# Who you are, or who you’re with? Age predicts disease risk

By Juliet Lamb
Clemson University, USA

Picture 1 of 1

## We’re all unique… but how important is that when it comes to our chances of catching an infectious disease? According to scientists, the answer may lie in mathematics.

Meet Jenny, aged 20. A social butterfly, she moves from economics courses in the morning, to a restaurant job in the afternoon, to drinks with friends at night. In any given day she might have close contact with at least twenty people – hugging a friend, borrowing a pencil, or handing change to a customer.

Now meet Ben, also aged 20. He prefers to keep to himself or spend time in small groups. Between university courses, odd jobs as a computer programmer, and tennis games with a friend, he usually has close contact with only a few people each day.

Which of these two would you expect to be at greater risk of catching or transmitting an infectious disease? The answer might surprise you.

### The study

In their recent contribution to the journal PLOS Pathogens, a research team headed by epidemiologist Adam Kucharski of the London School of Hygiene and Tropical Medicine studied how age and social behaviour affected the spread of a 2009 influenza outbreak in Hong Kong. After collecting blood samples from a group of 762 individuals, they asked each participant to record his or her number of social contacts on a randomly-assigned day. Several months later, after the peak of the influenza pandemic, the researchers returned to collect blood samples from the same participants. They were looking for an increase in neutralizing antibodies, the compounds that cells produce during a disease attack to disarm unwanted invaders. A four-fold increase in antibodies between samples showed that the participant had recently contracted influenza.

Once the team had identified infected patients, they used a set of mathematical models to describe the relationship between infection risk, age, and social behaviour. Like a model railway, a mathematical model represents a simplified version of reality. Unlike a model railway, however, a mathematical model isn’t a physical object. Instead, it’s a series of equations that describe how certain variables, like age and social contacts, produce certain outcomes, like infection. A model generalises from a set of known data (here, influenza in Hong Kong) to find predictable patterns that can be applied to other, unknown situations.

The researchers compared three different ways of classifying participants: by number of contacts, by age, or by a combination of both. They then analysed which models best approximated the real data. If the strongest model included age alone, Jenny and Ben would be expected to have similar infection risk. If the strongest model was based only on social behaviour, Jenny would be at greater risk of infection. If both parameters were included, Jenny might be at greater risk than Ben but at less risk than a 35-year-old with similar social habits.

Surprisingly, age alone was the best predictor of infection risk, with infection rates highest among young children and middle-aged adults. Individual behaviour, whether alone or combined with age, actually weakened the model’s ability to predict infection. After slight adjustments – using only close contacts, and accounting for the probability of pre-existing immunity in older subjects – the team settled on a model that divided individuals into 20 age classes and averaged social behaviour across each age class. In other words, Jenny with 20 close contacts, and Ben with two, would both be assigned to an age group with equal risk and an average contact rate of 11. This grouping balanced parsimony (using as few groups as possible) with accuracy (replicating observed patterns).

### Why are these findings important?

The team’s results could help public health officials charged with maximising the effectiveness of limited resources during an epidemic. Rather than assessing transmission risk for each individual, which is costly and difficult, vaccination efforts could instead target certain age groups with predictably high levels of risky social contact, particularly young children and adults likely to be parents of young children. Identifying target groups helps to direct vaccines toward the people most likely to contribute to the spread of the epidemic, and to protect members of at-risk populations.

Of course, social behaviour in Hong Kong may not be representative of other societies, particularly ones with different age- and gender-specific social norms. The study also relied on one-day contact histories, which might change over time – for instance, in the case of students like Jenny or Ben, close contacts might increase during the academic year relative to summer months.

Nevertheless, this study provides a straightforward mathematical link between social behaviour and disease transmission that could easily be adapted for different societies or behavioural patterns. More importantly, it highlights predictable social patterns as a potential key to improving both the speed and effectiveness of response to future disease outbreaks.