Dietary Diversity Measures Derived from Food Frequency Methods Are Weak Proxies for Anthropometric Indicators of Child Nutrition in East Africa (P11-111-19)

Abstract

Abstract Objectives The 7-item child dietary diversity score (DDS) is used widely as a metric of dietary diversity (DD) for infant and young child feeding (IYCF) in low-middle income countries. Validation of the DDS was based on mean micronutrient density adequacy and not other measures of child nutrition. We aimed to develop a measure of DD more strongly associated with anthropometric indicators of child nutrition. Methods We used data from two Tanzanian surveys, IMMANA (n = 88; ages 6–23 mo.) and the Demographic and Health Survey (DHS) (n = 9,156; ages 6–59 mo.), and two Kenyan surveys, MAMA SASHA (n = 384; age 6–12 mo.) and DHS (n = 17,509; ages 6–59 mo.). For each survey, we calculated the DDS, minimum DD, a score using all 15 items found in the standard DHS, and a yes/no indicator of animal source food (ASF) consumption. In IMMANA and MAMA SASHA, we also used a count of all unique food items consumed and based each measure on both 24-hr and 7-day recall periods. Outcome indicators were stunting, wasting, and underweight as defined by WHO. We evaluated the performance of each DD measure in identifying child undernutrition by comparing the area under the ROC curve (AUC), and estimating odds ratios using logistic regression methods. AUC = 50% is no better than random assignment. Complex survey procedures were used in all analyses and alpha was set at 0.05. Results In the DHS and MAMA SASHA data, all measures of DD had similar AUC values that did not exceed 61% when compared to the anthropometric indicators. In the IMMANA data, no measure exceeded 70% AUC for wasting, 64% for underweight, or 58% for stunting. For most DD measures, increasing the recall period to 7 days marginally improved its AUC (< 3%), while it decreased the AUC for wasting in the IMMANA data. The ASF indicator had a similar AUC to the DDS. In the Tanzania DHS, a higher DD, by all four measures, was associated with decreased odds of stunting, wasting, and underweight. In contrast, no measure of DD was associated with growth indicators for Kenya. The DD indicators based on the 15-item list and the count of unique foods were negatively associated with odds of wasting in the IMMANA data. Conclusions In two east African settings, measures of DD derived from existing food frequency questionnaires were weak proxies for measures of child undernutrition based on anthropometry. Funding Sources Innovative Methods and Metrics for Agriculture and Nutrition Actions.