The Minimum Dietary Diversity (MDD) indicator for children is a simple population-level assessment method for tracking diet quality among children 6–23 month of age. Large-scale surveys routinely collect MDD using two types of 24-hour dietary recall methods: a list-based recall method in which a list of food items is read to the respondent who indicates what the child consumed the previous day and night, and a multiple-pass recall method that comprises an open recall, followed by a list of items in food groups not mentioned in the open recall, followed by “Other solid, semi-solid, or soft food?” probes.

However, it is not known to what extent MDD for children, when constructed from these methods, is accurate when compared to the MDD calculated from an observation. USAID Advancing Nutrition conducted a study in representative samples of children 6 to 23 months of age in the Feed the Future Phase I Zone of Influence in Cambodia and Zambia.

Data for each child participant were collected over two consecutive days. On the first day, a child’s food intake was observed for an entire day and recorded. On the subsequent day, the child’s primary caregiver was interviewed twice to recall the child’s dietary intake from day one using list-based and multiple pass open recall methods—one in the morning, the other in the afternoon, in a randomly assigned order. The MDD indicator was thus estimated three times. 

We found that the MDD estimations from both recall methods (list-based and multiple pass open recall) over-estimated the observed value. Estimations from both recall methods were statistically equivalent to the observed MDD in Cambodia (30.8 percent and 36.7 versus 29.4 percent) but neither were statistically equivalent in Zambia (62.3 percent and 68.4 versus 58.2 percent). Both methods were equivalent to the observation in identifying most food groups and highly sensitive, although the multiple-pass method accurately classified a higher proportion of children meeting MDD than the list-based method in both countries (93.7 percent versus 78.6 percent in Cambodia, 90.0 versus 80.4 percent in Zambia). Both methods were highly specific in Cambodia but not in Zambia. The list-based recall method produced slightly more accurate estimates of MDD at the population level, took less time to administer, and was less costly to implement.

Results were presented during a webinar in May 2023, and at NUTRITION 2023, the American Society for Nutrition’s annual flagship meeting. A manuscript including an innovative analysis of the cost-accuracy comparison of the two methods, is forthcoming.  

A second study examined what subsequent passes add in a multiple pass recall. The first pass overestimated MDD in Cambodia and underestimated it in Zambia. Subsequent passes did not add information that changed the MDD estimate in Cambodia, while changes to the estimates in Zambia led to overestimation of MDD. For both countries, most food groups added in subsequent passes were accurate in only half or fewer reports compared to the observation. Given the low percentage of respondents adding new food groups in subsequent passes and the low accuracy of those additions, organizations that prefer the comprehensive approach to dietary data collection offered by the open recall in the multiple pass method may be able to proceed with just an open recall and forgo subsequent passes to collect information. That approach could save time preparing data collection materials, training enumerators, and analyzing data, while supporting data quality with fewer data cleaning and analysis steps. However, researchers should test such an approach with a small sample before adopting it.