03 · What You Need to Know
How Eligibility Changes Affect Population Boundaries and Research Inference
How Are Eligibility Criteria Related to the Target Population?
The target population is the collection of individuals or other entities about which a study intends to make claims.
Eligibility criteria specify the conditions under which prospective participants may enter the study.
These concepts are closely related, but they are not necessarily identical.
For example, a study may target all undergraduate students nationally while restricting recruitment to students at selected universities.
Such restrictions can produce an eligible or accessible population narrower than the intended target population.
Researchers should distinguish the target population, accessible population, and study sample before interpreting eligibility amendments.
What Is the Difference Between Changing Eligibility and Changing the Target Population?
An eligibility change modifies the conditions used to determine whether someone can participate.
A target population change modifies the population about which the study intends to make inferences.
The two may occur together, but they need not.
For example, expanding a study explicitly targeting full-time undergraduate students to include part-time students changes the population if the intended inference now includes both groups.
By contrast, removing an unnecessary laptop ownership requirement from a study already targeting all undergraduate students may bring the eligible population closer to the original target.
Researchers should evaluate the substantive meaning of the change rather than assuming that every amendment necessarily creates a new target population.
Which Eligibility Changes Are Most Likely to Affect the Population?
| Eligibility change |
Possible population consequence |
Key question |
| Expanding age range |
Includes additional developmental or demographic groups. |
Was the original target age-specific? |
| Adding educational levels |
May expand the population to different learning contexts. |
Are those levels part of the intended inference? |
| Including additional disease stages |
May broaden the clinical population and treatment indication. |
Does the intervention remain appropriate? |
| Removing an experience requirement |
May introduce novice or previously unexposed participants. |
Was experience central to the question? |
| Expanding geographic eligibility |
May include additional jurisdictions or settings. |
Does the intended geographic population change? |
| Changing enrollment reference dates |
May include different individuals or cohorts. |
Which temporal population is intended? |
| Removing a device ownership restriction |
May improve coverage of an existing target population. |
Was device ownership scientifically necessary? |
| Improving eligibility verification |
May improve classification without redefining membership. |
Has the underlying criterion changed? |
These consequences depend on the original research question and population definition. The same amendment may represent a substantive population change in one study and an operational correction in another.
What Is the Difference Between the Target Population and the Eligible Population?
The target population defines the broader population of substantive interest.
The eligible population consists of individuals satisfying the study's operational inclusion and exclusion criteria.
For example, a trial may seek evidence relevant to adults with a particular condition but exclude patients with specified contraindications.
The eligible trial population is therefore narrower than the broader clinical population of interest.
Removing a scientifically unnecessary exclusion could expand eligibility without changing the original target population.
However, including individuals outside the originally intended clinical condition would require reconsidering the target population and treatment-effect estimand.
Researchers should explicitly identify which population level is affected by a proposed amendment.
Can Changing Inclusion Criteria Expand the Target Population?
Yes, when the new criteria deliberately broaden the population of inference.
For example, a study initially designed to estimate AI literacy among first-year undergraduate students may be expanded to include students from all undergraduate year levels.
If the intended conclusion now concerns all undergraduates, the target population has changed.
The expansion may also introduce different academic experiences, baseline knowledge, and technology use patterns.
Researchers should determine whether the measurement instrument and sampling design remain suitable for the broader population.
A larger eligible pool does not automatically produce a valid estimate for the expanded target population.
Can Removing Exclusion Criteria Change the Target Population?
Sometimes. Removing an exclusion may either broaden the intended population or improve access to an already defined population.
For example, a study of all undergraduate students that initially excludes part-time students without justification may remove that restriction while preserving its original target population.
By contrast, a study explicitly targeting full-time students may change its target population if it begins making claims about full-time and part-time students together.
The difference lies in the intended population of inference, not merely the number of people newly eligible.
Researchers should examine whether the original exclusions unnecessarily restricted the eligible population.
Can Changing Screening Methods Alter Population Membership?
Yes, even when the written eligibility criteria remain unchanged.
For example, switching from self-reported diagnosis to documented clinical confirmation may classify some participants differently.
The conceptual population may remain patients with a specified condition, but the operationally eligible population may change because the assessment methods have different error characteristics.
Researchers should distinguish an improvement in classification accuracy from a substantive change in the eligibility definition.
They should also evaluate whether the revised method systematically excludes individuals whose records are incomplete.
The implications of different eligibility assessment methods should be considered before implementation.
How Do Geographic Eligibility Changes Affect the Population?
Geographic restrictions may define the population of interest or merely reflect where recruitment is possible.
For example, expanding a study from one province to several provinces may change the target population if the original question specifically concerned residents of the first province.
However, if the original target was national and recruitment initially occurred in only one province, adding other provinces may improve coverage without redefining the target.
Researchers should distinguish geographic population boundaries from study site locations.
Additional sites may also introduce differences in policies, institutional resources, and intervention implementation.
These differences may require attention even when the target population remains unchanged.
How Do Changes in Time-Based Eligibility Affect the Population?
Temporal eligibility determines when individuals must satisfy the population definition.
For example, expanding eligibility from students enrolled during one academic semester to students enrolled across two academic years may change the temporal population.
However, extending recruitment into a later month does not necessarily change the population if eligibility continues to be determined using the original reference date.
Researchers should distinguish population membership periods from recruitment and measurement periods.
Changes may be particularly consequential when technologies, policies, or treatment practices evolve over time.
The principles of defining temporal target population boundaries remain relevant.
How Do Disease Stage and Severity Changes Affect the Population?
Expanding clinical eligibility to additional disease stages or severity categories may change the population for which treatment effects are intended.
For example, a trial initially targeting patients with moderate disease may be expanded to include patients with severe disease.
The intervention may have different safety, effectiveness, or measurement implications across these groups.
Researchers should evaluate whether the treatment remains appropriate and whether the intended estimand is still coherent.
Changes to clinical eligibility should be supported by relevant evidence and reviewed under applicable safety and ethics procedures.
They should not be introduced solely to increase recruitment numbers.
What Happens to the Estimand When the Population Changes?
An estimand specifies the quantity a study intends to estimate, including the population to which the quantity applies.
For example, an average treatment effect among novice AI users is not necessarily the same as an average treatment effect among novice and experienced users combined.
If eligibility expands to include experienced users and the intended effect now concerns both groups, the population component of the estimand changes.
However, the original subgroup effect may remain scientifically relevant and estimable if the design provides sufficient information.
Researchers should identify whether the primary estimand is being replaced, expanded, or supplemented.
This decision should be documented before outcome-driven analysis whenever possible.
How Can a Population Change Affect an Average Treatment Effect?
Suppose a study initially targets novice users of an educational technology but later includes experienced users.
If treatment effects differ between these groups, the average effect for the combined population depends on their relative proportions.
This decomposition applies when the two groups partition the target population and the subgroup effects are defined on a common additive outcome scale.
It illustrates why changing population composition may change the overall estimand even when the intervention remains identical.
The subgroup effects and population proportions must be appropriately estimated or otherwise supported by evidence.
Does Every Population Change Alter the Estimated Effect?
Not necessarily. Changing the target population changes the definition of the population-specific estimand, but its numerical value may remain the same.
For example, if an intervention has the same average effect among novice and experienced users, expanding the population may not change the combined average effect.
However, this equality should not be assumed without justification.
Researchers should distinguish a conceptual change in the estimand from an observed numerical difference in its estimate.
The absence of statistically significant subgroup differences does not necessarily establish that treatment effects are identical.
Can Eligibility Changes Affect Internal Validity?
Yes, depending on how the changes are implemented.
For example, changing eligibility after treatment assignment or using outcome information to determine who qualifies may introduce selection bias.
In randomized trials, prospective amendments applied consistently before randomization may preserve the validity of assignment within the revised eligible population, subject to other design assumptions.
However, differences between recruitment periods, intervention versions, or study sites may complicate interpretation.
Researchers should distinguish eligibility changes made independently of outcomes from changes influenced by observed treatment responses.
Not every population amendment invalidates a study, but its consequences require evaluation.
Can Eligibility Changes Affect Generalizability?
Yes. Expanding eligibility may improve direct coverage of the intended population, while narrowing eligibility may reduce it.
However, broader eligibility does not automatically establish representativeness.
For example, adding part-time students to a convenience sample from one university does not make the sample nationally representative.
Generalizability also depends on sampling, recruitment, participation, relevant effect modifiers, and the assumptions supporting population inference.
Researchers should evaluate both the new eligibility boundaries and the actual composition of the recruited sample.
What Happens When Different Participants Were Recruited Under Different Criteria?
Participants recruited before and after an amendment may differ because the eligibility rules changed.
They may also differ because of calendar time, institutional conditions, recruitment channels, or changes in the intervention.
Researchers should document the criteria applicable to each participant and the date of recruitment.
Where appropriate, analyses may assess whether conclusions are sensitive to eligibility version or recruitment period.
However, comparing early and later participants does not isolate the causal effect of the eligibility amendment because other conditions may have changed simultaneously.
Interpretation should remain consistent with the observational nature of such comparisons.
Should the Original and Expanded Populations Be Analyzed Separately?
Sometimes. Separate or subgroup analyses may be informative when the original and newly eligible populations differ in characteristics relevant to outcomes or treatment effects.
For example, novice and experienced AI users may be analyzed separately if experience is a plausible effect modifier.
However, subgroup analyses require adequate information and should preferably be specified before outcome examination.
Small subgroup samples can produce imprecise estimates, and multiple exploratory comparisons can increase the risk of misleading findings.
Researchers should define whether the primary estimand concerns the original population, expanded population, or both.
Can Weighting Address Changes in Population Composition?
Weighting may help estimate a specified target population quantity when relevant population information and suitable assumptions are available.
For example, if novice and experienced users are recruited in proportions different from their target population distribution, appropriate weighting may help estimate a population average.
However, weighting requires sufficient overlap and reliable information about relevant characteristics.
It cannot automatically recover effects for groups absent from the study.
Researchers should also distinguish weighting for sampling differences from methods intended to address confounding.
Adjustment should be selected according to the estimand and design rather than used as an automatic remedy for eligibility amendments.
How Can Eligibility Changes Affect Statistical Power?
Changing eligibility can alter sample size, outcome variability, event rates, and expected treatment effects.
For example, expanding an educational intervention to additional year levels may increase recruitment while introducing greater variation in baseline knowledge.
The effect on statistical power depends on the analytical design and the quantity being estimated.
Researchers should revisit planning assumptions when changes are substantial.
However, revisions should not be driven by unblinded outcome comparisons without an appropriate prespecified or independently reviewed procedure.
Increasing the number of eligible participants does not automatically guarantee more informative results.
When Does an Eligibility Amendment Require a New Research Question?
A new or revised research question may be necessary when the amended population changes the phenomenon or comparison being investigated.
For example, a study of novice researchers' first experiences using generative AI cannot simply include experienced users without reconsidering its experiential focus.
Similarly, a clinical intervention developed for one disease stage may require a different question when expanded to substantially different stages.
However, some eligibility expansions preserve the original question while improving coverage.
The decision depends on whether the revised population supports the same conceptual construct and estimand.
Should Researchers Redefine the Target Population After Recruitment?
Researchers should avoid retrospectively redefining the target population merely to make the observed sample appear representative.
For example, a national study that recruits only students from selected urban universities should not silently redefine its target as those universities after examining results.
A narrower study may be defensible, but the revision should be transparent and substantively justified.
Researchers should distinguish the original research objective from the population directly supported by the available evidence.
Where appropriate, the final report may describe both the original target and the limitations of the achieved study population.
What Ethical Issues Arise When Eligibility Changes?
Eligibility changes may affect participant safety, fair selection, consent information, and the distribution of research burdens and benefits.
For example, expanding clinical eligibility to patients with additional comorbidities may require renewed risk assessment.
Broadening an educational intervention to participants with different accessibility needs may require appropriate adaptations.
Researchers should evaluate whether revised criteria create new ethical requirements or alter the approved risk-benefit assessment.
Substantive amendments should follow applicable ethics and regulatory procedures before implementation.
Recruitment convenience should not override participant protection or equitable selection.
How Should Eligibility Changes Be Documented?
Researchers should preserve the original and revised eligibility criteria, the rationale for the amendment, and the implementation date.
Documentation should identify whether the target population, estimand, intervention, or measurement procedures changed.
Where applicable, record ethics approvals, protocol amendments, and registration updates.
Participant-level records should identify the eligibility version under which enrollment occurred.
Statistical analysis plans should be revised when necessary and appropriately documented.
This audit trail helps distinguish planned methodological changes from outcome-driven modifications.
How Should Population Changes Be Reported in the Final Paper?
Methods sections should describe the original eligibility criteria, substantive amendments, and the periods during which each version applied.
Explain the reason for the changes and whether they affected the target population or estimand.
Report participant flow and relevant differences between recruitment periods where feasible.
In the results and discussion, distinguish estimates for the original population from those for the revised population when appropriate.
Reporting guidance such as CONSORT, SPIRIT, and STROBE provides relevant principles for transparent documentation of study populations and protocol changes.
Readers should be able to determine exactly whose outcomes support the final conclusions.
