What is the purpose of Stage 2?
To check: From our stage 1 findings, we have established what our main boosters and barriers to coverage are. We can use stage 2 as a ‘checkpoint’ to ensure that our findings from stage 1 are a true reflection of the programme.
To deep dive and inform the prior: Stage 2 is also an opportunity to collect additional information on an area of interest that may have an indirect impact on programme coverage, especially indirectly. Information collected in stage 2 can also be used during the formulation of the prior, ahead of the wide are survey in stage 3.
A common hypothesis tested is the impact of distance on coverage. In other words, testing what we already know, despite it providing little additional information. While this can confirm our findings from stage 1, it usually acts as only a repetition of information already highlighted in stage 1. Instead, descriptive aspects of distance that influence access to services, such as impacts this has on Community Health Worker (CHW) activity could be tested if stage 1 has not concluded the information. More usefully, investigation of factors that could inform the building of the prior in stage 3 should be considered.
Although results of stage 2 can highlight extreme variations in coverage within the assessment area (patchiness), this does not mean that you should not go ahead with stage 3. Stage 3 may be approached differently, for example, dividing the sample area to provide two coverage estimates, or simply to report the range of coverage within the area.
Stage 1 results in a triangulated and exhaustive set of findings, usually in the form of a list of boosters and barriers. Stage 2 is, therefore an ideal opportunity to conduct further investigation and enrich programmatic or contextual understanding, allowing to supplement stage 1 data, rather than imitating it.
Concept mapping is an ideal technique to explore linkages between boosters and barriers to coverage, and often identifies areas to usefully investigate further. The team may be tempted to focus on the strongest influences, but if the data indicates that this is triangulated by source and method, investigating it will not yield any further information.
See page 55 of the FANTA SQUEAC manual for further guidance on creating a concept map.
Anything can be investigated, as there are many methods available to conduct Stage 2 investigations:
1. For geographic and village level hypothesis, small area surveys can be used, e.g:
‘CHW activity is higher where public transport links are frequent and reliable’.
- Define what CHW activity you are looking at, and what is ‘high’ – target CHW activities may have been described in programme proposals or in the national CMAM guidelines, for example screening frequency.
- Define what frequent and reliable transport is in your context.
- Sample carefully (i.e. 8 villages with 4 of each combination of these conditions).
CHW screening data from the health centres (quantitative) and data from interviews with health workers, community members and CHWs (qualitative) indicate that in villages with good transport links, in this case classified as 2 buses a day, CHWs are more active. High activity of CHWs is classified as, in this context, screening each child U5 in their community once a month. It is hypothesised that this leads to high coverage in villages that meet this criteria.
- Two villages with good transport links are selected. Exhaustive case finding is conducted for all covered and non-covered cases in both villages and the results are combined.
- Two villages with poor transport links are selected. Exhaustive case finding is conducted for all covered and non-covered cases in both villages and the results are combined.
Data can be analysed using the formula:
d = decision rule; n= number of cases found ; p= coverage standard defined.
Have the decision rules been met? Is coverage above or below the defined standard?
More information and technical guidance can be found in the FANTA SQUEAC manual from page 68.
2. Technically, any method of testing a hypothesis can be used as long as it relates to ideas generated from Stage 1 data, e.g.:
Malnourished children with a sibling/other family member to act as a caregiver have high coverage, malnourished children without a sibling/other family member to act as a caregiver have low coverage.
In this example, cases found in the community are sorted according to population type. Specifically here, whether they have a sibling or other family member who is acting as a caregiver, or not. The results can be analysed according to whether the defined decision rule has been met within each population group.
A full write up of this particular example from South Sudan can be found here.
Other similar examples include:
Children who have been in the programme before, are more likely to be in the programme due to higher programmatic awareness, and therefore have high coverage. Children who have never been in the programme, are less likely to be in the programme due to lower programmatic awareness, and therefore have low coverage.
Data to test this could be tested through looking at records and checking for relapses (where access is challenging, this method of data collection may be more feasible), or through community case finding and taking a case history of each case found.
3. For investigating associations between factors affecting coverage, a case control study can be used, and an odds ratio calculated, e.g:
‘Women have higher odds of knowing about IYCF’
- Parameters do not necessarily have to be directly related to coverage.
- There will be higher odds of the second parameter in cases with the presence of the first parameter. I.e. In this case, there are higher odds of knowing about IYCF if the case is a woman.
- Sample sizes and confidence intervals can be calculated online and we advise referring to an expert in statistics for more guidance in how to analyse associations.
- Take time to step back and look at an overview of your boosters and barriers, try a concept map to identify linkages and areas for further investigation.
- Consider making more than one Stage 2 study / survey (why not add a couple of days extra), what additional information will be useful for future programming?
- Findings from stage 2 should be used to inform the prior, however SAM/MAM cases found should not be used in stage 3 findings (e.g. remove villages used from sampling frame).
- Be creative in design and clear in the rationale you follow, your interpretations of results and thorough reporting of findings.
- If you’re stuck or would like to share ideas and experiences, contact us at email@example.com