Opinion leader interventions in social networks

Abstract

Can change HIV risk behaviour in high risk communities

Disadvantaged populations who have poor access to health care are particularly vulnerable to the effects of HIV and sexually transmitted diseases. They are less likely to be routinely tested for HIV and sexually transmitted diseases. If they test positive they are often unable to afford the required treatments. Efforts have been made to curtail the spread of HIV and sexually transmitted diseases in many populations. Interventions have evolved from traditional classroom-type presentations that focus on the individual to those that involve couples and aim to improve their communications. Other interventions have used street theatre and novellas broadcast on television or radio. Most of these interventions have had only modest effects on behavioural change.

A study in this week's BMJ by Kelly and colleagues reports the effects of a social network intervention designed to reduce risky behaviour that can lead to HIV and sexually transmitted diseases in a high risk population of Roma (Gypsy) men in Bulgaria.¹ In the intervention arm of the study, leaders of Roma men's social networks counselled their own network members about reducing the risk of HIV and sexually transmitted diseases. The study found that people receiving the intervention had lower rates of unprotected intercourse over 12 months and had higher scores on knowledge related to AIDS, attitudes, and motivations to change behaviour.

This study is important for several reasons. Firstly, by using social network nominations it is a more sophisticated form of the peer opinion leader model used to great effect in many other studies.^{2 3 4} It also shows that opinion leaders should be identified by interviews with members of the social network rather than through observations by staff carrying out the programme. Such a social network method is a valid and reliable way of identifying opinion leaders and also provides a standardised protocol that can be replicated.²

Secondly, the study shows that even in marginalised communities, who may distrust government and other institutions, peer opinion leaders and programme staff can build trust through mutual understanding and can produce positive outcomes. If the approach works here, it is likely to be feasible in other places and with other populations.⁵

Thirdly, the study showed effects across a range of outcomes, from improving knowledge of HIV and sexually transmitted diseases and attitudes towards them to more objective outcomes like reducing the incidence of these diseases in the long term. Thus, the behavioural change is likely to be self sustaining and self propagating. Indeed, effects are likely to be an underestimate as people in the intervention group may have shared knowledge with people outside the group, thereby diluting the effect.

Although other opinion leader interventions have been effective in both clinical and community settings,^{4 6} this one may have been more effective because of the small size of the networks. Leaders were identified within small groups so the intervention was essentially a peer leader network intervention. So while the effectiveness of leaders may vary according to relationships within networks (tie strength, closeness, etc), in this study the social distance between leaders and group members was small. In other applications, it may be necessary to use network data and algorithms that are designed to minimise social distances between leaders and group members.^{7 8}

The social network information (the interpersonal links between participants in the study) can also be used to estimate the effect of interventions and understand the mechanics of behavioural change. Diffusion of innovations and other behavioural change theories⁹ suggest that information is transmitted more effectively between people with strong social ties. Kelly and colleagues can test this theory by examining whether behavioural change varied according to the characteristics of the leaders, their participation in intervention training, and the strength of their relationship with participants. Of particular interest might be whether properties of network groups—such as density (the number of links) or centralisation (the degree to which those links are concentrated toward one or few people)—mediate leaders' effectiveness. Leaders in dense or centralised groups may have more power, influence, and control than those in sparse or decentralised groups.

So what are the challenges for future programmes? Scaling up of such interventions requires dedicated funding. Although the programme described by Kelly and colleagues reached hundreds of people at high risk, considerably more people are at risk.

A more widespread and cost effective method of communicating information on changing behaviour could be through the media, either by mass media broadcasting (television and radio) or the internet.¹⁰ However, it is unclear whether attempting to change behaviour via the internet is as effective as face to face interaction.¹¹ Although communication over the internet can be supplemented with extra material (references, links to testimonials), it is often unavailable for the most marginalised communities. Most people still prefer face to face interaction for behavioural change.⁹

Accumulated evidence from studies of behavioural change among doctors,¹² HIV and sexually transmitted diseases, tobacco use, and substance misuse suggest that network data (surveys measuring who is connected to who) can be used to promote behavioural change.

The Roma men in Kelly and colleagues study were given the tools to promote behavioural change within their communities, and they showed that they were willing and able to do so. Healthcare professionals do their best work when they help communities realise their own potential for change. This study provides encouraging evidence of how that can be done.