1. First and foremost, the questionnaire should match the vaccination schedule of the country. Additionally, the objectives may differ from collecting data on just children age 1, so this may require adjustment. There are also a number of detailed customization pieces needed to verify, such as where (which body part) shot was given that require attention.

Second, objectives may or may not including where (which type of facility) shots were given, campaign participation and reasons for incomplete vaccination (which I am not a fan of).

Third, the questionnaire includes a module on birth registration, which may be deleted if not an objective. Additionally, the household level information is rather detailed in an effort to compute a wealth index, education levels and socio-cultural background, which may not be an objective to capture. Particularly, education should rather be captured from the child’s proxy respondent (mother/caretaker) him/herself, as should age (of respondent). I would probably also choose to find mothers/caretakers to confirm ages of children just out of eligibility in order to confirm information from household respondent (which we know often is not the best proxy).

Fourth, there are a few cover page details that should not be necessary to summarise, such as number of women age 15-49. This leads me to also think that we could probably reduce the size of list of household members to target relevant and helpful information only.

2. See file

NICS HH module codebook

3. If each individual cannot be uniquely identified the data collected on the individual is basically lost. There is nearly no way to place the individual record in the sample. As a minimum each individual should be identified in stratum, cluster and household or to put it differently: In the sample. It makes sense to make the numbering system as simple as possible, with each variable running from 1 to n, i.e. serial numbers, as this eliminates the need for very complex constructs that can always be elaborated afterwards. The only objections to this simplified approach would be if the implementing agency has a long-running standard that everyone is acquainted with perfectly and a new system will confuse.

4. I would create a sheet to which the identifiers are copied and place the vaccination card on the sheet before taking the picture. As always, the picture needs quality control and this would be part of this protocol (and is part of the protocol). There are more practical solutions depending on the typical card and the customized protocol for getting the best shot. It would be best to create the same for each photo, so perhaps a small card with identifiers can be clipped to the vaccination card.

5. This is a HUGE question. I think it is necessary to include 3 levels of controls:

a) Interviewer: The questionnaire should of course automate skips and provide range checks, but can also provide warning messages on potential inconsistencies, e.g. a child has participated in a polio campaign, but has no polio shots listed on card and none recalled. It is important to not go overboard, as we do not want interviewers to invent data to eliminate warnings, e.g. a child can have 0 polio doses while having attended the campaign – maybe he/she was deemed to sick to get the shot.

b) Supervisor: Should be equipped with summary data that allows progress tracking and minor performance monitoring. Again, we should not go overboard as some data is difficult to interpret and we do not want an overzealous supervisor to give problematic cleaning instructions to interviewers. It could be something as simple as length of interviews and similar process indicators. Additionally, the supervisor needs to do forced spot-checks.

c) Central office: The accumulated data should be held against certain standards: Do teams and individuals find expected results?

Malfunction is common – I have heard 5-10% total device failure in surveys. Teams should be equipped with back-up devices and take back-up of data (cloud, laptop, memory stick) – all solvable issues. The bigger challenge is however on application problems: How do we fix a wrong skip in the application? It used to be visits to each team, but now it can be done over the internet, so that is a solution that must be used. With this in mind, a data processing expert leading the design should be on call 24/7 during fieldwork.