How to: Predict vaccine coverage by modelling
Child on far right in Nigeria who contracted polio after his father prevented vaccination
LONDON, 22 August 2012 (IRIN) - Researchers in Canada are combining mathematics with “social learning” to predict how epidemics are affected by fears - usually unfounded - that a vaccine can harm, so as to improve the design of vaccination campaigns. Social learning - how behaviours are learnt and decisions are made in group settings - has long been examined in economics and now, increasingly, in healthcare.
“We’re giving 110-120 million courses of vaccine out every year around the globe, but how much do we know about why people take these up?” asked John Edmunds, of the London School of Hygiene and Tropical Medicine and contributor in a forthcoming book
on modelling and disease control. “How much do we invest in understanding what drives mothers to accept these vaccines, and sometimes not to?”
Payoffs and penalties
Chris Bauch and colleagues at the University of Guelph in Ontario, Canada, set out to create a mathematical model that would show how much a person’s decision to be vaccinated was influenced by disease prevalence, and how much by peer pressure.
They tested the model using data from a 1990s measles-mumps-rubella (MMR) vaccine scare and a 1970s pertussis (whooping cough) vaccine scare in England and Wales to see how well their model predicted vaccine coverage and disease outbreaks in those instances.
The researchers grouped people into “vaccinator” and “non-vaccinator” categories. The mathematical formula tried to calculate how people judged a vaccine's risks and rewards. The risk was the perceived chances of getting infected, multiplied by the cost of infection - the cost of medicine and doctors or clinic visits, being unable to work and perhaps losing income, and the discomfort of being ill. The perceived "payoff" was whether a person judged the vaccine would do more good than harm.
Social learning was included by measuring how often people switched from one group to the other, based on observing others’ vaccine decisions and whether their health improved.
Bauch reported that their model did well
in foretelling disease outbreaks and vaccination coverage for both MMR and pertussis.
When enough people in a community are vaccinated against a particular disease, outbreaks of that illness become rare, or no longer occur. Medical evidence has shown that this “herd immunity
” makes some people perceive certain vaccines as unnecessary or risky.
“When a disease becomes more rare, people begin to forget what it was like. As a result, they become less scared of the disease and against this backdrop - where the disease risk is perceived to be zero - even a small vaccine risk suddenly looks very large,” said Bauch.
His model assumes that low disease prevalence means less vaccine demand among people who feel protected by herd immunity. As more diseases are wiped out, the authors speculate that vaccine scares will become more common.
In 2003, polio immunization programmes were suspended for over a year in the northern Nigerian state of Kano
after religious leaders falsely told people that the vaccine could cause infertility, HIV and cancer.
Polio spread from Nigeria to 20 other African countries between 2006 and 2010, said a report
by the Independent Monitoring Board of the Global Polio Eradication Initiative in July 2011, which noted Kano as a “major worry” with “low routine immunization coverage”.
The model is still years away from having “predictive value” for vaccine campaigns, said Bauch. His team notes that the model needs data from the first years of a vaccine scare to predict what happens in subsequent years, and that “the model cannot predict when a vaccine scare will occur, since this presumably depends on...historical events”.
“It’s basic research and there are always surprises in basic research… It’s too early to base policy on [our model] in a very explicit way,” said Bauch. Data on vaccine coverage and subsequent disease outbreaks collected during vaccine campaigns could be useful to modellers.
“Modelling is useful in terms of trying to make public health decisions,” said Louise Ivers of the US-based international NGO, Partners in Health (PIH). “[It’s] very interesting - the addition of social learning into a model.”
PIH, in collaboration with the Haitian Ministry of Health and a local medical NGO, GHESKIO
, rolled out a cholera vaccine pilot project earlier this year that is evolving by means of ongoing data collection.
If models are to inform policy and guide vaccine campaigns, consistent real-time data collection and analysis are needed. Edmunds of the London School of Hygiene and Tropical Medicine noted: “Otherwise [policy decisions are] based on hunches - effectively, people doing models in their heads - and that’s how things have been traditionally been done.”