When to Throw Out a Test Question (and When Not To): Tips for Nursing Faculty

You've pulled your item analysis report.

You've found the question that students have already email you about and you stare at the ugly discrimination score. But no one ever taught you how to decide what to actually do about it.

Do you throw it out? Give everyone credit? Leave it alone and hold the line?

This is where a lot of faculty freeze. You either over-correct by tossing every question a student complains about, or you under-correct, defending a flawed item because changing your mind after the exam feels like losing. Neither one is sustainable, and neither one is actually about the data.

If you haven't already, it's worth reading the post on how to read and actually use a test item analysis report first. This post picks up where that one leaves off, and guides you to know when to throw out a test question as a nursing faculty. Once you know what the numbers mean, the real skill is knowing what to do with them.

---

Start With the Data

The idea of admitting you wrote a bad question is uncomfortable at first. The instinct to keep or toss a question is usually emotional before it's analytical. Student pushback lands, and suddenly you're re-reading the stem for the tenth time trying to talk yourself into one answer or the other.

Skip that step. Go straight to the report.

The decision to throw out a question should rest on two things: discrimination and distractor performance, not on how convincing the student's argument sounded in your inbox.

When to Throw Out the Question

These are the patterns that justify pulling an item, and whether that means removing it from scoring or adjusting acceptable answers.

• Negative or near-zero discrimination. This means that your strongest students missed it and your weakest students got it right. That's a flawed item, not a knowledge gap. If the question addressed a knowledge gap, the stronger students would have answered correctly, correctly discriminating between high-scoring and low-scoring learners.

• Negative point-biserial. This is also a measure of discrimination. You want to review the point-biserial for not just the question overall, but for each answer choice. If the point-biserial is negative, that means that high performers are consistently choosing that distractor (and not the correct answer). This usually means more than one answer is defensible, or your "best" answer isn't actually the best.

• A distractor is quietly correct. Sometimes on closer read, under a specific but reasonable interpretation, the "wrong" answer isn't wrong. If a substantial cluster of high-performing students chose it, take that seriously.

• The stem is ambiguous, not just hard. Difficulty is fine. Confusion isn't. If two faculty members read the item and land on different answers, students will too.

• The content is outdated or was taught incorrectly. If your slides said one thing and the correct answer reflects something else, that's on the exam, not the student.

In these cases, throwing out the item isn't lowering your standards but is actually protecting the validity of the exam.

A flawed item doesn't just mismeasure one question; it drags down your reliability for the whole test. Just make sure the data supports your decision.

When Not to Throw It Out

This is the harder skill, honestly, because it means holding steady when students are unhappy and the room feels tense.

• The question was simply hard. A 0.35 difficulty on a higher-level prioritization question is okay. That's what a well-built rigorous item looks like. Difficulty alone is never a reason to toss something.

• Discrimination is positive, even if modest. If your strong students are still outperforming your weaker ones on the item, it's doing its job, even if not everyone loved it.

• One or two students are unhappy, but the data isn't. A vocal complaint is not the same as a flawed item. If discrimination and point-biserial both look healthy, the issue is more likely content mastery than question quality.

• It's foundational content that was heavily covered. A low score on something you emphasized repeatedly is worth flagging for teaching, not for the exam key. That's useful information, just not the kind that changes the answer.

• The "obviously correct" argument doesn't hold up next to your objectives. Students are skilled at constructing a plausible case for the wrong answer under pressure. If the correct answer clearly aligns with your terminal objective and the evidence base, trust it.

Holding the line here isn't about being unmovable. It's about making sure your decisions are driven by evidence instead of the loudest voice in the room.

The Middle Path: Revise Instead of Delete

Not every flawed item needs to disappear. Sometimes the fix is smaller than a full toss. This is the version of test-question management that improves your question bank over time instead of shrinking it.

• Swap a weak distractor. If one option is pulling less than 5% of responses, it's not doing any work. Replace it with something built from a real student misconception, and the item often performs completely differently next semester.

• Tighten the stem. A single ambiguous phrase can tank discrimination. Reword it, keep the content and cognitive level intact, and re-test it in the next cohort.

• Reclassify the answer. Occasionally a "best answer" question needs to become a "select all that apply," or the key needs updating to reflect current evidence. That's a revision, not a question failure.

---

A Quick Decision Framework

When you're staring at a flagged item and need to figure out where to start:

1. Check discrimination and point-biserial for the question first. Negative or near-zero on either? Lean toward pulling it.

2. Look at distractor spread. Is a "wrong" answer pulling a suspicious number of strong students (evidenced by negative point biserials)?

3. Ask whether the confusion is about content or clarity. Content confusion = teaching note. Clarity confusion = likely a flawed item.

4. Decide: keep, toss, or revise for next time.

5. Document your reasoning in one line in your notes. Future you will thank present you.

That last step matters more than it seems. When you can point to discrimination and point-biserial data instead of "a student made a good point," your decision becomes defensible to your course lead, your curriculum committee, and (just as important) to yourself.

You don't have to get every item right the first time. You just have to know how to tell the difference between a hard question and a broken one.

Jacklyn DelPrete, EdD, CRNP, FNP-C, CNE is a family nurse practitioner, full-time graduate nursing faculty member, and founder of The Elevated NP. Questions or topics you'd like covered? Reach out at elevatednpteam@gmail.com.