Opening the Blackbox of Treatment Interference: Tracing Treatment Diffusion through Network Analysis

Area of Science:

• Social Epidemiology
• Causal Inference
• Network Analysis

Background:

• Treatment interference complicates causal inference in public health interventions.
• Previous studies often rely on strong assumptions about interference structures.
• Understanding how interventions spread (treatment diffusion) is key.

Purpose of the Study:

• To analyze treatment diffusion as a form of treatment interference in a school-based smoking prevention program.
• To model treatment diffusion networks and identify influencing factors.
• To provide empirical insights for causal inference and intervention design.

Main Methods:

• Analysis of data from 4,094 students across six middle schools in China.
• Measurement of treatment interference through tracking brochure sharing to construct treatment diffusion networks.
• Application of exponential random graph models to analyze network data.

Main Results:

• Identified key covariates and network processes significantly correlated with treatment diffusion.
• Provided an empirical basis for evaluating assumptions on treatment interference structures.
• Demonstrated the utility of network analysis for understanding intervention spread.

Conclusions:

• Findings inform data imputation for causal inference under treatment interference.
• The study offers insights for optimizing future intervention designs to enhance treatment diffusion.
• This research bridges the gap between theoretical causal inference and practical intervention implementation.