Case Study : Application of
Simple Spatial Sampling Method (S3M) Niger National CMAM coverage and IYCF practices survey
October 2011 to February 2012
Step 1 : Find a map
The first step in a S3M survey is to find a map of the survey area.
For a survey over a very large area such as for five regions of Niger, it will be practical and useful to have instead:
A small-scale map of the entire survey area. This map does not need to show the location of all towns and villages in the survey area.
A collection of larger scale maps of each of the five regions of Niger showing the locations of all towns and villages.
The small-scale map will be useful for identifying initial sampling locations.
The large-scale maps will be useful for identifying the precise location of sampling points and for selecting the communities to be sampled.
Small-scale map of five regions of Niger
Large-scale map of Maradi region
Step 2: Decide area represented by sampling point
The easiest way of thinking about this is as a function of the intended maximum distance (d) of any community from the nearest sampling point.
What value of d to use?
The value of d should be small enough that homogeneity of the area defined can be assumed.
For the Niger survey, a d of 15 km was judged to be small enough to assume homogeneity.
This will mean that no child will live more than 15 km away from a sampling point.
The distance between each sampling point is:
The area of the △ area formed: 292 km2
The area of the ⬡ area formed: 584 km2
Step 2 : Draw a grid over the maps
The next step is to draw a grid over the maps. The size of the grid is determined by the distance (d) decided in Step 2. In this case, d = 15 km.
The grid is rectangular.
The width of the grid in the east-west (x) direction is calculated using
The height of the grid in the north-south (y) direction can be calculated using
For the case of the Niger survey with d = 15 km:
Small-scale map with 22.5 x 13 km grid
Large-scale map of Maradi region with 22.5 x 13 km grid
Niger survey technical team composed of representatives from the Nutrition Department, Institute of National Statistics and UNICEF drawing grids on a large-scale map of Dosso region (above) and Tillaberi region (right).
Step 4 : Create an even spread of sampling points
Sampling points are located at the intersections of the rectangular grid in a staggered fashion. Alternate intersections of the grid are used such as that shown below:
Identified sampling points based on small-scale map
Corresponding sampling points for Maradi region
Survey technical team identifying and labelling sampling points on large scale map of Tillaberi making sure that sample points go right to the edge (or even over the edge) of the survey area.
Step 5 : Select communities to sample
Select the communities closest to the sampling points identified in Step 4.
About three (3) communities per sampling point were selected.
The position of the sampling point was moved into the middle of the three (3) communities selected.
Moving sampling points based on location of selected villages / communities
Step 6 : Label each sampling point
Gave each sampling point a unique identifying label:
The label may be a number or a name.
The label must be unique.
The label was used to identify which community belongs to which sampling point.
The label was used when collecting, organising, and analysing data.
Sampling points labeled and triangles drawn
Step 7 : Within-community sampling
Two within-community sampling strategies were used to select a sample from the selected communities
Coverage Survey – active and adaptive case finding was done in most communities selected. Door- to-door screening was used were in very small rural communities and in urban communities.
IYCF practices – QTR + EPI3, a variant of standard EPI within-community household sampling method
The first step in the method, QTR, divides the community into four (quarters hence QTR) areas each of which have roughly equal volumes.
The second step utilises the standard EPI strategy to select the first household in each of the quarters and selecting the third nearest house in a random direction (third nearest house hence EPI3).
EPI3
Survey Implementation
Step 1 : Identify and train supervisors
Thirty persons were initially hired as surveyors and underwent training phase for the survey.
After training, 20 trainees were retained and seven (7) teams were formed.
The training of surveyors was done through learning-by-doing approach.
Some classroom sessions conducted but on-the-job training was given emphasis during this phase.
There was no separate training period per se as the survey started alongside the training phase.
However, the training phase of the survey was done slower to allow for the surveyors to be trained and learn the survey skills and process.
Step 2: Identify the region to start the survey
The training phase was conducted in Dosso region because it had the least number of sampling points.
The smallest region was chosen to begin with so that the teams can be closely supervised during the training
It also allowed for end-of-day debriefs of the teams with the lead trainer and surveyors (right).
Step 3: Adjust, adapt, revise and re-train
The survey mechanics were adjusted as appropriate during the early stages based on learnings from the training phase
Adapted and refined the case-finding question for active and adaptive case finding (right)
Shifted from EPI3 to QTR + EPI3 approach
Door-to-door approach added as a sampling method for very small villages
Team structure and hierarchy changed based on optimal team composition based on surveyor dynamics
Re-training and refresher training for surveyors done routinely
Mind-mapping exercise with surveyors to refine the case-finding question for active and adaptive case finding
Step 4: Segmentation of regions
Segmentation done similar to that described in the SLEAC session
Regions segmented into 7 by route accessible by road. Each segment assigned to 1 team
List of sampling communities in each segment listed and provided to the team covering that segment
Teams started out in the farthest sampling point in the segment and then worked their way back to the regional capital
Above: Technical team segmenting and planning routes for Dosso region.
Top and bottom right: Technical team members explaining to team leader the route and providing the list of sampling points and communities within that route.
Step 5: Select the next region and continue the survey
The choice of the next region to survey was made based on the following principle:
Results
Barriers to service uptake and access - Maradi
Barriers to service uptake and access – Maradi region
Service problems
Difficulties / challenges faced by mother or caregiver
Screening and referral problems
Lack of awareness of programme and how it works
Lack of knowledge about malnutrition
Rejection
Access issues
Others
0 50 100 150 200 250
ICFI : Infant and Child Feeding Index
Based on an index devised by Mary Arimond and Marie Ruel of the International Food Policy Research Institute for the 2000 DHS survey of Ethiopia and developed by FANTA as a KPC2000+ indicator :
Age-group (months)
6 - 9 9 - 12 13 - 36 36 +
Value Score Value Score Value Score Value Score
Breastfed(24 Hours)
Yes + 2 Yes + 2 Yes + 1 Yes + 0
Food groups(24 Hours)
1
≥ 2
+ 1
+ 2
1 or 2
≥ 3
+ 1
+ 2
2 or 3
≥ 4
+ 1
+ 2
3 or 4
≥ 5
+ 2
+ 3
Meals(24 Hours)
1
≥ 2
+ 1
+ 2
1 or 2
≥ 3
+ 1
+ 2
2
3
≥ 4
+ 1
+ 2
+ 3
2
3
≥ 4
+ 1
+ 2
+ 3
The ICFI score is a measure of appropriate child feeding practices …
ICFI = Breastfeeding + Dietary diversity +Meal frequency… using age-speci,c weighting for each item.
All children are scored 0 – 6 … results presented as a numerical summary (e.g. median) or in a histogram.
Requires only simple and standardised question sets and small sample sizes.
Learnings from S3M pilot
S3M produces:
high resolution maps of indicator of interest over very wide area
S3M can:
serve as a base mapping method for the assessment of other indicators
infant and young child feeding (IYCF)
prevalence of childhood illnesses
water, sanitation and hygiene (WASH)
other typical indicators assessed through standard large-scale surveys like the Multiple Indicator Cluster Survey (MICS)