Systematic Review
A structured synthesis of all available evidence on a specific research question, following a pre-registered protocol to minimize bias — the highest-quality evidence type in evidence-based medicine and policy.
What it means
A systematic review is a structured research synthesis that identifies, selects, and critically appraises all available evidence relevant to a specific research question, following a pre-specified protocol. It differs from a narrative review (which is selective and reflects the authors’ judgment) by requiring exhaustive search, transparent inclusion/exclusion criteria, and reproducible methods.
The key components:
- Protocol registration — the research question, search strategy, and inclusion criteria are registered in advance (see Pre-Registration)
- Systematic search — multiple databases are searched using controlled vocabulary to capture all relevant evidence
- Screening — titles, abstracts, and full texts are screened independently by at least two reviewers
- Data extraction — key data from each included study is extracted into a standardized form
- Quality assessment — the risk of bias in each study is assessed using a validated tool
- Synthesis — results are combined, either qualitatively or quantitatively (see Meta-Analysis)
The process is standardized by reporting guidelines — PRISMA is the most widely used — and tools like the Cochrane Handbook.
Why AI matters here
AI tools can accelerate the most time-consuming parts of a systematic review without replacing the methodological rigor:
- Elicit automates structured data extraction across hundreds of papers, reducing the time to populate a data extraction table from days to hours
- Rayyan uses AI to suggest inclusion/exclusion decisions during the title-and-abstract screening phase, reducing reviewer effort without eliminating human oversight
- Semantic Scholar and OpenAlex support the systematic search component
What AI tools do not do: guarantee methodological compliance. The PRISMA checklist, risk-of-bias assessments, and final synthesis judgments still require human expertise and judgment. Using AI tools does not make a review systematic if the underlying protocol is weak.