Assessing the performance of national sentinel food lists at subnational levels in six countries.

Objective: To assess how well national sentinel lists of the most frequently consumed foods in each food group capture data at subnational levels to measure minimum diet diversity (MDD).

Design: We analysed data from seven surveys with 24-h open dietary recalls to evaluate: (1) the percentage of reported foods that were included in each sentinel food list; (2) whether these lists captured consumption of some food groups better than others and (3) differences between estimates of dietary diversity calculated from all food items mentioned in the open 24-h recall . only food items included in the sentinel lists.

Setting: Seven subnational areas: Bangladesh (2), Benin, Colombia, Kenya, Malawi and Nepal.

Participants: 8094 women 15-49 years; 4588 children 6-23 months.

Results: National sentinel food lists captured most foods reportedly consumed by women (84 %) and children (86 %). Food groups with the highest variability were 'other fruits' and 'other vegetables.' MDD calculated from the sentinel list was, on average, 6·5 (women) and 4·1 (children) percentage points lower than when calculated from open 24-h recalls, with a statistically significant difference in most subnational areas.

Conclusion: National sentinel food lists can provide reliable data at subnational levels for most food groups, with some variability by country and sub-region. Assessing the accuracy of national sentinel food lists, especially for fruits and vegetables, before using them at the subnational level could avoid potentially underestimating dietary diversity and provide more accurate local information for programmes, policy and research.