Rapid sociometric mapping of community health workers to identify opinion leaders using an SMS platform: a short report

Using opinion leaders to accelerate the dissemination of novel, innovative, or evidence-based practices in health care is a promising strategy for closing the gap between evidence and practice. Sociologists have observed for a generation that new ideas and practices—from adoption of contraceptives to new farming practices—spread through social networks [1]. Social networks also play a role in the spread of behaviors among health care workers. A seminal study in diffusion science conducted in the 1950s and still cited today found that opinion leaders in a community of physicians drove uptake of a new antibiotic [2]. Since then, numerous studies have documented that social forces can promote uptake of novel diagnostic and treatment strategies [3–6]. In a Cochrane review of 18 randomized trials, the use of opinion leaders to improve the behavior of health workers led to an average of 12% increase in compliance with desired behavior as compared to no use of an opinion leader [7].

Network interventions, defined by Valente as “the process of using social network data to accelerate behavior change or improve organizational performance” (e.g., the use of opinion leaders to catalyze dissemination) are promising [8]. Research to test network-based interventions could potentially play an important role in the global treatment response to HIV [9, 10]. Yet, socio-metric surveys can be difficult to carry out repeatedly in all settings where an intervention is desired. Therefore, despite their promise, the feasibility of network-based interventions in resource poor settings is unclear.

Use of popular opinion leaders and sociometric peer leader selection strategies may improve service provider engagement in HIV care [11, 12]. In addition, the rapid increase in electronic communication and social media, especially mHealth applications, could enable more widespread use of network interventions [13]. In this short report, we describe a case study of application of mobile phone technology to map the network of social influence and identify opinion leaders with regard to professional behavior among a cadre of community health workers in Kenya where mobile phone networks are robust and cell phone ownership is common. This sociometric survey was carried out as part of a project that sought to promote more personalized counseling practices among peer counselors supporting people living with HIV. The larger project (not reported in this Short Report) will evaluate the effect of using this mobile phone-based sociometric survey on adoption of more personalized counseling practices among identified opinion leaders and on patient outcomes such as retention in care and HIV virologic suppression. This Short Report is focused on healthcare worker behavior that could potentially be addressed through opinion leaders.

Leveraging the increasing access to SMS technology, we mapped the network of influence in a cadre of peer health workers associated with a HIV care and treatment program in Kenya. Using mobile phones, we identified opinion leaders by carrying out a sociometric survey in Kenya over the course of 3 days and at a cost of 8.46 USD.

This work represents a novel application of mobile phone technology to rapidly reveal network characteristics and identify influential individuals in a cadre of peer health workers, with direct implications for network interventions in global health. To our knowledge, this is the first ever application of mobile phone technology to do a sociometric survey to identify health care worker opinion leaders in Africa. Mobile phone penetration into the general public has exceeded 80% in Kenya, and in the health care workforce, it is estimated to be 100%, with 99% of health workers using SMS [17]. Indeed, despite the fact that peer health workers are a relatively low-paid cadre, all peer health workers included in this study had a personal mobile phone. Overall, 97% of participants completed the survey—indicating high acceptance and feasibility. The rapidity of the exercise demonstrates the relative ease of a survey delivered over the mobile phone platform, which allows a large number of respondents to be reached and respond despite being geographically dispersed. This is especially relevant for health workers in resource-limited settings where health facilities are widely dispersed and the terrain often difficult to navigate to bring groups of people together. The short time taken also meant that the survey could be conducted without placing an extra time burden on the already strained system.

Efficient mapping of social networks opens the door to reproducible, feasible, and efficient empirically based network interventions that seek to spread novel practices and behaviors among health care workers [9]. We found clearly influential individuals within a network of peer health workers in Kenya. The existence of influential persons, if widespread, can potentially be leveraged to catalyze dissemination of novel practices. Opinion leaders have been used in many settings as agents of behavior change in their communities. In this survey, we found a clear hierarchy of influence as measured by eigenvector centrality. This asymmetry indicates that it may be possible to target specific individuals to achieve change on a larger scale. Interventions can be more efficient if training, teaching, or capacitation is targeted toward a few individuals. Interventions can be more effective if the leaders of change will bring the community along with them through informal leadership. Of note, the persons of influence did not always coincide with positions of official authority—of the top five persons identified, only two held positions of overall authority, suggesting that informal network relations are crucial and must be measured empirically to be identified. Administrative or formal authority is an imperfect proxy of real influence in the network of health workers in Kenya.

This study had a number of limitations. First, the network mapping was not comprehensive in that we limited responses to only one “influencer” per topic. It would have been feasible to include more than one influencer and thereby have a denser network, but we sought to restrict our analysis to identify the top most influential individual. Second, a number of individuals reported no one as the answer to the questions posed. This could be interpreted in a number of ways. For example, it may indicate that they are otherwise confident in their own abilities to tackle the problem or it may also mean that the person they would turn to was not among the group of peer health workers included in the study. Third, the results on feasibility of the approach may not apply to other cadres and other settings. In our context, the peer health workers were paid by a PEPFAR-funded program and were approached following a training supported by a research study. Fourth, while the survey questionnaire was delivered using SMS, the survey respondents still met physically at one place and a paper list was used with names and codes to aid survey completion. We could potentially have delivered the entire survey using SMS without having the respondents meet physically. Of note, peer health workers in this region have periodic in-person meetings (sometimes as frequent as weekly) to review work plans and share experiences. We leveraged these already existing meetings to achieve high survey response and completion rates. However, we believe that the questionnaire can be delivered entirely using mobile phone technology without a physical meeting. Finally, we used a convenience sample, which may lead to biased responses.

In conclusion, we demonstrate the use of SMS technology using open-source software to carry out a rapid sociometric survey in Kenya among a cadre of community health workers. The rapid rise in access to mobile technology in resource-limited settings can be leveraged to efficiently conduct sociometric mapping among health workers at low cost.