Despite the fact that the fraudulent (and now retracted) Wakefield paper from 1998 has been soundly refuted by modern medicine, vaccines are still socially controversial among some parents and communities. This has led to an increase in the number of vaccine-exemptions, delays in vaccinations, the erosion of herd immunity, and outbreaks of vaccine-preventable diseases. How can scientists, medical doctors and scientific skeptics effectively respond to parents hesitant about vaccines?
How do you convince parents who are hesitant to vaccine their children? Do you debunk common myths about vaccines causing autism with facts? Do you explain that vaccine-preventable diseases are dangerous? Do you show them graphic imagery of children suffering from vaccine-preventable diseases? Do you present a gripping narrative about a child who almost died from a vaccine-preventable disease like measles? As an added complications, some effort to correct irrational beliefs are counterproductive. Due to various backfire effects, correction can actually increase confidence in mistaken beliefs. Therefore, it is enormously important to research effective strategies for vaccine promotion.
A recent paper by Nyhan, Riefler, Richey and Freed (2014) examined this issue by randomly assigning a nationally representative sample of parents to one of the four interventions above or to a no-information control group for comparison. First, parents were asked to complete a pre-intervention baseline survey on child health, vaccine attitude, trust in medical authorities and child vaccine coverage. After they had been exposed to their respective interventions, they were asked questions to gauge their level of vaccine misinformation about side-effects of the MMR vaccine as well as their level of intention to vaccinate their children.
More specifically, the four interventions were as follows:
(1) “correction”: presenting evidence debunking the alleged link between MMR/autism.
(2) “disease risk”: describing symptoms and complications associated with measles, mumps and rubella.
(3) “disease narrative”: providing parents with a narrative of a mother nearly losing her child to measles.
(4) “disease imagery”: showing graphic pictures of children with measles, mumps or rubella.
The control condition involved information on the pros and cons of bird-feeding.
The correction intervention lead to a considerable reduction in the belief that vaccines can cause autism (adjusted odds ratio 0.55, 0.38-0.79 95% CI), but a much smaller reduction when it comes to parental concerns about side-effects (adjusted odds ratio 0.81, 0.57-1.15 95% CI). The researchers dismiss this second reduction as “not significant”. However, a lack of statistical significance does not mean equivalence. It just means that the probability of obtaining at least as extreme results (given the null hypothesis) is not as unlikely as one would wish. The error bar is only slightly above 1 (1.15). The effect size is 0.81 which represents a moderate decrease, although the data is consistent with no difference or a slight increase.
The disease risk intervention had an adjusted odds ratio near 1 for both the belief that vaccines caused autism and the MMR side-effects question (1.15 and 0.93) with 95% CI error bars extending too far in both directions to say anything definitive from these results.
The disease image and the disease narrative approach were both counterproductive. It increased both the acceptance of the vaccines-cause-autism myth (adjusted odds ratio 1.47, 1.02-2.13 95% CI; adjusted odds ratio 1.35, 0.91-2.01 respectively) and concerns about side-effects (adjusted odds ratio 1.18, 0.82-1.69 95% CI; adjusted odds ratio 1.92, 1.33-2.77 95% CI). Some of these reached statistical significance, some did not. However, the consistent effect sizes and the fact 95% CIs stretch far above 1 makes it clear that these two approaches are harmful.
Surprisingly, receiving the correction intervention made parents less likely to intend to vaccinate their children. This reduction was centered in the subgroup of parents that were least favorable to vaccines. For parents that were somewhat favorable or most favorable, correction was effective increasing intent to vaccinate. The 95% CI error bars were too wide to make any pronouncements on the other three interventions.
There are a number of limitations to this research project. The sample size was large (N = 1759), but the proportion that finished all the requirements was only moderate (1759 / 4462 = ~40%). The messages tested were taken almost word-for-word from the CDC. It is possible that other ways of phrasing or delivering the same message could be more effective. The source of the message was not given to parents, so it may be the case that mentioning that the information is taken from the CDC can moderate the observed differences in any direction. Another issue was that this study did not measure actual vaccines given, only the intent to vaccinate.
This study reached several important conclusions:
—> Messages designed to promote vaccines must be well-tested before deployed. Otherwise, it may be counterproductive and increase misconceptions about vaccines.
—> Factual corrections and attempting to scare parents with images and narratives might make vaccine hesitant parents more hesitant.
—> It is important to measure both belief and intent when it comes to health interventions. Otherwise one might miss situations when strength in belief is reduced, but so is intent.
—> None of the interventions tested increases the intent to vaccinate.
It seems that denialists are extremely difficult to convince. They are trapped in a distorted world were information against their beliefs only strengthen their resolve.
Nyhan, B., Reifler, J., Richey, S., & Freed, G. (2014). Effective Messages in Vaccine Promotion: A Randomized Trial PEDIATRICS, 133 (4) DOI: 10.1542/peds.2013-2365