PaperLabHow do you do a systematic literature review?
Define a focused question, write and ideally register a protocol, then search several databases with a documented query. Remove duplicates, screen titles and abstracts against your criteria, then screen full texts and record a reason for every exclusion. Snowball through the references of included papers, extract data in a fixed form, appraise quality, synthesise the findings, and report the counts at each stage in a PRISMA flow diagram.
By Malek Khannoussi · Updated25 September 2026
1. Frame the question
A focused question decides everything after it. In health research, PICO (population, intervention, comparison, outcome) is the usual frame; in software engineering and the social sciences, a clear scope with inclusion and exclusion criteria does the same job.
2. Write a protocol
Record the question, databases, search strings, criteria and planned analysis before you start. Health reviews register it on PROSPERO; others can publish it on OSF. A protocol protects you from changing the rules halfway.
3. Search, and document the search
- Search several sources: subject databases, plus broad ones such as Crossref, Semantic Scholar or OpenAlex.
- Save every query string, date and result count: reviewers will ask.
- Remove duplicates by DOI first, then by title.
4. Screen in two stages
- Title and abstract: quick include, exclude or maybe against your criteria.
- Full text: read the paper and record the reason for every exclusion.
5. Snowball
Snowballing: for every included paper, check its references (backward) and the papers citing it (forward). Screen what you find with the same criteria, and count it separately in your flow diagram.
6. Extract, appraise, synthesise
- Extract into a fixed form, and keep the page each number came from.
- Appraise quality with a tool suited to the study design.
- Synthesise narratively, or with a meta-analysis when studies are comparable.
7. Report with PRISMA
Follow the PRISMA 2020 checklist and include the flow diagram. The numbers are easy if you counted as you went, and painful to rebuild at the end. See also PaperLab for systematic reviews.
Questions people ask
How long does a systematic review take?
Often six months to over a year for a full review with two reviewers. Rapid reviews cut steps to finish in weeks, and should say which steps they cut.
What is a PRISMA flow diagram?
A diagram from the PRISMA 2020 reporting guideline showing how many records were found, removed as duplicates, screened, excluded (with reasons) and included. Most journals expect one.
What is the difference between a systematic review and a literature review?
A systematic review follows a predefined, reproducible method for searching, selecting and appraising studies. A narrative literature review is shaped by the author and doesn’t need to be reproducible.
What is snowballing in a literature review?
Following the references of included papers backwards, and the papers that cite them forwards, to find studies the database search missed.
Do I need two reviewers?
For a full systematic review, independent screening by two people reduces missed studies and bias. When that isn’t possible, say so and have a second person check a sample.