A systematic review is a structured assessment of research on a clearly defined question. Researchers set eligibility criteria, search specified sources, select relevant studies, assess their limitations, and synthesise findings using documented methods. AnswerThis can support the workflow by helping teams organise evidence, generate a PRISMA flow diagram, and draft from their verified research.
TL;DR
A systematic review answers a defined question using documented search, selection, appraisal, and synthesis methods.
Set eligibility rules and a protocol before examining study findings.
Keep exact search strategies, screening decisions and full-text exclusion reasons.
Meta-analysis is optional. PRISMA 2020 guides reporting, not methodological quality.
Check each included study and reconcile records, reports and studies before drawing the flow diagram.
What is a systematic review?
A systematic review uses explicit methods to find, assess, and synthesize studies that answer a defined question. For a clinical team, that means deciding what evidence qualifies before interpreting the results, then keeping a record of the searches and decisions that shaped the final evidence set.
This systematic review definition is about method, not the number of papers summarized. Readers need to see where the team searched, why studies were included or excluded, how study limitations were assessed, and how the findings were combined. For a more detailed walkthrough of the work, see the guide to conducting a systematic literature review.
How does a systematic review differ from a meta-analysis or narrative review?
Systematic review
What it is: A review using explicit methods to identify, select, assess and synthesize studies
Key distinction: It does not have to pool results statistically
Meta-analysis
What it is: A statistical method for combining suitable study results
Key distinction: It can be part of a systematic review, but is not a substitute for one
Narrative review
What it is: An interpretive overview of a topic
Key distinction: It need not use prespecified searches and selection criteria
Scoping review
What it is: A review that maps the extent and characteristics of available evidence
Key distinction: Its question and purpose differ from estimating a treatment effect
PRISMA 2020
What it is: A guideline for reporting a systematic review
Key distinction: It does not certify that the review’s methods were sound
The distinction matters when studies address different populations, interventions or follow-up periods. A team should not combine results merely because the papers share a broad topic. It can instead explain the findings in a structured narrative synthesis.
The PRISMA 2020 statement sets out a 27-item reporting checklist and flow diagrams. For guidance on conducting intervention reviews, use the Cochrane Handbook. PRISMA guides reporting; it is not a methods-quality score.
How to do a systematic review
1. Define the clinical question and write a protocol
Start with a question precise enough to determine which studies qualify. For an intervention review, PICO helps specify the population, intervention, comparator, and outcomes. Decide which study designs, settings, and follow-up periods are relevant.
Record eligibility rules, planned sources, screening methods, extraction fields, risk-of-bias approach, and synthesis plan in a protocol. If the plan changes, document the amendment and its reason.
Keep: A dated protocol and version history. These show which decisions were made before the team saw the study results.
2. Search and preserve the exact strategies
Choose databases, registers, and other sources that fit the clinical question. Adapt the search syntax to each source rather than assuming one query works everywhere. Record the full strategy, filters, run date, and result count for every search.
Database, register, or other source
Full query and filters: Copy the exact strategy used
Date searched: Enter the actual date
Records returned: Enter the actual count
Save revised strategies as separate versions. If an AI research assistant or citation search identifies additional papers, log that discovery route too. An AI answer does not replace the searches required by the protocol.
3. Deduplicate and screen the results
Combine search records and identify duplicates. Screen titles and abstracts against the written eligibility rules, then retrieve and assess potentially relevant full-text reports.
Record each stage’s decision. At full-text assessment, give a primary reason for each exclusion, such as wrong population, intervention, outcome or study design. Follow the protocol’s plan for independent screening and resolving disagreements.
A usable selection log contains:
A record or report identifier
The screening stage
Each reviewer’s decision, where applicable
The final decision
The primary reason for a full-text exclusion
These records support the eventual flow diagram. Do not reconstruct its counts from memory.
4. Extract data and assess risk of bias
Set extraction fields before working through all included studies. For a clinical intervention question, use a sheet that keeps results traceable to the report:
Add fields your protocol needs, such as setting or adverse effects. Include the page, table or passage behind a consequential value. If reviewers extract data independently, reconcile differences against the original report.
Assess risk of bias with an approach suited to the included study designs. Keep that judgment distinct from the certainty of the body of evidence for an outcome. Neither can be inferred from polished writing or a journal’s reputation.
5. Synthesize the findings
Group studies addressing comparable populations, interventions and outcomes. Check differences in design, follow-up and risk of bias before deciding whether statistical pooling makes sense.
If pooling is inappropriate, a structured narrative synthesis can still answer the question. State which studies support each finding, explain conflicting results, and make limitations visible. Check AI-assisted draft claims against the extraction sheet and original reports before including them.
6. Report what the team did
Describe the planned and actual searches, selection decisions, study characteristics, assessments and synthesis. Reconcile every count in the selection flow with the search and screening logs. Complete the applicable PRISMA 2020 checklist alongside the manuscript.
A flow diagram is one part of transparent reporting. It cannot establish that the question, searches or appraisal methods were appropriate.
A clinical example: remote blood-pressure monitoring
This is an illustrative question and planning template, not a completed review or a claim about treatment effects.
Suppose a team asks: Among adults with hypertension, how does remote blood-pressure monitoring compare with usual care for blood-pressure outcomes?
Before searching, the team must decide:
Population: Which age and diagnosis criteria identify eligible adults?
Intervention: Does monitoring require clinician feedback, or does self-monitoring alone qualify?
Comparator: What counts as usual care?
Outcomes: Which blood-pressure measures and follow-up times matter?
Study designs: Which designs can answer the question?
Exclusions: How will mixed populations or interventions with several inseparable components be handled?
The team then runs its planned searches and screens records using those rules. In the extraction sheet, it records the outcome and follow-up time separately. A result measured shortly after treatment begins should not be presented as though it answers a longer-term question.
The question and prespecified decisions determine the evidence set. The result the team hopes to find does not.
How to read a PRISMA 2020 flow diagram
A flow diagram tracks evidence from identification through inclusion. It distinguishes three units:
A record is an indexed search entry, often a title and abstract.
A report is a document about research, such as a journal article or trial-register entry.
A study is the investigation itself. One study can have several reports.
The diagram accounts for records found and removed, records screened, reports sought and assessed, reasons for full-text exclusions, and studies included. Reports that could not be retrieved are recorded separately from reports excluded after their content was assessed.
Keep a report-to-study mapping. Otherwise, multiple publications about one investigation can inflate the included-study count. Also reconcile reports found through more than one search route rather than counting the same report twice. For more details, explore here.
Common mistakes that weaken a clinical review
Changing eligibility rules without a record. Date and explain protocol amendments.
Saving database names but not queries. Preserve each full strategy, filter, and search date.
Losing full-text exclusion reasons. Record one primary reason for each excluded report.
Confusing reports with studies. Map all reports to the investigations they describe.
Accepting a citation without checking the claim. Verify its population, outcome, and result in the source.
Treating PRISMA as proof of quality. Transparent reporting lets readers inspect a review; it does not guarantee sound methods.
AnswerThis can help clinical teams find papers, organize evidence, and draft cited text. Its clinical research workflow is relevant once the team has defined its question and review process. Researchers remain responsible for search coverage, screening decisions, appraisal, and checking every consequential claim.
FAQs
Is a systematic review the same as a meta-analysis?
No. A systematic review is a structured process for finding, assessing and synthesizing evidence. A meta-analysis is an optional statistical method for combining suitable results.
Does every systematic review need a PRISMA flow diagram?
A review reported under PRISMA 2020 should use the applicable flow diagram to show how evidence was selected. The diagram does not replace the reporting checklist or validate the review’s methods.
What is the difference between a systematic and a narrative review?
A systematic review uses explicit, documented methods for searching, selecting, and assessing studies addressing a defined question. A narrative review can interpret a topic without the same prespecified selection process.
How do researchers decide which studies to include?
Researchers apply eligibility criteria set before screening. They assess search records and full-text reports against those rules, document decisions, and give reasons for full-text exclusions.
Can AI conduct a systematic review on its own?
No. AI can assist with discovery, organization, extraction, or drafting. The research team must verify sources and decisions, assess the evidence, and report what it actually did.
Next step: Once your search and screening logs reconcile, build a PRISMA flow diagram using the actual counts.
