The project would be a secondary analysis of the data of the most recent iodine survey from Madagascar to compare the results and deductions of composite samples per cluster against those of single samples per households. Main aim is to determine if the composite sampling methodology produce similar mean/median iodine contents per strata level as those obtained using single samples and determine the proportion of samples below the threshold of 15 mg I/kg from the two methodologies.

Important: As in Madagascar results of composite salt samples per cluster are not available, mathematical average of the single salt samples per cluster (i.e., enumeration area) will be used as a proxy for the iodine content of a “theoretical” composite sample of each cluster.

Methodological suggestions:

A. Population strata:

• National,
• Geographic,
• Residence (*urban/rural),
• Wealth quintiles*, and
• Educational levels*:

* This requires the estimation of the corresponding indexes of each cluster, based on the categorization of each studied household in the cluster.

B. Statistical parameters:

Average, SD, SE

95% Limits of Agreement (LOE) with the available results, and 95% CI, Proportion of samples above 15 ppm iodine

Preparation of frequency distribution and cumulative frequency distributions of the salt iodine content at the national level, as well as per geographic (region and district) and residence strata, using the values of single household sample

Same as above but using the average values per cluster.

C. Special analysis: For each strata, prepare correlations (and estimate linear regression equations) between the averages calculated with single samples and the averages calculated with cluster averages.

For the wealth quintiles and educational level strata, it is necessary to estimate the corresponding indexes per cluster following this procedure: estimate the wealth index of each cluster by averaging the wealth quintile value of each household (1 to 5). Estimate the educational level index of each cluster using the same procedure (household educational level may be from 0 to 4 or from 1 to 5).

Estimate the average (SD, SE, and 95%CI, and proportion of samples above 15 ppm iodine) for each level of the wealth-quintile and educational-level based on results of single households.

Prepare linear correlations between the wealth quintiles (or educational levels) in the x-axis vs the average iodine content in salt per cluster in the y-axis. Using the linear equations of these correlations, estimate the y-values (i.e., iodine content in salt) for each level of the two stratum types (wealth quintiles or educational levels). Compare the results with those estimated in the prior step using single samples.

D. Iodine content in the salt samples of different quality:

Similar to the preparation of indexes per cluster, the proportion of fine salt might be estimated per cluster. If there are more than two (coarse or fine) types of salt, the approach might be estimating salt type indexes, giving a number of each type of salt.

Estimate average (SD, SE, and 95% LOE, and 95%CI, and proportion of samples above 15 ppm iodine) for each type of salt using single values.

Prepare linear correlations between the proportion of fine salt per cluster (or quality of salt index) in the x-axis vs the average iodine content in salt per cluster in the y-axis. Using the linear equations of these correlations, estimate the y-values (i.e., iodine content in salt) for each level of type of salt (for example, if only two types of salt, intercept in x-0 is going to be the approximated value for coarse salt, and intercept in x-1 is going to be the approximated value for fine salt). Compare the results with those estimated in step (2) with single samples.

E. Comparison of single results vs duplicate results

As the data set of Madagascar has values per duplicate of each salt sample, estimate average (SD, SE, 95% LOA, and 95%CI, and proportion of samples above 15 ppm iodine) for each one of the two duplicates (A and B), and for the average of the duplicates. Prepare frequency distribution and cumulative frequency distribution for each replicate (A and B), and for the average of the two replicates, using the data of the whole country.

F. Any other calculation:

The consultant is asked to suggest any other alternate calculations require to complete the proposed scope of work.

Travel, if applicable: (include estimated duration and potential locations).

The consultant is not expected to make any travel under this project.

Terms of Reference / Key Deliverables:

• Detailed workplan: By 20 Aug 2024
• Draft protocol for data analysis: By 25 Aug 2024
• Final protocol: By 05 Sep 2024
• Document with key findings: By 30 Sep 2024
• Draft report: By 20 Oct 2024
• Final report: By 15 Nov 2024
• Draft peer-reviewed paper: By 20 Nov 2024
• Final paper: By 15 Dec 2024

Requirements: 

Education:

At least a master’s degree.

Disciplines: Statistics, Data Science, Mathematics, Economics, Public Health, Quantitative Social Sciences, Information Technology, Operations Research, or any other relevant discipline.

Experience:

• Expertise in applying mathematical principles to nutrition, health and/or food security issues, with a focus on data analysis and modeling.
• Previous experience of atleast 3 years of working as an international consultant in food fortification/ salt iodization/other public health nutrition initiatives.
• Experience in designing and monitoring and evaluating nutrition/food fortification/ food security programmes is an asset.
• Previous record of development and publication of peer-reviewed papers/ articles is preferred.

Competencies/Knowledge:

• Demonstration of a strong academic foundation and research capabilities will be an advantage.
• Demonstrated ability to analyze large datasets and identify areas for improvement.
• Strong skills in documenting data findings and in communicating complex data and findings to diverse stakeholders.