Should You Recruit More Participants Than Planned to Account for Unusable Data?

If Your Study Needs 300 Participants, Should You Recruit 330 or 350 Just in Case?

You have calculated that your research requires 300 participants. However, you expect that some students may submit incomplete questionnaires, provide responses that fail prespecified quality checks, or withdraw before the required outcome is measured.

To be safe, you consider recruiting 350 participants instead of 300. Your reasoning is straightforward: collecting extra data now might prevent the study from falling below its required sample later.

But how many additional participants are justified? Should you automatically add 10% or 20%? What if some incomplete questionnaires can still contribute to the analysis? And what happens if nearly everyone provides usable data, leaving substantially more participants than the original calculation required?

The issue is not whether researchers should prepare for data loss. They should. The methodological question is how to make that preparation proportionate, evidence-based, and consistent with the intended analysis.

Recruiting Extra Participants Can Be Justified, but the Allowance Should Reflect Expected Data Loss

Planning Additional Recruitment Without Inflating the Sample Unnecessarily

What Does Overrecruitment Mean in Research?

Overrecruitment generally refers to enrolling more participants than a specified target or requirement. However, the term can describe different situations.

A researcher may deliberately plan additional enrollment to compensate for anticipated losses. Alternatively, a study may exceed its target because more participants enroll than expected.

These situations should be distinguished from approaching more people than the required sample. Issuing 600 invitations to obtain 300 participants is not equivalent to enrolling 600 participants.

When additional enrollment is prospectively justified by expected data loss, it is often more precise to describe the adjustment as sample-size inflation or a recruitment allowance.

The important question is whether the enlarged enrollment target serves a legitimate methodological purpose.

What Makes Research Data Unusable?

Data may be unusable for a particular analysis when they fail to provide information required by the prespecified statistical or methodological procedure.

Examples include missing primary outcomes, measurements invalidated by documented instrument failures, records that cannot be linked to the appropriate participant, or observations that violate prespecified eligibility requirements.

However, unusability is analysis-specific. A participant missing one questionnaire item may still contribute substantial information to other analyses.

Researchers should therefore avoid defining every incomplete record as unusable.

How Do You Calculate an Allowance for Unusable Data?

When a study requires a specified number of usable observations and researchers anticipate that a proportion of enrolled participants will not provide them, the expected enrollment target can be calculated using the usable-data rate.

The calculation assumes that the required sample is defined in terms of usable observations and that the anticipated loss proportion is applicable to the enrolled group.

It is an expected-value calculation, not a guarantee that the final sample will contain at least 300 usable participants.

Why Is Adding 10% Not the Same as Allowing for 10% Data Loss?

A common shortcut is to add 10% to the required sample.

For 300 participants, that produces an enrollment target of 330.

However, if 10% of 330 participants provide unusable data, the expected usable sample is 297, slightly below the required 300.

Dividing by the expected usable-data proportion gives 334, whose expected usable count is 300.6.

The difference becomes more consequential as the anticipated loss proportion increases.

Researchers should use the appropriate denominator rather than applying an arbitrary percentage increase.

How Much Additional Recruitment Is Needed at Different Loss Rates?

The required allowance depends on the proportion of enrolled participants expected not to contribute usable data.

These values are mathematical illustrations, not recommended loss rates. The appropriate assumption must be justified using evidence relevant to the study.

Where Should the Expected Unusable-Data Rate Come From?

Relevant estimates may come from previous studies using the same instrument, institutional data-collection records, comparable research populations, or a preliminary feasibility assessment.

For example, previous administration of an online questionnaire may provide information about incomplete submissions and documented technical failures.

However, a completion rate from a short anonymous survey may not transfer to a longitudinal study requiring repeated assessments.

Researchers should evaluate whether the prior evidence resembles the proposed recruitment procedures, measurement requirements, and participant population.

Where uncertainty remains substantial, alternative plausible assumptions should be examined.

Should Different Causes of Data Loss Be Added Together?

Not automatically. Different losses may occur sequentially or overlap.

For example, some participants may withdraw before completing a questionnaire, while others complete the questionnaire but have an invalid primary measurement.

If 10% withdraw and 5% of the remaining participants provide invalid measurements, the expected usable proportion is 0.90 × 0.95 = 0.855.

The combined expected loss is therefore 14.5%, not 15%.

When the causes overlap, simply adding their percentages may count some participants more than once.

The recruitment model should follow the actual sequence of data collection and use consistent denominators.

Is Incomplete Data Always Unusable?

No. The appropriate treatment of incomplete observations depends on the research design, estimand, statistical method, and assumptions about missingness.

For example, a participant with a missing secondary outcome may still contribute fully to the primary analysis.

Longitudinal models may use partially observed repeated measurements under specified assumptions. Multiple imputation may also be appropriate in certain settings.

Little and colleagues (2012) emphasize the importance of preventing missing data and planning appropriate methods for handling them.

Recruiting additional participants should not replace a scientifically appropriate missing-data strategy.

Can Researchers Exclude Responses That Appear Careless?

Potentially, when exclusion follows valid, transparent, and preferably prespecified quality criteria.

Examples may include confirmed duplicate submissions, demonstrated ineligibility, or failures of appropriately justified attention checks.

However, unusually fast completion, unexpected opinions, or extreme scores do not automatically establish careless responding.

Overly aggressive screening can exclude genuine participants and introduce selection bias.

Researchers should distinguish evidence of invalid measurement from responses that merely differ from expectations.

Should Outliers Be Replaced With Additional Participants?

No, not simply because they are statistically unusual.

An extreme observation may be a legitimate part of the population distribution.

Removing such observations and recruiting replacements can distort estimates and reduce the transparency of the analysis.

Researchers should investigate potential recording or measurement errors and use appropriate analytical methods when extreme values are genuine.

Exclusion should follow defensible criteria rather than a desire to obtain more convenient results.

Does Recruitment Inflation Solve Missing-Data Bias?

No. Increasing enrollment may improve the number of observed outcomes, but it does not necessarily correct systematic missingness.

Suppose students experiencing the greatest difficulty with online learning are less likely to complete a follow-up survey.

Recruiting additional students may restore the numerical target while leaving those experiences underrepresented.

The resulting estimates may remain biased despite an apparently adequate final sample.

Prevention, retention procedures, examination of missingness, and appropriate analysis remain important.

How Is Additional Enrollment Different From Additional Invitations?

Researchers may need to approach more people than the enrollment target because not everyone is eligible or willing to participate.

For example, a study requiring 334 enrolled participants may need considerably more than 334 invitations.

The invitation requirement depends on screening completion, eligibility, and consent rates.

The calculation of how many people must be approached should be distinguished from the allowance for unusable data after enrollment.

Combining the two without identifying the recruitment stages can produce incorrect targets.

What If More Participants Than Expected Provide Usable Data?

The study may achieve a larger usable sample than originally required.

This is not automatically a methodological problem when the additional enrollment was justified and authorized.

However, researchers should consider whether continued recruitment remains necessary once the intended information has been obtained.

For studies with fixed enrollment procedures, stopping decisions should follow the approved protocol and avoid introducing outcome-dependent selection.

Researchers should not exclude valid observations solely to make the final sample match a previously calculated number.

Can Additional Recruitment Affect Statistical Analysis?

Yes. A larger achieved sample may improve precision or power under appropriate assumptions.

However, changing the sample size during data collection can create methodological concerns when decisions are influenced by emerging treatment effects or significance tests.

Recruitment adjustments based on blinded operational information may be treated differently from changes driven by observed comparative outcomes.

In confirmatory research, unplanned outcome-dependent sample-size changes require particular statistical care and may affect error-rate control.

The protocol and statistical analysis plan should distinguish operational recruitment monitoring from inferential decisions.

Should Replacement Participants Be Recruited?

Replacement may be appropriate when the protocol anticipates unusable observations and permits additional enrollment.

However, replacing participants is not equivalent to erasing their participation from the study record.

For example, randomized participants who withdraw may remain relevant to participant-flow reporting and intention-to-treat principles, depending on the trial and the data available.

Researchers should not silently remove enrolled participants and substitute new individuals as though the original enrollment had never occurred.

Replacement procedures should be defined and reported transparently.

What Are the Ethical Limits of Additional Recruitment?

Additional participation involves time, effort, disclosure of information, and potentially discomfort or risk.

The anticipated scientific benefit should justify the additional burden.

For a brief anonymous questionnaire, the burden may be modest. For a lengthy intervention or repeated sensitive interviews, it may be substantially greater.

The approved protocol may also specify an enrollment ceiling or other restrictions.

The question of when overrecruitment creates unnecessary participant burden should be considered before increasing the target.

How Should an Allowance for Unusable Data Be Reported?

The methodology should distinguish the required analytical sample from the inflated enrollment target.

Researchers should state the anticipated unusable-data rate, the evidence supporting it, and the calculation used.

After data collection, they should report the actual enrollment, exclusions, missing data, and analyzed sample as appropriate to the design.

The STROBE statement recommends transparent reporting of participant numbers at relevant stages of observational studies.

For randomized trials, applicable CONSORT reporting requirements should be followed, including transparent participant flow.

Planning an Allowance for Incomplete Student Questionnaires

The additional 27 enrollments represent a prospective allowance under the stated assumptions, not a recommendation to recruit 9% extra participants in every study.

If the planned statistical method can appropriately use some incomplete observations, the required allowance may differ.

Common Misconceptions About Recruiting Extra Participants

Use a Justified Allowance Rather Than a Routine Percentage Increase

Start by defining the information required for the primary analysis and determining which missing or invalid observations would genuinely reduce the usable sample.

Then estimate expected losses, calculate the appropriate enrollment allowance, and ensure that the resulting target is consistent with the approved protocol.

Additional enrollment should protect the study against foreseeable information loss, not compensate for weak measurement procedures or create an unnecessarily large dataset.

Before Increasing Enrollment to Account for Unusable Data

Frequently Asked Questions About Recruiting Extra Participants

Recruit Extra Participants Only When Expected Information Loss Justifies Them

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.