A penalized likelihood approach to estimate within-household contact networks from egocentric data

TL;DR

This paper introduces a penalized likelihood method to infer detailed within-household contact networks from egocentric data, revealing age-related contact patterns crucial for epidemic modeling.

Contribution

It develops a novel statistical approach to estimate household contact networks from survey data, addressing a gap in modeling infectious disease transmission.

Findings

Contact behavior depends on age composition.

Older households have fewer contacts.

Estimated networks inform epidemic models.

Abstract

Acute infectious diseases are transmitted over networks of social contacts. Epidemic models are used to predict the spread of emergent pathogens and compare intervention strategies. Many of these models assume equal probability of contact within mixing groups (homes, schools, etc.), but little work has inferred the actual contact network, which may influence epidemic estimates. We develop a penalized likelihood method to infer contact networks within households, a key area for disease transmission. Using egocentric surveys of contact behavior in Belgium, we estimate within-household contact networks for six different age compositions. Our estimates show dependency in contact behavior and vary substantively by age composition, with fewer contacts occurring in older households. Our results are relevant for epidemic models used to make policy recommendations.