Is a systematic review the right method?

Choose a systematic review when you need a defensible answer to a focused question and can commit to an explicit protocol. A traditional or narrative review may be more appropriate when the goal is broad interpretation, theory-building, or a selective intellectual history. A scoping review is often better when the goal is to map concepts, evidence types, or an emerging field rather than estimate a narrowly defined effect.

Completion test: You can state the decision question in one sentence and explain why systematic methods are necessary.

Systematic review vs. traditional literature review

Both approaches synthesize prior work. The difference is procedural accountability. A systematic review normally uses a protocol, explicit eligibility criteria, reproducible searches, documented screening, appropriate critical appraisal, and a declared synthesis method. A traditional literature review can be rigorous, but it does not necessarily attempt exhaustive retrieval or provide the same audit trail.

PRISMA 2020 improves complete reporting of systematic reviews; it is not itself a conduct methodology. Use a discipline-appropriate conduct guide—such as the Cochrane Handbook for intervention reviews—alongside the relevant PRISMA reporting guidance.

Step 1: Formulate the review question

Do: Translate the research problem into a framework appropriate to the discipline. PICO is common for intervention questions: population, intervention, comparator, and outcome. Other questions may require different frameworks.

Input: Population or context, intervention/exposure/phenomenon, comparator if relevant, outcomes, and acceptable study designs.

Output: A focused question and an operational definition for every concept.

Illustrative example: “Among first-year university students, do structured retrieval-practice programs, compared with usual study guidance, improve course assessment performance?”

Common error: Writing a broad topic—“retrieval practice in education”—instead of an answerable question.

Completion standard: Two reviewers can apply the definitions consistently without guessing what the author meant.

Step 2: Write and, when appropriate, register a protocol

A protocol prevents the method from silently changing after results are seen. Specify objectives, eligibility criteria, information sources, search methods, screening, extraction, risk-of-bias assessment, synthesis, subgroup analyses, and amendment handling. Registration options depend on discipline and review scope; confirm that the registry accepts your review type.

Record every amendment with its date, reason, and likely effect. A protocol is not evidence that the review was executed correctly, but it makes deviations visible.

Illustrative protocol excerpt

This excerpt is instructional. A real protocol must follow the relevant registry, review method, discipline, and journal requirements.

Step 3: Define eligibility criteria before searching

Create inclusion and exclusion rules for:

• population, setting, and phenomenon;

• interventions, exposures, and comparators;

• outcomes or concepts;

• study designs and publication types;

• date, language, and geography;

• publication status and gray literature.

Avoid criteria chosen only because they are convenient. Language and date restrictions may be justified, but report them and consider the bias they introduce.

Copyable eligibility table

Step 4: Design a reproducible multi-database search

Separate concepts into synonym blocks, combine synonyms with OR, and combine concepts with AND. Translate the strategy for each database rather than pasting one syntax everywhere. Record the database and platform, complete search string, filters, search date, and result count. PRISMA-S offers reporting guidance for literature searches in systematic reviews.

Example concept logic:

("retrieval practice" OR "practice testing" OR "test-enhanced learning")
AND
(undergraduate* OR "college student*" OR "university student*")

Search more than one relevant source when a single database cannot reasonably cover the field. Consider trial registries, dissertations, conference proceedings, citation chasing, and gray literature when omission could distort the conclusion. Involve an information specialist for high-stakes or complex reviews.

Completion standard: Another researcher can rerun each strategy and understand every filter.

Step 5: Deduplicate and screen transparently

Export records with stable identifiers and provenance. Deduplicate conservatively; do not assume similar titles are identical records. Pilot the eligibility criteria on a sample before full screening. For team reviews, define how disagreements are resolved and document reasons for full-text exclusion.

Maintain counts for identification, deduplication, title/abstract screening, full-text assessment, exclusion, and final inclusion. These counts feed the appropriate PRISMA flow diagram.

Step 6: Extract data and assess study quality

Pilot a structured extraction form. Typical fields include citation, design, sample, setting, intervention/exposure, comparator, outcomes, measurement timing, findings, limitations, and reviewer notes. Preserve page or table locations for important claims.

Select a critical-appraisal or risk-of-bias tool that matches the study design and review question. Do not let AI make final bias judgments: those decisions require domain knowledge, methodological interpretation, and documented reviewer agreement.

Match appraisal to the evidence

The exact tool depends on the review question and methodology. Record the version, reviewer training, judgment rationale, and conflict-resolution process.

Step 7: Choose the synthesis method

Meta-analysis is not a synonym for systematic review. Pool results only when the studies, outcomes, designs, and effect measures are sufficiently compatible and the statistical model is justified. Otherwise, use a structured narrative synthesis or another appropriate evidence-synthesis method.

Before synthesis, group studies by meaningful dimensions such as population, intervention, design, outcome, or follow-up. Explain contradictory findings instead of averaging them rhetorically. Distinguish absence of evidence from evidence of no effect.

Step 8: Report with current guidance

Use the current PRISMA materials relevant to the review type. Report the complete methods, flow of records, study characteristics, risk-of-bias findings, synthesis, limitations, certainty where applicable, funding, conflicts, and protocol information. Check journal and discipline requirements before submission.

Worked example and audit trail

This example is illustrative; it does not report real study results.

1. Input:

   A broad interest in retrieval practice among university students.

2. Decision:

   Use a focused systematic review because the goal is to compare an intervention with usual practice.

3. Question:

   The PICO question stated above.

4. Eligibility:

   Controlled studies, first-year university students, assessment outcomes.

5. Search log:

   Database/platform, full query, date, filters, and result count recorded separately for every source.

6. Screening output:

   One decision per record, two reviewers where required, full-text exclusion reason retained.

7. Extraction output:

   A row per study plus source-located evidence notes.

8. Synthesis decision:

   Pool only if outcome and design compatibility are defensible; otherwise use structured narrative synthesis.

9. Audit check:

   Every conclusion traces to included studies and every included study traces to the search and screening log.

Acade workflow: build a traceable evidence record

Acade is an AI academic writing and research workspace. Its most relevant role here is connecting the research question to candidate literature and then organizing verified evidence notes. A practical, bounded handoff is:

1. Enter the question, discipline, date range, source requirements, and exclusions.

2. Review search terms and candidate literature.

3. Export or record candidate bibliographic details, then verify each record and lawful access against the original source.

4. Add only reviewer-approved studies to the evidence matrix, with page or table locators.

5. Ask for comparisons only across this approved evidence set.

6. Have human reviewers confirm eligibility, appraisal, synthesis, interpretations, and citations.

Use Acade within the evidence workflow after defining the protocol. Acade does not register protocols, guarantee comprehensive retrieval, independently resolve screening disagreements, perform accountable risk-of-bias assessment, or replace a statistician, librarian, subject expert, or review team.

Systematic review checklist

• [ ] The question and method are aligned.

• [ ] A dated protocol and amendment log exist.

• [ ] Eligibility criteria are operational.

• [ ] Every search is saved exactly as run.

• [ ] Screening decisions and exclusion reasons are retained.

• [ ] Extraction has been piloted.

• [ ] Appraisal tools match the included designs.

• [ ] The synthesis method is justified before conclusions are written.

• [ ] PRISMA is used as reporting guidance, not misrepresented as the conduct method.

• [ ] Sources, citations, and AI-assisted outputs receive human verification.

FAQ

What is a systematic literature review?

It is a review that uses predefined, transparent, and reproducible methods to identify, select, appraise, and synthesize evidence for a focused question.

Is every systematic review a meta-analysis?

No. Meta-analysis is a statistical synthesis that is appropriate only when pooling is methodologically defensible.

What is the difference between a systematic review and a literature review?

The key difference is the explicit, reproducible method and audit trail. A traditional literature review may be rigorous without attempting the same level of systematic retrieval and documentation.

Does PRISMA tell me how to conduct the review?

PRISMA primarily guides reporting. Pair it with a relevant conduct methodology and discipline-specific standards.

Can one person conduct a systematic review?

Requirements vary, but independent screening, appraisal, and conflict resolution commonly benefit from multiple reviewers. Follow the intended journal, institution, and methodology.

Can AI complete a systematic review automatically?

No. AI can support discrete tasks, but human researchers remain responsible for the protocol, eligibility decisions, appraisal, synthesis, source verification, and final report.

Conclusion

A credible systematic literature review is defined by decisions that are explicit, reproducible, and open to audit. Start with the question and protocol, preserve the search and screening trail, appraise studies with design-appropriate methods, and choose synthesis based on the evidence—not on convenience. Support your next review task with Acade, then verify every methodological and scholarly decision with the responsible human team.