Please Note: This seminar will be given online.

Novel contributions to statistical methods for meta-analyses of diagnostic or screening test accuracy studies

Diagnostic or screening tests vary from the noninvasive rapid strep test used for identifying the presence of a bacterial sore throat, or a patient-reported questionnaire for screening to detect a mental health condition, to the much complex and invasive biopsy test used for examining the presence, cause, and extent of a severe condition, such as cancer. Meta-analysis of diagnostic or screening test accuracy studies is a widely used statistical approach that synthesizes evidence from multiple studies that aim to quantify the accuracy of a common diagnostic or screening test. Researchers commonly use the hierarchical or bivariate random-effects model as standard models to analyze diagnostic test accuracy data. However, these standard models are not designed for asymmetrical data or when there are outlying, or influential studies, and may lead to biased statistical inferences in those situations. Additionally, the standard methods may result in biased accuracy estimates when used to quantify the accuracy of screening tools evaluated against various reference standards that have different accuracy for diagnosing a target condition. In this talk, I will mainly discuss novel bivariate random-effects meta-analytic models and objective statistical approaches that I have recently developed and validated to address the issues of identifying and accommodating skewness, outlying, and influential studies in aggregate data meta-analyses (ADMA) of diagnostic test accuracy data. I will then briefly introduce the challenges of comparing depression screening tools against multiple reference standards that have different diagnosing capabilities, and the statistical approaches that I am proposing to tackle the issues in the context of an individual participant data meta-analyses (IPDMA). I will conclude the talk by describing some of my future works to tackle statistical problems in conducting both the ADMA and IPDMA of diagnostic or screening test accuracy studies.