Sensitivity of Food-Based Recommendations Developed Using Linear Programming to Model Input Data in Young Kenyan Children

Abstract

Food-based recommendations (FBR) developed using linear programming generally use dietary intake and energy and nutrient requirement data. It is still unknown to what extent the availability and selection of these data affect the developed FBR and identified problem nutrients. We used 24 h dietary recalls of 62 Kenyan children (4–6 years of age) to analyse the sensitivity of the FBR and problem nutrients to (1) dietary intake data, (2) selection criteria applied to these data and (3) energy and nutrient requirement data, using linear programming (Optifood©), by comparing a reference scenario with eight alternative scenarios. Replacing reported by estimated consumption frequencies increased the recommended frequencies in the FBR for most food groups while folate was no longer identified as a problem nutrient. Using the 10–90th instead of the 5–95th percentile of distribution to define minimum and maximum frequencies/week decreased the recommended frequencies in the FBR and doubled the number of problem nutrients. Other alternative scenarios negligibly affected the FBR and identified problem nutrients. Our study shows the importance of consumption frequencies for developing FBR and identifying problem nutrients by linear programming. We recommend that reported consumption frequencies and the 5–95th percentiles of distribution of reported frequencies be used to define the minimum and maximum frequencies.