03 · What You Need to Know
How to Determine Whether an Exclusion Criterion Is Necessary
What Are Exclusion Criteria?
Exclusion criteria are conditions that disqualify otherwise potentially eligible individuals or other units from participating in a study or receiving particular study procedures.
They may concern safety, contraindications, incompatible exposures, measurement limitations, or other design-specific requirements.
For example, a clinical trial may exclude patients with a documented contraindication to the investigational intervention.
An educational intervention may exclude individuals who are not eligible to participate in the instructional activity being evaluated.
However, a criterion should not be considered justified merely because it can be expressed as an exclusion.
The underlying scientific or ethical reason must be established.
How Do Exclusion Criteria Differ From Inclusion Criteria?
Inclusion criteria establish the characteristics necessary for eligibility.
Exclusion criteria identify additional circumstances that prevent participation among individuals who would otherwise satisfy the inclusion requirements.
For example, a clinical study might include patients with a specified diagnosis but exclude those with a contraindication to the study treatment.
However, the distinction can sometimes be expressed differently without changing the substantive eligibility rule.
The important issue is whether the complete eligibility framework is coherent and consistently applied.
Researchers should first identify the factors determining inclusion criteria before deciding which additional exclusions are necessary.
Which Factors Can Justify Exclusion Criteria?
| Factor |
Potential justification |
Important caution |
| Participant safety |
A defined intervention poses unacceptable risk under specified conditions. |
Assess whether safeguards can address the risk. |
| Clinical contraindication |
A treatment is inappropriate for a particular medical condition. |
Use evidence-based and condition-specific criteria. |
| Intervention incompatibility |
The intervention cannot meaningfully be delivered under a defined circumstance. |
Distinguish necessity from logistical convenience. |
| Measurement validity |
The intended outcome cannot be meaningfully assessed under specified conditions. |
Consider accessible or alternative measurement methods. |
| Competing treatment or exposure |
A concurrent intervention prevents interpretation of the intended contrast. |
Consider measurement or analytical alternatives. |
| Research design requirements |
A condition conflicts with the defined estimand or comparison. |
Avoid restrictions that unnecessarily change the population. |
| Ethical or legal requirements |
Participation is not permissible under applicable approvals or regulations. |
Do not assume that protected or vulnerable groups must always be excluded. |
| Data suitability |
Essential eligibility or outcome information cannot be established. |
Distinguish true ineligibility from missing data. |
These factors are possible grounds for exclusion, not automatic rules. Each must be evaluated in the context of the study.
When Does Participant Safety Justify Exclusion?
Safety may justify exclusion when participation creates a risk that cannot be adequately managed within the approved research procedures.
For example, a clinical intervention may be contraindicated for patients with certain medical conditions or treatment histories.
However, researchers should distinguish actual risks from generalized assumptions about participant vulnerability.
Some concerns can be addressed through monitoring, modified procedures, additional support, or other safeguards.
The appropriate response depends on the nature and magnitude of risk, the expected benefits, and applicable ethical requirements.
Safety exclusions should be specific, evidence-informed, and proportionate rather than unnecessarily broad.
Should Every Safety Concern Become an Exclusion Criterion?
No. A safety concern requires assessment, but exclusion is only one possible response.
For example, an educational study involving participants with disabilities may require accessible interfaces, alternative assessment formats, or additional support rather than exclusion.
Similarly, clinical research may use monitoring or adapted procedures when risks can be appropriately managed.
However, researchers should not minimize genuine contraindications or risks that make participation inappropriate.
The decision should be based on the actual study procedures and approved risk-management arrangements.
The distinction is central to deciding whether safety concerns should become exclusion criteria.
Can Clinical Comorbidities Justify Exclusion?
Comorbidities may justify exclusion when they create unacceptable safety risks, make an intervention inappropriate, or prevent meaningful interpretation of the intended outcome.
However, excluding every participant with additional medical conditions can produce a highly selected clinical population.
Such a population may differ substantially from patients encountered in routine practice.
Researchers should identify which comorbidities genuinely affect safety or study validity and whether they can instead be measured and addressed analytically.
Clinical exclusions should be supported by relevant evidence and appropriate professional assessment.
Should Potential Confounders Determine Exclusion Criteria?
Not automatically. Confounding occurs when an exposure-outcome association is distorted by relevant common causes or other causal structures.
Restriction can sometimes reduce variation in a confounder, but it also changes the eligible population.
For example, excluding students with low prior achievement from an educational intervention study may remove individuals for whom the intervention is particularly relevant.
Researchers should consider measuring baseline achievement and using suitable design or analytical approaches instead.
Potential confounders should be evaluated according to the causal question rather than automatically converted into exclusion criteria.
This distinction is explored further in whether every potential confounder should become an exclusion criterion.
When Can Concurrent Treatments or Exposures Justify Exclusion?
Concurrent treatments may complicate interpretation when they influence the outcome or interact with the intervention.
For example, a clinical trial may restrict certain concomitant medications because of safety concerns or known interactions.
An educational intervention study may encounter students receiving additional tutoring outside the study.
However, concurrent support is not automatically grounds for exclusion.
If additional tutoring is common in the intended population, excluding all students who receive it may reduce practical relevance.
Researchers should distinguish safety-related incompatibility from variation that can be measured or incorporated into the intended treatment-effect estimand.
When Does Measurement Validity Justify Exclusion?
Measurement limitations may justify exclusion when the outcome cannot be meaningfully assessed for particular individuals under the study's intended design.
For example, an assessment requiring performance of a specific task may not be applicable to individuals for whom that task has no substantive meaning.
However, researchers should consider whether measurement adaptations can preserve the construct while improving accessibility.
Excluding participants because the available instrument is inconvenient or inaccessible may unnecessarily restrict the population.
Researchers should distinguish construct incompatibility from limitations of the measurement procedure.
Should Disability Be an Exclusion Criterion?
Disability should not be treated as an automatic reason for exclusion.
Researchers should determine whether a particular study procedure presents a specific risk or whether reasonable adaptations can support meaningful participation.
For example, a digital learning study might provide keyboard navigation, screen-reader compatibility, accessible materials, or alternative assessment formats.
However, adaptations must preserve the validity of the intended measurement or be accounted for appropriately.
Where participation remains incompatible with essential procedures, the rationale should be specific and documented.
Broad exclusions based on disability labels alone may undermine equity and generalizability.
Should Age Determine Exclusion?
Age may be relevant when the research concerns a particular developmental population, intervention indication, or regulatory requirement.
However, age-based exclusion should not be introduced without justification.
For example, a study of university students should not automatically exclude older adult learners if they belong to the intended population.
Similarly, research involving children may be possible under appropriate consent, assent, and safeguarding arrangements.
Researchers should distinguish substantive age eligibility from assumptions about vulnerability or convenience.
Should Pregnancy Automatically Lead to Exclusion?
No universal exclusion rule is appropriate across all studies.
Pregnancy may introduce specific safety considerations for certain interventions, particularly where fetal or maternal risks are uncertain or established.
However, automatically excluding pregnant participants from unrelated or low-risk research may be scientifically and ethically unjustified.
Researchers should assess the intervention, available safety evidence, applicable regulations, and ethical requirements.
Where exclusion is necessary, the criterion should be precise and supported by a documented rationale.
Can Consent Requirements Justify Exclusion?
Participation must comply with applicable ethical and legal requirements concerning informed consent or other authorized permission arrangements.
However, inability to provide standard independent adult consent does not automatically make every person ineligible for all research.
Some studies involving children or individuals with impaired decision-making capacity may proceed under approved permission, assent, or other safeguards.
Researchers should follow the relevant ethics review and legal framework rather than adopting blanket exclusions without justification.
Respect for persons and equitable selection should both inform the decision.
Can Language Barriers Justify Exclusion?
Language restrictions may be necessary when language competence is central to the construct or intervention.
For example, a study specifically assessing English academic writing may require participants to undertake English writing tasks.
However, an unrelated survey should not automatically exclude participants who do not understand the researcher's preferred language.
Translated materials, interpreters, and appropriately adapted instruments may improve inclusion where feasible.
Researchers should evaluate measurement equivalence and the practical requirements of informed participation.
Language-based exclusion should be justified by the research rather than convenience alone.
Should Limited Digital Access Be Grounds for Exclusion?
Not necessarily. In educational technology research, digital access may be central to the phenomenon under investigation.
Excluding students without personal devices or reliable connectivity may remove those most affected by digital inequality.
Researchers should consider providing devices, alternative access arrangements, or suitable offline procedures when these are compatible with the intervention and study design.
If continuous connectivity is genuinely necessary for the intervention, the requirement may define the population in which that intervention can currently be delivered.
However, conclusions should then be restricted accordingly.
The distinction between intervention feasibility and population relevance should be explicit.
When Does Missing Information Justify Exclusion?
Missing information does not always mean that a participant is ineligible.
For example, an incomplete administrative record may make eligibility uncertain rather than establish that the person fails the criteria.
Researchers should consider verification, clarification, or prespecified adjudication procedures.
Similarly, missing outcome data after enrollment should generally be treated as a missing-data problem rather than retrospective ineligibility.
Excluding participants because their outcomes are incomplete may introduce bias when missingness is related to the exposure or outcome.
Researchers should distinguish eligibility assessment from data completeness and analytical inclusion.
Why Is Excluding Participants After Treatment Assignment Problematic?
Post-assignment exclusions can change the comparison established by randomization.
For example, excluding students who did not complete an AI tutoring program may select individuals based on motivation, access, early performance, or perceived usefulness.
These characteristics may be related to the outcome.
If the intended estimand concerns the effect of assignment, restricting analysis to adherent participants may no longer estimate that effect.
Researchers should distinguish baseline eligibility from post-assignment adherence, withdrawal, and missing outcomes.
Appropriate analytical methods depend on the estimand and assumptions, not merely on whether complete data are available.
How Can Exclusion Criteria Introduce Selection Bias?
Selection bias can arise when exclusion depends on variables associated with the exposure and outcome in ways that distort the intended relationship.
For example, restricting a study to students who remain enrolled throughout an academic year may exclude those whose withdrawal relates to academic difficulties or intervention experiences.
Such restrictions can change the composition of the analyzed population.
However, not every exclusion creates selection bias.
The consequences depend on the causal structure, timing, and intended estimand.
Researchers should assess why a characteristic leads to exclusion rather than assuming all restrictions are harmless or all are biased.
How Do Exclusion Criteria Affect Generalizability?
Exclusion criteria determine which otherwise relevant population members cannot contribute evidence.
Restrictive criteria may produce a study population that differs from the population encountered in ordinary educational, clinical, or organizational practice.
For example, an AI tutoring trial restricted to students with reliable connectivity may provide limited evidence about effectiveness among students with intermittent access.
However, excluded groups do not necessarily have different outcomes or treatment effects.
The relevant question is whether the excluded characteristics influence the intended estimand or intervention mechanism.
Researchers should evaluate the implications of eligibility restrictions for generalizability rather than assuming that broader eligibility always guarantees external validity.
Can Too Many Exclusion Criteria Make a Sample Unrealistic?
Yes. Multiple individually plausible exclusions can collectively produce a highly selected population.
For example, a clinical trial excluding patients with common comorbidities, concurrent medications, older age, and functional limitations may study a population unlike many patients who would receive the treatment in practice.
Similarly, an educational intervention restricted to students with high achievement, reliable connectivity, and extensive prior technology experience may not reflect ordinary implementation conditions.
Researchers should examine the cumulative effect of exclusion criteria on population composition.
Each criterion should be justified independently and evaluated in combination with the others.
Can Exclusion Criteria Improve Statistical Precision?
Restriction may sometimes reduce outcome variability or remove circumstances that complicate measurement.
However, it can also reduce the eligible sample, alter event rates, and limit the intended population.
For example, excluding participants with low baseline achievement might reduce some variability in an educational outcome while removing a group for whom the intervention may be especially relevant.
Researchers should not introduce exclusions merely because they appear likely to improve statistical significance.
Sample size planning should reflect the scientifically justified population rather than allowing desired analytical results to determine eligibility.
What Alternatives to Exclusion Should Researchers Consider?
Possible alternatives depend on the concern.
For safety issues, monitoring, additional safeguards, or adapted procedures may be appropriate.
For potential confounding, randomization, stratification, measurement, or analytical adjustment may be preferable.
For measurement difficulties, accessible instruments or validated alternative procedures may improve participation.
For population heterogeneity, subgroup analysis or interaction modeling may preserve relevant variation.
These alternatives are not universally sufficient. Their appropriateness depends on the design, available evidence, and ethical requirements.
How Should Exclusion Criteria Be Operationalized?
Each criterion should describe a condition that can be assessed consistently using appropriate evidence.
For example, "serious medical condition" may be too vague unless the relevant contraindication or risk is defined.
Similarly, "poor attendance" requires a specified reference period and operational threshold if it is genuinely necessary.
Researchers should identify the assessment method, relevant timing, and procedures for uncertain cases.
Where clinical judgment is required, the protocol should describe the decision process and applicable professional standards.
Eligibility rules should be clear enough to minimize arbitrary differences between recruiters or sites.
How Can Researchers Audit Exclusion Criteria?
A useful audit asks four questions about every proposed exclusion.
First, what specific risk or inferential problem does the criterion address?
Second, what evidence supports the concern?
Third, could the concern be addressed through a less restrictive safeguard, measurement procedure, or analytical strategy?
Fourth, which population members would be excluded, and how might their absence affect the intended findings?
Researchers should also examine whether several exclusions disproportionately affect particular groups.
This audit helps distinguish necessary restrictions from assumptions that have become embedded in the protocol without sufficient justification.
When Should Exclusion Criteria Be Finalized?
Exclusion criteria should generally be specified before recruitment and outcome assessment.
Changes may sometimes be necessary because of new safety information, ethical requirements, or substantive methodological concerns.
However, modifications can alter the eligible population and intended estimand.
Researchers should document the timing, rationale, and implications of revisions and obtain relevant protocol or ethics approvals where required.
Post hoc exclusions selected because they produce more favorable results can undermine the credibility of the study.
How Should Exclusion Criteria Be Reported?
Methods sections should identify exclusion criteria and explain their scientific or ethical rationale.
Where relevant, report the number of otherwise eligible individuals excluded for each reason, while respecting privacy and avoiding misleading double counting.
Describe assessment procedures, classification thresholds, and the timing of eligibility decisions.
Distinguish exclusions made before enrollment from withdrawals, protocol deviations, missing outcomes, and analytical exclusions after enrollment.
Reporting guidance such as STROBE and CONSORT supports transparent descriptions of eligibility and participant flow.
Clear reporting allows readers to evaluate whether the study population appropriately represents the intended research question.
