Social networks research and challenges to causal inference.

摘要

Epidemiologic studies of the social determinants of health present investigators with significant challenges to causal inference. Social networks are considered to be an important social determinant of health, but studies involving social networks present additional unique challenges to causal inference [1,2]. In this issue of the Journal, Valente and colleagues [3] present the results of a study of obesity and friendship networks in 562 students, 11-13 years old, attending school in the greater Los Angeles area. The investigators nicely demonstrated that the body mass index (BMI) of a student is associated with the average BMI of his or her named friends. To gather the data needed to conduct their analysis, the research team asked each participating student to nominate up to five friends who were also students in the school, and measured the height and weight of all participating students. These data were then assembled such that the BMI of a student's friendship network became attributes of the individual themselves. This relatively simple-sounding methodology allowed the investigators to begin the exploration of how network factors that may operate outside of the student's direct control may nonetheless contribute to the BMI of the student.