Lesson 11 of 11 · 20 min
Analysing outbreak data
Analyse epidemic curves, attack rates and case maps to predict outbreaks, find the source, infer the mode of transmission and judge how well a control strategy worked, and see how outbreak models guide vaccination.
What you will be able to do
- Identify point-source, continuous-source and propagated outbreaks from epidemic curves.
- Calculate and compare attack rates to determine the source of an outbreak.
- Infer the mode of transmission from the pattern of cases.
- Use data to judge the effectiveness of a control strategy, including vaccine effectiveness.
- Explain how outbreak models, built by people from many disciplines, make vaccination programs more successful.
When an outbreak begins, public health teams have the same questions every time: Where did it come from? How is it spreading? Will it get bigger? Is what we are doing working? Each is answered by comparing groups of people in the data, and the syllabus asks you to do the same.
Definition
Epidemic curves
An epidemic curve is a histogram of the number of new cases by date (or hour) of onset of symptoms. Its shape suggests how the outbreak is spreading:
- Point source: everyone was exposed at one time and place (a contaminated meal). Cases rise sharply and fall within about one incubation period.
- Continuous common source: exposure goes on (a contaminated water supply). Cases rise and stay high until the source is removed.
- Propagated: spread from person to person. Waves of cases about one incubation period apart, often growing at first.
Cases of gastroenteritis by time of onset after a school camp lunch.
What kind of outbreak does the epidemic curve show?
Key idea
Finding the source with attack rates
For each food (or other exposure), compare the attack rate, the percentage who became ill, in people who were exposed and people who were not:
The source is the exposure with a high attack rate among those exposed, a low attack rate among those not exposed, and so the biggest difference (or ratio) between the two. It should also explain most of the cases.
Worked example
Find the source of a food-poisoning outbreak
Of 90 students and staff at a school camp, 54 developed gastroenteritis after lunch (the epidemic curve in the lesson). Investigators asked everyone what they ate.
| Food | Ate: ill | Ate: total | Did not eat: ill | Did not eat: total |
|---|---|---|---|---|
| Chicken salad | 50 | 60 | 4 | 30 |
| Fried rice | 42 | 70 | 12 | 20 |
| Fruit salad | 26 | 45 | 28 | 45 |
| Juice | 49 | 80 | 5 | 10 |
(a) Calculate the attack rate for each group. (b) Determine the most likely source, with reasons. (c) Explain why the four people who did not eat chicken salad but became ill do not rule it out.
Step 1: Attack rates
Food Ate Did not eat Chicken salad Fried rice Fruit salad Juice Step 2: Compare
Only chicken salad shows a big difference: of those who ate it became ill against of those who did not, about times the risk. For the other foods the attack rates are similar whether or not people ate them.
A food eaten by many of the ill people is not enough on its own: juice was drunk by almost everyone, so it is common among the ill too. What counts is the difference between eaters and non-eaters.
Step 3: Coverage of cases
Chicken salad accounts for of the cases ().
Step 4: The four exceptions
They may have tasted it without remembering, eaten food contaminated from it (shared utensils or serving spoons), caught it from an ill person, or had gastroenteritis from another cause. Recall surveys are never perfect.
Try it yourself first, then check the working one step at a time.
Answer
(a) Chicken salad 83 % vs 13 %; fried rice 60 % vs 60 %; fruit salad 58 % vs 62 %; juice 61 % vs 50 %. (b) Chicken salad: the highest attack rate in eaters, a low rate in non-eaters, and it explains 50 of 54 cases. (c) Imperfect recall, cross-contamination, secondary spread or another cause.
Method
Attack rate = ill ÷ total in the group × 100, for eaters and non-eaters of each food.
The source has a high rate in eaters, a low rate in non-eaters, and explains most cases.
Watch out for
Choosing the food with the most ill eaters (juice, 49) without comparing with the non-eaters.
Comparing numbers of cases instead of rates when the groups differ in size.
Use the map, the case table and the epidemic curve to identify the source and the mode of transmission. Then try the food, vector and contact outbreaks.
History
John Snow and the Broad Street pump
In late August 1854 cholera killed more than 500 people within ten days in Soho, London. Most doctors blamed "miasma", bad air. The physician John Snow marked each death on a street map and found them clustered around the public water pump in Broad Street. The exceptions fitted too: workers at a nearby brewery, who drank beer and water from the brewery's own well, mostly escaped, and some victims who lived further away were found to have fetched water from Broad Street because they preferred its taste. Snow persuaded the local authorities to remove the pump's handle on 8 September 1854. The source (one pump) and the mode (contaminated water) were inferred from the pattern of cases, nearly 30 years before Robert Koch identified Vibrio cholerae as the cause.
| Pattern in the data | Suggests |
|---|---|
| Cases cluster around one water supply; people with a different supply are not ill | Contaminated water |
| Cases share one meal or food; attack rate highest in those who ate it | Contaminated food |
| Cases rise in the wet season, in areas with standing water, and in people who were outdoors at dusk; no spread within households | A vector such as a mosquito |
| Successive waves of cases within households, classrooms and workplaces | Direct contact or droplets, person to person |
| Cases mostly among people who shared needles, had transfusions or had sexual contact with a case | Body fluids |
After heavy summer rain in a coastal Queensland town, cases of a virus disease with joint pain and fever rise over six weeks. Cases are spread across the town, are highest among people who work outdoors in the evenings, and rarely occur in the household contacts of cases. Infer the most likely mode of transmission and justify it.
Key idea
Is the strategy working?
To judge a control measure, compare cases before and after it, or in a population that had it and a similar one that did not, over the same period. Use the percentage change in cases, and look for other explanations (the season changing, more or less testing, immunity from earlier infection).
For a vaccine, compare attack rates in vaccinated (ARV) and unvaccinated (ARU) people in the same outbreak:
Like the herd immunity threshold, this formula is not printed in the syllabus, so a task that needs it should give it to you. It measures how much the vaccine cut the risk of illness: means vaccinated people had one-tenth the attack rate of unvaccinated people.
In a measles outbreak at a school, 4 of 400 vaccinated students and 30 of 50 unvaccinated students caught measles. Calculate the vaccine effectiveness as a percentage.
Key idea
Predicting outbreaks
Outbreaks are predicted from surveillance data: notifications of cases, laboratory tests, and early signals such as virus in wastewater, mosquito trapping and rainfall, or rising emergency department visits. A sustained rise in weekly cases (or a reproduction number above 1), seasonal patterns (influenza in winter, Ross River virus in wet summers) and falling immunity (a drop in vaccination coverage) all point to an outbreak ahead.
Weekly cases are shown against a baseline: the mean of past years plus two standard deviations. Decide in which week an outbreak is starting.
History
Models guide vaccination
In 1927 William Kermack, a chemist, and Anderson McKendrick, a military physician turned mathematician, published the SIR model, which divides a population into Susceptible, Infectious and Recovered (immune) people and predicts how an epidemic rises and falls. Their result that an epidemic cannot take off once enough people are immune is the herd immunity threshold you have used.
Models like this make mass vaccination programs more successful because they predict how much coverage is needed, which groups to vaccinate first (those who spread the most, or those most at risk), and when. Smallpox eradication switched from mass vaccination to ring vaccination around every case after field data from west Africa in the late 1960s showed that it stopped outbreaks with far fewer doses. In 2021 modelling by the Doherty Institute for Australia's National Cabinet linked coverage targets of 70 % and 80 % of people aged 16 and over fully vaccinated against COVID-19 to how much the public health measures could be eased.
No one person builds such a model. It takes epidemiologists, mathematicians and statisticians, virologists and immunologists who measure how a pathogen behaves, clinicians who supply case data, computer scientists, and social scientists who know how people actually mix and travel.
Summary
Epidemic curves: one sharp peak means a point source; a plateau, a continuing source; successive waves, person-to-person spread.
The source is the exposure with a high attack rate in those exposed and a low one in those not exposed.
The mode of transmission is inferred from who becomes ill, where, when and who does not.
Judge a strategy by comparing rates before and after, or with and without it, and consider other explanations; vaccine effectiveness = (ARU − ARV) ÷ ARU × 100.
Outbreak models, built by many disciplines, set vaccination targets, priorities and timing.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.