Systematic Review Checklist for Beginners

A systematic review is a long project with a lot of moving parts, and the danger is not that any single step is hard. It is that you reach stage six and realize you skipped something in stage two that you can no longer fix. A checklist is the antidote. Work through the stages in order, confirm each item before moving on, and the project stays repairable.

This is a working checklist, not a reporting standard. When it comes time to write up, you will follow PRISMA, which has its own checklist for what your manuscript must report. This one is about what you actually do, in the order you do it.

Stage 1: Question and feasibility

• Stated the question in one specific sentence, not as a broad topic
• Structured it with a framework (PICO for interventions, or a suitable variant)
• Checked no recent systematic review already answers it
• Ran a preliminary scoping search to confirm enough evidence exists
• Confirmed the scope is manageable for your team and timeline
• Decided the review type (systematic, scoping, rapid) and confirmed it matches the question

Do not proceed until the question is specific enough that you could decide, for any given study, whether it belongs.

Stage 2: Protocol and registration

• Written the protocol before running the formal search
• Defined explicit inclusion and exclusion criteria
• Specified the databases and other sources you will search
• Drafted the search strategy itself, not just the database names
• Stated how many reviewers will screen, and whether independently
• Named who resolves disagreements
• Chosen the risk-of-bias tool matched to your study designs
• Pre-specified the synthesis plan, including any subgroup analyses
• Registered the protocol on a suitable register

Do not proceed until the protocol is registered. Everything after this point is influenced by what you see, which is why the plan has to come first.

Stage 3: Search

• Built the search from your question's concept blocks
• Included synonyms, controlled vocabulary, and truncation where appropriate
• Had the strategy checked by an information specialist or librarian if possible
• Tested the search against a few papers you know should be found
• Searched multiple databases, not one
• Checked reference lists of key papers
• Searched grey literature and trial registries if relevant
• Recorded the exact search string, database, date, and number of hits for each source

Do not proceed until your search is documented well enough that someone else could rerun it and get your numbers.

Stage 4: Screening

• Imported all records into one place
• Deduplicated and recorded the counts before and after
• Piloted the eligibility criteria on a sample with both reviewers
• Refined any criteria that produced disagreement in the pilot
• Screened titles and abstracts with two independent reviewers
• Retrieved full texts for everything that passed
• Screened full texts with two independent reviewers
• Recorded a specific reason for every full-text exclusion
• Resolved all disagreements through discussion or a third reviewer

Do not proceed until every record is accounted for, from the initial search down to the final included set.

Stage 5: Data extraction

• Designed the extraction form during the protocol stage
• Piloted the form on a few studies and revised it
• Extracted study characteristics, participant details, methods, and outcome data
• Extracted in duplicate where feasible, and reconciled differences
• Documented any data conversions (for example, median to mean)
• Contacted authors for missing data where it mattered
• Recorded what data were unavailable

Do not proceed until you know exactly which outcome data you have and which you are missing.

Stage 6: Risk of bias

• Applied the appraisal tool you pre-specified, matched to each study design
• Assessed every included study
• Had assessments done or checked by two people
• Recorded the judgment and the supporting reason for each domain
• Considered how the appraisal affects your confidence in the findings

Do not proceed until you can say, for each included study, how much weight it deserves.

Stage 7: Synthesis

• Judged honestly whether the studies are similar enough to pool
• If not pooling, planned a structured narrative synthesis
• If pooling, chosen the effect measure and model deliberately
• Quantified heterogeneity and interpreted it rather than ignoring it
• Run only the subgroup analyses you pre-specified, labelling any others as exploratory
• Assessed publication bias if you have enough studies to do so meaningfully
• Rated the certainty of the evidence, for example with GRADE

Do not proceed until your synthesis approach is justified by the data, not by the desire for a single number.

Stage 8: Writing and reporting

• Followed the PRISMA checklist for reporting
• Produced a PRISMA flow diagram accounting for every record
• Reported the full search strategy for at least one database
• Reported reasons for full-text exclusions
• Reported the risk-of-bias assessments
• Disclosed and justified every deviation from the registered protocol
• Stated the limitations honestly, including anything you streamlined
• Cited your registration number
• Chosen a journal that publishes systematic reviews in your field

The checks that matter most

If you only internalize a few items, make them these.

Register the protocol before searching. Almost every serious criticism of a review traces back to decisions made after the results were visible. This single check prevents most of them.

Screen with two independent reviewers. The redundancy is the error-correction mechanism. Removing it removes the reason anyone should trust your included set.

Search more than one database. A single-database search reliably misses studies and reliably gets noticed.

Only pool studies that belong together. A meta-analysis of incomparable studies is worse than no meta-analysis, because it looks authoritative while meaning nothing.

Disclose everything you changed or skipped. Deviations and limitations are acceptable. Hiding them is not.

Frequently asked questions

Is this the same as the PRISMA checklist? No. PRISMA is a reporting checklist covering what your manuscript must state. This is a working checklist covering what you do, in order. You need both, at different stages.

Can I skip steps if my review is small? The steps scale, they do not disappear. A small review still needs a protocol, a documented search, and independent screening. If you must streamline, run an explicit rapid review and disclose what you compressed.

What is the most commonly skipped step? Writing and registering the protocol before searching. It feels like paperwork and it is the step that protects the whole review.

Do I have to pilot my criteria? It is strongly advisable. Ambiguities you cannot see in the armchair show up immediately when two people apply the criteria to real records, and fixing them early is far cheaper.

When do I decide whether to run a meta-analysis? You plan the approach in the protocol and confirm it once you see whether the included studies are similar enough to pool. Planning to pool does not commit you to pooling if the data say otherwise.

The bottom line

Work the stages in order and confirm each block before moving on: question, protocol, search, screening, extraction, appraisal, synthesis, reporting. The gates that matter most are registering before you search, screening in duplicate, searching widely, pooling only when justified, and disclosing everything you changed.

The value of a checklist is not that the steps are difficult. It is that skipping one quietly is easy, and discovering it at the end is expensive.

Working through the checklist? Verflux covers the search import, deduplication, dual screening, appraisal, and PRISMA diagram in one place.