How to Write a Systematic Review, Step by Step

A systematic review answers a focused question by finding, appraising and synthesizing all the relevant research using an explicit, reproducible method.

That method is what separates it from a traditional literature review. Another researcher following your protocol should be able to find the same studies and reach similar conclusions.

Students meet systematic reviews in health, nursing, psychology, education and social science programs, as dissertation projects or assignments. The work is demanding because every step must be documented. This guide walks through each stage, from the question and protocol to reporting with PRISMA 2020, and explains the tools reviewers rely on.

Types of Evidence Review

Student projects often call themselves systematic reviews but use a single reviewer and a limited search. Many programs accept this as a "systematic literature review" or "systematic approach". Describe honestly what you did.

Framing the Question

Writing and Registering a Protocol

The protocol sets out your methods before you begin, which protects against bias from changing decisions once you see results. It covers the question, eligibility criteria, information sources, search strategy, screening process, data extraction, risk of bias assessment, synthesis methods and any planned subgroup analyses.

Published reviews are usually registered on PROSPERO (for health topics) or the Open Science Framework. Student reviews may not require registration, but writing a protocol and following it is good practice, and examiners notice when methods appear to have changed midway.

Setting Inclusion and Exclusion Criteria

Clear criteria make screening consistent and are what allow another reviewer to reproduce your selection.

Building the Search Strategy

• Concepts: break the question into two or three key concepts, such as diabetes, text messaging and glycaemic control.
• Terms: list synonyms and subject headings for each (MeSH in PubMed, CINAHL Headings in CINAHL).
• Combine: use OR within a concept and AND between concepts; use truncation (remind*) and phrase searching.
• Databases: search at least two or three relevant databases, plus gray literature where appropriate.
• Document: record every database, date, search string and number of results, so the search can be reproduced.

Example Search String

Each line combines synonyms and a subject heading for one concept with OR, and the lines are joined with AND. Translate the strategy for each database, because subject headings and syntax differ.

Screening and the PRISMA Flow Diagram

Screening happens in stages: remove duplicates, screen titles and abstracts against your inclusion criteria, retrieve full texts of possible studies, and assess them in detail, recording a reason for every full-text exclusion.

Ideally two reviewers screen independently and resolve disagreements by discussion. Tools such as Rayyan or Covidence make this manageable. The PRISMA 2020 flow diagram reports the number of records at each stage: identified, screened, assessed for eligibility and included, with reasons for exclusion.

Data Extraction

Use a piloted extraction form so the same information is captured from every study.

Quality Appraisal and Risk of Bias

Report the appraisal for each study, usually in a table, and use it in the synthesis: findings from high-risk studies should carry less weight. Use GRADE, where appropriate, to rate the overall certainty of evidence for each outcome.

Synthesis: Narrative or Meta-Analysis

If studies are similar enough in population, intervention, comparison and outcome, a meta-analysis can pool their results statistically, reporting a combined effect size with a confidence interval, a forest plot and a measure of heterogeneity such as I².

If they are too different, use a structured narrative synthesis: group studies by intervention type or outcome, describe patterns, explain differences in light of study quality and context, and avoid "vote counting" (simply tallying positive and negative studies). Qualitative reviews use approaches such as thematic synthesis or meta-aggregation.

Writing the Results and Discussion

The results section reports the study selection (with the PRISMA flow diagram), the characteristics of included studies (usually in a table), the risk of bias assessment, and the synthesis for each outcome, with the certainty of evidence where GRADE was used.

The discussion summarizes the main findings, compares them with previous reviews, considers the strengths and limitations of both the evidence and your review methods, and states the implications for practice, policy and future research.

Keep interpretation proportionate to the quality of the evidence; a review of small, high-risk studies cannot support strong recommendations.

Reporting with PRISMA 2020

PRISMA 2020 is a 27-item checklist covering the title, abstract, introduction, methods (eligibility, sources, search, selection, extraction, risk of bias, synthesis), results (study selection, characteristics, risk of bias, results of syntheses, certainty of evidence), discussion, and other information such as registration and funding.

Submit the completed checklist as an appendix if your program asks for it, and include the full search strategy for at least one database.

A Realistic Timeline

Screening and extraction take longer than most students expect. A narrower question is the most effective way to keep the review manageable.

Common Mistakes

• A question that is too broad, producing thousands of irrelevant results.
• Searching only one database or only Google Scholar.
• Undocumented searches that cannot be reproduced.
• No reasons recorded for full-text exclusions.
• Appraisal done but not used in the synthesis.
• Summarizing studies one by one instead of synthesizing.
• Calling a narrative review "systematic" without the methods to support it.

How We Help

Support is available for each stage: framing the question and protocol, building and running search strategies, screening, data extraction tables, quality appraisal, narrative synthesis or meta-analysis of the data you have extracted, PRISMA reporting and editing.

Send your protocol or brief, your search results or included studies, and your deadline. Revisions within the scope of your original request are free.

Software That Helps

Several tools make a systematic review manageable. Reference managers such as EndNote, Zotero and Mendeley store results and remove duplicates. Rayyan (free) and Covidence (subscription, often provided by universities) support blinded title, abstract and full-text screening by two reviewers.

RevMan and R packages such as metafor run meta-analyses and draw forest plots. A shared spreadsheet works well for data extraction in small reviews. Whatever you use, export and keep records of every stage, because examiners may ask to see your screening decisions.

Publication Bias and How to Address It

Studies with positive or significant results are more likely to be published, so a review based only on journal articles can overestimate an intervention's effect. Reduce the risk by searching trial registries and gray literature, including studies in any language you can translate, and contacting authors about unpublished data.

In a meta-analysis with enough studies (often ten or more), a funnel plot and statistical tests can suggest whether small negative studies are missing. Discuss publication bias as a limitation whenever you cannot rule it out.