اثرات مصاحبه کننده در نظرسنجی های خرید مواد غذایی Interviewer effects in food acquisition surveys

Food acquisition surveys have been used to track habitual food and drink consumption, examine patterns of food acquisition by different demographic subgroups and understand food security(1–3). Estimates produced by these surveys provide valuable insight for the drafting and modification of policies and programmes designed to promote healthy eating and better lifestyles, and for assisting groups of people who are identified as being food insecure. A recent survey by the US Department of Agriculture (USDA), the National Household Food Acquisition and Purchase Survey (FoodAPS), represents such an effort to generate these important estimates(4). FoodAPS is the first nationally representative survey in the USA to collect detailed week-long information on all household food acquisitions along with other related indicators such as food security, income, employment and participation in food assistance programmes. FoodAPS is a comprehensive survey that involved both face-to-face interviewing and a 7 d diary survey component to be completed by all members of the sampled households. The data collection protocol for FoodAPS required an extensive amount of involvement from human interviewers. Interviewers were responsible for training the household to record their food acquisitions throughout the survey period using the diary. They were also responsible for conducting the initial and final face-to-face interviews. It is therefore important to understand if survey outcomes in FoodAPS (e.g. measures of food security, missing income reports) tend to have distributions that vary across interviewers, above and beyond any variance introduced by the quality of the instruments or the features of the respondents. Despite nearly a century of research into the potentially negative effects of human interviewers on survey data collection(5), we are unaware of any prior research that has evaluated these effects in the context of food acquisition surveys. If these effects exist, they may have an impact on the conclusions drawn from using these data. The objective of the present study was to examine these systematic effects of interviewers on several different data collection outcomes in the FoodAPS.