Validation of the Cluster-Compositing Method to Assess the Performance of the Salt Iodization Program in Ethiopia

Traditionally, thousands of single salt samples, which have highly variable iodine content, are usually analyzed in population surveys requiring costly and lengthy laboratory work. The adoption of composite samples may overcome this constraint.

To assess the validity of using composite salt samples for simpler and more efficient monitoring of iodized salt in Ethiopia.

We collected 6148 household single salt samples from 576 clusters during the 2021/2022 National Food and Nutrition Strategy Survey of Ethiopia. Equal amounts of single samples from each cluster were mixed, regardless of the type of salt (fine or coarse), to make a cluster-composite sample. Iodine determination was conducted using the titration method. Heterogeneity of the iodine content was also studied in samples collected from markets.

Iodine contents were comparable between composite and single samples: 39 ± 4 parts per million (ppm) compared with 35 ± 15 ppm for fine salt, and 24 ± 11 ppm compared with 26 ± 17 ppm for coarse salt, respectively. Composite samples accurately identified underperforming regions and provided useful information at the subnational level and sociodemographic categories. Higher iodine contents were found among urban and more educated households, and to a lesser degree among wealthier groups. Iodine content was heterogeneous in market salt, even within the same package, with 23% of samples showing >2 ppm differences among 5 replicates, regardless of the salt type. This within-sample heterogeneity shows that the performance of the salt iodization programs should be based on measures of central tendency (median or means) rather than the proportion of single samples around a fixed threshold, as the latter is highly influenced by single-sample heterogeneity.

The cluster-compositing method provides a cost-effective alternative to assess the performance of salt iodization programs; it provides better precision and reduces the laboratory workload by 10-fold. The methodology can be applied to different sociodemographic characteristics without the need to analyze thousands of single samples.