What is a data interpretation station?
In a Multiple Mini Interview (MMI) circuit you rotate through short stations. At a data station you are handed a table, graph or trial summary, then asked to explain what it shows and what it means for patients. Assessors are looking at structure, numeracy, honesty about uncertainty and clear communication. They do not expect medical knowledge you have not been taught.
The four-step framework (say it out loud)
Use the same order every time. It gives you something to say in the first ten seconds, which calms nerves.
Mock station: the Cardiostat trial
All data below is invented for practice. It is not from a real trial.
Candidate instructions
You are a medical student on a cardiology placement. A patient aged 62 with high blood pressure asks whether he should take a new tablet, Cardiostat, after reading a headline: "New pill cuts heart attacks by 30%." You have the summary table below.
Tasks: (1) Describe the results. (2) Explain whether the 30% figure is fair. (3) Explain the benefits and risks to the patient in plain English. (4) Say what else you would want to know.
| Measure | Placebo (n = 1,000) | Cardiostat (n = 1,000) |
|---|---|---|
| Mean age (years) | 61 | 62 |
| Heart attacks (number) | 80 | 56 |
| Dizziness needing review | 30 | 90 |
| Withdrew from trial | 50 | 120 |
| Deaths from any cause | 22 | 20 |
Reported by the trial team: difference in heart attack rate 2.4 percentage points, 95% confidence interval 0.2 to 4.6, p = 0.03.
Model answer and worked marking
Attempt the station aloud first. Then open each part to compare. Several sections can stay open at once so you can cross-reference.
Step 1: Orient (about 30 seconds)
"This is a randomised, double-blind trial comparing Cardiostat with placebo in 1,000 patients per group, with heart attack as the main outcome over two years. The groups look similar in age, 61 and 62, so the comparison seems fair on that measure."
Step 2: Describe (about 45 seconds)
"In the placebo group 80 of 1,000 had a heart attack, which is 8.0%. In the Cardiostat group 56 of 1,000 did, which is 5.6%. Dizziness needing review happened in 3% on placebo and 9% on Cardiostat. More people left the Cardiostat arm, 12% against 5%. Deaths were similar, 22 against 20."
Note: always state the denominator. "80 heart attacks" means little without "out of 1,000".
Step 3: Analyse (about 2 minutes)
- Absolute risk reduction (ARR) = 8.0% − 5.6% = 2.4 percentage points.
- Relative risk reduction (RRR) = 2.4 ÷ 8.0 = 30%. So the headline is arithmetically correct, but it describes the proportion of the placebo risk removed, not how many people are helped.
- Number needed to treat (NNT) = 1 ÷ 0.024 = 41.7, so about 42 people treated for two years for one to avoid a heart attack.
- Natural frequencies: "Out of 1,000 people like you, about 80 would have a heart attack without the tablet and about 56 with it, so roughly 24 fewer."
- Harms: dizziness rose by 6 percentage points (3% to 9%). Number needed to harm is 1 ÷ 0.06 = about 17. So more people are likely to be made dizzy than to be saved from a heart attack, although a heart attack is a far more serious outcome.
- Uncertainty: the 95% confidence interval for the ARR runs from 0.2 to 4.6 percentage points. It does not include zero, which fits p = 0.03, but the true benefit could be very small (0.2 points would mean an NNT of about 500) or fairly large.
- Drop-outs: 12% against 5% withdrew. If people left because of side effects, and their outcomes are not counted, the benefit could be over- or under-estimated. I would ask whether the analysis was intention-to-treat.
- Deaths: 22 against 20 is not meaningfully different, so there is no sign here that the tablet saves lives, and the trial is probably too small to tell.
Step 4: Conclude, with a patient-friendly sentence (about 1 minute)
"The headline of 30% is true as a relative figure, but the real-life difference is smaller: roughly 24 fewer heart attacks per 1,000 people over two years, with more dizziness. I would want to know how long the benefit lasts, whether the trial included people like this patient, who funded it, and whether the result has been repeated. I would not tell him what to do. I would explain these numbers, ask what matters most to him, and encourage a shared decision with his doctor."
Examiner follow-up questions to practise
"Would you prescribe it?"
As a student, you would not decide. Say so, then explain that the decision depends on the patient's other risks, preferences and a senior clinician's judgement. This tests professional boundaries, not pharmacology.
"Why is the relative figure used in headlines?"
It is a bigger number (30% versus 2.4 points), so it makes better news and marketing. It also ignores baseline risk. The same 30% RRR would mean far fewer people helped if the baseline risk were 1% instead of 8%.
"Does p = 0.03 prove the drug works?"
No. It means that if the drug truly had no effect, a difference this large or larger would occur about 3% of the time by chance. It does not give the chance the drug works, and it says nothing about the size or importance of the effect.
"What if the trial was funded by the manufacturer?"
Funding is a possible source of bias in design, reporting and which results are published. It does not make the results wrong. You would check for independent replication and a pre-registered protocol.
"The patient says: 'So it only helps 2 in 100?' How do you respond?"
Agree that is a fair summary, avoid being dismissive, and acknowledge that avoiding a heart attack matters a lot to some people. Offer to go through the harms too and invite his questions. This tests empathy and honesty.
Self-marking rubric (practice only)
| Theme | Strong (4-5) | Weak (1-2) |
|---|---|---|
| Structure | Clear order, signposts steps, finishes in time | Jumps about, runs out of time |
| Accuracy | Correct numbers with denominators and units | Misreads table, mixes % and counts |
| Insight | Absolute vs relative, harms, drop-outs, uncertainty | Only repeats numbers |
| Communication | Plain English, natural frequencies, checks understanding | Jargon, one-way delivery |
| Professionalism | Tentative, shared decision, knows limits | Gives orders or overclaims |
Quick risk calculator
Enter your own numbers (try the Cardiostat figures) to check ARR, RRR and NNT.
Seven common mistakes
- Quoting counts without denominators (80 against 56 of what?).
- Treating RRR as the whole story and ignoring absolute change.
- Ignoring harms and drop-outs because the headline is positive.
- Confusing percentages and percentage points (8.0% to 5.6% is 2.4 points, not 2.4%).
- Reading p-values as "probability the drug works".
- Overclaiming causation from a single study.
- Telling the patient what to do instead of supporting a shared decision.
Frequently asked questions
What is an MMI data interpretation station?
One timed station in a Multiple Mini Interview where you describe and interpret a table, graph or trial summary, then explain what it means for patients. It tests numeracy, structure and honest communication.
What is the difference between absolute and relative risk reduction?
ARR is the difference in event rates (8.0% minus 5.6% = 2.4 percentage points). RRR is that difference as a proportion of the control rate (2.4 ÷ 8.0 = 30%). RRR can sound larger and mislead if quoted alone.
How do you calculate the number needed to treat?
NNT = 1 ÷ ARR (as a decimal). With ARR 0.024, NNT is about 42. Round up to a whole person.
Do I need to do calculations without a calculator?
Often yes. Practise with simple percentages and fractions, and check each university's own guidance on what is allowed.
How long should I spend on a data station?
Formats vary. A common pattern is a couple of minutes to read and about five to seven minutes to discuss. Practise with a timer and confirm the format with your chosen medical schools.
This page is independent practice material written by a tutor. It is not produced or endorsed by any medical school, and all trial data is fictional.