B1286
Title: Parametric and nonparametric methods for outlier detection and accommodation in diagnostic test meta-analyses
Authors: Zelalem Negeri - University of Waterloo (Canada) [presenting]
Abstract: Outlying studies are prevalent in meta-analyses of diagnostic test accuracy studies. Statistical methods for detecting and downweighting the effect of such studies have recently gained the attention of many researchers. However, these recent methods dichotomize each study in the meta-analysis as outlying or non-outlying and focus on examining the effect of outlying studies on the summary sensitivity and specificity only. A parametric random-effects bivariate mixture model is developed and evaluated for meta-analyzing diagnostic test accuracy studies by accounting for both the within- and across-study heterogeneity in diagnostic test results. Instead of dichotomizing the studies in the meta-analysis, the proposed model generates the probability that each study is outlying and allows assessing the impact of outlying studies on the pooled sensitivity, specificity, and between-study heterogeneity. A nonparametric bivariate random-effects model will also be developed and evaluated for accommodating outlying studies. The performance of the developed statistical methods is illustrated using real-life and simulated meta-analytic data.