Alternative Explanations for Negative Findings in the Community Popular Opinion Leader Multisite Trial and Recommendations for Improvements of Health Interventions Through Social Network Analysis

Schneider, John A MD, MPH*; Laumann, Edward O PhD†

JAIDS Journal of Acquired Immune Deficiency Syndromes:
doi: 10.1097/QAI.0b013e318207a34c
Letter to the Editor
Author Information

*Departments of Medicine and Health StudiesUniversity of Chicago Chicago, IL; †Department of SociologyUniversity of Chicago Chicago, IL

Article Outline

To the Editors:

We read with great interest the report of the Results of the National Institute of Mental Health Collaborative HIV/Sexually Transmitted Disease Prevention Trial of a Community Popular Opinion Leader (C-POL) Intervention.1 This trial tested whether a well-described and effective community level intervention2 could be imported globally in diverse settings. Unfortunately, the community level intervention did not produce greater behavioral risk and disease incidence reduction than the comparison condition, and the authors suggest that this is likely due to the potent and unanticipated effect that the control condition offered in reducing behavioral risk and HIV/STI incidence. We raise a few aspects of the described study that might provide alternative explanations for the lack of effect observed, and in so doing provide recommendations for future HIV interventions. Although there were several settings with culturally diverse populations, we use wine shops in India as a case example, because of our experience in this setting, but more importantly, because this setting best highlights the plurality of points outlined below, several of which would also be pertinent to social mixing within other venues such as trade school dormitories in Russia or vendor markets in China.

First, C-POL interventions are most effective when directed to an identifiable target population in well-defined community venues and where the population's size can be estimated.2 Wine shops in Chennai, however, may not have been the best venue and wine shop patrons the best target population. Wine shops are well defined in the sense that they are easily identifiable and accessible to investigators. However, before this study, wine shops and wine shop patrons in India as a community were rather ill defined-that is, there are no ethnographic, descriptive, or other assessments of the social networks linked to such venues. Additionally, estimation of the wine shop patron population size has not been conducted and might have been feasible through capture recapture sampling that has been conducted in other difficult to identify populations in India.3 The authors state that venues for the study were chosen where high behavioral risk was identified through epidemiologic studies. In the case of wine shops, there are no such studies that assess behavior risk amongst wine shop patrons or the wine shop venue before the current study. This is in contrast to the initial venue/population that was initially proposed for India-young men in slums of Chennai.4 This setting did fulfill all 3 criteria suggested for a C-POL population/venue and the local partner had previous experience working with this group, however, actual prevalence and incidence of HIV/STIs in this group was lower than expected and determined through well conceived and carried out field estimations of relevant STIs in this population.5 Despite these limitations, wine shops do seem to be an important candidate venue for interventions and may be particularly attractive for combination HIV and alcohol abuse prevention interventions, however, perhaps not a C-POL based type of intervention.

Another challenge for vendor markets or wine shops as a C-POL intervention setting, but perhaps less so for more homogeneous trade school dormitories, is the heterogeneity of the population. C-POL has been found to be most effective in settings in the United States where homogeneous populations are targeted.6 Certainly, within a heterogeneous population, distinct strata (age, sexuality, gender) can be identified and C-POLs selected from these groups, however, the “target population” becomes a bit more diffuse and follows more nominative boundary specification principles that are set by the investigators a priori.7 Additionally, many of the patrons socializing at vendor markets or wine shops may be mobile, visit other similar settings, or be positioned peripherally within the network. Ethnographic methods, observation, and interviews of key informants are central to identifying and recruiting community popular opinion leaders, however, may not be appropriate for identifying complex and heterogeneous groups or individuals at the confluences of such groups.8,9 Formal baseline social network characterization and analysis can provide clear social structural characteristics for venues, and importantly, determine the position of group members that may be at the periphery, or bridging between components of heterogeneous networks, or between intervention and control arm.

There have been some suggestions from several scholars that contamination between condition and control arms might have limited the observed efficacy of the opinion leader intervention, at least in India, perhaps through message diffusion from opinion leaders to control condition community members. Such “bridging actors” are likely to be missed during preliminary work that characterizes the target population or networks to be studied through ethnography and observation alone. These bridging actors play important epidemiologic roles for connecting groups that may be dissimilar with respect to behavioral risk, HIV prevalence, or HIV virus type. Bridging actors may be in the strategic position to play an important role for fostering diffusion of information within networks and between groups. They may serve to be more efficient diffusion agents than opinion leaders because they have fewer relationships over which to persuade others10; they can devote more energy to persuading and hence be a more effective change agent. Additionally, bridging actors may be more receptive to behavior change as they have fewer direct contacts than opinion leaders and less pressure to support prevailing norms and behaviors11 or create a reputation cost for inappropriate behavior.12 Finally, occupying a bridging position may be indicative of attitudinal and behavioral dispositions such as being open to new ideas and practices.13,14 If the macro level of these C-POL venues were characterized, one might discover that an opinion leader within 1 venue might also be a bridging actor within the larger network to other groups/venues, and thus potentially strengthen the HIV prevention messages at these other venues.

In sum, we feel the methods used to identify C-POLs could have been strengthened through formal network characterization and that interpretation of the results would also have likely benefitted through such an approach. Perceptions are often that formal social network analysis is tedious and not feasible, however, this is generally not the case.14 Additionally, newer methods to rapidly assess social networks using digital communication technology are currently being tested in the field, which will make formal network characterization less onerous. Future interventions should not conclude that the C-POL approach is ineffective based on these results. Indeed, there is reason to be optimistic about the effectiveness of network-based interventions.

John A. Schneider, MD, MPH*

Edward O. Laumann, PhD†

*Departments of Medicine and Health StudiesUniversity of Chicago Chicago, IL

†Department of SociologyUniversity of Chicago Chicago, IL

Back to Top | Article Outline


1. NIMH. Results of the NIMH collaborative HIV/sexually transmitted disease prevention trial of a community popular opinion leader intervention. J Acquir Immune Defic Syndr. 2010;54:204-214.
2. Kelly JA. Popular opinion leaders and HIV prevention peer education: resolving discrepant findings, and implications for the development of effective community programmes. AIDS Care. 2004;16:139-150.
3. Vadivoo S, Gupte MD, Adhikary R, et al. Appropriateness and execution challenges of three formal size estimation methods for high-risk populations in India. AIDS. 2008;22(Suppl 5):S137-S148.
4. Solomon S, Kumarasamy N, Ganesh AK, et al. Prevalence and risk factors of HIV-1 and HIV-2 infection in urban and rural areas in Tamil Nadu, India. Int J STD AIDS. 1998;9:98-103.
5. NIMH, HIV/STD, Prevention, Working, Group. The community popular opinion leader HIV prevention programme: conceptual basis and intervention procedures. AIDS. 2007;21(Suppl 2):S59-S68.
6. Kelly JA, St Lawrence JS, Diaz YE, et al. HIV risk behavior reduction following intervention with key opinion leaders of population: an experimental analysis. Am J Public Health. 1991;81:168-171.
7. Laumann EO, Marsden PV, Prensky D. The boundary specification problem in network analysis. In: Freeman LC, White DR, Romney AK, eds. Research Methods in Social Network Analysis. Fairfax, VA: George Mason University Press; 1989.
8. Laumann EO, Gagnon JH, Michael RT, et al. Sexual Networks. The Social Organization of Sexuality. Chicago, IL: University of Chicago Press; 1994:225-268.
9. Laumann EO, Gagnon JH, Michael RT, et al. Epidemiological Implications of Sexual Networks. The Social Organization of Sexuality. Chicago, IL: University of Chicago Press; 1994:269-282.
10. Holme P, Ghoshal G. The diplomat's dilemma: maximal power for minimal effort in social networks. Understanding Complex Systems. 2009;DOI: 10.1007/978-3-642-01284-6_13 269-288.
11. Cancian F. The Innovator's Situation: Upper-Middle-Class Conservatism in Agricultural Communities. Palo Alto, CA: Stanford University Press; 1979.
12. Burt RS. Neighbor Networks. Oxford, United Kingdom: Oxford University Press; 2010.
13. Valente TW, Fujimoto K. Bridging: locating critical connectors in a network. Soc Networks. 2010;32:212-220.
14. Valente TW. Social Networks and Health. Models, Methods and Application. Oxford, United Kingdom: Oxford University Press; 2010.
© 2011 Lippincott Williams & Wilkins, Inc.