Causal Estimation of Intervention Spillover Effects in Egocentric Network-Randomized Trials in the Presence of Network Misclassification

• Activity Number: 528 - Advances in Social Network Analysis for Public Health Solutions
• Type: Invited
• Date/Time: Thursday, August 11, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ariel Chao* and Laura Forastiere and Ashley Buchanan and Sten Vermund and Donna Spiegelman
• Companies: Yale School of Public Health and Yale School of Public Health and University of Rhode Island and Yale School of Public Health and Yale School of Public Health
• Keywords: Causal inference; Measurement error; Misclassification; HIV; Networks; Spillover effect

Abstract

Many HIV prevention trials are network- or cluster-randomized, where interference is often present. The resulting spillover effect is a causal effect of interest as it measures the impact of the intervention on study participants who do not directly receive treatment, but whose outcomes can be affected by others who do. The assessment of the spillover effect relies on the specification of interference sets; however, in the presence of mismeasured interference sets, causal effects estimated by existing approaches may be biased. As treatment contamination is often of concern in cluster-randomized trials, such as in HIV prevention trials where risk networks play an important role in disease transmission, correcting effect estimates for bias due to misclassified networks is critical when quantifying spillover effects. We propose bias-correction methods for the estimation of spillover effects in the presence of network misclassification, when a surrogate network is observed in place of a true network, and validation data that relate the misclassified to the true networks are available, under an egocentric network-randomized study design.