The designs it covers
Parallel, crossover, cluster and stepped-wedge randomised trials; quasi-experimental studies; prospective and retrospective cohorts; case-control studies; cross-sectional surveys; diagnostic accuracy studies; prediction models; qualitative studies; systematic and scoping reviews. Each comes with its reporting guideline from the EQUATOR Network.
Questions
- Is the randomised trial always the best design?
- For questions about whether an intervention works, when it is feasible and ethical, yes. For prevalence, diagnostic accuracy, prognosis or experience, other designs answer the question better, and a trial cannot study harmful exposures.
- Cohort or case-control?
- Start from the outcome. If it is rare or takes years to develop, a case-control study is far more efficient. If it is common, a cohort gives incidence and relative risk and measures the exposure before the outcome.
- What is a reporting guideline?
- A checklist of what a paper of that design must report, so readers can judge it. Journals usually ask you to submit the completed checklist. They are collected by the EQUATOR Network.
- Does the chooser use AI?
- No. It is a fixed decision tree written from standard epidemiology teaching, so the same answers always give the same design. Discuss the result with your supervisor or a methodologist.
Also free: the sample size calculator and the power calculator.