Picking the right chart, and spotting patterns and trends in epidemiologic data
A pile of numbers doesn't tell a story until it's displayed the right way. Every chart type has a job it's best at, and part of being a Disease Detective is matching the data to the display, then reading what it's actually saying.
Shows a trend in numbers or rates over time. An arithmetic-scale line graph works for most data; a semilogarithmic (semi-log) scale is used when values span two or more orders of magnitude.
👆 Tap for an exampleShows the frequency distribution of a continuous variable, like age or day of onset. An epi curve is a special histogram: case counts by date of onset.
👆 Tap for an exampleCompares the size of different categories of a single variable. Grouped or stacked bars can compare two or three variables at once.
👆 Tap for an exampleShows how the parts of a whole compare to each other: like the percentage of cases from each different symptom or exposure.
👆 Tap for an examplePinpoints the exact location of each case or event: useful for spotting geographic clustering.
👆 Tap for an exampleUses shading or color to display rates or numbers across whole regions, rather than individual points.
👆 Tap for an example👆 Tap a chart type above to see a Disease Detective example of it in action.
Read each scenario and pick the chart type that fits it best. There are six rounds: one for each chart type above.
A secular trend: a gradual rise or fall over years or decades, often reflecting slow changes like an aging population or better diagnostic tools.
A pie chart shows how suspected exposure routes break down as parts of one whole outbreak.
Beyond secular trends, watch for these recurring patterns in disease-frequency data over time:
Rises and falls on a predictable yearly cycle.
Repeats regularly, but over a period longer than one year.
One sharp peak from a single shared exposure (covered in Topic 2).
Tests hand you a real graph and ask several questions about it. Work through the same checks every time:
Strength: show a lot of data for many places at once, and make regional patterns easy to see. Weakness: one color per region hides differences inside it, the borders are arbitrary, and big regions draw the eye more than small crowded ones.
A stacked area or bar chart where each column adds to 100% shows how the mix changes over time, like the share of cases caused by each virus variant. Read one layer's thickness, not its top edge.
Early in an outbreak, case counts can grow exponentially: they double in a fixed time. Read the doubling time straight off the graph (25 → 50 → 100 every 4 days = a 4-day doubling time). Flattening the curve means slowing the spread (masks, distancing, isolation) so the doubling time gets longer and fewer people are sick at the same time, keeping hospitals from being overwhelmed.
A graph axis where each equal step stands for multiplying by 10 (1, 10, 100, 1,000…) instead of adding a fixed amount. It's used when values span a huge range, so a graph can show a rise from 2 cases to 2,000 cases on the same chart without squashing the small numbers flat.