Respondent Fatigue Reduces Dietary Diversity Scores Reported from Mobile Phone Surveys in Ethiopia during the COVID-19 Pandemic

BACKGROUND: The computer-assisted telephone interview (CATI) has been used extensively during the COVID-19 pandemic, but the effects of respondent fatigue during these interviews on responses to questions about diet are unknown. OBJECTIVES: We designed an experiment that randomized the placement of a survey module on the dietary diversity of rural Ethiopian women and assessed whether responses were altered by placing this module earlier or later in a phone survey. METHODS: Two CATIs were implemented; in the second, women were randomly assigned to answer questions on diet diversity either earlier or later in the interview. Women’s Dietary Diversity Scores were the primary outcome. Secondary outcomes were dichotomous measures of consumption from four or more and five or more food groups and consumption of food groups consumed frequently, often, and rarely. Impacts were assessed using a respondent fixed effects model. RESULTS: Delaying the food consumption module by 15 min in the interview led to an 8%–17% (P < 0.01) decrease in reported Dietary Diversity Scores, a 28% (P < 0.01) decrease in the number of women who consumed a minimum of four dietary groups, and a 40% (P < 0.01) and 11% (P < 0.01) decrease in the reporting of consumption of animal source foods and fruits and vegetables, respectively. Moving the food consumption module closer to the beginning of the interview increased the number of reported food groups consumed by older women, women with a below-median education level, and women in larger households. CONCLUSIONS: Our findings suggest that comparisons of descriptive statistics across studies and countries on metrics such as food security and dietary quality may be confounded by where these modules are placed in the interview, thus highlighting trade-offs between volume of information collected and data quality when designing CATI surveys.