Most nursing faculty have had this moment: the exam average comes back in the high 80s, everyone is relieved, and a quiet voice asks whether the class really knows the material that well or whether the exam simply didn't ask much of them. That voice is worth listening to. An exam is the main evidence a program has that students can think like nurses, and a few habits can quietly drain that evidence away.
This article looks at four of them, and then at how to check your own exam using numbers you already have.
Are we failing students by not making exams hard enough?
"Hard" is the wrong goal. A hard exam full of trick questions measures test anxiety and reading speed as much as nursing knowledge. The right goal is an exam that tells students apart: the ones who can apply what they learned should do noticeably better than the ones who can't.
An exam that's too easy fails at that. When 95% of the class gets a question right, the question contributes almost nothing. It doesn't separate anyone, and it adds a point to every score. Stack enough of those questions and you get a comfortable average that says little about readiness. Students leave believing they're on track, and the first honest signal they get may be a standardized exam or the NCLEX itself.
There's also a quieter cost. Easy questions are usually easy because they ask students to recognize a fact: a normal lab value, a definition, a drug class. The Next Generation NCLEX, introduced in April 2023, is built around clinical judgment: recognizing cues, forming hypotheses, choosing and evaluating actions. An exam made mostly of recall items gives students far too little practice with that kind of thinking.
Are we asking the right questions?
A question can be well written and still ask for the wrong thing. Two checks help.
What level of thinking does it ask for? Bloom's taxonomy is the usual frame. Remember and understand questions ask a student to recall or explain. Apply and analyze questions put the student in a situation and ask what to do or what matters most. NCLEX-style items sit mostly at apply and above. If most of your exam is remember-level, the exam is measuring something different from what students will be held to.
Does it match what you taught? Every question should trace back to a course objective and to something covered in class or assigned reading. Questions that test material nobody taught frustrate students and make the exam unfair. Objectives that never appear on any exam are the other half of the problem: content you said mattered, with no evidence anyone learned it.
Have your students already seen your test-bank questions?
Publisher test banks save faculty enormous time, and many programs use the same ones. That's also their weakness. A test-bank question has been given to many students at many schools, and copies circulate. Some show up on study-sharing sites like Quizlet, sometimes word for word, sometimes with the answer attached.
There's nothing wrong with using test-bank questions. The point is not to assume they're secure. Signs a question may be known:
- It gets much easier from one term to the next with no change in how you taught it.
- Students who struggle on the rest of the exam get it right at about the same rate as your strongest students.
- A search for a distinctive phrase from the stem turns it up on a public site.
The fix is usually simple: rewrite the scenario, change the client details, or move the correct answer to a different choice. A rewritten question tests the same objective without rewarding memorized answers.
Are AI chatbots sharing exam questions with students?
Not in the sense of one chatbot passing your questions to a student. General chatbots don't hand one user's conversation to another. But two things are true and matter more:
- Any question a student can see, they can paste into a chatbot. That includes practice questions, questions from a previous term, and anything a student wrote down or remembered after an exam. The chatbot will give an answer and often a convincing explanation.
- Anything posted publicly can be found and reused. Once a question is on a study site, it can be searched, copied, and shared indefinitely.
The practical takeaway is the same as with test banks: treat any question that has ever left a secure exam as possibly known. Keep a pool of questions you rotate and revise, and watch for the signs above. The same data that shows a question got suspiciously easy will catch it, whatever the cause.
How do I know if I am asking the right questions?
You already have most of the answer in your item-analysis export from ExamSoft, Canvas, Blackboard or a similar system. Four numbers per question tell most of the story.
1. Difficulty (p-value): the share of students who got it right. Most well-functioning items fall somewhere between about 0.30 and 0.90. Above 0.90, the question is likely too easy to tell students apart. Below 0.30, check whether the key is right, whether the content was taught, or whether the wording is confusing.
2. Discrimination (point-biserial): did stronger students get it right more often? As a rough guide, 0.30 and up is good, 0.20 to 0.30 is acceptable, below 0.20 is weak, and a negative number means your strongest students tended to miss it. A negative value almost always points to a problem with the question: a miskeyed answer, two defensible answers, or wording that punishes students who know more.
3. Distractors: how the wrong answers were chosen. Each wrong answer should attract at least a few students, mostly weaker ones. A choice nobody picks isn't working and makes the question easier. A wrong answer that pulls more of your top students than the correct answer deserves a hard look.
4. Reliability (Cronbach's alpha) for the whole exam. This tells you how consistently the exam measured. Weak and negative items pull it down, and fixing them is usually the fastest way to raise it.
Put those together with the level of thinking each question asks for and which objective it covers, and you have a clear picture: which questions to keep, which to revise, which might have leaked, and what your exam actually measured.
A short checklist for your next exam
- Remove or rewrite questions almost everyone gets right, unless they cover a safety-critical fact you want every student to know.
- Fix any question with a negative point-biserial before you use it again.
- Replace distractors nobody chooses.
- Raise a few remember-level questions to apply-level by adding a client scenario.
- Map every question to an objective, and note objectives the exam never tested.
- Compare each question with how it performed last term, and look closely at any that got much easier.
Where MapMyExam fits
Reading all of this for a 50-question exam takes hours most faculty don't have. MapMyExam turns your item-analysis export into an Exam Health Report: which questions to fix first and why, a suggested rewrite for each, how the exam lines up with your objectives, and, when you upload the next version, which flagged questions were changed and which weren't. Every report is reviewed by hand before it reaches you. Exam Protection checks the web each month for your questions on study-sharing sites. A match shows where a question appears, not who posted it.
The goal isn't a harder exam. It's an exam whose score means what you need it to mean.