Identifying the Impact of Social Influences in Health-related Discrete Choice Experiments

"[T]he maternal decision for childhood vaccination was essentially almost completely socially driven, suggesting that the potential impact of social network influences can and should be considered..., particular[ly]...where there are likely to be strong underlying social norms dictating decision maker behaviour."

Social context is important to individual choices. However, relatively little research has been focused on integrating the effect of social influence into choice models, especially given the importance of such choices in healthcare. This study developed and empirically tested a choice model that accounts for social network influences in a discrete choice experiment (DCE). It focuses on maternal choices for childhood vaccination in Australia, using a model that (i) incorporates vaccine schedule characteristics, benefits, and costs and (ii) represents up to ten different identifiable key influencer types (e.g., partner, parents, friends, healthcare professionals), allowing for the attribution of directional importance of each influencer on the woman's decision to adhere to or reject childhood vaccination.

The context of the study is that, in Australia: (i) subjects' choices for childhood vaccination are likely highly sensitive to social influences due to the strong underlying social norm (including Australian policy response to vaccine hesitancy); (ii) declines in childhood vaccination rates in Australia (and worldwide) have received media coverage in the last few years, contributing to the relevance of the study; and (iii) understanding broader influences on vaccination decisions may help determine targeting of appropriate education, behaviour change strategies, and/or policy.

Pregnant women (N = 604) aged 18 years and older completed a survey, including a DCE and questions about key influencers. A two-class ordered latent class model was conducted to analyse the DCE data, which assumes that the underlying latent driver (in this case, the World Health Organization (WHO) vaccine hesitancy scale) is ordered, to give a practical interpretation of the meaning of the classes.

Seventy-eight percent of the respondents only opted for the recommended schedule, whereas less than 3% of the respondents only opted for the delayed schedule or only opted out for vaccination. Essentially, 117 out of 604 respondents (19.4%) could be persuaded to trade between the alternatives. When the model considered both childhood vaccination attributes and key influencers, a very high model fit was reached. That is, the impact of key influencers on maternal choice for childhood vaccination was was massive (in magnitude and significance) compared to the impact of childhood vaccination attributes. For example, if the respondent believes that a friend with children would recommend vaccination, she would be more likely to vaccinate; this influencer's impact would be largely mitigated if the respondent is deemed more knowledgeable about vaccination, but the impact would be increased if the respondent is deemed more susceptible to social influences. This result suggests that for these respondents and this choice context, the key influencers were so important in making their choices in the DCE that the decision for childhood vaccination can be safely interpreted as being purely socially driven.

In terms of types of key influencers, the partner of the pregnant woman had by far the largest impact on the maternal choice for childhood vaccination, followed by the woman's mother, and thirdly, friends with children. This result was partly confirmed by the direct survey question regarding how important the view of a certain key influencer was (convergent validity), where the partner of the pregnant woman was also mentioned to be the most important key influencer. Respondents mentioned that all key influencers would likely (80% or more) recommend the government recommended childhood vaccination schedule profile. Except for the partner, the pregnant woman's utility for government-recommended childhood vaccination schedule profile changed a little bit due to the level of support for the government schedule she perceived from a certain key influencer. However, if there was a high consensus between key influencers (i.e., a high group support), the odds ratio impact for government recommended childhood vaccination schedule profile increased substantially and showed the value of social consensus.

Thus, these results "suggest the existence of healthcare decisions (this first-year childhood vaccination schedule among them) which may be wholly driven by social influences rather than service attributes. Hence, to increase the uptake of the childhood vaccination it is more effective to convince the social network than to reduce costs or clinic visits, for example. The influence of a policymaker via the social network will work very strongly if one can reach and convince the partner and mother of the gravid woman as these are the two persons who are most efficacious increasing the mother's likelihood to undertake childhood vaccination in Australia."