How to Conduct a Systematic Review: A Step-by-Step Guide

A systematic review is only as strong as its methodology. Unlike a narrative literature review, it follows a pre-defined, reproducible process designed to minimize bias and capture the full body of relevant evidence. Here's how the process actually breaks down, stage by stage.

1. Define a focused research question

Most systematic reviews start with a PICO (Population, Intervention, Comparison, Outcome) or similar framework to turn a broad topic into an answerable question. A question that's too broad leads to an unmanageable number of studies; too narrow, and you risk having nothing to synthesize. This step alone often determines whether the rest of the review goes smoothly.

2. Write and register a protocol

A protocol documents your research question, eligibility criteria, search strategy, and planned analysis before you start screening studies. Registering it — commonly with PROSPERO for health-related reviews — creates a public record of your plan and reduces the temptation to quietly change your criteria after seeing the results.

3. Design a comprehensive search strategy

A defensible search strategy is built with Boolean logic across multiple databases relevant to your field (e.g. MEDLINE, Embase, CINAHL, PsycINFO, Cochrane Library), not just a single search engine. The strategy should be documented in enough detail that someone else could reproduce it exactly.

4. Screen titles, abstracts, and full texts

Screening typically happens in two passes: a fast title/abstract pass against your eligibility criteria, followed by a full-text review of anything that survives. Many high-quality reviews use two independent reviewers at this stage and report their level of agreement, since single-reviewer screening is a common source of error.

5. Extract data using a standardized form

A structured extraction form — built before screening begins — keeps the data you pull from each study consistent: study design, sample size, outcomes, effect estimates, and anything else your analysis will need.

6. Assess risk of bias

Each included study is appraised using a tool matched to its design — RoB 2 for randomized trials, ROBINS-I for non-randomized studies, or the Newcastle-Ottawa Scale for cohort and case-control studies, among others. This step is what lets you weigh the strength of your evidence base, not just its size.

7. Synthesize the findings

Depending on how comparable your included studies are, synthesis can mean a statistical meta-analysis (see our guide to forest plots) or a narrative synthesis when pooling isn't appropriate.

8. Report to PRISMA 2020 standards

The PRISMA 2020 checklist and flow diagram are now the expected reporting standard for systematic reviews across most health and social science journals. See our PRISMA 2020 checklist guide for what each item actually requires.