When Does Expected Participant Attrition Make a Longitudinal Study Infeasible?

At What Point Does Expected Dropout Make a Longitudinal Study Unrealistic?

You are planning a longitudinal study that will follow participants for two years. Your sample size calculation indicates that 300 participants must contribute information to the primary analysis. However, comparable studies suggest that a substantial proportion of participants may be unavailable at final follow-up.

Should you simply recruit more people? What if the required initial sample exceeds the accessible population, recruitment budget, or time available?

Even if recruitment is possible, another concern remains. Participants who discontinue may differ systematically from those who remain, particularly when dropout is related to the outcomes being studied.

The challenge is determining when attrition represents a manageable planning assumption and when it threatens the scientific or operational feasibility of the entire design.

Attrition Makes a Study Infeasible When Its Consequences Cannot Be Managed Credibly

Evaluating Whether Longitudinal Research Remains Scientifically and Operationally Feasible

Feasibility Is More Than Achieving the Recruitment Target

Researchers sometimes define feasibility primarily in terms of whether enough participants can be enrolled.

For longitudinal research, recruitment is only one component. Participants must also provide sufficient information at scientifically meaningful time points.

A study might recruit its target sample successfully but fail to obtain enough primary outcome measurements. Another might achieve the planned number of final observations while experiencing selective attrition that seriously limits interpretation.

Feasibility should therefore be assessed across several connected dimensions.

These dimensions should be evaluated together. A technically achievable recruitment target does not necessarily establish that the study is feasible as designed.

There Is No Universal Attrition Threshold for Infeasibility

It may be tempting to classify a study as feasible below a particular attrition percentage and infeasible above it.

However, such a rule would ignore important differences among research designs.

A study with 25% missing final outcomes may remain informative when the available repeated measurements support an appropriate analysis and the missing-data assumptions are defensible.

Another study with only 5% missing outcomes could experience consequential bias if nearly all missing participants belong to a small subgroup with substantially different outcomes.

Neither percentage independently determines feasibility.

The relevant issue is whether the study can still estimate its intended quantity with acceptable uncertainty and credibility.

Researchers should therefore avoid treating familiar attrition allowances as universal quality standards.

Start With the Primary Estimand and Essential Measurements

An estimand describes the quantity the research intends to estimate.

In longitudinal research, this may involve change over time, an outcome at a particular endpoint, or differences in trajectories between groups.

Attrition feasibility depends on which observations are necessary for that estimand.

For example, a study investigating students' academic engagement at graduation may depend heavily on obtaining measurements near graduation.

A study investigating average engagement trajectories across several semesters may use information from multiple assessment points, even when some participants miss particular measurements.

Researchers should distinguish missing one observation from losing all information contributed by a participant.

They should also consider whether later measurements can answer the original research question or whether a shorter follow-up would require changing the scientific objective.

Calculate the Recruitment Required Under Plausible Attrition Scenarios

A simplified planning calculation can help determine whether expected attrition creates an unrealistic recruitment requirement.

When a specified number of complete final outcomes is needed, the initial recruitment target can be estimated by dividing that number by the expected retention proportion.

The calculation indicates that recruiting 500 participants would produce an expected 300 complete outcomes if the assumed retention proportion is realized.

It does not guarantee 300 outcomes, establish adequate power under every analytical model, or correct selective missingness.

For repeated-measures designs, clustered studies, survival analyses, or complex missing-data patterns, analysis-specific calculations or simulation may be necessary.

The distinction between a planning allowance and a realistic prediction is explained by estimating participant attrition using comparable evidence.

Compare Recruitment Requirements With the Accessible Population

A calculated recruitment target is useful only when it can plausibly be achieved.

Suppose a study requires 500 enrolled participants, but the participating institutions collectively have only 420 potentially eligible individuals.

Even if every eligible individual agreed to participate, the proposed recruitment target would remain unattainable.

In practice, eligibility confirmation, refusal, scheduling difficulties, and other recruitment barriers may further reduce enrollment.

Researchers should therefore distinguish the total population, the accessible eligible population, the number expected to receive invitations, and the number likely to enroll.

Expanding recruitment sites or channels may help, but only when those changes are feasible, approved, and compatible with the research question.

Additional recruitment should not be assumed to solve every limitation, particularly when new channels reach substantially different participant populations.

Recruitment Time and Follow-Up Time Must Be Considered Together

Longitudinal studies require enough time both to recruit participants and to observe them through the planned endpoint.

Staggered recruitment can create a feasibility problem when participants enrolled late cannot complete follow-up before funding or staffing ends.

For example, a study requiring eighteen months of observation cannot necessarily continue enrolling participants until the final month of a two-year project.

The timeline must accommodate the last participant's required follow-up, data processing, and analysis.

Researchers should examine recruitment rates, enrollment deadlines, expected assessment completion, and available project extensions.

A study may be feasible in terms of population size but infeasible within the funded period.

High Attrition Can Reduce Precision and Statistical Power

Missing outcomes generally reduce the amount of information available for estimating study parameters.

The effect on statistical power depends on the analysis, outcome variability, correlation among repeated measurements, group allocation, and missing-data pattern.

For a simplified analysis requiring complete final outcomes, fewer observed participants may directly reduce precision.

In repeated-measures models, partial observations may still contribute information, although the extent depends on the model and assumptions.

Researchers should not assume that every participant missing a final assessment contributes no useful information.

Likewise, they should not assume that modern missing-data methods preserve the original power automatically.

Simulation can be useful when simple calculations cannot adequately represent the anticipated data structure.

Selective Attrition Can Threaten Validity Even When Enough Participants Remain

A study may retain enough participants numerically while losing important information systematically.

Suppose researchers investigate changes in student well-being over two years. Students experiencing worsening difficulties may be less likely to complete later questionnaires.

If these students are disproportionately missing, the observed trajectories may appear more favorable than those of the original cohort.

Alternatively, participants whose circumstances improve may become less interested in follow-up, potentially producing a different distortion.

The direction of bias cannot be inferred from attrition alone.

Researchers should examine plausible relationships among missingness, observed characteristics, and unobserved outcomes.

The National Research Council's report on missing data in clinical trials emphasizes prevention, appropriate analytical assumptions, and sensitivity analyses. These principles are relevant when assessing whether longitudinal findings remain credible.

Recruiting more participants cannot automatically recover the outcomes of individuals who are systematically missing.

Missing-Data Methods Do Not Make Every Study Feasible

Likelihood-based models, multiple imputation, inverse probability weighting, and other methods may address incomplete data under particular assumptions.

For example, methods relying on missing-at-random assumptions require that the relevant missingness mechanism can be adequately explained by observed information, conditional on the model.

Such assumptions are generally not fully testable using observed data alone.

If missingness depends strongly on unobserved outcomes, the analysis may remain sensitive to assumptions that the available evidence cannot resolve.

Sensitivity analyses can investigate how conclusions change under alternative plausible assumptions.

However, if reasonable assumptions lead to substantially different substantive conclusions, the study may have limited ability to answer its primary question decisively.

Statistical sophistication cannot manufacture information that the research design has systematically failed to obtain.

Differential Attrition Can Complicate Group Comparisons

In comparative longitudinal studies, participants in different groups may have different probabilities of completing follow-up.

An intervention requiring additional activities may be more burdensome than the comparison condition. Participants experiencing unfavorable outcomes may also be more likely to discontinue particular interventions.

These mechanisms can create differential attrition.

However, identical attrition rates do not guarantee unbiased comparisons, and unequal rates do not independently prove bias.

Researchers should examine the missing-data mechanisms and whether the analysis remains appropriate for the intended estimand.

Planning comparable retention opportunities across study groups may reduce avoidable differences in outcome ascertainment.

Participant Burden Can Make Long-Term Follow-Up Unsustainable

A longitudinal design may require repeated questionnaires, interviews, physical assessments, or intervention-related activities.

Each procedure may be reasonable individually, but the cumulative burden can become substantial.

Participants may face competing responsibilities, financial costs, emotional demands, or accessibility barriers that make continued participation difficult.

Researchers should evaluate whether every measurement is necessary and whether the schedule can be simplified without compromising the research question.

Reducing unnecessary burden may improve feasibility more effectively than increasing the recruitment target.

However, reducing assessment frequency or shortening follow-up can change the scientific information obtained.

Such modifications should be justified by the study's objectives rather than introduced solely to achieve more favorable retention statistics.

Retention Strategies Can Improve Feasibility, but Their Effects Are Uncertain

Researchers may use reminders, flexible appointments, accessible assessment methods, reimbursement, or appropriate incentives to support continued participation.

Evidence reviews indicate that some retention interventions improve follow-up completion, while evidence for other strategies remains limited or context-dependent.

Researchers should not assume that introducing several retention strategies will reduce attrition to a predetermined level.

Instead, the study should use relevant evidence and, where feasible, pilot information to estimate what improvement is plausible.

Retention procedures should also preserve participant autonomy and avoid excessive contact or undue influence.

These considerations belong within the research protocol's retention plan.

When Does a Pilot Study Become Particularly Valuable?

A feasibility or pilot study may be especially useful when the population is difficult to follow, the proposed assessments are demanding, or comparable retention evidence is limited.

Such a study can examine recruitment rates, assessment completion, communication procedures, participant burden, and operational constraints.

However, small pilot samples produce uncertain attrition estimates.

A pilot showing 20% missing outcomes does not establish that the definitive study will experience exactly 20% attrition.

Researchers should use feasibility findings to refine procedures and evaluate plausible scenarios rather than treating preliminary estimates as precise predictions.

Prespecified progression criteria may help determine whether to proceed, modify the design, or discontinue development.

These criteria should reflect the study's scientific and operational requirements rather than arbitrary universal thresholds.

When Should Researchers Redesign Rather Than Recruit More?

Recruitment inflation may be reasonable when attrition is predictable, the required population is accessible, and the remaining missing-data concerns can be managed credibly.

Redesign becomes more appropriate when the required recruitment target is unattainable, follow-up demands are excessive, or essential outcomes are unlikely to be observed with sufficient validity.

Possible modifications include reducing unnecessary assessments, changing follow-up arrangements, expanding approved recruitment access, or reconsidering the duration of observation.

However, these changes may alter the research question or estimand.

For example, replacing a two-year outcome with a six-month outcome does not simply make the same study easier. It changes the period over which the phenomenon is evaluated.

Researchers should be explicit about the scientific trade-off rather than presenting a shorter study as equivalent to the original design.

When Might Discontinuing the Proposed Design Be Justified?

Sometimes no feasible modification can preserve the intended research objective.

A study may require a rare population that cannot be recruited in sufficient numbers. Essential long-term outcomes may occur beyond the available funding period. Participants may face burdens disproportionate to the expected scientific value.

Under such circumstances, proceeding with the original design may consume resources and impose participant demands without a reasonable prospect of producing informative evidence.

Researchers should consider alternative questions, designs, collaborations, or funding arrangements.

Abandoning an infeasible design is not necessarily abandoning the research problem. It may be the more responsible methodological decision.

When Attrition Makes a Two-Year Student Study Difficult to Complete

Evaluating Alternative Attrition Scenarios

These are hypothetical planning scenarios, not empirical predictions or universal feasibility thresholds.

At 30% expected attrition, the simplified calculation requires 429 participants, exceeding the accessible recruitment capacity.

At 40% expected attrition, the requirement increases to 500.

Deciding What to Change

Suppose the team develops approved follow-up procedures that make 20% to 25% attrition plausible, although uncertain. The study may become operationally feasible within the available recruitment capacity.

However, if credible evidence continues to suggest attrition near 40%, recruiting 400 participants would be expected to yield only 240 complete final outcomes under the simplified assumptions.

The researchers should not describe that scenario as adequately powered merely because the initial enrollment target was reached.

Nor should they assume that the remaining 240 participants are free from selection bias.

The example illustrates why feasibility requires joint consideration of recruitment, retention, statistical information, and missing-data validity.

Common Misconceptions About Attrition and Longitudinal Feasibility

Make Feasibility a Decision About Credible Evidence, Not Just Enrollment Numbers

Begin with the primary research question and identify the follow-up information required to answer it.

Then compare plausible attrition scenarios with recruitment capacity, available resources, statistical requirements, and assumptions about missing outcomes.

Researchers should document the assumptions supporting their progression decision, particularly when attrition estimates are uncertain.

Where participant burden is substantial, feasibility also requires considering whether the expected scientific value justifies asking individuals to undertake prolonged follow-up.

A study that cannot credibly answer its intended question should be reconsidered before participants and resources are committed unnecessarily.

Before Deciding Whether a Longitudinal Study Is Feasible

Frequently Asked Questions About Attrition and Study Feasibility

Attrition Becomes a Feasibility Problem When the Study Can No Longer Answer Its Question Credibly

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.