Household-level consumption data can be redistributed for individual-level Optifood diet modeling: analysis from four...

A barrier to using Optifood linear programming (LP), which identifies nutrient gaps and supports population-specific food-based recommendation (FBR) development, is the requirement for dietary intake data. We investigated whether Household Consumption and Expenditure Surveys (HCESs) could be used instead of individual-level 24-h recalls (24HRs). The 24HR data from 12- to 23-month-old breastfeeding children in rural Kenya, Uganda, Guatemala, and Bangladesh were paired with HCES food consumption data from similar areas (n = 8) and time periods. HCES food intakes (g/week) were estimated using adult male equivalents, adjusted for breastfeeding. Paired HCES- and 24HR-defined LP inputs and outputs were compared using percentage agreement. Mean overall percentage agreements were 42%, 63%, and 80%, for food, food subgroup, and food-group model parameters, respectively. HCES food lists were on average 1.3 times longer than 24HR. Similar nutrient gaps (77-100% agreement), food sources of nutrients (71-100% agreement), and FBRs (80-100% agreement) were identified. The results suggest that HCES data can be used in Optifood analyses for 12- to 23-month-old children, despite recognized challenges of using it to estimate dietary intakes of young children compared with older age groups. Further analyses, however, are required for different age groups and locations to confirm expectations that it would perform equally well.