The Cochrane Rapid Reviews Methods Group is leading an important piece of methodological research.
Does it make a difference whether abstracts are screened by one or two people to find relevant studies for a systematic review? How many relevant studies will be missed when abstracts are reviewed by only one person? Can the Cochrane RCT-classifier, a machine tool for classifying RCTs, increase this accuracy? The Cochrane Rapid Reviews Methods Group and the Cochrane Crowd Team have developed a study to answer these questions.
Details on the study
The study is made up of three ‘arms’ – three different groups that participants to the study could be assigned to. If you already have experience screening literature for a review, you will be randomly assigned to review a set of 100 abstracts and classify them as "Relevant" or "Not Relevant" to the research question. Then you will be randomly assigned to screen a set of abstracts retrieved from a search for either an pharmacological review or for a public health review that has broader study design inclusion criteria. If you have no experience screening for reviews, you will be assigned to the "RCT-classifier" arm where you can screen as many records as you like for the pharmacological review. In this arm, the "RCT Classifier" is tested by one person in combination with the literature review.
We kept the "screening phase" of study running until 28th June. Currently we are analysing the data. Thanks to all of you who participated! We will post results as soon as they are available.