Software for meta-analysis with correlated outcomes

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Meta-analysis is the statistical procedure employed to synthesize data from a series of studies. As such, it falls at the core of evidence-based policy in such fields as alcohol and substance abuse, medicine, social science, ecology, education, gerontology, among many others. The current generation of meta-analysis software is designed for the situation where each study contributes one (and only one) effect size to the analysis. However, many meta-analyses do incorporate studies that report more than one effect size. In particular, studies sometimes report effects for more than one outcome, for more than one time-point, and/or for more than one intervention. In these cases the standard formulas and software cannot be used, and there are no good options for performing the analysis. The only software that can be used in these cases requires a high level of statistical expertise, and additionally requires the user to provide information, such as the correlation matrix among outcomes, that the user is not likely to have. Therefore, as a practical matter, there are no good options for performing a meta-analysis involving more than one effect size. Under a series of earlier grants we developed a program for meta-analysis that is currently in use at some 10,000 institutions around the world, including thousands of hospitals, universities, and agencies such as the FDA, CDC, NIH, WHO, among many others. Our plan is to expand this to work with studies that report more than one effect size, and that will not require the user to supply a correlation matrix. PUBLIC HEALTH RELEVANCE: The current generation of software for meta-analysis can only be used when each study provides a single effect size. The goal of this project is to develop software for meta-analysis that can be used when some (or all) studies report an effect for multiple outcomes, time-points, or treatments. This software will not require the user to provide a correlation matrix among effects.