Eligibility criteria turn a review question into decisions that reviewers can apply consistently. This guide shows how to draft and pilot those rules, handle unclear information and record the reason for a decision. The examples cover several fields, with particular attention to the difference between inclusion in a review and contribution to a specific synthesis.
What Are Inclusion and Exclusion Criteria?
Eligibility criteria define which studies address the review question. State the population, intervention or exposure, comparator and eligible designs as appropriate, and justify any additional restrictions in the protocol. An unclear abstract calls for clarification or retrieval; it is not evidence that a study fails a criterion.
Keep eligibility for the review separate from contribution to a particular synthesis. A study can meet the review criteria yet lack a usable result, uncertainty estimate or time point for one analysis. Record that limitation and how the study’s relevant evidence will be handled. Do not automatically exclude a study simply because the result you hoped to extract is absent from its report. The Cochrane eligibility chapter explains this distinction.
Predefine the rules before screening so decisions are tied to the question rather than the results encountered. Pilot them on a small, varied set of records, including ambiguous cases, and resolve differing interpretations. If a rule changes, retain its earlier version, the reason and date of the change, and identify records that need reassessment.
Developing PICOS Eligibility Criteria (Step-by-Step)
For an intervention question, PICOS is a useful way to organize the population, intervention, comparator, outcomes and study design. These elements do not all serve the same purpose: outcomes guide extraction and synthesis, but are seldom used as eligibility restrictions in intervention reviews. Translate each relevant element into a rule and an action for missing information.
- Population: Define the relevant characteristics and setting. For example, “adults aged 18 years or older with type 2 diabetes treated in primary care” includes older adults unless the protocol supplies a reason for an upper age limit. State how mixed populations will be handled rather than silently dropping them.
- Intervention or exposure and comparator: Specify the intervention, dose or delivery features that matter to the question. For an exposure review, define the exposure of interest. State eligible comparators when a comparison is required. An uncontrolled study is ineligible only when its design falls outside the documented review criteria.
- Outcomes: List the outcomes and time points to extract and the rules for grouping them in syntheses. Missing reporting does not establish that an outcome was never measured. If measurement of a particular outcome is genuinely part of eligibility, justify that choice and explain how unclear measurement will be resolved; do not replace it with a requirement to report an extractable result.
- Study design: Describe eligible design features, such as random assignment or a concurrent comparator, rather than relying only on authors’ labels. For example, a review restricted to randomized trials should state whether cluster and crossover trials are eligible and how they will be handled.
- Additional restrictions: Explain any date, setting, language or publication restriction in relation to the review question and practical limits. Small sample size is not a general reason for ineligibility; precision and risk of bias need their own assessment.
Put the rules in the protocol and screening form, including examples, the criterion version and the action for unclear information. Record reasons for exclusions after eligibility assessment. PRISMA 2020 items 5, 16a and 16b address reporting criteria, selection results and apparently eligible studies excluded with an explanation.
Other Frameworks (Beyond PICOS)
Other frameworks organize different kinds of questions. The JBI methods manual discusses question-specific frameworks; Cooke and colleagues’ SPIDER paper introduces one for qualitative evidence synthesis. The examples below show how the elements differ.
- PEO (Population, Exposure, Outcome): Organize an exposure question, for example urban adults exposed to air pollution, with asthma incidence as the outcome of interest. Decide separately which elements define eligibility and which guide synthesis.
- SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type): For a qualitative question about inclusive education, the sample might be high-school students; the phenomenon, experiences of inclusive education; the design, interviews or focus groups; the evaluation, perceived inclusion and barriers; and the research type, qualitative or mixed-methods research.
- CoCoPop (Condition, Context, Population): For a prevalence question, specify the condition, setting or context and population. Choose a question framework that fits the intended evidence, then turn it into explicit rules rather than treating the acronym as a complete method.
Examples of Inclusion/Exclusion Criteria in Different Fields
These fictional examples show question-specific boundaries. They are not default rules for an entire discipline. The synthesis column deliberately separates the planned result from review eligibility.
| Field and question | Review eligibility boundary | Separate synthesis decision |
|---|---|---|
| Clinical intervention | Adults with condition X; randomized trials comparing drug A with placebo or usual care. Animal studies and non-randomized designs fall outside this example’s scope. | Plan a four-week symptom analysis. If a review-eligible trial does not supply that result, record the missing result and assess what other evidence it contributes. |
| Community public health | Community interventions in the specified region; randomized or specified controlled quasi-experimental designs. Exclude a laboratory-only study if it does not match the population and setting. | Define how six-month incidence results will be grouped. Different follow-up or missing uncertainty affects that synthesis, not automatically review inclusion. |
| School education | K–12 school-based evaluations of the specified teaching method versus usual teaching, using the designs listed in the protocol. Adult education is outside this example’s population. | Plan achievement and attendance syntheses separately. Assess a small pilot on the same eligibility rules; do not impose an unexplained minimum sample size. |
| Social sciences | Primary empirical studies of the specified exposure and population, with eligible longitudinal, cross-sectional or experimental designs defined in advance. Editorials are outside this example’s scope. | State the measure, context and time point needed for each synthesis. Keep unclear reporting separate from a confirmed failure to meet the review criteria. |
Common Pitfalls and How to Avoid Them
- Vague scope: Replace terms such as “young people” or “usual treatment” with definitions relevant to the question. Pilot those definitions on difficult records.
- Undocumented changes: Amend the protocol transparently and revisit affected records instead of silently moving the boundary after seeing results.
- Unjustified exclusions: Do not exclude a study merely because it is small, unpublished, in another language or has an unfavorable result. Explain any necessary scope restriction and its consequences.
- Confusing missing results with ineligibility: Keep the study-level decision and the analysis-specific data limitation as separate fields.
- Treating appraisal as a screening shortcut: Apply the planned design criteria, then assess risk of bias with an appropriate tool. A reporting omission or low numerical quality score is not a substitute for that assessment.
- Inconsistent application: Calibrate reviewers and document who screened independently and how disagreements were resolved. Cochrane requires at least two people independently to make final eligibility decisions; describe your actual process at each screening stage.
- Incomplete records: Retain the report identifier, criterion version, reviewer, decision and reason. Separate reports not retrieved from reports excluded after assessment.
| Situation | Screening action | Record |
|---|---|---|
| Age is unclear in the abstract | Retrieve or clarify according to the protocol | Missing information and next action |
| A review-eligible study lacks the planned 12-week result | Keep review eligibility and synthesis contribution separate | Missing result, attempted clarification and synthesis decision |
| A report is an editorial and the protocol excludes editorials | Exclude with the specified reason | Publication type, criterion version and reviewer |
Language, Publication-Type, and Quality Filters
- Language: Plan relevant language coverage and translation resources. If a restriction is necessary, state and justify it in the protocol and report the limitation. Language alone does not indicate methodological quality.
- Publication status: Consider unpublished and grey sources appropriate to the question. A conference abstract may be one report of an eligible study, so look for linked reports or additional information before treating a document’s format as a study-level decision.
- Publication type: Distinguish primary research from editorials, narrative discussions and secondary reviews. A study-design restriction should follow the question; it should not label qualitative research as inherently poor-quality evidence.
- Risk of bias: Keep appraisal judgments separate from eligibility. Prespecify how they will inform the synthesis and interpretation. Avoid arbitrary sample-size thresholds or unexplained evidence hierarchies as a shortcut to validity.
An eligible study with unusable outcome data remains part of the review record. Explain whether data could be obtained, what remains missing and why a result does or does not enter a particular synthesis. The Cochrane selection guidance explicitly distinguishes study inclusion from the availability of usable data. Do not translate a risk-of-bias label into an arbitrary analysis weight.
Handling Non-English Studies and Grey Literature
Plan how the team will identify, obtain and interpret relevant reports across languages and publication formats:
- Non-English reports: Record the language and translation approach. Use an appropriate language reviewer or verified translation for eligibility decisions when needed; an uncertain automated translation calls for clarification.
- Grey literature: Select relevant registers, dissertations, conference proceedings, government reports or organizational sources. Apply the same question-based criteria and an appropriate appraisal method, without presuming that publication status determines quality.
- Unobtainable reports: Record retrieval attempts and keep these separate from exclusions based on assessed eligibility. An inaccessible report does not prove that its study fails a criterion.
- Unresolved information: Retain an explicit pending or awaiting-classification record when eligibility cannot be determined, following the review’s methods. Document the uncertainty and its consequences rather than inventing a negative decision.
Aligning with PRISMA, Cochrane, and Best Practice
Use reporting and methods guidance for different purposes:
- PRISMA 2020: Report the actual eligibility criteria, grouping for syntheses, selection process and results. A diagram and checklist need to agree with the underlying decisions.
- Cochrane Handbook: Use the relevant methods chapters when planning study selection and synthesis. Outcome reporting should not quietly become a condition of review eligibility.
- Review protocol: Keep the rules and amendment history accessible to screeners. Explain changes and record which earlier decisions were reconsidered.
Citing a guideline does not establish that the decisions were appropriate. Readers need the operational rules, their rationale and the record of how they were applied.
Conclusion
The useful output is a set of rules another reviewer can apply to the same difficult records. Pilot them, resolve ambiguity and preserve both the initial decision and any correction. Keep review eligibility, missing information, risk-of-bias assessment and synthesis contribution distinct so that a limitation in one step does not silently become an exclusion in another.
Use these criteria to pilot screening and extraction tools, preserving the proposed decision, human correction and reason for each difficult case.