How to Write Inclusion and Exclusion Criteria (With Examples)

What Criteria Are Actually For

An examiner reading your literature review has one quiet question running the whole time: how did you decide what to include? Without stated criteria, the honest answer is usually “whatever came up first and looked useful”, and that is the answer they will assume. Criteria replace that with something you can point at. They turn a pile of papers into a sample, and a sample is something you can defend.

They also do something kinder, which is to protect your time. A criterion set is a decision you make once and then apply two hundred times. Screening an abstract stops being a small act of judgement that costs you five minutes of second guessing and becomes a lookup: does this meet the rule, yes or no. The system does the remembering, so you do not have to hold your reasoning in your head across three weeks of reading.

How formal your criteria need to be depends on the kind of review you are writing. A full systematic review needs them specified in a protocol before you search. A scoping review needs them stated up front but broader. A narrative review still benefits from them, even if you describe them in a paragraph of prose rather than a table. If you have not settled that question yet, start with which review type you need, because it sets how strict this has to be.

Inclusion and Exclusion Are Not Mirrors

This is the most common misunderstanding, and it produces tables that say things like “include: peer reviewed. exclude: not peer reviewed.” That row carries no information. If every exclusion criterion is the negation of an inclusion criterion, you have written one list twice.

Inclusion criteria define the boundary of your topic. Exclusion criteria remove studies that fall inside that boundary but still cannot be used. A paper can be about your population, your concept, and your context, and you still exclude it because it is a conference abstract with no method section, or because it reports the same dataset as a study you have already included, or because the full text does not exist in a language you read. Those are exclusions. They are not the opposite of anything.

The Six Dimensions to Build Them On

Most criteria sets that go wrong are missing a dimension rather than getting one wrong. Work through these six in order and you will not leave a gap that surfaces halfway through screening.

If your field uses PICO (population, intervention, comparison, outcome) or PCC (population, concept, context), you will recognise the first three. Use whichever framework your discipline expects and treat the last three as the additions that make it operational. The time window is the one students most often state without justifying. “2015 to 2026” is not a criterion until you say what happened in 2015 that makes earlier work a different conversation.

A Six Step Process

1. Write your research question on one line. Every criterion should trace back to a word in it. If a criterion has no anchor in the question, either the question is incomplete or the criterion is your preference in disguise. A well built research question does most of this work for you.
2. Fill in the six dimensions, inclusion side first. One line each, written as a property a paper either has or does not have. Avoid words that need a judgement call, like “recent”, “relevant”, or “high quality”.
3. Add exclusions that survive the mirror test. Duplicate reports of one dataset, full text unavailable, editorials and commentary, wrong outcome measured. Only rows that catch something your inclusion list would otherwise let through.
4. Write the reason next to each one. A single clause is enough. You will paste these into your methodology later, and writing them now is how you notice that one of your criteria has no reason behind it.
5. Pilot on twenty abstracts. Screen a sample before you commit. If you hesitate on more than two or three, a criterion is ambiguous and needs rewording. This step takes half an hour and saves days.
6. Lock them, then log every change. Criteria can be revised, but the revision has to be visible. Note what you changed, when, and why. Silent changes midway through screening are what make a selection look shaped to fit the argument.

Weak and Strong, Side by Side

Take a question a lot of dissertations look like: what is known about the effect of remote work on employee wellbeing in the public sector since the pandemic?

The strong version is longer, and that is the point. Every line either answers a question an examiner would ask or stops you relitigating a decision. Notice too that the exclusions catch papers the inclusions would have admitted, which is exactly what exclusions are for.

How to Word Them in Your Chapter

Criteria live in your methodology section, usually right after the search strategy and before the screening process. A table plus a short paragraph is the standard treatment, and a narrative review can compress the whole thing into one honest paragraph.

Two habits of phrasing are worth keeping. Write criteria as properties of the study rather than instructions to yourself: “studies published from 2020 onward” instead of “I only looked at papers after 2020”. And keep the reason attached to the criterion, not gathered into a separate paragraph at the end, so a reader never has to hold an unexplained rule in their head.

Screening Without Losing the Paper Trail

Criteria only pay off if you record what they did. Screen in two passes: titles and abstracts first, then full texts for whatever survives. The first pass is fast and generous, because an abstract often will not tell you the sample size or the outcome measure. When in doubt at that stage, keep it and let the full text decide.

Keep one spreadsheet with a row per record and these columns: identifier, author, year, title, source, screening decision, and the criterion that excluded it. That last column is the one people skip and then regret, because a count of exclusions with no reasons cannot become a flow diagram. If you already have a system for organising your research sources, add the columns there rather than starting a second file.

Record four numbers as you go: records identified, duplicates removed, records screened and excluded at title and abstract, and full texts assessed with exclusions by reason. Those four numbers are the whole flow diagram, and reconstructing them afterwards from memory is close to impossible. A systematic review reports them formally against PRISMA; even a narrative review looks much stronger for a sentence like “of 214 records identified, 38 met the criteria”. The difference between review types on this point is covered in literature review vs systematic review.

Five Ways This Goes Wrong

• Writing the criteria after the reading. Criteria reverse engineered from the papers you already like will always fit perfectly, and a supervisor can usually tell. Write them first, even roughly, and revise them in the open.
• Criteria that need a judgement call. If applying a criterion requires you to decide whether something is “important” or “good”, two people will screen the same abstract differently, and so will you on two different days.
• Criteria so narrow that nothing survives. Ending up with four papers is usually a criteria problem, not a literature problem. Widen the context or the design before you widen the concept, since the concept is what your review is about.
• Excluding by convenience and calling it method. Language and full text availability limits are legitimate and extremely common. State them as limitations rather than dressing them up as methodological choices, and note what they might have cost you.
• Criteria that never reappear. If your discussion says nothing about how your boundaries shaped your findings, the criteria were paperwork. Come back to them when you write limitations. Several of the common literature review mistakes trace back to this one.

Do This Before Your Next Search

Open a blank page and write six lines, one per dimension, with a reason clause on each. Add three or four exclusions that survive the mirror test. Then screen twenty abstracts against them and fix whichever line made you hesitate. That is under an hour, and it is the difference between a folder of PDFs and a defensible sample.

Once your criteria are settled, the papers that pass are the ones you actually have to read, and the work shifts to what they mean together. That is where grouping them into themes and synthesising them begins, and where a well specified set of criteria quietly makes a research gap easier to see.