Influenza is a global health problem that affects 3-5m people a year and causes fatalities among the very old, the very young, and those with existing medical conditions. The virus spreads through contact. When we look at people’s social networks, an inter-connected web emerges that can help disease spread.
People are also influenced by the others around them and imitation in a social network can lead to groups of people all acting in a similar way. While this behaviour has positive effects such as creating social bonds, in some cases it can have negative implications. If a group all choose not to vaccinate, for example, it not only puts these people at a greater risk of disease as it can quickly spread through their social cluster but also perpetuates illness in the general population.
Failing to take up vaccination is often levelled at so-called “antivaxers” but some obvious candidates also fail to get their jabs. Healthcare workers are especially at risk of illness and it’s important to keep these staff on the frontline in the event of an epidemic. But only half of these are vaccinated against flu despite the government wanting three quarters of NHS staff to be protected. So why aren’t even healthcare workers – who are surely more aware of the benefits of vaccination than most – persuaded of the benefits of the flu jab?
As medical students will one day be the NHS staff required to deal with any flu epidemic, we decided to look for patterns in the influenza vaccination of medical students at Lancaster Medical School and the effects of this on influenza outbreaks, which we published in The Lancet.
Tapping into the network
We collected social network data by asking 253 students in our medical school to rate the strength of their relationship with all the other students. These ratings ranged from not recognising names, to living, having or socialising with other named people on a regular basis. We also asked the students about their vaccination habits.
We used the data to test whether vaccinated students were more likely to spend time with other vaccinated students – even if they didn’t realise they were doing so. We found that students did not show signs of clustering according to their vaccination status. Students were randomly mixed regardless of their seasonal influenza vaccination status. This is beneficial because it would provide a natural defence against the spread in infection by breaking up transmission pathways. For example, a 2008 measles outbreak in San Diego, US, in a highly vaccinated population was attributed to the intentional non-vaccination of clusters of individuals.
Using the data we developed a model to simulate the spread of influenza through the medical student social network. We used this model to assess the effects of preferentially vaccinating according to the social network analysis data.
In network analysis there are multiple ways of measuring well-connected individuals. We wanted to test whether vaccinating the most well-connected students would reduce the spread of influenza in our network. We also wanted to know which measure of connectivity was the best for doing this.
We used our model to simulate the spread of influenza in the network under the following conditions: when 20 extra students were vaccinated but we selected these at random; when 20 extra students were vaccinated that had a high degree of connections to the most other students; and when 20 extra “gatekeeper” students were vaccinated (located between many groups of students).
After running our simulation model 1,500 times for each test we discovered that vaccinating the most connected students was much better than randomly vaccinating students. So we suggest that vaccination strategies should be targeted – and we hope to develop tools to do this.