RESEARCH DESIGN GUIDE
Choose a review design for your question
Define what you want to estimate, then choose a review type, effect measure, and eligible study designs. This is general methods guidance; the final protocol depends on the question, available data, and expert input.
In this article
1. Types of clinical questions
Intervention
Estimate the effect of a treatment or intervention.
Common direction：Comparative effectiveness review
Diagnosis
Assess how accurately a test identifies a condition.
Common direction：Diagnostic test accuracy review
Frequency
Estimate prevalence or incidence in a defined population.
Common direction：Prevalence or incidence review
Prognosis
Describe future outcomes or their predictors.
Common direction：Prognostic factor review
Safety
Assess adverse events and other harms.
Common direction：Benefits and harms review
Multiple interventions
Compare several interventions across a connected evidence network.
Common direction：Network meta-analysis
2. Review types
Intervention effect review
Does an intervention change an outcome?
- Typical measure
- MD or SMD for continuous outcomes; RR or OR for binary outcomes
- Possible designs
- Randomized trials; observational studies when appropriate
- Example question
- Among adults with condition X, does intervention Y improve outcome Z?
Caution：Choose the effect measure and eligible designs before screening.
Diagnostic accuracy review
How well does an index test classify a target condition?
- Typical measure
- Sensitivity, specificity, likelihood ratios, diagnostic odds ratio
- Possible designs
- Diagnostic accuracy studies using an appropriate reference standard
- Example question
- How accurately does test X detect condition Y?
Caution：Consider threshold effects and variation in reference standards.
Prevalence review
How common is a condition in a defined population?
- Typical measure
- Proportion or prevalence with a confidence interval
- Possible designs
- Cross-sectional and population-based studies
- Example question
- What is the prevalence of condition X in population Y?
Caution：Define the population, time period, and case definition.
Network meta-analysis
How do several interventions compare, directly and indirectly?
- Typical measure
- Relative effects within a connected network
- Possible designs
- Randomized trials with suitable treatment links
- Example question
- Which of treatments A, B, and C performs best for outcome X?
Caution：Assess transitivity, consistency, and network connectivity.
3. Study designs and evidence
| Study design | Evidence characteristics | Typical use | Reporting guidance |
|---|---|---|---|
| Randomized controlled trial | Random allocation helps balance known and unknown confounders. | Intervention effects when randomization is feasible. | CONSORT |
| Cohort study | Follows groups over time; confounding may remain. | Prognosis, exposure, and outcomes that develop over time. | STROBE |
| Case-control study | Efficient for rare outcomes; vulnerable to selection and recall bias. | Rare outcomes or long latency periods. | STROBE |
| Cross-sectional study | Measures a population at a point or period in time. | Prevalence and distribution of characteristics. | STROBE |
| Diagnostic accuracy study | Compares an index test with a reference standard. | Sensitivity and specificity of tests. | STARD |
4. Protocol checklist
0 of 6 complete