Optimization of a Food List for Food Frequency Questionnaires Using Mixed Integer Linear Programming: A Proof of Concept Based on Data from the Second German National Nutrition Survey

Abstract

Food Frequency Questionnaires (FFQs) are important instruments to assess dietary intake in large epidemiological studies. To determine dietary intake correctly, food lists need to be adapted depending on the study aim and the target population. The present work compiles food lists for an FFQ with Mixed Integer Linear Programming (MILP) to minimize the number of foods in the food list. The optimized food lists were compared with the validated eNutri FFQ. The constraints of the MILP aimed to identify food items with a high nutrient coverage in a population and with a high interindividual variability. The optimization was based on data from the second German National Nutrition Survey. The resulting food lists were shorter than the one used in the validated eNutri FFQ.