Identifying Key Influencers using an Egocentric Network-based Randomized Design

TL;DR

This paper introduces a novel method to identify key influencers in social networks using an egocentric randomized trial design, enhancing the evaluation of spillover effects in behavioral health interventions.

Contribution

It develops a new testing approach, the Multiple Comparison with Best, to detect subgroups with the strongest spillover effects and provides power calculations for designing effective ENRT studies.

Findings

Identified subgroups with significant spillover effects in HIV prevention.

Demonstrated the method's application in a peer education program.

Provided guidelines for selecting influential individuals in network interventions.

Abstract

Behavioral health interventions, such as trainings or incentives, are implemented in settings where individuals are interconnected, and the intervention assigned to some individuals may also affect others within their network. Evaluating such interventions requires assessing both the effect of the intervention on those who receive it and the spillover effect on those connected to the treated individuals. With behavioral interventions, spillover effects can be heterogeneous in that certain individuals, due to their social connectedness and individual characteristics, are more likely to respond to the intervention and influence their peers' behaviors. Targeting these individuals can enhance the effectiveness of interventions in the population. In this paper, we focus on an Egocentric Network-based Randomized Trial (ENRT) design, wherein a set of index participants is recruited from the population and randomly assigned to the treatment group, while concurrently collecting outcome data on their nominated network members, who remina untreated. In such design, spillover effects on network members may vary depending on the characteristics of the index participant. Here, we develop a testing method, the Multiple Comparison with Best (MCB), to identify subgroups of index participants whose treatment exhibits the largest spillover effect on their network members. Power and sample size calculations are then provided to design ENRTs that can detect key influencers. The proposed methods are demonstrated in a study on network-based peer HIV prevention education program, providing insights into strategies for selecting peer educators in peer education interventions.