Optimizing peer-led health interventions: A social network analysis approach for identifying influential fishermen in Malawi

Social network interventions (SNIs) can leverage influential individuals within personal networks to amplify health behavior adoption and intervention diffusion. While SNIs are potentially effective in enhancing uptake of HIV prevention and other interventions, identifying optimal peer leaders remains challenging. This study assessed how social network-based approaches can improve peer leader identification compared to non-structural approaches (i.e., those that do not use network data for promoter selection). Between 8 th October 2024 and 31 st March 2025, we mapped the social connections among fishermen in two communities in Mangochi, Malawi. The communities had previously participated in a cluster-randomized trial (FISH), which aimed to create demand among fishermen for HIV and schistosomiasis services via peer-nominated leaders. Using Network Canvas and a photographic census, we conducted a network survey, capturing ties, support roles and interaction frequency. Whole-network maps were constructed, centrality measures and the key player problem positive (KPP-POS) algorithm were applied to identify the highest-ranking individuals as potential promoters. We compared the reach (i.e., proportion of nodes within two steps of a promoter) of peer-nominated leaders selected by FISH to these sociometric methods . We recruited 370 of 397 eligible fishermen (mean age 34.8years [SD 13.11]). Both communities exhibited sparse, low-reciprocity networks with long path lengths. One network had two dense cores, while the other featured a single core. There was no evidence of assortative mixing by education, age or village of residence. FISH trial peer leaders were not consistently central in the constructed maps. The KPP-POS algorithm identified alternative, more dispersed nodes, achieving the highest reach (66% in F001; 75% in F024) versus in-degree (58%, 67%), closeness (58%, 68%), betweenness (58%, 69%) and eigenvector (60%, 68%) promoter sets. Our findings highlight that strategically identified promoters can achieve good reach within social networks, crucial for effective public health programming in complex settings like fishing communities.