Historical measles data reveals early warning signal for outbreak size

Before vaccines arrived in Britain, measles moved through the population with a rhythm as predictable as the school calendar. Children got sick in waves, epidemics peaked and subsided, and then, a year or two later, another wave would crest. Researchers studying decades of case records from England and Wales between 1948 and 1968 have now found something hidden in that pattern: a signal that could tell us, a full year in advance, how severe the next outbreak will be.

The key lies in counting susceptible people. Every time a child is born, the pool of people who can catch measles grows. Every time someone recovers from the disease, that pool shrinks—measles infection confers lifelong immunity. The researchers built a mathematical model that reconstructed how many susceptible individuals existed at the start of each year, using nothing but birth records and weekly case reports. Then they asked a simple question: does that number predict how many people will get sick in the year ahead?

The answer was yes, and strongly so. The model, published in Scientific Reports, found that the reconstructed susceptible population at the start of a year showed a very strong relationship with the actual attack rate—the proportion of people infected—in the following year. This held true not just for England and Wales as a whole, but also when the researchers zoomed in on five major cities: London, Birmingham, Liverpool, Manchester, and Leeds. The relationship was clearest at the national level, but it persisted everywhere they looked.

The researchers tested five different ways of structuring their model and settled on one that balanced fit with simplicity. This winning version treated the basic reproduction number—how many people one infected person will infect—as constant across years and regions, while allowing the susceptible population and initial infections to vary annually. The model incorporated school holidays, when children scatter and transmission drops by roughly 20 to 33 percent. It used a generation-time distribution drawn from the literature, with a maximum infectious period of 21 days. When they compared what the model predicted to what actually happened in the historical records, the fit was strong across all six regions studied.

But the researchers are careful about what this means. The susceptible population was not measured directly; it was inferred from case counts and birth data, which means it carries the fingerprints of whatever biases and gaps existed in the original reporting. The model also had limitations: it was deterministic rather than stochastic, it lacked detailed age structure, and it did not capture the full complexity of social mixing patterns. In some cities—Liverpool, Manchester, and Leeds—the agreement between what the model reconstructed and what it could verify was less clean than in others.

What matters now is whether this approach can travel forward in time. The historical data came from an era when measles was nearly universal, when almost every child was infected by age 15, and when vaccination did not yet exist. Modern surveillance systems are different. Vaccination rates vary by region and community. Reporting practices have changed. The researchers acknowledge that extending this forecasting method to contemporary settings would require further methodological work. But the principle is clear: if you know how many susceptible people exist at the start of a year, you have a foothold for predicting what comes next. In a world where measles is resurging in under-vaccinated populations, that foothold could matter.