Designing a Study That Could Be Appraised

In practice

A team has a question no existing study answers and decides to generate the evidence. Appraising and designing are the same knowledge viewed from opposite ends, which is why unit 6 was the preparation for this. The design follows the question rather than the team's preference: an intervention question wants randomisation, because random allocation distributes the factors nobody thought of and nothing else does. Outcomes are chosen that matter to someone rather than that are easy to collect. Inclusion and exclusion criteria are set before recruitment, and the variables and measures are specified so that a measure actually measures what is intended. The ethical questions are settled first — a question has to be ethical to ask, risk cannot exceed benefit, and review happens before the study runs rather than as a submission after it. Then the threats are addressed by design instead of in the discussion section: attrition is planned for, because uneven dropout changes the answer rather than shrinking the sample, and maturation and observer drift are anticipated. What is built is a study that a reader applying unit 6's three questions would accept. The test is worth running explicitly before recruitment starts: read your own protocol as though appraising someone else's, and note where you would have objected. Every objection found then is one a reviewer will not have to find later.