1.Data cleaning plan

Data cleaning will be process by the data manager, using R software.

• Duplicate, missing, or conflicting data

Duplicates entries should be checked by data manager. We consider as duplicates entries, participants for which all the ID variables are the same. ID variable should be unique to distinguish every participant. Moreover no data should be missing for these ID variables. The data manager has to highlight all entries with one of these problems.
Multiple interviewers enter data into dataset, so you have multiple versions of the dataset (usually two versions). If data differs from the two versions, data manager has to refer to original data collection forms and/or photos of home-based records to determine which entries are correct.

• Implausible or illogical responses

Data manager has to check many variables to check if value reported is plausible. For example :

Zone name (variable level2name) has to be one of the following zone names of Nigeria: NORTH EAST or SOUTH SOUTH.
State name (variable RI02 and level3name) has to be one of the following state names of Nigeria: Adamawa, Bauchi, Borno, Gombe, Taraba, Yobe, Akwa Ibom, Bayelsa, Cross River, Delta, Edo, Rivers.
Dates variables have to be valid :

- Years should be lower or equals to the year when participants have been interviewed
- Months should be between 1 (JAN) to 12 (DEC).
- Day of date should be between 1 and 31 (except for month which do not have 31 days).

- Date of vaccination should be between date of birth date of the interview. One child can not be vaccinated before his own birth, and the interviewer only collected past immunization coverage data, not the ones to come.
-  All variables with a “yes/no” answer are coded 0/1. Data manager have to check that no other values than 0 or 1 are entered.

If the data manager finds implausible or illogical values, he has to correct them by consulting the photos of home-based records. Every correction made by the data manager must be justified, documented clearly, applied consistently, and noted in the final report.

• Skip patterns

Data manager should include in his check a step concerning skip questions. For example, if the value reported for variable RI26 (“Did you ever receive or were given a vaccination card or a family folder for (name)?”) is “No” or “Do Not Know”, all the following variables concerning vaccination card should be skipped. Data manager has to check that is the case for all these variables. If a question should have been skipped but data was recorded and entered, data manager has to change the response to “missing” and document the change.

2. Weighting plan

In this Immunization coverage survey, data were collected by a Multiple Indicator Cluster Survey (MICS) team, so the quality of household listing should be high
The three steps involved in weighting are 1) calculating the design, 2) adjusting for nonresponse and 3) post-stratifying to match population totals.
The first step should be done in all vaccination coverage surveys. The other two are applied as needed. We know the probability of selection of the clusters and the households and every eligible respondent should have been interviewed, so we have good data for calculating design weights and for adjusting for non-response at the household level. The design weight is equal to (1/P(Selection))
The second step ensure that the survey result represent the target population by taking in transferring the sampling weight of the non-respondents to the respondents of the survey. The adjustment is calculated as the ratio of the sum of all the design weights in the class to the sum of the design weights of the respondents. The adjust weight for nonresponse is equal to the adjustment factor multiplied by the design weight of the respondent.
The second step is sufficient to estimate coverage within each state, but not to calculate the national coverage estimate or to look at a sub-group, that’s why we process the third step depending to the objectives of the surveys (especially if we want a national coverage estimate). Post-stratified weights are adjusted to make the sum of weights in each stratum proportional to the known eligible population.

In this coverage surveys, sponsors are very interested in equity between sexes, which is the reason why statistician will have to post-stratified weights.

4. Table shells

Table 1 : Percentage of children age 12-23 months currently vaccinated against vaccine preventable childhood diseases with crude doses, Nigeria, 2016-2017

Table1 contains the percentage of crude doses administered to children aged 12-23 months for all vaccine. Results are presented separately in the different states of Nigeria, and according to the sex. 95% confidence interval will be calculated for each state and for each sex. Children weights were used to estimate coverages. Numerator corresponds to the sum of weights of children (boys/girls/all) who received a crude dose of a specific vaccine in a specific state. Denominator corresponds to the sum of weight of the total number of children (boys/girls/all) in a specific state. Coverage is estimated by the ratio between numerator and denominator.

Table 2: Percentage of children age 12-23 months with a dropout between the following vaccines/doses/antigen combination, Nigeria, 2016-2017

Table 2 present the dropout rates between different doses/vaccine. It allows evaluating the performance of vaccination program: if children are vaccinated for an early dose of a vaccine, but not for a later dose, it means that vaccination program has performance issues. Children weights were used to estimate dropout rates.

Table 3: Percentage of children age 12-23 months currently vaccinated against vaccine preventable childhood diseases with valid doses, Nigeria, 2016-2017

Table3 contains the percentage of valid doses administered to children aged 12-23 months for all vaccine. Valid doses means doses administered at the appropriate age (for example at birth for BCG and HepB0). Results are presented separately in the different states of Nigeria, and according to the sex. 95% confidence interval will be calculated for each state and for each sex. Children weights were used to estimate coverages. Numerator corresponds to the sum of weights of children (boys/girls/all) who received a valid dose of a specific vaccine in a specific state. Denominator corresponds to the sum of weight of the total number of children (boys/girls/all) in a specific state. Coverage is estimated by the ratio between numerator and denominator.

Table 4: Percentage of children aged 12-23 months for which immunization card were available and showed by interviewer, Nigeria, 2016-2017

Table 4 shows the percentage of children aged 12-23 months which presented their vaccination card to the interviewer during the interview. If there are too many children for who immunization card is not available, it could biases the survey. Children weights were used to estimate coverages. Numerator corresponds to the sum of weights of children (boys/girls/all) who showed their vaccination card in a specific state. Denominator corresponds to the sum of weight of the total number of children (boys/girls/all) in a specific state. Coverage is estimated by the ratio between numerator and denominator.

Table 5: Percentage of children age 12-23 months currently vaccinated against all vaccines preventable childhood diseases with all doses, Nigeria, 2016-2017

Table 5 present results for fully immunized children. It means the percentage of children aged 12-23 months who received crude doses for all vaccines, and the percentage of children aged 12-23 months who received valid doses for all vaccines. Children weights were used to estimate coverages. Numerator corresponds to the sum of weights of children (boys/girls/all) who received all doses (crude/valid) in a specific state. Denominator corresponds to the sum of weight of the total number of children (boys/girls/all) in a specific state. Coverage is estimated by the ratio between numerator and denominator.

5. Completed table shell

Tables are filled using R software. Data needed to create table will be calculated, and a program will automatically create the different completed tables with the appropriate calculated data. Please find the commented syntax in the attached R file. Titles and footnote are manually added.

The completed table is table 5.

6. Graphical summary

This graph show estimates of DTP3 coverage among children aged 12-23 months from the different states of Nigeria, according to sex. The crosses correspond to the DTP3 coverage estimate, and the lines correspond to the 95% confidence interval. States from NORTH EAST and from SOUTH SOUTH are presented with different colors. For each state and sex, the total number of children is displayed on the right of the graph. This graph have been created using R software.

7. Methods

- Data were collected using tablet. Interviewer took photo of home-based records, for children where they found them, and there were no visits to health facilities.
Data were stratified in 12 different strata (one for each states of Nigeria).The survey sponsors and survey manager were especially interested in equity between sexes, so analysis have been processed comparing boys and girls in each stratum.
- Analyses were processed using R software version 3.4.3. All our analyses are weighted using post-stratified weights.
- Denominator, numerator and coverage were calculated using the weight of each child aged 12-23 months. No data were missing, so for each antigen/doses denominator was calculated as the sum of weight of boys (respectively girls) in each state. Numerator was calculated as the sum of weight of boys (respectively girls) who received the doses/antigen in each state. The coverage was calculated as the ratio between numerator and denominator. 95% confidence interval was calculated.

8. Results

Results presented concerned DTP3 crude coverage among children aged 12-23 months in each state, classified by sex. The number of boys is approximately the same than the number of girls (444 vs 465). Children aged 12-23 months with the highest DTP3 crude coverage are girls from Edo: 78% (more or less 9%) of girls aged 12-23 months from Edo who were eligible for the survey are estimated to have received DTP3 crude dose, as documented by card or recall. For boys aged 12-23 months the highest DTP3 coverage is for them from Cross River: 77% (more or less 12%) of boys aged 12-23 months from Cross River who were eligible for the survey are estimated to have received DTP3 crude dose, as documented by card or recall. The lowest DTP3 crude coverage are in Yobe for both boys and girls aged 12-23 months: : 6% (more or less 5%) of boys and 11% (more or less 7%) of girls aged 12-23 months from Yobe who were eligible for the survey are estimated to have received DTP3 crude dose, as documented by card or recall.

9. Caveats or concerns

Data collected only came from home-based records. No data were collected by visits in health facilities, which can induce a selection bias.

10. Strengths and limitations

One strength in this analysis is that there were no missing data which would have biases the results. But due to the number of strata, and because we separately analyzed boys and girls, confidence intervals were quite large due to the low number of people in each state for each sex.