COMPUTER-BASED STRUCTURED REPORTING OF RANDOMIZED TRIALS
Ida Sim,1,2 D K Owens,1,2 and G D Rennels2
Section on Medical Informatics, MSOB X-215, Stanford University School of Medicine, Stanford, CA 94305, USA
Objective: Methodologists want structured and standardized reporting of randomized controlled trials (RCTs), but non-methodologists dislike reading structured text reports. We propose that RCTs be reported as both text, and as entries into structured databases. Such dual reporting already exists: results of genomic sequencing are published in databases (eg, GenBank) while discussions are published in prose. For integrating RCT databases worldwide, we specified a core set of trial information that all RCT databases, or trial banks, should contain and share.
Design: We based this core data set on trial-reporting recommendations from the literature, and on our analysis of 24 randomized trials for meta-analysis. We specify this core data set as an object-oriented database schema of 219 concepts (eg, endpoint, or allocation concealment) that can be modified incrementally to suit future needs. We used this schema as the blueprint for a Web-accessible trial bank containing 2 randomized trials from the Veterans Affairs Cooperative Studies Program. We hyperlinked the CONSORT reporting checklist to the appropriate information about each trial.
Results: Using the original trial execution records, it took about 15 hours to enter 1 study into our trial bank. The core database schema was able to express all trial information needed for meta-analysis and critiquing, as determined by published trial scoring and reviewing instruments.
Conclusions: Direct reporting of RCTs into trial banks is not prohibitively time-consuming, and may assist peer review and meta-analysis.
Return to Session Information
