Can You Change Your Research Design After Data Collection Has Started?

Can a Research Design Still Be Changed Once the Study Is Underway?

You planned the study, obtained the necessary approvals, and started collecting data. Then something stops working as expected. Recruitment is slower than anticipated. An instrument performs poorly. A comparison group becomes difficult to maintain. Perhaps an assumption that seemed reasonable during planning no longer holds.

At that point, continuing exactly as planned may weaken the study. Changing the design, however, creates another concern: can you modify a study after seeing what is happening without undermining its credibility?

The answer depends less on whether data collection has technically begun than on what you want to change, why you want to change it, what information influenced the decision, and how the change affects the original research question and the interpretation of the evidence.

Yes, but a Mid-Study Change Is Not Methodologically Neutral

The Real Issue Is Why, When, and How the Design Changes

Starting Data Collection Does Not Make Every Design Decision Irreversible

Research plans are made before investigators know exactly how a study will unfold. Some problems become visible only during implementation. Participants may be harder to recruit than expected, procedures may prove impractical, external circumstances may change, or a measurement problem may emerge.

Changing course is therefore not automatically evidence of poor research practice. Formal research protocols can be amended after a study begins. In randomized trials, for example, SPIRIT 2025 explicitly treats the protocol as a living document that may be formally amended during a trial and calls for an audit trail of protocol versions and changes. CONSORT guidance likewise recognizes that important changes to trial methods can occur after commencement and requires important changes and their reasons to be reported.

The methodological question is not simply, “Did the researchers change something?” It is whether the change preserves a defensible relationship among the research question, design, data, and conclusions.

Not Every Change Has the Same Consequences

Changing the wording of an instruction after discovering that participants consistently misunderstand it is different from changing the primary outcome because another outcome is producing more favorable results. Extending recruitment is different from replacing the target population. Correcting a procedural problem is different from redefining the hypothesis after examining the findings.

These categories can overlap. A seemingly operational decision may have methodological consequences. Expanding recruitment to a different setting, for example, may solve a practical problem while simultaneously changing the composition of the sample.

Ask Whether the Change Was Driven by the Results

One of the most important distinctions concerns what researchers knew when they decided to make the change.

A modification prompted by information independent of the study's emerging outcomes may be easier to defend. Examples could include a recruitment channel becoming unavailable, new external evidence showing that an assumption was unrealistic, or a technical failure affecting an instrument.

A change made after researchers have examined outcome patterns requires greater scrutiny. If investigators see which outcome produces a desirable result and then make that outcome primary, or alter inclusion rules because particular participants weaken an observed association, the analysis is no longer independent of the evidence it is intended to evaluate. This can introduce bias and make confirmatory claims difficult to defend.

This distinction is not always perfectly clean. Researchers may know something about study progress without examining comparative outcomes. The appropriate response is therefore not to pretend that no information was available, but to record what was known when the decision was made.

Consider Whether Data Collected Before and After the Change Remain Comparable

A mid-study modification can create two versions of the same study: observations collected under the original design and observations collected under the revised one.

Suppose a questionnaire is replaced halfway through data collection. Even if both instruments appear to measure the same construct, their scores may not be directly interchangeable. Similarly, changing recruitment criteria can mean that early and later participants come from meaningfully different populations.

Before implementing a modification, ask whether it changes the conditions under which observations are generated. If it does, the analysis may need to distinguish data collected before and after the modification, evaluate comparability, conduct appropriate sensitivity analyses, or limit particular claims. Sometimes the incompatibility is substantial enough that combining all observations as though nothing changed would be misleading.

If the problem concerns a measure or instrument that is not working as expected, the decision should therefore consider not only how to fix the instrument but also what the modification means for data already collected.

Check Whether the Research Question Still Matches the Revised Design

A design is useful because it generates evidence capable of answering a particular research question. A modification should therefore be evaluated against that question rather than merely against the practical goal of keeping the project alive.

Ask: if the revised study had been proposed from the beginning, would you still say that its design appropriately addresses the same research question?

If yes, the modification may represent an adaptation within the existing study. If the answer is only partly, the scope of the conclusions may need to change. If no, you may be confronting the more fundamental problem of a method that no longer fits the research question.

There is no universal threshold at which a modification automatically becomes a new study. The significance of the change depends on what it does to the study's objectives, population, interventions or exposures, outcomes, comparisons, data-generating procedures, and inferential logic.

Ethics and Protocol Requirements Are Separate From Methodological Justification

A change can make methodological sense and still require formal approval before you implement it.

For research subject to Institutional Review Board or Research Ethics Committee oversight, researchers should follow the requirements of the approving body and their institution. Under the U.S. Common Rule, for example, IRBs must have procedures ensuring that changes to approved research are not initiated without IRB review and approval except when necessary to eliminate apparent immediate hazards to participants. Requirements in other jurisdictions and institutions may differ.

Changes can also affect consent materials, study registrations, protocols, statistical analysis plans, sponsor requirements, or preregistrations. SPIRIT 2025 recommends that important amendments to randomized-trial protocols be reported to ethics committees and trial registries as they occur. Preregistration does not mean that a study can never change; its value partly lies in making the original plan and subsequent deviations distinguishable.

The exact oversight implications depend on the study and governing requirements, so determine which research changes require revisiting ethics approval, the protocol, or preregistration before proceeding.

Preserve the History of the Study Rather Than Rewriting It

Once data collection has begun, a revised plan should not erase the original one. Keep the original protocol or plan, record the revised version, date the change, state when it took effect, identify what prompted it, and document who made or approved the decision.

This distinction matters because a final manuscript written entirely in the language of the revised design can make an adaptation look as though it had always been planned. Transparent reporting instead allows readers to distinguish prespecified decisions from decisions made during study conduct.

APA's preregistration guidance similarly emphasizes documenting amendments and disclosing deviations from the original plan in the final report. For studies using formal protocols, SPIRIT 2025 recommends version control and a transparent audit trail of changes.

A separate treatment of how to document unplanned changes to a study can help when several modifications accumulate over the course of a project.

Suppose Recruitment Reveals a Problem With the Planned Design

The example does not have one automatic answer. Adding sites might be a defensible adaptation if the revised recruitment still samples the population relevant to the question. It could also materially alter the study if the new participants represent a different population or if site is closely related to the phenomenon being investigated.

The important point is that “we needed more participants” is not, by itself, a sufficient methodological justification. The researchers must examine what the solution does to the study. If recruitment difficulty is the immediate problem, the options for responding to participant recruitment that does not go according to plan should be considered before altering the design more substantially.

Common Mistakes When Changing a Study That Has Already Started

Evaluate the Proposed Change Before You Implement It

When something goes wrong during data collection, resist two equally problematic reactions: following the original plan mechanically even though it no longer works, or changing the design immediately because a different approach seems more convenient.

First identify the problem. Then evaluate the proposed solution against the research question, validity of the resulting evidence, comparability of existing and future data, participant protections, and applicable oversight requirements.

The objective is not to preserve the original plan at any cost. It is to preserve the integrity of the inference. Sometimes that means continuing as planned. Sometimes it means making a carefully controlled amendment. Occasionally, the more defensible choice is to acknowledge that the original design cannot be salvaged and consider whether the study should be stopped or redesigned because the design no longer works.

Whatever you decide, keep the chronology visible. A transparent adaptation can be evaluated by readers. An undocumented adaptation cannot.

Before Changing a Research Design Mid-Study, Check These Points

Questions About Changing a Research Design During Data Collection

You Can Change the Design, but You Cannot Pretend It Never Changed

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.