(P17-055-23) Accuracy and Cost of Two Assessment Methods for Estimating Diet Diversity of Children 6 to 23 Months in Cambodia and Zambia

Objectives: Two common approaches are used to assess minimum dietary diversity (MDD) and food group consumption for children 6-23 months: a sentinel list-based and a hybrid (open recall followed by a list) method. We assessed the accuracy of both proxy dietary data collection methods compared to an observation. We also estimated the costs associated with the accuracy of each method in estimating MDD.

Methods: Participants were selected for this 2-day study using two-stage probability sampling to represent Feed the Future Phase I Zones of Influence in Cambodia and Zambia.

On day 1, an interviewer observed all food/beverage items the selected child consumed while awake and interviewed the caregiver about the child’s consumption before the observation began or after they went to sleep for the night. The following morning and afternoon two interviewers, in random order, administered the sentinel list and hybrid methods to the caregiver to recall food/beverage consumption during the previous day. Equivalence testing was used to compare each proxy method to the observation for estimating MDD and food group consumption using a 10% equivalence margin.

An activity-based costing approach was used to calculate the costs to collect, clean, and analyze proxy dietary data to estimate MDD per method.

Results: In Cambodia (n=638), both recall methods were equivalent to the observation in assessing MDD and consumption of all food groups except breastmilk. The sentinel list, but not the hybrid method was equivalent in assessing any breastmilk consumption. The cost per participant to estimate MDD was $75 using the sentinel list method and $82 using the hybrid method. The sentinel list method was more cost-accurate (lower cost per percent agreement with the observation for MDD). The hybrid method’s higher costs were primarily driven by personnel costs for data collection preparation. In Zambia (n=608), cost and accuracy results are emerging.

Conclusions: In the Cambodian context, the sentinel list and hybrid methods are both acceptable options for assessing MDD, but the sentinel list is more accurate at assessing any breastmilk consumption. The cost per participant and per percentage of children accurately classified was lower using the sentinel method and therefore may be preferred.

Funding Sources: This project was funded by USAID under a contract awarded to JSI Research & Training Institute, Inc.