AI Thesis Writer
Guide

How to Do a Systematic Review (PRISMA Step by Step)

How to do a systematic review step by step with PRISMA 2020: question, protocol, search, screening, data extraction, risk of bias, synthesis and reporting.

A systematic review answers a precise research question by finding, selecting, evaluating and combining all the relevant studies with a method that is planned in advance and reported in full, so that others can repeat it. Most systematic reviews follow the PRISMA 2020 reporting guideline. The main steps are: define the question, write and register a protocol, search several databases, screen the results, extract data, assess risk of bias, synthesize the findings and report them with a PRISMA flow diagram.

Systematic review vs literature review

Traditional literature reviewSystematic review
QuestionOften broadNarrow and precise
SearchFlexiblePre-planned, comprehensive and documented
SelectionAuthor's judgmentExplicit inclusion and exclusion criteria
Quality assessmentInformalFormal risk-of-bias assessment
ReportingNarrativeFollows a guideline such as PRISMA
ReproducibleRarelyYes

See how to write a literature review for the traditional kind.

What PRISMA is

PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) is a reporting guideline. The PRISMA 2020 statement, published by Page and colleagues in 2021, includes a 27-item checklist and a flow diagram showing how many records were found, screened, excluded and included. The checklist and diagram templates are available on the PRISMA website.

Step 1: Define the question

Use a framework to make the question precise. For health research, PICO is the most common:

  • Population: adults with type 2 diabetes
  • Intervention: text-message reminders
  • Comparison: usual care
  • Outcome: medication adherence

In adults with type 2 diabetes, do text-message reminders improve medication adherence compared with usual care?

Other fields use frameworks such as SPIDER (for qualitative evidence) or PEO (population, exposure, outcome). The research question generator can help you draft and narrow the question.

Step 2: Write and register a protocol

A protocol sets out your methods before you start: the question, eligibility criteria, databases, search strategy, screening process, data to extract, risk-of-bias tools and synthesis plan. Health-related reviews are often registered in PROSPERO; other fields can use the Open Science Framework. Registering prevents duplication and shows that you did not change your methods after seeing the results.

Step 3: Search systematically

  • Search at least two or three databases relevant to your field, such as PubMed/MEDLINE, Embase, CINAHL, PsycINFO, Scopus, Web of Science or ERIC.
  • Build the search from your question's concepts, combining synonyms with OR and concepts with AND, and use each database's subject headings.
  • Search grey literature (reports, theses, trial registers) if relevant.
  • Record every search string, database, date and number of results.

A librarian can help you build and test the search.

Step 4: Screen the results

  1. Remove duplicates, using a reference manager or screening software.
  2. Title and abstract screening: exclude clearly irrelevant records.
  3. Full-text screening: apply your inclusion criteria and record the reason for every exclusion.

Ideally, two reviewers screen independently and resolve disagreements by discussion or with a third reviewer.

Step 5: Extract data

Use a standard form to record, for each included study: author and year, country, design, participants, intervention or exposure, comparison, outcomes, results and funding. Pilot the form on a few studies first.

Step 6: Assess risk of bias

Judge how much each study's design and conduct could have biased its results, with a recognized tool:

  • RoB 2 for randomized trials;
  • ROBINS-I for non-randomized studies of interventions;
  • CASP checklists or the JBI tools for many other designs, including qualitative studies.

Step 7: Synthesize the findings

  • Narrative synthesis: describe and compare the studies, grouped by outcome, design or population.
  • Meta-analysis: combine numerical results statistically when the studies are similar enough. Report heterogeneity and, where relevant, assess publication bias.
  • Qualitative synthesis: methods such as thematic synthesis or meta-ethnography combine qualitative findings.

Grade the certainty of the evidence for each outcome where your field expects it, for example with GRADE.

Step 8: Report with PRISMA

Write the review following the PRISMA 2020 checklist, and include the flow diagram showing records identified, screened, excluded (with reasons) and included. Report the limitations of the evidence and of your review.

How long does it take?

A full systematic review often takes many months and usually needs at least two reviewers. For a bachelor's or master's thesis, a rapid or scoping review with a smaller scope may be more realistic. Ask your supervisor.

Can AI tools help?

AI tools can speed up searching, deduplication and first-pass screening, and some platforms support whole systematic-review workflows. You still need a documented method, human judgment at each step and an honest report of how you used the tools. To see quickly what the literature on your topic looks like before you plan a review, try the free literature review generator.

Frequently asked questions

How many studies should a systematic review include?

As many as meet your criteria. Some reviews include very few studies; that is itself a finding.

What is the difference between a systematic review and a meta-analysis?

A meta-analysis is a statistical method for combining results. It is often part of a systematic review, but not every systematic review includes one.

Do I need two reviewers?

It is the recommended standard for screening and data extraction. If you work alone, as in many theses, say so and explain how you reduced the risk of error.

Is PRISMA required?

Many journals and universities require PRISMA for systematic reviews. Even when it is not required, it is the best guide to what to report.

Write your thesis from your own sources

Build a source library, approve an outline and draft chapter by chapter, with every reference checked against its DOI record.

Start your thesis free