How to Write a Systematic Review: A Step-by-Step Guide
On this page
- 1. Define your question with PICO or PICOS
- 2. Register your protocol
- 3. Build and test your search strategy
- 4. Screen in two independent stages
- 5. Extract data with a piloted form
- 6. Assess risk of bias
- 7. Synthesize your findings
- 8. Report against the PRISMA 2020 checklist
- The one thing that trips up most first-time reviewers
Writing a systematic review is not the same task as writing a literature review, and treating it that way is the most common reason reviews get sent back by supervisors or journal reviewers. A systematic review follows a defined, reproducible methodology from the first day of planning to the final manuscript -- every step needs to be documented well enough that another researcher could repeat it and land on the same set of included studies.
Example PRISMA flow diagram, showing how records narrow to included studies.
Here is what that process actually looks like, stage by stage.
1. Define your question with PICO or PICOS
Before you write anything, your review question needs to be specific enough to build a search strategy from. The PICO framework -- Population, Intervention, Comparison, Outcome -- forces that specificity. Some review types add a fifth element, Study design, making it PICOS.
A vague question like "does exercise help depression" is not reviewable. A PICO-formed version -- "In adults with major depressive disorder (P), does structured aerobic exercise (I) compared to standard pharmacological treatment (C) reduce depressive symptom severity (O)" -- is.
2. Register your protocol
Most journals and institutions now expect a registered protocol before screening begins, typically on PROSPERO. Registration locks in your methodology in advance, which protects you from accusations of changing your inclusion criteria to fit whatever results you found. It also surfaces duplicate reviews already underway on the same question.
Your protocol should specify your PICO(S), your planned databases, your inclusion and exclusion criteria, your risk-of-bias tool, and your planned synthesis approach.
3. Build and test your search strategy
A search strategy that misses key studies undermines everything built on top of it. This step usually involves an information specialist or a researcher experienced with database syntax across multiple platforms (MEDLINE, Embase, CINAHL, PsycINFO, and others each use different search logic).
Build your Boolean strings, test them against a set of studies you already know should be included, and document every database, every date range, and every filter applied. This documentation is what makes your search reproducible -- and reproducibility is what a methods reviewer checks first.
4. Screen in two independent stages
Screening happens in two passes: title/abstract screening, then full-text screening. Both should be done independently by at least two reviewers, with disagreements resolved by discussion or a third reviewer. Track your numbers at every stage -- how many records identified, how many duplicates removed, how many excluded and why. This is what becomes your PRISMA flow diagram.
5. Extract data with a piloted form
Design a structured data extraction form before you start pulling numbers from studies, and pilot it on a handful of included studies first. Inconsistent extraction is one of the most common sources of error in systematic reviews, particularly when multiple reviewers are extracting from the same set of studies without a shared template.
6. Assess risk of bias
Match your risk-of-bias tool to your study designs -- Cochrane's RoB 2 for randomized trials, ROBINS-I for non-randomized studies. This step should also be done independently by two reviewers, since risk-of-bias judgments are one of the places methods reviewers most often push back.
7. Synthesize your findings
Depending on your data, this is either a meta-analysis (pooling effect sizes statistically) or a narrative synthesis (organizing findings thematically when studies are too heterogeneous to pool). If you're running a meta-analysis, you'll need to choose between fixed-effect and random-effects models based on your actual heterogeneity, not by default.
8. Report against the PRISMA 2020 checklist
PRISMA 2020 isn't something you check against at the end -- it should shape your manuscript structure from the first draft. Go through the checklist item by item before submission; missing PRISMA items is one of the most common reasons reviews are sent back for revision.
The one thing that trips up most first-time reviewers
Every stage above needs to be decided *before* you start screening, not adjusted partway through based on what you're finding. A review whose methodology shifts mid-process loses its defensibility -- and that's usually the difference between a review that sails through peer review and one that gets stuck in revision cycles.
If you're working through any of these stages and want a second set of eyes on your methodology -- not someone to do it for you, but someone to check your PICO framing, your search strategy, or your risk-of-bias judgments before you submit -- that's exactly the kind of consulting we do.
**[Get a quote](/systematic-reviews#complete-systematic-reviews)** for methodology consulting on your own review.
Related articles