WHY ENGLISH?
A systematic review becomes trustworthy when every include, exclude and maybe can be traced to a written criterion and a specific record.
This guide helps students and research teams turn Covidence screening actions into reproducible decisions, consistent exclusion reasons and an honest PRISMA account.
Covidence supports parts of a systematic-review workflow, including title-and-abstract screening, full-text review, conflict resolution and reporting. English matters because every vote is an interpretation. A reviewer must connect a short record to written eligibility criteria without inventing information that the title or abstract does not provide.
People searching for Covidence screening, systematic review software, exclusion reasons or PRISMA flow usually need more than instructions for clicking include or exclude. They need decision language that different reviewers can apply consistently. Precise definitions reduce avoidable conflicts. Hierarchical exclusion reasons prevent several labels from competing for the same study. Clear notes make later auditing possible.
This is practical English for research projects and evidence reviews. It does not claim that software guarantees completeness, removes bias or replaces a librarian, subject expert, statistician or review protocol. It respects multilingual evidence by asking teams to plan how non-English records will be screened. Current Covidence support pages checked on 4 November 2026 describe title-and-abstract and full-text screening, conflict resolution, custom exclusion reasons and an updating PRISMA flow diagram.
Begin With a Protocol Question, Not a Button
Screening is the application of a prior rule to a candidate record. The software records a choice; it does not supply the rule.
Translate the review question into population, concept or intervention, comparator where relevant, outcome, design, setting, date, language and publication-status criteria. Operational definitions should tell a reviewer what counts and what evidence is enough at each stage.
Screening scenario: Begin With a Protocol Question, Not a Button
‘School students’ is replaced with ‘learners enrolled in primary or secondary education; exclude university-only samples unless school results are reported separately.’ Two reviewers can now apply the boundary.
Criteria should be justified by the review purpose, not rewritten after attractive results appear.
Team move: Begin With a Protocol Question, Not a Button
Create a one-page eligibility guide with definitions, examples, counterexamples and escalation rules before screening begins.
Calibrate Reviewers Before the Main Screen
Shared criteria can still produce different interpretations until reviewers test them together.
Run a pilot set that includes obvious includes, obvious excludes and ambiguous edge cases. Compare decisions, discuss wording and revise definitions. Keep the pilot record so later changes are visible.
Calibration case: Calibrate Reviewers Before the Main Screen
One reviewer treats mixed-age samples as eligible; another excludes them. The team adds a rule requiring separable school-age data and reruns the pilot.
Agreement is not the only goal. Reviewers can agree consistently on a badly written rule, so a content expert should inspect the decisions.
Team move: Calibrate Reviewers Before the Main Screen
Set a pilot size appropriate to the project, document disagreements and approve the final decision guide before unlocking the full workload.
Use Include, Exclude and Maybe for Different States
A title-and-abstract vote should represent evidence available at that stage, not the reviewer’s private guess about the full paper.
Use include when the record appears eligible, exclude when it clearly fails a criterion and maybe when current information cannot decide. Covidence’s current help explains how records move through screening and how a maybe vote functions within review settings.
Vote example: Use Include, Exclude and Maybe for Different States
An abstract names the right population and outcome but omits study design. The reviewer chooses maybe or include for full-text assessment rather than inventing an ineligible design.
An overuse of maybe increases workload; an overuse of exclude risks missed evidence. Calibration should address both.
Team move: Use Include, Exclude and Maybe for Different States
Add stage-specific examples to the guide and require a note for recurring ambiguity.
Resolve Conflicts by Returning to the Criterion
A conflict is evidence that two interpretations met. It is not proof that one reviewer was careless.
During conflict resolution, compare the record, the exact criterion and each reviewer’s rationale. Decide the record, then decide whether the guide needs clarification. Distinguish factual uncertainty from policy uncertainty.
Conflict case: Resolve Conflicts by Returning to the Criterion
Reviewers disagree over a programme delivered partly online. The protocol criterion concerns educational setting, not delivery mode. The resolver includes it and adds that delivery format alone does not exclude.
A senior person’s preference is not a reproducibility method. Unexplained overrides create hidden rules.
Team move: Resolve Conflicts by Returning to the Criterion
Record the decisive criterion, resolver and any wording change. Recheck earlier records when a rule changes materially.
Design Full-Text Exclusion Reasons as a Hierarchy
Full-text exclusion reasons should be mutually understandable and applied in a consistent order.
The Covidence exclusion-reasons guide says teams can customise and reorder reasons and recommends ordering them by importance. A hierarchy avoids assigning whichever reason a reviewer notices first.
Exclusion case: Design Full-Text Exclusion Reasons as a Hierarchy
A study has the wrong population and wrong outcome. If population is first in the approved hierarchy, the team records wrong population. Notes can retain the secondary issue without double-counting the PRISMA reason.
Reasons that are too broad hide decisions; reasons that overlap produce unstable reporting.
Team move: Design Full-Text Exclusion Reasons as a Hierarchy
Define each reason, exclusion test, priority and example. Pilot the list on full texts before screening at scale.
Write Reasons That a Reader Can Reproduce
‘Not relevant’ is an opinion, not an auditable exclusion reason.
Use noun phrases tied to criteria: ineligible population, wrong study design, no eligible outcome, duplicate report, not primary research or full text unavailable after documented attempts. Keep wording stable across reviewers.
Reason-writing example: Write Reasons That a Reader Can Reproduce
Instead of ‘wrong paper’, record ‘commentary with no original empirical data’. The reason shows what was inspected and which boundary failed.
Do not use absence of a reported result as proof the outcome was not measured without checking supplementary material or protocols when the review requires that depth.
Team move: Write Reasons That a Reader Can Reproduce
Test each label by asking whether an outside reviewer could apply it to a new paper.
Distinguish Records, Reports and Studies
One study can produce several reports, and one imported record can duplicate another database record.
Maintain relationships among conference abstracts, protocols, registry entries, preprints and journal articles. Deduplication removes duplicate records; linking reports prevents the same participants from being counted as separate studies.
Study-family case: Distinguish Records, Reports and Studies
A trial protocol, primary outcome paper and follow-up paper belong to one study family. The review may use all three for different information while counting one underlying study.
Titles and author lists can change. Automatic duplicate detection needs human verification.
Team move: Distinguish Records, Reports and Studies
Assign a study-family identifier and record which report supplies each outcome or method detail.
Retrieve Full Text Through a Documented Route
Full-text screening cannot be reproducible when access attempts disappear from the record.
Record library searches, DOI links, repositories, author contact and dates. Distinguish not found, inaccessible under current access, non-digitised and unavailable after author contact.
Retrieval case: Retrieve Full Text Through a Documented Route
A paper is absent from one database but found in an institutional repository. The team screens the lawful copy and records the route.
Do not treat paywall status as an eligibility criterion unless the protocol explicitly and defensibly does so.
Team move: Retrieve Full Text Through a Documented Route
Create a retrieval log with record ID, action, date, result and next action.
Let PRISMA Flow Describe the Review, Not Decorate It
A PRISMA flow diagram tells how records moved from identification to inclusion and exclusion.
The current Covidence PRISMA guide says the diagram updates as a review progresses and includes imported references, duplicates, screening stages and full-text exclusions. It also explains that downloaded output may need manual updating for items such as studies not retrieved.
PRISMA check: Let PRISMA Flow Describe the Review, Not Decorate It
A team downloads the DOCX, checks database-source counts, confirms duplicate totals and manually updates not-retrieved studies before submission.
An automatically populated diagram can still be wrong when imports, merges or exclusion decisions are wrong.
Team move: Let PRISMA Flow Describe the Review, Not Decorate It
Reconcile every box with exports and the decision log before publication.
Keep Numbers and Nouns Aligned
PRISMA language distinguishes records, reports and studies because their counts can differ.
Use the noun matching the workflow stage. Imported database entries are records; retrieved documents are reports; linked reports can represent one study. The current Covidence guidance explains that merged references can change study counts while reference counts remain.
Counting example: Keep Numbers and Nouns Aligned
Three reports describe one trial. The diagram may show three reports sought and one included study. Calling all three ‘studies’ would inflate the evidence base.
Interfaces and templates can use shorthand. The review team remains responsible for accurate terminology.
Team move: Keep Numbers and Nouns Aligned
Create a glossary beside the flow counts and use it consistently in the abstract, methods and results.
Document Deduplication Without Hiding Decisions
Deduplication is a methodological action, not housekeeping outside the review.
Record sources imported, software used, automatic and manual duplicate handling and verification rules. Inspect close matches before removal, especially titles translated across languages or conference and journal versions.
Duplicate case: Document Deduplication Without Hiding Decisions
Two records share title and authors but have different years. Reading shows one is an erratum, not a duplicate. It remains linked to the study record.
Overaggressive removal can lose unique reports; weak removal can double-screen and double-count.
Team move: Document Deduplication Without Hiding Decisions
Audit a sample of removed duplicates and preserve an export of the pre- and post-deduplication sets.
Plan Non-English Screening Before It Appears
Language decisions belong in the protocol, not in an improvised response to an unfamiliar title.
Specify eligible languages, translation routes, reviewer competence and how uncertainty is escalated. Preserve original titles and record translated working titles separately.
Language case: Plan Non-English Screening Before It Appears
A Malay abstract appears eligible but needs a full-text reviewer with appropriate language competence. The team routes it without treating English visibility as a quality filter.
Machine translation may support triage but may not safely resolve technical eligibility on its own.
Team move: Plan Non-English Screening Before It Appears
Create a language-support roster and a decision rule before importing records.
Use Notes Without Creating a Shadow Protocol
Notes should explain application of criteria, not introduce private criteria visible to only one reviewer.
Use concise evidence notes: population mixed and not separable; protocol only; outcome measured but not reported at eligible time point. When a note recurs, update the shared guide.
Note example: Use Notes Without Creating a Shadow Protocol
Several reviewers write ‘school setting unclear’. The team adds a standard check for recruitment site and participant age.
Free-text notes can contain copyrighted text or sensitive information. Quote only what is needed and follow project governance.
Team move: Use Notes Without Creating a Shadow Protocol
Review note patterns weekly and convert recurring ambiguity into shared guidance.
Manage Amendments and Date Them
Eligibility criteria sometimes need clarification or amendment after screening begins.
Record what changed, why, when, who approved it and which decisions require reassessment. Distinguish clarification of intended meaning from a substantive expansion or restriction.
Amendment example: Manage Amendments and Date Them
A criterion originally says ‘secondary students’ but the protocol question includes upper-primary transition. The team documents a substantive amendment and rescreens excluded primary records.
Undated changes make outcome-driven selection impossible to assess.
Team move: Manage Amendments and Date Them
Maintain an amendment table and link every change to affected record sets.
A Worked Screening Chain From Import to Inclusion
Imagine a review of peer-feedback interventions in secondary English writing.
The team imports database records, deduplicates, pilots criteria, screens titles and abstracts independently, resolves conflicts, retrieves full texts and applies ordered exclusion reasons. Reports from the same study are linked before extraction.
End-to-end case: A Worked Screening Chain From Import to Inclusion
One paper appears eligible by title but is a university study; it receives the full-text reason ineligible population. Another has secondary pupils but reports attitudes only; the protocol accepts engagement outcomes, so it remains included.
The final set is not ‘all research everywhere’. It is the result of named searches, dates, languages and criteria.
Team move: A Worked Screening Chain From Import to Inclusion
Write one sentence for every flow transition and confirm that the PRISMA count follows it.
Parents and Students Can Practise Screening on Small Projects
School projects can borrow the discipline of systematic screening without pretending to be full systematic reviews.
Write a narrow question, two include criteria, two exclude criteria and a small source log. A parent can ask the student to justify one include and one exclude from the text.
School-project case: Parents and Students Can Practise Screening on Small Projects
A student excludes a blog from an empirical-evidence table but keeps it as contextual reading. The source has not become worthless; its role has been named accurately.
Do not label a classroom search ‘systematic’ unless its method meets that standard.
Team move: Parents and Students Can Practise Screening on Small Projects
Use the honest phrase ‘structured source selection’ and describe what was actually done.
Transfer Screening Language to Work Decisions
Eligibility criteria resemble many workplace gates: vendor shortlisting, job applications, incident triage and document review.
Define mandatory criteria, desirable criteria, evidence requirements and conflict resolution before reviewing candidates. Record the reason for each decision.
Workplace case: Transfer Screening Language to Work Decisions
A team screens grant proposals. ‘Not a fit’ becomes ‘outside the published funding scope: no direct education component.’ Applicants and auditors can understand the boundary.
Systematic-review software should not be repurposed for personal data without governance and authority.
Team move: Transfer Screening Language to Work Decisions
Carry the rule ‘criterion, evidence, decision, reason’ into any high-volume review process.
Finish With a Reproducibility Audit
Before exporting results, verify that another trained reviewer could follow the recorded route.
Check protocol version, search sources and dates, imports, duplicates, calibration, votes, conflicts, full-text retrieval, exclusion reasons, study families, amendments and PRISMA counts.
Audit case: Finish With a Reproducibility Audit
The audit finds three exclusions labelled ‘other’. The team reopens them, applies specific reasons and updates the flow before writing the results section.
Perfect agreement is not the goal; transparent, defensible handling of disagreement is.
Team move: Finish With a Reproducibility Audit
Freeze dated exports and the decision guide used for the final analysis.
Define Borderline Populations Before Screening
Mixed and transitional populations often create avoidable conflicts.
Specify age, education stage, diagnostic status or setting and whether separable subgroup data are required. Give examples for transition years and mixed samples.
Population case: Define Borderline Populations Before Screening
A study includes 60 secondary and 20 university students. The rule includes it only if secondary results are reported separately.
An arbitrary threshold needs justification.
Team move: Define Borderline Populations Before Screening
Pilot the rule on three mixed-sample papers and record the outcome.
Distinguish Study Design Labels From Actual Methods
Titles and abstracts may use design terms loosely.
Check allocation, comparison, timing and analysis before deciding that a record is randomised, quasi-experimental, cohort or cross-sectional.
Design case: Distinguish Study Design Labels From Actual Methods
A paper calls itself an experiment but assigns intact classes without randomisation. It fits the protocol’s quasi-experimental category, not its randomised category.
Do not exclude from a title label alone when full text can resolve the design.
Team move: Distinguish Study Design Labels From Actual Methods
Add observable design tests to the eligibility guide.
Handle Conference Abstracts and Protocols Consistently
Conference abstracts and protocols can signal studies without reporting complete results.
State whether they are eligible reports, sources for ongoing-study tracking or routes to later publications. Link them to study families.
Report-type case: Handle Conference Abstracts and Protocols Consistently
A conference abstract is excluded from effect synthesis but retained to search for the completed trial.
Discarding it without tracking can hide publication bias; including it as full evidence can overstate detail.
Team move: Handle Conference Abstracts and Protocols Consistently
Create separate statuses for awaiting full report, ongoing and eligible report.
Screen Outcomes at the Correct Stage
Abstracts often omit eligible outcomes even when full texts report them.
The protocol should state whether outcome absence can justify early exclusion. Many reviews carry plausible records to full text to avoid false negatives.
Outcome case: Screen Outcomes at the Correct Stage
An abstract focuses on attendance but the full paper contains the eligible writing score. Early exclusion would have lost the study.
Overly broad inclusion at the first stage increases workload but may protect recall.
Team move: Screen Outcomes at the Correct Stage
Calibrate on known full texts and choose a transparent trade-off.
Record Reasons for Unavailable Full Text Carefully
‘No full text’ should mean documented retrieval attempts failed, not that one link was broken.
Use library holdings, DOI resolution, repositories, interlibrary services and author contact as appropriate. Record dates and results.
Access case: Record Reasons for Unavailable Full Text Carefully
A broken publisher URL is resolved through a DOI and repository copy, preventing a false exclusion.
Access limits can bias a review toward well-resourced publishers and English-visible literature.
Team move: Record Reasons for Unavailable Full Text Carefully
Define a retrieval minimum and report unavailable studies separately.
Audit Reviewer Workload and Blinding Rules
Fatigue and uneven allocation can change screening consistency.
Plan batches, breaks, progress checks and any blinding of votes. Covidence can hide another reviewer’s vote until both decide under common settings.
Workload case: Audit Reviewer Workload and Blinding Rules
Conflict rates rise late in long sessions. The team shortens batches and recalibrates rather than blaming reviewers.
Conflict rate alone does not diagnose the cause.
Team move: Audit Reviewer Workload and Blinding Rules
Monitor patterns by stage and criterion while protecting reviewer privacy.
Export Decisions Before Major Changes
A reproducible review needs recoverable checkpoints.
Before changing criteria, merging reports or resolving a large conflict set, export records and decisions where permitted. Name the protocol version and date.
Checkpoint case: Export Decisions Before Major Changes
After a rule amendment, the team can identify exactly which earlier exclusions need rescreening.
Exports can contain personal information and copyrighted abstracts; store them securely.
Team move: Export Decisions Before Major Changes
Create dated, access-controlled checkpoints and a short change note.
Reconcile Included Counts With Extraction
The number entering extraction should follow from screening and study-family merging.
Compare included reports, underlying studies and extraction forms. Resolve orphaned forms and linked publications before synthesis.
Reconciliation case: Reconcile Included Counts With Extraction
Twenty included reports become fourteen studies because several trial families have follow-ups. The methods and PRISMA text use both numbers correctly.
A mismatch can signal duplicate extraction or a lost report.
Team move: Reconcile Included Counts With Extraction
Run an ID-based reconciliation before analysis begins.
Use Research Integrity Alerts as Prompts
Current Covidence guidance lists research-integrity alerts among screening support topics.
An alert should lead to official correction, expression-of-concern or retraction records. Determine scope and protocol handling with content experts.
Integrity case: Use Research Integrity Alerts as Prompts
A retracted report is linked to a corrected follow-up. The team records both and applies the protocol’s publication-status rule.
An alert is not itself the complete editorial record.
Team move: Use Research Integrity Alerts as Prompts
Add notice URL, date, affected report and decision to the study-family note.
Report Screening Methods in Complete Sentences
Readers need to know who screened, how many reviewers, what stages occurred and how conflicts were resolved.
Name software version or access date when relevant, calibration, independence, criteria, retrieval and exclusion reporting. Avoid the empty sentence ‘Studies were screened in Covidence.’
Methods-writing case: Report Screening Methods in Complete Sentences
The methods state that two reviewers independently screened each stage and a third resolved unresolved conflicts using the prespecified guide.
Software branding cannot substitute for method description.
Team move: Report Screening Methods in Complete Sentences
Use a reporting checklist and match every sentence to a process record.
Define a Conflict Escalation Path
Some conflicts are routine; others expose uncertainty requiring a subject expert, methodologist or information specialist.
Specify who resolves ordinary eligibility disagreements and when questions escalate. Record the advice and its effect on criteria.
Escalation case: Define a Conflict Escalation Path
A biomedical terminology conflict cannot be resolved from the abstract. The team asks a clinician, documents the definition and applies it consistently.
Escalation should not become an informal route for changing criteria.
Team move: Define a Conflict Escalation Path
Add issue type, responsible expert and response time to the screening plan.
Separate Eligibility From Quality Appraisal
A study can meet inclusion criteria and still have serious risk of bias.
Screening asks whether the study belongs; appraisal asks how much confidence to place in it. Excluding weak-looking studies at screening can hide evidence and introduce judgement not specified in the protocol.
Appraisal-boundary case: Separate Eligibility From Quality Appraisal
A small uncontrolled study meets the design criterion and is included, then rated appropriately during appraisal.
Some protocols do set minimum design or report criteria; those must be explicit before screening.
Team move: Separate Eligibility From Quality Appraisal
Keep include/exclude fields separate from risk-of-bias fields and reasons.
Review the Maybe Queue Before It Becomes a Dumping Ground
Maybe is a temporary state for genuine uncertainty, not a way to avoid decisions.
Review maybe records in planned batches, identify recurring ambiguity and update guidance. Resolve with the same independence and conflict rules as other records.
Maybe case: Review the Maybe Queue Before It Becomes a Dumping Ground
Many abstracts omit participant age. The team creates a full-text rule and moves plausible records forward instead of leaving them indefinitely.
Forcing a premature decision can lose studies; leaving everything unresolved blocks progress.
Team move: Review the Maybe Queue Before It Becomes a Dumping Ground
Set a review date and owner for every maybe batch.
Preserve Reasons When Exclusion Categories Change
Changing the exclusion list midway can orphan or misclassify earlier decisions.
Map old reasons to new reasons, identify records requiring manual review and preserve the original label in the audit trail.
Migration case: Preserve Reasons When Exclusion Categories Change
‘Wrong context’ is split into school level and country-policy context. Earlier records are rechecked rather than automatically recoded.
Mechanical mapping can create false precision.
Team move: Preserve Reasons When Exclusion Categories Change
Create a migration table and verify a sample plus every ambiguous record.
Write the Results Narrative From the Flow
The methods explain how screening worked; the results explain what moved through it.
Report identified records, duplicates, screened records, full texts sought and assessed, exclusions with reasons and included studies using the same nouns as the diagram.
Reporting case: Write the Results Narrative From the Flow
The text says 1,240 records were screened, 86 reports sought and 21 studies included, matching the reconciled flow.
Do not round counts or switch records and studies casually.
Team move: Write the Results Narrative From the Flow
Generate the narrative from the final reconciled table and cross-check every number.
Did You Know? A Flow Count Can Change After Report Merging
Covidence’s current PRISMA guidance explains that merging references into studies can change study counts while reference counts remain visible.
This is why screening teams must know whether a number names records, reports or studies. Merging is an interpretation about research identity, not cosmetic tidying.
Identity case: Did You Know? A Flow Count Can Change After Report Merging
Two articles report different outcomes from one trial. They remain two reports but become one study family for the included-study count.
Incorrect merging can hide independent studies or inflate one study into several.
Team move: Did You Know? A Flow Count Can Change After Report Merging
Require title, authors, sample, registration and setting evidence before merging, then reconcile the flow and extraction list.
Frequently Asked Questions
Does Covidence make a review systematic?
No. Systematic quality comes from a justified protocol, comprehensive search, reproducible screening, appraisal and synthesis. The software supports workflow.
Should exclusion reasons be recorded at title-and-abstract stage?
Practices vary by protocol and platform configuration. Full-text exclusions require specific reasons for transparent reporting; early-stage notes should follow the approved method.
What should happen when reviewers disagree?
Return to the exact criterion, compare evidence, resolve the record and update shared guidance if the wording caused the conflict.
Can a study have two exclusion reasons?
It may fail several criteria, but teams normally report one primary full-text reason using an agreed hierarchy while preserving useful notes.
Are duplicate records the same as multiple reports?
No. Duplicate records describe the same report; multiple reports can describe one underlying study and may each contain unique information.
Does the Covidence PRISMA diagram need checking?
Yes. Reconcile imports, duplicates, merges, exclusions and not-retrieved items. The downloaded diagram may require manual updates.
Can machine translation decide eligibility?
It can assist triage, but technical decisions may require a competent reader or qualified support. Record the method and uncertainty.
What is a good exclusion reason?
A stable, criterion-linked phrase such as ineligible population or wrong study design, defined well enough for another reviewer to apply.
Can a school project use this method?
Yes, on a smaller scale, while naming it honestly as structured source selection rather than claiming a full systematic review.
What is the final quality question?
Can every included and excluded full text be traced to the protocol version, evidence inspected and reason recorded?
Useful Next Reading
- Covidence screening guidance — verify the current screening workflow.
- PRISMA 2020 flow diagrams — choose the appropriate official reporting template.
- How systematic reviews work — connect screening to protocol, search, appraisal and synthesis.
- Read Elicit screening tables and supporting quotes — compare AI-assisted discovery without competing with Covidence’s workflow intent.
Closing thought: Covidence becomes educationally valuable when clicks remain attached to criteria, evidence and reasons. English makes the review’s invisible judgement visible enough to question, correct and reproduce.