Learning how to read medical graphs starts with resisting the urge to pick the tallest bar. Suppose one clinic reports 30 patients with a headache at follow-up and another reports 20. The first clinic has more headache reports, but did a larger share of its patients report a headache? You cannot tell until you know how many patients each clinic assessed. On an exam, those are two different questions, and they can have opposite answers.
On scratch paper, try starting your answer with “The graph shows ___, measured in ___, for ___.” If you cannot fill in one of those blanks, return to the labels or caption before doing any arithmetic.
Imagine a fictional figure with two bar charts side by side. Both panels show Clinic A on the left and Clinic B on the right. Each patient assessed at a one-week follow-up answers the same question about whether they currently have a headache. Panel 1 has a vertical axis labeled “Patients reporting a headache,” with ticks from 0 to 40 in steps of 10. Panel 2 has a vertical axis labeled “Percent of assessed patients reporting a headache,” with ticks from 0% to 50% in steps of 10 percentage points. The table gives the bar heights and the totals from the figure caption.
| Clinic | Assessed patients | Panel 1 bar height | Panel 2 bar height |
|---|---|---|---|
| Clinic A | 150 | 30 patients | 20% |
| Clinic B | 50 | 20 patients | 40% |
In Panel 1, Clinic A has the taller bar: 30 reports compared with 20. In Panel 2, Clinic B has the taller bar: 40% compared with 20%. To see why, divide each count by its own clinic’s total. For Clinic A, (30 / 150) × 100 gives 20%. For Clinic B, (20 / 50) × 100 gives 40%. Clinic A assessed three times as many patients, so it could have more reports even though a smaller share reported a headache.
If an exam asks how many more patients reported a headache at Clinic A, subtract the counts: 30 minus 20 gives 10 patients. If it asks how much higher the percentage was at Clinic B, subtract the percentages: 40% minus 20% gives 20 percentage points. If it asks how many times as high Clinic B’s percentage was, divide 40% by 20% to get two. The wording tells you which calculation to use.
In the fictional figure, both vertical axes start at zero, but the tick marks mean different things. One tick interval is 10 patients in Panel 1 and 10 percentage points in Panel 2. Comparing the physical heights of bars across the panels tells you nothing about the size of the difference. Read each bar against its own axis.
A bar chart that starts above zero can make a small difference look much larger. Imagine bars for 48% and 50% on an axis that starts at 45%. The visible bars extend 3 and 5 percentage points above that baseline, even though the actual values are only 2 percentage points apart. A line chart may start above zero to make small changes easier to see. In either case, use the labeled values to measure the difference instead of relying on how dramatic the picture looks.
The headache figure shows one follow-up point, so it cannot show whether headaches became more or less common over time. It also says nothing about headache severity. And because the percentages use only assessed patients, you cannot assume they describe everyone who was expected to attend follow-up. If some patients missed the visit, the figure does not tell you how those patients would have answered.
A defensible answer is: “At the one-week follow-up, a larger share of assessed patients at Clinic B reported a headache than at Clinic A, 40% versus 20%.” That does not show that Clinic B caused more headaches or gave worse care. The figure does not tell you whether the patients differed before follow-up or whether they received the same care.
Try the four checks on a figure from the IMedC learning resources or while working through training and practice problems. Before reading the explanation, write one sentence naming the groups, the outcome, and the units. If the graph reports percentages, add the group totals beside your answer when they are available. Then underline any word that makes an extra claim, such as “improved,” “caused,” or “steadily.” Keep the word only if the figure and the information that comes with it support that claim.
Why can a higher count go with a lower percentage?
The group with the higher count may also have a larger total. Compare each count with its own group total to find the percentage.
Is 40% minus 20% a 20% increase?
No. The difference is 20 percentage points. Relative to 20%, going from 20% to 40% is a 100% increase.
Does every graph have to start at zero?
No. Bars generally should start at zero because their lengths represent values, but a line chart can use a narrower scale to show small changes clearly.
About the International Medicine Competition: an online medicine competition for high school students from around the world. Qualification Round 2027: submissions are open until Sunday, 7 February 2027. How the rounds work