3007 — Applied Statistical Methods and Emerging Trends for Performing High Impact Meta-Analyses
Lead/Presenter: Donna White,
COIN - Houston
White DL, COIN-Houston; Baylor College of Medicine; Hysong SJ, COIN-Houston, Baylor College of Medicine;
Meta-analysis is a powerful and widely used tool for quantitatively summarizing research findings within a systematic review. Policy makers increasingly rely on meta-analytic findings to inform clinical decision-making and health policy guidelines, with meta-analyses among the most highly cited publications. This workshop's purpose is teaching participants how to pool diverse commonly reported effect metrics reported in the literature, as well as how to perform statistical tests to identify and quantify sources of between-study heterogeneity and to assess for potential biases including publication bias. We will also briefly overview an emerging form of meta-analysis, network meta-analysis, and demonstrate use of metabus, freely available cloud-based platform allowing customized rapid online synthesis of large corpus of curated social science research. Our primary goal is to help participants be able to independently use these statistical methods to perform a meta-analysis using widely available software. The instructors for this workshop are highly experienced in conducting, presenting and publishing high impact meta-analyses in the fields of clinical epidemiology and outcomes research and industrial and organizational psychology, respectively; several cited over 150 times.
Specific examples/exercises will teach attendees to: 1. Understand applied steps in obtaining pooled estimators and assessing their reliability and validity including use of statistical tests to assess for and identify sources of heterogeneity and publication bias (Higgins I-square, analysis of influence, meta-regression, cumulative meta-analysis, Egger's test, number needed to treat) and graphical assessment of funnel plots. 2. Practice interpreting meta-analytic findings that pool varied commonly pooled effect metrics (e.g., odds ratios, Cohen's d, diagnostic test findings like positive predictive value) in clinical epidemiology, comparative effectiveness and psychological research. We will also review key biases to be aware of in performing and appropriately reporting meta-analytic findings. In addition to copies of worked examples and syntax, participants will also receive a list of relevant software packages/useful readings/weblinks as well as additional background information on how to perform rigorous meta-analyses.
Researchers interested in independently conducting or supervising performance of meta-analyses or in appropriately interpreting and using meta-analytic findings.
Assumed Audience Familiarity with Topic:
We will assume little direct knowledge or experience in using statistical methods to perform meta-analysis.