Selection Bias and Estimates of Social Quality of Life in Friend Comparison Groups

• Activity Number: 423
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Susan Stewart*+ and Joan Bloom
• Companies: University of California at Davis and University of California at Berkeley
• Keywords: quality of life ; multiple imputation ; breast cancer

Abstract

When comparing the quality of life of cancer survivors with that of their cancer-free peers, survivors' friends may be recruited in order to obtain a socioeconomically similar comparison group. However, this method of recruitment is likely to oversample individuals with larger social networks, and the selection bias may in turn lead to biased estimates of mean differences in social quality of life (QOL) variables. However, it is possible to regard hypothetical acquaintances of non-referring survivors as members of a population of cancer-free individuals with smaller social networks. Then the association between the social QOL of participating survivors and friends can be used as a basis for multiple imputation (MI) of QOL values for the hypothetical acquaintances who did not participate. This approach is illustrated with a study of 448 10-year breast cancer survivors and 395 women without cancer who were referred by 214 survivor study participants and 27 ineligible survivors. Initial between-group comparisons indicated that survivors experienced lower quality of life than women without cancer. However, smaller differences were found using bias reducing methods of analysis.