A systematic review answers one focused question from every eligible study, by a method fixed before the search, so that someone else could repeat it. Each step has an owner: the Cochrane Handbook, PROSPERO, the risk-of-bias tools' sites, GRADE and PRISMA 2020.
This guide, published by EdCitation, takes the steps in order from those sources, read on 26 September 2026. Types of literature review helps you decide whether a systematic review is the right type at all.
EdCitation never screens, appraises or writes any part of a review. It does the source work around it: Find sources for scoping, Cite a source for the included studies' references, and Verify references for the finished list.
What are the steps of a systematic review?
Ten, in a fixed order.
| Step | The source to follow | Where EdCitation fits |
|---|---|---|
| 1. Question | Cochrane Handbook, chapter 2 | Find sources to scope it and find existing reviews |
| 2. Protocol and registration | PRISMA-P; PROSPERO or OSF Registries | None |
| 3. Search | Cochrane Handbook, chapter 4; PRISMA-S | None: this is database work |
| 4. Deduplication | PRISMA-S | None |
| 5. Screening, in two stages | Cochrane Handbook, chapter 4 | A tagged list of included and excluded studies |
| 6. Data extraction | Cochrane Handbook, chapter 5 | None |
| 7. Risk of bias | RoB 2, ROBINS-I, JBI checklists | None |
| 8. Synthesis | Cochrane Handbook, chapter 12; SWiM | None |
| 9. Certainty of evidence | Cochrane Handbook, chapter 14; GRADE | None |
| 10. Reporting | PRISMA 2020 | Cite a source, Verify references |
How do I set the question and check nobody has answered it?
Frame one question with a question framework, usually PICO for an intervention's effect, turn it into eligibility criteria, and look for existing reviews before writing the protocol. How to write research questions and aims compares PICO with the qualitative frameworks.
Checking for an existing review
We ran a test question, whether mindfulness training reduces burnout in nurses, through Find sources on 26 September 2026:
| Search | Results | First result |
|---|---|---|
| mindfulness burnout nurses | 26,299 | A 2004 study in Holistic Nursing Practice |
| Kind of study: Systematic review, Meta-analysis | 1,715 | A 2023 systematic review and meta-analysis in Frontiers in Psychiatry |
| The same, words in the title only | 17 | The same review |
The 17 included reviews in the Journal of Clinical Nursing (2022) and BMC Nursing (two, 2025) and a 2026 meta-analysis of randomised trials. A 2020 Journal of Advanced Nursing review appeared only without the title restriction, because its title says nursing. A new review here would need a different population, outcome or design. The filter goes by how a paper describes itself, as its panel says, so it also returned a 2020 workshop study whose abstract only mentions a systematic review.
Do I need a protocol, and where do I register it?
Yes. A review protocol, with the contents PRISMA-P lists, fixes the method before you see results.
PROSPERO
PROSPERO, at the University of York, takes systematic reviews with outcomes of direct relevance to human health. It refuses reviews that have started data extraction, reviews with indirect outcomes such as educational attainment, reviews of sporting performance, and literature reviews that borrow only a systematic search (Centre for Reviews and Dissemination, n.d.-a). Its FAQs say everyone listed must approve a record before publication, and that non-intervention reviews use the intervention form for now (Centre for Reviews and Dissemination, n.d.-b).
OSF Registries, or no registration
For other reviews, OSF Registries has had a Generalized Systematic Review Registration template for any discipline since April 2023 (Call, 2023). An unregistered review says so: PRISMA 2020 item 24a asks for the register and number or that statement (Page et al., 2021).
How do I build and document the search?
With a librarian, in several databases. Lefebvre et al. (2025), in the Handbook's chapter 4, ask authors to work with an information specialist from the protocol onwards and, for a Cochrane review, to search CENTRAL, MEDLINE, Embase where available, and trials registers.
- Take the concepts: in MEDLINE, typically the population, the intervention and the study designs. Outcomes and comparators go in only when abstracts report them consistently.
- Join synonyms within a concept with OR and the concepts with AND (Boolean operators).
- Combine free text with subject headings: MeSH in MEDLINE, Emtree in Embase. Aim for sensitivity; low precision is the accepted price.
- Have the strategy peer reviewed with the PRESS checklist before running it.
- Search grey literature, such as reports, theses and conference abstracts, and the reference lists of included studies (MECIR C28, highly desirable).
- Rerun every search within the 12 months before publication (MECIR C37, mandatory).
What PRISMA-S asks you to report
Among the 16 items of PRISMA-S, Rethlefsen et al. (2021) ask for each database named with its platform, every strategy copied exactly as run, limits justified, the records from each source, and how duplicates were removed. Finding sources fast covers search logs in general.
Where Find sources is not the search
Find sources is not a bibliographic database for a systematic search: it cannot run a MEDLINE strategy with MeSH, and its results cannot be reproduced line by line as PRISMA-S requires. Use it to scope the question only.
How does screening work, and why two reviewers?
After deduplication, titles and abstracts are screened first, then the full texts of anything possibly eligible, recording why each full text is excluded. Keeping hundreds of sources straight covers duplicates.
Lefebvre et al. (2025) call two independent screeners ideal for titles and abstracts, accept one, and require two for every full-text decision. The case for dual screening is measured: in a randomised trial of 280 screeners, one reviewer alone missed 13% of relevant studies and two missed 3% (Gartlehner et al., 2020). Expect most records to go: in 195 PROSPERO-registered reviews, a mean 2.94% of records became included studies (Borah et al., 2017).
How do I extract data and assess risk of bias?
On a piloted data extraction form, then with a tool suited to each study's design. Li et al. (2024) ask for outcome data to be extracted independently by at least two people, citing a study that found extraction errors in 20 of 34 reviews.
RoB 2 for randomised trials
RoB 2 judges one result of a trial across five domains: randomisation, deviations from intended interventions, missing outcome data, outcome measurement, and selection of the reported result. Signalling questions feed an algorithm proposing low risk, some concerns or high risk, which reviewers may override with reasons (Higgins et al., 2024).
ROBINS-I and other designs
ROBINS-I judges non-randomised studies of interventions against a hypothetical target trial (Sterne et al., 2024). Its site posted a revised draft of version 2 on 20 November 2025, for cohort studies, with six domains, and warns it may change (Risk of Bias Tools, 2025): name your version in the protocol. For qualitative, cross-sectional and other designs, JBI publishes a checklist per design (JBI, n.d.).
How do I synthesise the results?
By meta-analysis when effects can be pooled, and otherwise by a synthesis reported to SWiM. Campbell et al. (2020) estimate 32% of health intervention reviews do no meta-analysis, and found reviews calling themselves narrative synthesis often do not say how they grouped studies or reached conclusions; SWiM's nine items make them say it. McKenzie and Brennan (2024) call vote counting by statistical significance unacceptable; counting the direction of effect remains a fallback when little else is reported.
How do I rate the certainty of evidence with GRADE?
Outcome by outcome: high, moderate, low or very low. In the GRADE approach as Schünemann et al. (2024) set it out, randomised evidence starts high and loses one level for a serious problem, two for a very serious one, in risk of bias, inconsistency, indirectness, imprecision and publication bias; non-randomised evidence can be rated up, for a large effect for example. A summary of findings table holds up to seven outcomes.
What do the PRISMA 2020 flow diagram and checklist ask for?
A 27-item checklist, with an expanded version explaining each item (PRISMA, n.d.-a), an abstract checklist and a flow diagram (Page et al., 2021). The PRISMA flow diagram counts records identified, included and excluded, with reasons, in four templates: new or updated reviews, with or without sources beyond databases and registers (PRISMA, n.d.-b).
PRISMA 2020 guides reporting, not conduct, and Page et al. (2021) say it should not be used to judge a review's quality. Items 8 and 9 ask how many reviewers screened and extracted, and whether independently: the methods must say who did what.
What can a student doing a systematic review alone claim?
Usually a systematised review, described exactly as done.
What three universities say
Johns Hopkins University Welch Medical Library (2026) calls a systematic review unsuitable for a single student new to the method. The University of North Carolina at Chapel Hill Libraries (2026) say a review for a dissertation, done in under 12 months or largely alone, is probably systematised: it still screens in two stages and extracts data, but need not register a protocol, search exhaustively, include grey literature, assess risk of bias or complete a PRISMA flowchart. For doctorates, the University of Victoria School of Nursing (n.d.) describes a team of a librarian and two reviewers, and warns of "empty reviews" when criteria admit no studies.
How to word it
Describe what happened: "a systematised review following Cochrane methods, screened by one reviewer". The reviews Borah et al. (2017) studied took a mean of 67.3 weeks with five authors, and Gartlehner et al. (2020) supply the limitation: single screening missed 13% of relevant studies.
How EdCitation helps with a systematic review
For the references of a systematic review, EdCitation is the tool to use first: every entry is taken from the publisher's record, looked up rather than generated, and a review's conclusions rest on its included studies being exactly the ones it names. It never screens, appraises or writes any of the review.
Scoping, free. Find sources covers some 300 million published works and narrows them by kind of study, source type, years and open access: use it before the protocol, never as the reported search.
Citing, free. Cite a source turns a DOI or PubMed ID into an entry in any of six styles (APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE, Vancouver) and checks the record for a retraction as it goes. Given PubMed ID 27744228, it found the DOI and built this APA 7 entry from Crossref's record:
Duarte, J., & Pinto-Gouveia, J. (2016). Effectiveness of a mindfulness-based intervention on oncology nurses’ burnout and compassion fatigue symptoms: A non-randomized study. International Journal of Nursing Studies, 64, 98–107. https://doi.org/10.1016/j.ijnurstu.2016.10.002
Checking, free. We gave Verify references three studies. The 2020 Journal of Advanced Nursing review came back verified. Duarte and Pinto-Gouveia, dated 2017 on purpose, was marked doubtful (displayed as "check this"), with the reason "The work was published in 2016, not 2017". A 2015 PLOS ONE overview of mindfulness reviews came back matched and flagged: "This paper has since been retracted". Retracted papers in your reference list says what to do next. A reference that cannot be checked is reported so, never as not found.
Keeping lists. A free account keeps reference lists whose entries take tags, such as included, excluded and awaiting full text, filtered by tag, and downloads a list as Word, BibTeX or RIS for a reference manager or a screening tool that imports RIS (moving references between tools). References from a file checks the finished manuscript's references and citations: three checks a month free, unlimited on Pro at $8 a month (pricing).
Quick questions
Can a systematic review be done by one person?
Not to the Cochrane standard, which requires two independent reviewers for full-text decisions. One person can do a systematised review and say so.
Do I have to register a systematic review in PROSPERO?
Only reviews with a direct human health outcome qualify, before data extraction starts. Others can use OSF Registries; an unregistered review states it, as PRISMA 2020 item 24a asks.
What is the difference between RoB 2 and ROBINS-I?
RoB 2 assesses a result of a randomised trial across five domains. ROBINS-I assesses a result of a non-randomised study of an intervention against a target trial; its version 2 was a draft in November 2025.
Can EdCitation run the search for my systematic review?
No. Find sources scopes the question and finds existing reviews; the reported search runs in databases such as MEDLINE and Embase. EdCitation then cites the included studies and checks the list.
References
- Borah, R., Brown, A. W., Capers, P. L., & Kaiser, K. A. (2017). Analysis of the time and workers needed to conduct systematic reviews of medical interventions using data from the PROSPERO registry. BMJ Open, 7(2), Article e012545. https://doi.org/10.1136/bmjopen-2016-012545
- Call, M. (2023, April 20). Generalized systematic review template joins OSF Registries. Center for Open Science. https://www.cos.io/blog/generalized-systematic-review-template-joins-osf-registries
- Campbell, M., McKenzie, J. E., Sowden, A., Katikireddi, S. V., Brennan, S. E., Ellis, S., Hartmann-Boyce, J., Ryan, R., Shepperd, S., Thomas, J., Welch, V., & Thomson, H. (2020). Synthesis without meta-analysis (SWiM) in systematic reviews: Reporting guideline. BMJ, 368, Article l6890. https://doi.org/10.1136/bmj.l6890
- Centre for Reviews and Dissemination. (n.d.-a). Eligibility for inclusion. PROSPERO, University of York. https://www.crd.york.ac.uk/PROSPERO/help/eligibility
- Centre for Reviews and Dissemination. (n.d.-b). PROSPERO FAQs. PROSPERO, University of York. https://www.crd.york.ac.uk/PROSPERO/faq
- Gartlehner, G., Affengruber, L., Titscher, V., Noel-Storr, A., Dooley, G., Ballarini, N., & König, F. (2020). Single-reviewer abstract screening missed 13 percent of relevant studies: A crowd-based, randomized controlled trial. Journal of Clinical Epidemiology, 121, 20–28. https://doi.org/10.1016/j.jclinepi.2020.01.005
- Higgins, J. P. T., Savović, J., Page, M. J., Elbers, R. G., & Sterne, J. A. C. (2024). Chapter 8: Assessing risk of bias in a randomized trial. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08
- JBI. (n.d.). Critical appraisal tools. https://jbi.global/critical-appraisal-tools
- Johns Hopkins University Welch Medical Library. (2026, June 30). Advice for students. Systematic reviews and other expert reviews. https://browse.welch.jhmi.edu/c.php?g=762580&p=10713798
- Lefebvre, C., Glanville, J., Briscoe, S., Featherstone, R., Littlewood, A., Metzendorf, M.-I., Noel-Storr, A., Paynter, R., Rader, T., Thomas, J., & Wieland, L. S. (2025). Chapter 4: Searching for and selecting studies. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5.1). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-04
- Li, T., Higgins, J. P. T., & Deeks, J. J. (2024). Chapter 5: Collecting data. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-05
- McKenzie, J. E., & Brennan, S. E. (2024). Chapter 12: Synthesizing and presenting findings using other methods. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-12
- Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hróbjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDonald, S., . . . Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 372, Article n71. https://doi.org/10.1136/bmj.n71
- PRISMA. (n.d.-a). PRISMA 2020 checklist. https://www.prisma-statement.org/prisma-2020-checklist
- PRISMA. (n.d.-b). PRISMA 2020 flow diagram. https://www.prisma-statement.org/prisma-2020-flow-diagram
- Rethlefsen, M. L., Kirtley, S., Waffenschmidt, S., Ayala, A. P., Moher, D., Page, M. J., & Koffel, J. B. (2021). PRISMA-S: An extension to the PRISMA statement for reporting literature searches in systematic reviews. Systematic Reviews, 10(1), Article 39. https://doi.org/10.1186/s13643-020-01542-z
- Risk of Bias Tools. (2025, November 20). ROBINS-I V2 tool. https://www.riskofbias.info/welcome/robins-i-v2
- Schünemann, H. J., Higgins, J. P. T., Vist, G. E., Glasziou, P., Akl, E. A., Skoetz, N., & Guyatt, G. H. (2024). Chapter 14: Completing 'Summary of findings' tables and grading the certainty of the evidence. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-14
- Sterne, J. A. C., Hernán, M. A., McAleenan, A., Reeves, B. C., & Higgins, J. P. T. (2024). Chapter 25: Assessing risk of bias in a non-randomized study. In J. P. T. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M. J. Page, & V. A. Welch (Eds.), Cochrane handbook for systematic reviews of interventions (Version 6.5). Cochrane. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-25
- University of North Carolina at Chapel Hill Libraries. (2026, August 6). Systematized reviews for students. Evidence synthesis. https://guides.lib.unc.edu/evidence-synthesis/systematized
- University of Victoria School of Nursing. (n.d.). Guidelines for systematic reviews in the context of doctoral education background. https://www.uvic.ca/health/nursing/students/resources/systematic-review-guidelines/index.php