We analyzed cross-sectional data from the National Cancer Institute's Food Attitudes and Behaviors (FAB) Survey, a mail survey of US adults that was conducted from October through December of 2007. The study was approved by the National Cancer Institute's institutional review board.
Participants of the FAB Survey were recruited via household quota sampling through Synovate Global Opinion Panels (N = 450,000) (Synovate, Inc, Chicago, Illinois). A stratified random sampling of the panel was used to identify 5,803 potential respondents. The sample was balanced by region, annual household income, population density, age, and household size, and non-Hispanic blacks were oversampled (target, 25%). A total of 3,418 surveys were returned, 20 of which were incomplete, yielding a response rate of 59%. This sample was approximately 48% male and 28% non-Hispanic black, and most respondents had a high school degree (58%), an annual household income less than $50,000 (52%), and an age approximately evenly split between the categories of 18 to 34 years, 35 to 54 years, and 55 years or older.
Participants of the FAB Survey were asked "Who is the primary food shopper(s) in your household?" We obtained our sample (N = 1,994) by including only those respondents who identified themselves as their household's primary grocery shopper. Two follow-up questions were, "Where does the primary food shopper go grocery shopping?", and "In the summer, how often does the primary food shopper get fruits and vegetables from a farmers' market, roadside stand, pick-your-own produce farm, or community-supported agriculture program?"
Analyses were conducted by using SAS version 9.2 (SAS Institute, Inc, Cary, North Carolina). Data on sex, race/ethnicity, age, education, and annual household income were weighted based on the 2000 US Census to ensure that statistical estimates were more representative of the US household population; therefore, we report prevalence and standard errors. We used χ
2 tests to evaluate use of shopping venues, and significance was set at P ≤ .001 (2-sided α) to account for multiple testing. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were computed for at least weekly summertime use (versus less often) by using multivariable logistic regression and controlling for sex, age, race/ethnicity, education, annual household income, and region.