Design of egocentric network-based studies to estimate causal effects under interference

Area of Science:

• Public Health Research
• Epidemiology
• Biostatistics

Background:

• Public health interventions often occur in connected populations, leading to potential spillover effects.
• Assessing individual, spillover, and overall effects is crucial for understanding intervention impact in social networks.
• Egocentric network designs are common for interventions leveraging peer influence, such as those for HIV prevention.

Purpose of the Study:

• To clarify assumptions for identifying causal effects in egocentric network-based randomized designs.
• To develop sample size formulas for estimating individual, spillover, and overall effects.
• To provide a framework for designing and analyzing network-based public health interventions.

Main Methods:

• Utilized the potential outcomes framework to define and identify causal effects.
• Employed a regression model with a block-diagonal structure for joint estimation of effects.
• Derived sample size formulas for single and joint hypothesis tests concerning intervention effects.

Main Results:

• Established identification strategies under clarified assumptions for egocentric network designs.
• Developed practical sample size formulas applicable to various network-based intervention studies.
• Demonstrated the utility of the formulas using an example of a peer education intervention for HIV prevention.

Conclusions:

• The proposed methods enable robust estimation of individual and spillover effects in network settings.
• The sample size formulas are essential tools for optimizing the design and power of network-based public health studies.
• This work supports the rigorous evaluation of interventions that rely on social influence, such as peer education for HIV prevention.