Based on a previous request to support the selection of study clusters to be included in the evaluation study, WFP Sudan provided a list of the various programme sites that are implementing or will be implementing the various components of the MAM prevention programme package which includes food-based prevention of MAM (FBPM), home fortification (HF) and social and behavioural change communication (SBCC) from which to select the study clusters from.
Description of the list
The list was provided in a spreadsheet and the programme sites were organised based on the various contextual information of which the following are of greatest significance for the purposes of selecting study clusters:
- Locality in which programme site is located;
- Implementing partner responsible for the programme site;
- Type of services available from the site; and,
- Coordinate locations of the programme sites.
Observations about the list
On review of the list, we observed the following:
- There is a single row of data that is for Gedarif State (Al Fashakaga locality).
- Summarising the list by localities in Kassala state, we have the following distribution of programme sites by localities of Kassala:
- It seems that the food-based prevention of MAM interventions have already been implemented throughout all the WFP distribution sites in Kassala. Whilst there are still some localities where there are still much fewer distribution sites compared to the others (i.e., Wad Elhelew and West Kassala) it seems that the programme is in almost all localities of Kassala (with the exception of Seteet locality?). It would seem that this would be an issue as the study would need to select areas where there is no MAM prevention activities implemented at all. But after clarification by WFP, it was made clear that implementation of the MAM prevention programme package was suspended given pipeline issues. Therefore, it can be said that Kassala (and probably the rest of the other states) are starting from scratch with regard to MAM prevention activities which would mean that almost all the areas in Kassala can be considered for selection as study clusters.
We are assuming that this is probably a mistake when creating the list as we are guessing that the Kassala list was created from a longer list that contains all the states in Sudan and all the localities within each of the states of Sudan. We therefore took this data row out of the list.
|ID||Locality||Number of programme sites|
|5||Rural El Girba||9|
Selection of the study clusters
Given these observations and considerations regarding the list, we identify the following localities as the most feasible clusters for our study:
- Kassala (urban) – 20 sites; this most likely has the highest catchment population of all the localities; one partner delivers the services in this site (SMOH)
- Rural Kassala – 17 sites; 1 partner delivers the package in this site (TOD)
- North Delta – 34 sites; 2 partners work in this locality (SMOH and TOD); we can potentially break this into two clusters – one for the area that is covered by SMOH and the other for the area covered by TOD but this will depend whether this division is geographically well-demarcated and evenly divided.
- River Atbara – 9 sites; relatively small; we might consider joining this with the Rural Aroma locality and/or with Rural El Girba (8 sites); This used to be run by Plan but it seems they might not be renewed as partners; It might make more sense to combine this with Rural El Girba so that we can keep Rural Aroma as a cluster.
- Rural Aroma – 16 sites; two partners work in this locality (SMOH and TOD)
- Telkuk – 29 sites; three partners (SMOH, SRC, Sudan Vision)
Based on these first choice selections, we think the most reasonable study cluster selections are the following:
- Kassala Urban – 20 sites; 1 partner
- Rural Kassala – 17 sites; 1 partner
- North Delta – 34 sites; 2 partners
- River Atbara + Rural El Girba (9 + 8 sites); 1 partner
- Rural Aroma – 16 sites; 2 partners
- Telkuk – 29 sites; three partners
Given this choice, we will be working with 5 to 6 unique partners based on previous partnership agreements struck by WFP.
Randomisation of rollout in the selected localities
If we are to randomise the rollout of the programme into these six clusters, we just need to run a random number generator to generate 6 random numbers between 1 and 6 without replacement. We do this in R using the following commands:
sample(x = 1:6, size = 6, replace = FALSE)
What we are basically asking R to do is to randomly sample from the numbers 1 to 6 to find 6 numbers and to not replace the chosen number every time a selection is made. This means that we want to have 6 unique numbers between 1 to 6 chosen randomly.
The result of this random selection is the following series of numbers:
1, 5, 2, 6, 3, 4
This would mean that we will implement the rollout of the programme in the following order:
First: Kassala Urban
Second: Rural Aroma
Third: Rural Kassala
Fifth: North Delta
Sixth: River Atbara + Rural El Girba
Now, based on our revised design, we said we will implement in a 1, 1, 2, 2 fashion. This means we will go rollout first in 1 cluster, then in another 1 and then in 2 and then finally in 2.
However, thinking about this again and in light of the incidence study, it might be best that we have a 2, 1, 1, 2 rollout. This allows for us to have 2 clusters for the intervention arm of the incidence study and another 2 clusters in the control arm of the incidence study. This will afford us maximum possible sample size for the incidence study. This is also consistent with how the original incidence study was designed.
This means that we will first implement in Kassala Urban and Rural Aroma, followed by Rural Kassala, followed by Telkuk and then North Delta and River Atbara + Rural El Girba. The schematics of the study implementation is shown here. Figure 1 shows a map of the selected study clusters in Kassala grouped by study steps 1 to 4.
|Rollout Step||Study Cluster|
|Step 1||Kassala Urban|
|Step 1||Rural Aroma|
|Step 2||Rural Kassala|
|Step 4||North Delta|
|Step 4||River Atbara + Rural El Girba|
Issue of contamination and spill over
In the study design, we have articulated that selected study clusters should be separated from each other by clear boundaries and geographical gaps so as to avoid contamination and spill over effects. In Figure 1, it is clear that from a locality level, the study clusters share lots of borders with each other hence the potential for contamination and spill over effects particularly between control and intervention areas. However, as can be noted with the locations of the programme sites in each selected locality, the potential catchment area of each site is lesser than the entirety of the locality boundaries (and is most likely dictated by where the villages and populations are). This would indicate that there is a lot of room for adjusting the sampling within the locality that would avoid contamination and spill over effects. This can partly be done by avoiding sampling from villages in the borders of the localities. There are also additional information during data collection that can be used to ascertain possible contamination and spill over.