J Horton

Delayed submission (technical issues & travel)

Review the 2018 Vaccination Coverage Manual, Chapters 1, Section 6.1 and Annex A. - completed
Read carefully the Harmonia Survey Concept Note. Consider noting key concepts that you may want to remember and any concept that you don’t fully understand - completed
Review and learn to use the Excel -Sample Size Calculator Harmonia.
Read CAREFULLY each question, in order to reply what is asked.

For optional questions, you may try to reply with what you have read, but for more details consider sections 2.5-2.7, plus annex B – sample size; section 6.2 and annex J – weighting. Module A3 will introduce these concepts, but will not into details as this is part of Modules A1 and A2. - optional components not completed 

Questions

1.a How many strata does the Harmonia survey will have?

There will be 10 strata. One for each province.

Use the Excel tool provided to answer the next two questions.

1.b List and briefly describe at least two advantages and disadvantagesof having these many strata compared to having more, for example 100 (one for each district)? (MANDATORY)

Advantages of many strata:

More granularity to the estimates or classification of coverage possible to lower administrative subunit (e.g. more strata = lower administrative subunit
Facilitates estimates for different contexts or sub-groups (e.g. rural vs urban, attending school or not)

Disadvantages of many strata:

Costly to design and implement
Complex to perform high quality data collection and analysis (e.g. make choice of what strata and ensure that there is no overlap w- other strata, aka mutually exclusive)

 

3.a What kind of data do you need to collect to complete a table like the one below? (MANDATORY)

Unclear on what this question is about, as table lists a series of types of data already.

Type is quantitative (categorical (e.g. y/n¨), numeric, dates etc)

As per table, the data would include: census data, detail about the vaccination coverage implementers (e.g. surveyors details about training dates), and then detail from the actual implementation of the survey (e.g. household level data, and child eligible, mothers interviewed etc.).

 

3.b How does this table relate to potential selection bias and what cautions should you have when interpreting the vaccination indicators if response rate is not 100% and/or several households could not be interviewed? (MANDATORY)

It is not clear from the table if there is a policy in place to revisit unoccupied homes repeat times to see if there are any eligible children. Thus there is potential for there to be a response bias towards those families at home at time the surveyors circulate through areas. Additionally, I am unclear on the relation of the two columns in the table… is “questionnaires” actually intended to reflect “variable”.

Likewise for the category about households visited, there is not a differentiation between homes with more than one mother with children in the eligible age bracket. Unclear how a mother would be counted if she had two (or more) children between the age of 12-23 m (the age eligibility bracket), or a “household” (if the variable is indeed “children aged 12-23m”). The latter is particularly concerning as there could be many circumstances imaginable where it is quite common to have more than one child in this age bracket living it a single home. Thus one child does not equate to one home.

For these reasons, would anticipate that the response rates calculated between the two different columns would therefore be different.

4.a What variables do you need to collect or define in the analysis to estimate a [weighted] vaccination coverage with the first dose of diphtheria-tetanus-pertussis containing vaccine (DTP1) and with the third dose of diphtheria-tetanus-pertussis containing vaccine (DTP3) among children aged 12-23 months in each stratum by: a) ‘documented evidence of vaccination (home-based record – HBR OR facility-based record – FBR)’, b) ‘by recall’, and c) ‘by documented evidence of vaccination + recall’. [Do NOT include the data you need to calculate weights]. (MANDATORY)

a) Documented evidence

Age of child (DOB)

DTP1 received and date?

How verified?

DTP3 received and date?

How verified

Calculate the age that child was vaccinated for DTP1 and 3 (vs. what would be according to actual schedule).

Plus, in order to make a “weighted” determination: the total number of children in each strata that were i) estimated to be in the strata area, and ii) considered vaccinated based on the evidence (e.g. the probability that a child in each cluster might be sampled)

b) By recall

Age of child (DOB)

DTP1 vaccine received (e.g. injected vaccine around the age of XX weeks)

DTP3 vaccine received (e.g. follow up injected vaccine around the age of YY weeks)

Calculate the age that child was vaccinated for DTP1 and 3 (vs. what would be according to actual schedule).

Plus, in order to make a “weighted” determination: the total number of children in each strata that were i) estimated to be in the strata area, and ii) considered vaccinated based on the evidence (e.g. the probability that a child in each cluster might be sampled)

c) by documented evidence and recall?

All of the above, plus the level of agreement between the comparison of the two sets of data (as per detail above)

Plus, in order to make a “weighted” determination: the total number of children in each strata that were i) estimated to be in the strata area, and ii) considered vaccinated based on the evidence (e.g. the probability that a child in each cluster might be sampled)

4.b How does this table relate to potential information bias and what cautions should you have when interpreting the vaccination indicators in surveys with “low” percentage of documented evidence? (MANDATORY)

As anticipated, there are different types of information bias (bias arising from misclassification / measurement errors or observational bias).

If the one uses the occupied homes based on a single attempt to visit, there is an observational bias. Homes that are briefly unoccupied (e.g. people at work in fields) would be disproportionately less likely to be sampled.

There is measurement bias inherent to assumptions that there could only be one child in target age bracket per household (or that all children are measured and counted but to mistakenly assume this represents a physically separate household). Furthermore, there would be bias to assume that a mother could only have one child in the target age group (mothers that have two children of different ages OR mothers with multiples would be misrepresented).

Specific to “low” percentages, the rate of absences without complete re-visit strategy planned and implemented could indicate a deceptively high rate of “absence” and low rate of response.