Can Question Bank Percent Correct Predict Board Failure Risk?

Yes. But not cleanly, and not by itself.

The data shows question bank percent correct is a useful risk signal for board failure. Lower scores generally mean higher risk. Higher scores generally mean lower risk. That part is real. What is wrong is the lazy leap people make from “my bank average is X” to “therefore I will definitely pass” or “therefore I am doomed.” That is not how prediction works.

A single percent correct is a noisy metric. Its value depends on:

• Exam type: specialty board exams behave differently than shelf-style tests or in-training exams.
• Timing: a 58% six weeks before the exam means something very different from a 58% three days before the exam.
• Study volume: 58% over 120 questions is weak data; 58% over 2,400 first-pass timed mixed questions is much stronger data.
• Question source: some banks are harder, some are inflated by repeats, and some reward pattern recognition more than true mastery.

I have seen residents panic over a low early average that later turned into a comfortable pass. I have also seen the opposite: someone sitting on a smug 72% built on repeated tutor-mode questions, then getting punished on a timed standardized exam. Same number. Very different risk.

So the right analytical frame is probability, not prophecy. You are not looking for a magical pass/fail cutoff. You are looking for:

1. Relative risk
2. Direction of change
3. How stable the number is
4. Whether other metrics agree

Used that way, question bank percent correct is valuable. Used as a standalone scoreboard, it is overrated. The data shows the metric works best as part of a layered model: percent correct, trend over time, number of questions completed, and correlation with full-length practice tests. That is how you estimate failure risk like an adult, not like someone checking horoscope signs in a dashboard.

What the Data Usually Shows About Percent Correct and Board Outcomes

The broad relationship is consistent: as question bank percent correct rises, the probability of passing usually rises as well. That is not controversial. Performance on board-style questions reflects content knowledge, pattern recognition, and test execution. Those are the same ingredients boards measure.

But correlation is not destiny. Plenty of residents with middling question bank scores pass because they improve late, test well under pressure, or use harder banks. Plenty with superficially strong scores underperform because their number was inflated by repeats, cherry-picked topics, or slow untimed review.

A practical way to think about percent correct is by bands, not fantasies of precision:

• Below 50%: high concern zone
• 50% to 59%: elevated risk, especially if late in prep
• 60% to 69%: borderline to reasonably reassuring depending on context
• 70%+: generally favorable, though still not immunity

The chart is illustrative, not universal. Different specialties and different banks shift the absolute numbers. Still, the pattern holds. A resident averaging in the 40s late in the cycle is carrying substantially more risk than one averaging in the 70s. The gap is not subtle.

The data also shows that question volume matters. Early averages are unstable because baseline knowledge is low and the sample is thin. If you answer 80 questions and get 52% correct, that tells me very little. If you answer 1,800 first-pass mixed questions and stay at 52%, that tells me a lot more, and none of it is comforting.

There is also bank-specific variation. A 62% on a notoriously difficult bank may place you in a stronger peer position than a 68% on an easier one. Specialty matters too. Internal medicine, surgery, pediatrics, psychiatry, and subspecialty boards all carry different content breadth, question styles, and score distributions. Raw percent correct is not portable across all contexts.

This is why early numbers mislead so many residents. You start weak. Everyone does. The question is not whether your first 200 questions are ugly. The question is whether the curve bends upward as your volume increases. That trend is what separates normal early struggle from real failure risk.

Why a Single Percent Correct Can Mislead

A single number can lie to you. Often.

The first reason is small sample size. Percentage accuracy stabilizes only after enough observations. With too few questions, randomness dominates. A lucky streak in topics you like, or a bad run in nephrology and biostats, can swing the average wildly. That is noise pretending to be insight.

The second problem is selection bias. Residents do this all the time:

• Repeat familiar questions
• Stay in tutor mode
• Choose favorite subjects first
• Skip mixed timed blocks
• Review explanations so thoroughly that recognition replaces reasoning

Then they stare at a polished score and feel reassured. Bad idea. Inflated percent correct without true transfer to unfamiliar timed questions does not reduce board failure risk nearly as much as people think.

The third issue is mode of use. First-pass timed mixed blocks are much more predictive than second-pass untimed subject-specific sets. Why? Because boards are timed, mixed, fatiguing, and indifferent to your comfort. Tutor mode is good for learning. It is weaker for prediction. Repeated questions are good for reinforcement. They are bad for pretending you are seeing the material fresh.

The fourth issue is the oldest trap in test prep: memorization masquerading as mastery. I have seen residents score very well on familiar item styles, then crater when the same concept is tested from a new angle. They knew the answer shape, not the underlying medicine. High percent correct can overstate readiness if performance falls apart on unfamiliar or timed blocks.

So if you are using one summary percentage to estimate failure risk, at least ask whether that number comes from a fair test of readiness. If it does not, it is cosmetic.

Better Predictors Than a Single Number: Trends, Percentile Context, and Practice Exam Correlation

If I had to choose between a resident with a flat 68% and one who climbed from 52% to 64% across six weeks of serious timed practice, I would rather bet on the second profile. The data shows trajectory matters. Improvement means knowledge gaps are closing and test adaptation is happening. A static score, especially after heavy question volume, can signal stalled learning.

That is why trend over time is more informative than a snapshot. Look for:

• Rising weekly first-pass performance
• Better scores on mixed blocks, not just subject subsets
• Reduced swings between blocks
• Improved timing and fewer rushed final questions

Consistency matters more than people admit. A resident scoring 63%, 64%, 65%, and 66% on successive timed mixed sets is often in better shape than someone bouncing between 55% and 78%. Lower variance usually means the floor is rising. And your floor is what protects you on exam day.

Percentile rank can also outperform raw percent correct. Why? Because it provides context. A 61% may sound mediocre, but if it places you near the 70th percentile on a hard bank, that is a different story. Conversely, a 70% that sits near the middle of a softer resource is less impressive than it looks. The data shows relative standing against comparable users often predicts outcomes better than naked accuracy.

Then there are full-length practice exams. These usually beat question bank averages for prediction because they better simulate the real event:

• Longer duration
• Better fatigue modeling
• More realistic pacing pressure
• Broader content sampling
• Less contamination from repeats and piecemeal studying

When question bank trend and practice exam scores agree, your estimate gets stronger. If both are improving, failure risk is usually falling. If both are weak, risk is real. If they disagree, I trust the more exam-like metric. That usually means the practice test.

Here is the hierarchy I use in real-world advising:

1. Full-length standardized practice exam performance
2. Trend in first-pass timed mixed question bank blocks
3. Percentile rank within the bank
4. Raw percent correct
5. Second-pass or tutor-mode performance

That hierarchy is not perfect, but it is honest. Raw percent correct belongs in the conversation. It should not run the conversation.

How to Use Percent Correct to Estimate Board Failure Risk in Practice

Here is the framework I actually recommend.

1) Start with the score band, but do not stop there

Use percent correct as an initial screening signal:

• Low-risk / reassuring: usually 70%+, especially with strong volume and stable practice test performance
• Borderline: usually 60% to 69%, where context determines whether this is acceptable or dangerous
• High concern: usually below 60%, especially close to the exam or after large first-pass volume

This is not a universal cutoff table. It is triage. A way to decide whether to look closer.

2) Check whether the sample is big enough to trust

Ask:

• Have you completed at least several hundred first-pass questions?
• Are the questions mixed and timed?
• Is this mostly first exposure, or are repeats driving the number?

As a rule, the data becomes more interpretable once volume is substantial. Below that, avoid dramatic conclusions.

3) Add trend direction

This is where many residents misread their own risk.

• Flat and low: bad sign. You need intervention.
• Declining: worse sign. Fatigue, poor retention, or weak strategy may be in play.
• Improving but still below target: often salvageable. Keep pushing, but verify with better testing.
• Stable and high: reassuring, though still confirm with a practice exam.

A 57% that is rising steadily may carry less practical danger than a 64% that has stalled for three weeks. The data shows movement matters.

4) Add practice exam correlation

Now build a more serious estimate.

If your question bank percent correct is:

• Low, and your practice exam is also low: risk is high.
• Low, but your practice exam is acceptable: your bank may be harsh, or your exam execution may be stronger than your study sets suggest.
• Moderate, with a solid practice exam: risk is often lower than the bank score alone implies.
• High, but your practice exam is weak: that is a red flag for inflated bank performance.

5) Trigger action early, not late

This is the part residents avoid because it feels bad.

If your percent correct is low or stagnant, act. Do not “stay the course” out of pride. That is how people fail exams they could have passed.

Practical interventions:

• Increase timed mixed blocks
• Stop leaning on repeated questions
• Audit weak domains by topic
• Add a full-length practice test within 1 to 2 weeks
• Build a remediation plan with faculty, chief residents, or a learning specialist if the trend remains poor

I have seen this happen repeatedly: someone sits at 54% for weeks, keeps reviewing explanations, feels busy, and mistakes effort for progress. Then the first real practice test lands and exposes everything. Busy is not the same as improving.

6) Use simple decision rules

If your score is:

• Below 55% and flat after substantial volume
  Treat this as serious failure risk. Intensify study, seek help, and consider whether exam timing needs reevaluation.

• 55% to 64% and improving
  This is a recoverable zone. Double down on high-yield weaknesses and confirm progress with a standardized practice exam.

• 60% to 69% but inconsistent
  Focus on stability. Random strong blocks do not save you. You need a higher floor.

• 70%+ with good volume and a supportive practice exam
  Risk is generally low. Maintain discipline. Do not get cute and coast.

Key Takeaways

• The data shows question bank percent correct is a useful screening metric, but it is not a standalone board failure predictor.
• Trend over time, completed question volume, and practice exam performance provide a more reliable risk estimate than any single percentage.
• Low percent correct should prompt action, not panic: increase timed mixed practice, identify weak domains, and reassess with higher-quality data.

What to Do Next

If you want to use your question bank score intelligently, do this this week:

1. Record your true first-pass timed mixed average.
2. Count how many questions that average is based on.
3. Plot the last 3 to 6 weeks of performance.
4. Take or schedule a full-length practice exam.
5. Compare trend, volume, and practice score together.
6. Intervene early if the data is flat or falling.

That is the clean answer. Percent correct can predict board failure risk. Just not alone, not blindly, and not nearly as well as residents pretend. The number matters. The pattern matters more.