A System for Comprehensive Meta-Analysis

点击翻译按钮获取中文摘要

Description (provided by applicant): Meta-analysis has become a standard technique in evidence-based medicine for combining information from different studies testing the efficacy of similar medical treatments, procedures and diagnostic tools. As such, it is widely used by many government agencies, clinical advisory groups, policy makers, providers and users to make decisions about important clinical and public policy questions. Because meta-analytic reports have become so influential in clinical practice and because the techniques appear straightforward, if laborious, inexperienced researchers are attempting meta-analyses. Furthermore, policy needs have expanded many investigations from a narrow focus on a particular treatment for a disease to consideration of the optimal approach for treating the disease. This broader perspective greatly expands the scope of the review and analysis needed. Thus, the need for tools to efficiently perform the tasks involved is now great. Although new methods and new computing technology now make more thorough and robust meta-analysis technically feasible, many meta-analyses still employ outdated methods using simple tools like spreadsheets because users do not have either access to or guidance in the new technology. Researchers, both experienced and inexperienced, could benefit immensely from software geared to guide and assist the analyst from start to finish. Such software would keep track of abstracts screened, papers recovered and data extracted; provide standardized forms for these activities that the user could customize; offer sophisticated analysis with software designed to guide the user through the analytic process; provide immediate model diagnostics and sensitivity analysis; offer presentation quality graphics customized for meta-analysis; and simplify report construction by automatically linking data storage and analysis with tabular report formats. Existing software, even that designed explicitly for meta-analysis, lacks many of these features. We propose to remedy that deficit by 1) building menu-driven software for meta-analysis which consists of four independently functional modules (Abstract Screening Assistant, Data Extraction Assistant, Data Analysis Assistant, and Report Building Assistant) linked by a common front-end into a seamless program; 2) writing a manual (and online help) to accompany the software; and 3) incorporating software training into existing meta-analysis workshops.