When You Don't Need IRB Review (And When You Do)

Not every project involving people requires IRB review. Whether yours does depends on two federal definitions: whether the activity is "research" and whether it involves "human subjects." Both definitions live in 45 CFR 46.102, and both have enough ambiguity at the edges that your IRB — not you — makes the final call at most institutions.

The two defining questions

Is it research? The regulation defines research as "a systematic investigation, including research development, testing, and evaluation, designed to develop or contribute to generalizable knowledge." Three elements: systematic, generalizable, and knowledge-producing.

Does it involve human subjects? A human subject is a living individual about whom an investigator obtains data or biospecimens through intervention or interaction, or obtains, uses, studies, analyzes, or generates identifiable private information.

If both answers are yes, IRB review is required. If either is no, the activity is outside the federal regulatory definition — which does not mean it is outside ethical scrutiny, only that the federal IRB requirement does not apply.

Activities that are usually not research

Several common activities are typically not research in the regulatory sense, though you should still get an institutional determination:

• Quality improvement (QI). Internal projects designed to improve processes at your institution, without intent to generalize beyond the local setting, are usually QI, not research. The line blurs when QI results are submitted for publication.
• Program evaluation. Evaluating a specific program's effectiveness for internal decision-making is usually not research. If the evaluation is designed from the start to produce generalizable knowledge, it crosses into research.
• Journalism and oral history. Activities designed to produce a historical or journalistic account of individuals, rather than generalizable knowledge, are generally outside the definition.
• Case reports. A single or small-number clinical case write-up without systematic investigation is typically outside research.
• Classroom projects. Student projects conducted solely for course credit, where results are not intended for publication or presentation beyond the class, are usually not research. This is institution-specific — many universities still require a streamlined review.

Program evaluation at the boundary

Program evaluation is the most common source of "do I need IRB?" confusion. The test is intent plus audience:

• Is the project designed to improve this specific program? If yes and only yes, probably evaluation.
• Is the project designed to contribute to knowledge about this type of program in general? If yes, probably research.
• Will the results be presented at a conference, submitted to a journal, or used as thesis or dissertation data? If yes, treat it as research regardless of original intent.

For evaluators in community and public health settings, the line between evaluation and research has real operational consequences. The Public Health Practicum Logbook includes guidance on when a practicum evaluation project needs IRB review and when it does not, which maps cleanly to the general evaluation-vs-research distinction.

Secondary data analysis

Secondary analysis of existing de-identified data is usually exempt rather than "not research," which is a different thing. Exempt means the study is human subjects research under the regulation but meets one of the exemption categories in 45 CFR 46.104(d). You still need an IRB determination letter, just not a full review. The category most relevant is 46.104(d)(4) for secondary research with identifiable information or biospecimens.

If the data is already fully de-identified and publicly available (a public-use file, a data archive), some IRBs treat this as outside human subjects research altogether. Ask your IRB; do not assume.

Do not self-exempt

The universal rule across institutions is: you do not decide that your project is exempt or not-research. The IRB does. Self-exempting is one of the fastest ways to get a study in trouble, and it can render publications unpublishable if the journal requires documented IRB review.

Most institutions have a short "request for determination" form that takes less than an hour to fill out. File it early, even if you are confident the answer is "not research." The determination letter is the documentation you need later.

When in doubt, file

A short submission that returns a "not human subjects research" or "exempt" determination in two weeks is cheap insurance. The cost of finding out mid-study that you should have had IRB review — and may not be able to publish — is substantially higher.

For guidance on how to structure the request for determination, our protocol writing guide shows the short form most institutions use, and the review categories overview explains what happens after the determination.

The edge cases worth knowing

A handful of situations recur often enough to name:

• Dissertation pilot studies. Some students think pilot data does not require IRB because it is not "real" data. Any systematic data collection from human subjects intended to inform a future publication or a dissertation requires IRB review or a determination.
• Interviews with experts about their expertise. If you interview experts solely for information about a topic (journalism-style), you may be outside human subjects. If you ask about their experiences, attitudes, or practices, you are probably in.
• Public performance observation. Observing public behavior in genuinely public spaces without interaction is often exempt; observation with intervention or with identifiers is not.

The broader frame

IRB determination is about whether federal rules apply, not whether ethics apply. Every project with human participants benefits from a Belmont-style review of risks, benefits, and participant respect, whether or not it requires formal IRB oversight. The process of deciding what your project is forces exactly that reflection, which is part of why the determination step is worth taking seriously even when the answer is "not human subjects research."