A Comparative Evaluation of Dietary Indicators Used in Food Consumption Assessments of at-Risk Populations

• Published in: Food and nutrition bulletin 2008-06, Vol.29 (2), p.113-122
• Main Authors: Rose, Donald, Chotard, Sophie, Oliveira, Leila, Mock, Nancy, Libombo, Marcella
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Aged; Child; Child, Preschool; Diet - standards; Diet - statistics & numerical data; Diet Surveys; Energy Intake - physiology; Family Characteristics; Female; Food Supply - statistics & numerical data; Humans; Infant; Infant, Newborn; Male; Mental Recall; Middle Aged; Mozambique; Nutrition Assessment; Predictive Value of Tests; ROC Curve; Sensitivity and Specificity
• ISSN: 1564-8265; 0379-5721; 1564-8265
• DOI: 10.1177/156482650802900205

Background Easy-to-collect dietary indicators have been used increasingly for planning and evaluation of food security interventions. Various indicators have been employed, but rarely has a full set of indicators been compared using a common framework. Objective This paper evaluates the performance of five dietary indicators for the assessment of household energy consumption using a common framework and recent data from Mozambique. Methods Data were analyzed from a 2004 household survey, Current Vulnerability Analysis in Seven Provinces of Mozambique. Households (n = 4,358) were sampled from 42 rural districts using a two-stage design, and a quantitative 24-hour dietary recall was employed. Household energy intake ratios were calculated as the food energy consumed by household members divided by the sum of the members' recommended intakes. Five proxy indicators of household consumption in the previous day were developed: the number of meals, the number of food groups, the number of food items, a score based on a simple weighting of food groups consumed, and a predicted energy intake ratio based on weighting of food groups consumed with previously estimated regression coefficients. The performance of these indicators was assessed using correlations with energy intake, receiver operator characteristic analysis, efficiency of predictions, and prevalence estimate comparisons. Results Although the predicted energy intake ratio performed best on all these performance criteria, and the simple food-group-weighted score performed second best, differences among the indicators on many of the criteria were relatively small. Conclusions New assessment systems could take full advantage of easy-to-collect information by using one of these best-scoring indicators, although established systems could continue to use some of the other indicators explored here, such as the food item count, without much loss in accuracy.