How to Do a Systematic Review: 7 Essential Steps for a Rigorous, PRISMA-Ready Evidence Synthesis

How to Do a Systematic Review in 7 Essential Steps

Systematic reviews sit at the top of the evidence hierarchy because they aim to summarize all relevant research on a focused question using transparent, reproducible methods. But they can also feel intimidating: thousands of search results, complex screening decisions, and unfamiliar terms like “PICO,” “MeSH,” “risk of bias,” and “heterogeneity.” The good news is that a strong systematic review is less about genius and more about process. When you follow a structured workflow, you reduce bias, save time, and produce conclusions decision-makers can trust.

This guide walks you through seven essential steps to master the systematic review process, with brief explanations of key terms (and common synonyms) along the way. Whether you’re preparing a publication, a thesis chapter, or evidence for policy, these steps will help you deliver a review that is rigorous, PRISMA-ready, and genuinely useful.

1. Define the Research Question (and Make It Answerable)

Your research question determines everything: which studies you include, what you extract, and how you synthesize results. In systematic review methodology, the question must be specific enough to guide searching and screening, yet meaningful enough to address a real evidence gap.

A popular tool is the PICO framework—a shorthand for defining your question:

  • Population: Who is being studied?
  • Intervention/Exposure: What is being done or experienced?
  • Comparison: What is the intervention/exposure compared against?
  • Outcomes: What outcomes are measured?

Synonyms you may see: PICO is sometimes expanded or adapted (e.g., PICOS adds “Study design”; PICOT adds “Time”). For qualitative reviews, you may encounter SPIDER or PICo (Population–Interest–Context).

Before you move on, pressure-test your question:

  • Is the scope realistic? A narrowly defined population or outcome often improves feasibility.
  • Are outcomes measurable? Avoid vague outcomes like “improved well-being” unless you specify how it’s assessed.
  • Does it match the review type? Questions about effectiveness often fit meta-analysis, while questions about experiences may fit qualitative synthesis.

2. Draft a Protocol (Your Review’s Blueprint)

A protocol is a pre-specified plan describing exactly how you will run the review. It protects against “researcher degrees of freedom” (changing methods after seeing results) and improves credibility. Many journals and funders increasingly expect protocol-driven reviews.

At minimum, a good protocol includes:

  • Objectives and the full research question
  • Eligibility criteria (also called inclusion/exclusion criteria)
  • Databases and sources you will search, including grey literature
  • Screening process (who screens, how disagreements are handled)
  • Risk-of-bias/quality appraisal tools
  • Planned synthesis approach (meta-analysis vs narrative vs thematic)

Registering your protocol (often via PROSPERO for health-related systematic reviews) creates a public timestamp of your plan. If PROSPERO isn’t suitable for your topic, consider alternatives such as OSF registries or institutional repositories.

3. Conduct a Comprehensive Literature Search

The search is where systematic reviews earn the word “systematic.” A comprehensive strategy aims to capture all relevant evidence—not just the most visible or convenient papers. Missing studies can distort conclusions, especially if publication bias is present.

Choose the Right Sources

Common databases include PubMed/MEDLINE, Embase, Cochrane Library, Web of Science, and Scopus. Your field may also rely on domain-specific databases (e.g., PsycINFO, CINAHL, ERIC).

Grey literature refers to research not published in traditional journals—such as theses, conference abstracts, trial registries, government reports, and preprints. Including it can reduce publication bias and reveal emerging evidence.

Build Search Strings (Controlled Vocabulary + Keywords)

Most databases support controlled vocabulary—standardized indexing terms. In PubMed these are called MeSH (Medical Subject Headings); in Embase they’re Emtree. Pair controlled vocabulary with free-text keywords to catch newer terms and variations.

Synonyms matter. For example, “myocardial infarction” may also appear as “heart attack.” “Randomized controlled trial” may appear as “RCT,” “randomised,” or “clinical trial.” A strong search strategy lists synonyms, spelling variants (US/UK), and abbreviations.

Iterate and Document

Search strategies often improve through testing. Run a pilot search, confirm that known key studies appear, then refine. Document the final search date, databases, and exact strings—this is essential for reproducibility and PRISMA reporting.

4. Screen and Select Studies (Reduce Noise, Keep Evidence)

Screening turns your large search result set into the final evidence base. This is usually done in two stages:

  • Title/abstract screening: a fast filter to remove clearly irrelevant records
  • Full-text screening: a detailed eligibility check against your criteria

To reduce bias, best practice is to use two independent reviewers for screening (or at least for a subset), then resolve disagreements by discussion or a third reviewer.

Use a PRISMA flow diagram to report how many records were identified, deduplicated, excluded, and included. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) is a reporting guideline—not a method—but it helps readers see that your process was transparent.

Helpful synonym: “Study selection” is often used interchangeably with “screening.”

5. Assess Study Quality (Risk of Bias, Not Just “Good vs Bad”)

A systematic review is only as reliable as the studies it includes. That’s why you assess risk of bias—the possibility that a study’s design, conduct, or reporting systematically distorts results. This is different from general “quality” in the casual sense; it focuses on threats to validity.

Choose the tool that matches the study design:

  • RoB 2 (Cochrane Risk of Bias 2): for randomized controlled trials (RCTs)
  • ROBINS-I: for non-randomized intervention studies
  • QUADAS-2: for diagnostic accuracy studies

Risk-of-bias domains often include randomization, allocation concealment, blinding, missing data (attrition), selective reporting, and confounding. Summarize results in tables and, when possible, visual plots to help readers interpret your conclusions.

How it influences synthesis: You may run sensitivity analyses (re-running analyses excluding high-risk studies) or downgrade certainty when most evidence is biased.

6. Extract and Synthesize Data (From Papers to Evidence)

Data extraction is the step where you convert each included study into structured information. Use a standardized extraction form and pilot it on a few studies to ensure consistency across reviewers.

What to Extract

  • Study identifiers (authors, year, country)
  • Design and setting (e.g., RCT, cohort; primary care vs hospital)
  • Participant characteristics and sample size
  • Intervention/exposure and comparison details
  • Outcome definitions and time points
  • Effect estimates (e.g., risk ratio, odds ratio, mean difference) and uncertainty (CI/SE)

Choose the Right Synthesis Method

Meta-analysis is a statistical approach that combines comparable effect sizes into a pooled estimate. It is most appropriate when studies are sufficiently similar in PICO and outcomes. If studies are too diverse, a narrative synthesis (structured textual summary) may be more appropriate. For qualitative studies, a thematic synthesis (or similar qualitative synthesis method) can identify recurring themes and patterns.

Key term: Heterogeneity means variation between study results. It can be clinical (different populations), methodological (different designs), or statistical (different effect sizes). In meta-analysis, heterogeneity is often summarized using I² and sometimes tau².

Common software options include RevMan, Stata, or R (e.g., meta and metafor packages). Whichever you choose, document versions, settings, and models (fixed vs random effects).

7. Report and Present Your Systematic Review (PRISMA-Ready Output)

A systematic review becomes impactful when it is clearly reported. Follow a standard structure—Abstract, Introduction, Methods, Results, Discussion, Conclusions—and align with PRISMA to ensure completeness.

What Strong Reporting Looks Like

  • Transparent methods: full search strategy, screening process, and tools used
  • Clear results: PRISMA flow diagram, study characteristics tables, risk-of-bias summaries
  • Balanced discussion: interpret findings while acknowledging limitations (bias, heterogeneity, imprecision)
  • Actionable implications: for practice, policy, and future research

Where possible, share supplementary materials such as extraction forms, datasets, analytic code, and full search strategies. This strengthens trust and makes your work reusable.

FAQ: Common Systematic Review Terms (Quick Definitions)

What’s the difference between a systematic review and a literature review?

A systematic review follows a predefined, reproducible method to find and synthesize all relevant studies. A traditional literature review (sometimes called a narrative review) may be more selective and interpretive, but is often less transparent about search and selection methods.

What does “grey literature” mean?

Grey literature includes research outside peer-reviewed journals—theses, reports, trial registries, and conference proceedings. Including it can reduce publication bias.

What does PRISMA mean?

PRISMA is a reporting guideline (a checklist and flow diagram) that helps ensure systematic reviews are reported transparently and completely.

Conclusion

Mastering the systematic review process is about following a clear, defensible workflow: define an answerable question, write and register a protocol, run a comprehensive search, screen studies consistently, assess risk of bias, extract and synthesize data appropriately, and report transparently using PRISMA. If you start with a tight research question and a detailed protocol, the rest of the review becomes far more manageable—and your final product will be more credible, reproducible, and publishable.

If you’re beginning today, open a blank protocol document and write your PICO question and eligibility criteria first. That single step will clarify every decision that follows.