Low Vaccine Coverage and Factors Associated with Incomplete Childhood Immunization in Racial/Ethnic Minorities and Rural Groups, Central Brazil

Round 1

Reviewer 1 Report

Overall:  This is an interesting analysis of vaccination coverage data from an important and presumably understudied population in Brazil.  The outcomes are important.  I have some questions about the statistical methods and depending on the responses to those questions, I recommend either changing from regression to a simple descriptive paper, or clarifying to assure the reader that the regression is appropriate.

Major concerns:

1.      One unstated assumption of regression is that the respondents were selected by a rigorously random method and that the analysis is weighted by a factor associated with their probability of selection.  The manuscript does not say how communities and municipalities were selected for eligibility in the study.  If they were not selected by a random mechanism, but were selected with purpose, then regression is inappropriate and the confidence intervals and p-values in the regression are not meaningful.  In that case I recommend switching and simply describing the coverage of the convenience sample, or purposeful sample in descriptive terms.

2.      Even if the communities and municipalities were selected via random mechanism, the language about selection of families does not sound random:  …project drew lots of families residing in the municipalities eligible for the project as well as invited those families available at the time of data collection who expressed their willingness to participate in the study. This text makes it sound like a combination of a random selection and a convenience sample.  If so, then regression is not appropriate because respondents who expressed interest, but were not selected randomly may be quite different than a random sample.  If participants could self-select into the study then their outcomes should be reported separately from those who were randomly selected, and they should not be included in the regression analysis.

Medium concerns:

3.      The text does not clarify how the cutpoints were selected for counts and continuous variables that were dichotomized.  The mother’s age is cut at 30 years.  HH size is cut at 6.  Income at $254.60.  If these cut points were modified, the regression outcomes would be somewhat different.  How were they selected?

4.      The text does not clarify whether the analysis and its confidence intervals and p-values accounted for the two-stage sample design, or not.  It should.  (c.f., Heeringa et al., Applied Survey Data Analysis, 2^nd Ed, 2017; https://doi.org/10.1201/9781315153278 )  Every confidence interval and p-value in the manuscript should be adjusted for the sample design.   The analysis should use syntax that incorporates the cluster sample design.  If it did not, then the confidence intervals will be too narrow and p-value too small.  If it did adjust for the sample design, then the methods section should say so clearly.

5.      The discussion interprets one of the adjusted odds ratios with this sentence:  families that did not receive a home visit from a health professional in the last year were 2.4 times more likely to have incompletely vaccinated children than those who received a visit.  This is not the correct interpretation of an odds ratio.  This sounds like interpretation of a risk ratio, but with multiple logistic regression, we would say:  families that did not receive a home visit from a health professional in the last year had odds of being incompletely vaccinated children that were 2.4 times higher than those who received a visit.  The 2.4 is the ratio of the odds…but not the ratio of the probability or the risk.  This is an important point.

6.      The discussion attributes causal power to this analysis:  the home visit contributed significantly to the increase in the vaccination coverage of investigated children.  A cross-sectional survey is definitely not able to prove causation…only correlation.  The two factors are correlated, but we cannot say based on this evidence that the relationship is causal.

Smaller concerns

7.      The paper is not clear for how many children were home-based records seen?  For how many were the records obtained from the SI-PNI?  It says that records were analyzed for 94+133 children, but how many were interviewed who did not have documented evidence of vaccination and were therefor excluded from the analysis.  If there were any such children, the paper should list as a limitation the number and % of such children and mention that their vaccination status may have been worse, on average than that of children with documented records.

Very small concerns

8.      Chart 1 has some confusion with asterisks (*).  Those in the chart and those in the notes are not sync’d.  I would have found the chart easier to read if the cells were collapsed for the doses whose schedules did not change.

9.      The notes underneath Table 2 are probably not needed as the info in them has been provided elsewhere in the manuscript.

10.   Table 1 should include a vertical line to visually separate the bivariate analysis from the multiple analysis.

Author Response

Reviewer 1
Comment: Overall: This is an interesting analysis of vaccination coverage data
from an important and presumably understudied population in Brazil. The
outcomes are important. I have some questions about the statistical methods and depending on the responses to those questions, I recommend either changing from regression to a simple descriptive paper, or clarifying to assure the reader that the regression is appropriate.

Response: Thank you for your evaluation and comments that contributed to improve the quality of our manuscript. The answers about the statistical methods and the modeling used were answered in the following questions.

Comment: One unstated assumption of regression is that the respondents were selected by a rigorously random method and that the analysis is weighted by a factor associated with their probability of selection. The manuscript does not say how communities and municipalities were selected for eligibility in the study. If they were not selected by a random mechanism, but were selected with purpose, then regression is inappropriate and the confidence intervals and p-values in the regression are not meaningful. In that case I recommend switching and simply describing the coverage of the convenience sample, or purposeful sample in descriptive terms.

Response: We appreciate the reviewer's suggestion. We clarify the sampling in this comment. We used four sampling units: (i) municipalities; (ii) coimmunities; (iii) individuals and (iv) families. The quilombola municipalities and communities were not randomly selected. Initially, municipalities that had at least one certified quilombola community (n=45) were included. Of these municipalities, all quilombola and settled communities were included (n=106). Families were selected by systematic sampling and all individuals from families selected at this stage were included. Thus, we consider families as primary sampling units (PSU) through systematic selection and individuals as secondary sampling units (SUS). We also consider a stratum, the type of community (quilombola/settlement). The regression was performed, including the primary sampling unit (family), the platform (quilombola/settlement) and the sampling weight of selection of each individual that took into account the age group, gender and community. The regression was performed considering the complete sample design and using the “survey” package in STATA. The new results were shown in Table 3. The sampling was described as shown below:

“Sampling for the SanRural Project was carried out in multiple stages. Initially, the municipalities that had one or more quilombola communities certified and recognized in the state of Goiás were included, information checked in the official sources of accreditation [14]. Therefore, of the total number of municipalities in the state of Goiás (n=246), 45 (18.3%) met this criterion and were included in the study. In these 45 municipalities, all communities of recognized settlements were also included [15]. Thus, all quilombola communities and settlements in the selected municipalities were included in this study. This represents 44 quilombola communities and 62 settlements, in a total of 106 communities. Municipalities and communities were thus selected based on community certification criteria. Next, the SanRural Project encompassed the following sampling units: (i) families and (ii) individuals. Families were selected by systematic random sampling. In each community, the first individual was selected by lot, and, for every two households (k=2), one family was interviewed until reaching the n sample. The parameters considered for the sample calculation of the SanRural Project study were carried out so that the estimates of proportions of the main indicators were obtained with a confidence level of 95%, maximum margin of error per community of 10% and margin of error for totality. of communities of the same type of 2%. After selecting the family, information was collected from all individuals in the family, including the children to collect the vaccination card. Since the family was selected by systematic random sampling, we considered this sampling unit as the primary sampling unit (UPA) and the individuals as the secondary sampling unit (USA). In this study, we used data only from children born from January 2015 to December 2017. This included information from children in 36 municipalities (80% of the SanRural Project municipalities), 44 settled communities (71% of the total SanRural Project settlements) and 37 quilombola communities (84.1% of the total quilombola communities in the SanRural Project). Thus, data from 81 communities were analyzed, including information from 227 children (94 from settlements and 133 from quilombola communities).”

Comment: Even if the communities and municipalities were selected via random mechanism, the language about selection of families does not sound random: …project drew lots of families residing in the municipalities eligible for the project as well as invited those families available at the time of data collection who expressed their willingness to participate in the study. This text makes it sound like a combination of a random selection and a convenience sample. If so, then regression is not appropriate because respondents who expressed interest, but were not selected randomly may be quite different than a random sample. If participants could self-select into the study then their outcomes should be reported separately from those who were randomly selected, and they should not be included in the regression analysis.

Response: Thanks for the note. We clarified the sampling in the previous
comment, as well as rewrote the text for clarity. The new regression was
performed using sample probability weights.

Comment: The text does not clarify how the cutpoints were selected for counts
and continuous variables that were dichotomized. The mother’s age is cut at 30
years. HH size is cut at 6. Income at $254.60. If these cut points were modified,
the regression outcomes would be somewhat different. How were they selected?

Response: We appreciate the observation and note. We perform the cutoff based on the mean of the variable. After a new bivariate and multiple analysis, it was observed that there were no differences in the results. Changes are identified in the tables. We also add this information to the methods, as shown below:

“The independent variables were the sex of the child (male or female), the type of community (settlement or quilombola), housing zone (rural or urban/periurban), mesoregion of Goiás (Central Goiano, East Goiano, Northwest Goiano, North Goiano or South Goiano), mother's age (≤28 or ≥29 years), number of people in the house (≤5 or ≥6), has internet (yes or no), income (≤US$277.91 or ≥ US$277.92), health professional visit in the last year (yes or no) and, community health unit available in the community (yes or no). Quantitative variables (mother's age, number of people in the house and income) were categorized based on the mean of the variable (less than or equal to the mean versus greater than or equal to the mean).”

Comment: The text does not clarify whether the analysis and its confidence
intervals and p-values accounted for the two-stage sample design, or not. It
should. (c.f., Heeringa et al., Applied Survey Data Analysis, 2nd Ed, 2017;
https://doi.org/10.1201/9781315153278 ) Every confidence interval and p-value
in the manuscript should be adjusted for the sample design. The analysis should
use syntax that incorporates the cluster sample design. If it did not, then the
confidence intervals will be too narrow and p-value too small. If it did adjust for
the sample design, then the methods section should say so clearly.

Response: Thanks for the note. We performed the new analyzes (Tables 1, 2 and 3) considering the complex sampling design. Analyzes were performed using the STATA "survey" package. The method has been described as shown below:

“The data collected during the interview, information about the vaccines recorded on the vaccination card, and the vaccine data obtained from the SI-PNI were exported to statistical analysis software (IBM SPSS®, version 24 and Stata, "College Station," Texas, version 17.0). All analyzes were performed using the complex sample design. Stata's "survey" package was used. The selected families were included as UPA and the type of community (settlement/quilombola communities) were used as a stratum. In addition, individual selection sample weights were included for each child [23], considering the selection probability, according to community, sex and age group.”

Comment: The discussion interprets one of the adjusted odds ratios with this
sentence: families that did not receive a home visit from a health professional in the last year were 2.4 times more likely to have incompletely vaccinated children than those who received a visit. This is not the correct interpretation of an odds ratio. This sounds like interpretation of a risk ratio, but with multiple logistic regression, we would say: families that did not receive a home visit from a health professional in the last year had odds of being incompletely vaccinated children that were 2.4 times higher than those who received a visit. The 2.4 is the ratio of the odds…but not the ratio of the probability or the risk. This is an important point.

Response: Thanks for the observation and suggestion. We made the necessary correction. We believe that the error may have been during the English translation. We appreciate your careful reading. The text changed to: “Families that did not receive a home visit from a health professional in the last year had odds of being incompletely vaccinated children that were 1.96 times higher than those who received a visit from a health care worker.”

Comment: The discussion attributes causal power to this analysis: the home visit
contributed significantly to the increase in the vaccination coverage of investigated children. A cross-sectional survey is definitely not able to prove causation…only correlation. The two factors are correlated, but we cannot say based on this evidence that the relationship is causal.

Response: We thank you again for the suggestion and we agree with your
placement. We modify the writing to: “the home visit seem to contribute to the increase in the vaccination coverage of the investigated children.”

Comment: The paper is not clear for how many children were home-based
records seen? For how many were the records obtained from the SI-PNI? It says
that records were analyzed for 94+133 children, but how many were interviewed who did not have documented evidence of vaccination and were therefor excluded from the analysis. If there were any such children, the paper should list as a limitation the number and % of such children and mention that their vaccination status may have been worse, on average than that of children with documented records.

Response: Thanks for the note and we believe it will increase the quality of the manuscript. The following information has been added to the manuscript: ”In total, 227 children were part of the study and distributed in settled communities (n=94) and quilombola communities (n=133). For children who did not present the vaccination card at the time of the interview (n=80), the search for vaccination data was carried out in the Information System of the National Immunization Program (SI-PNI) in Brazil. Of the total number of children investigated (n=227), 23 had no vaccination records and were considered unvaccinated [16]. In this study, we included in the analysis children who did not have any vaccination records, since vaccination coverage through household surveys allows estimating the real proportion of vaccinated children, as well as measuring existing social inequalities in vaccination coverage, since all databases of information has been consulted [16].

Comment: Chart 1 has some confusion with asterisks (*). Those in the chart and
those in the notes are not sync’d. I would have found the chart easier to read if
the cells were collapsed for the doses whose schedules did not change.

Response: Thanks for your suggestion. We modified Chart 1 to make it easier for the reader to understand. Only changes in schemes were noted in the "years" column.

Comment: The notes underneath Table 2 are probably not needed as the info in
them has been provided elsewhere in the manuscript.

Response: Thanks for the suggestion. We removed the footnotes from Table 2. However, we left the meanings of the acronyms for a better understanding of the reader/editor without having to return to the text.

Comment: Table 1 should include a vertical line to visually separate the bivariate
analysis from the multiple analysis.

Response: Thanks for the suggestion. We added the vertical line separating the bivariate and multiple analyzes in Table 3. “Table 1” had the wrong name, we apologize for this and have corrected it.

Reviewer 2 Report

The manuscript entitled: Low vaccine coverage and factors associated with incomplete

childhood immunization in racial/ethnic minorities and rural 3 groups, Central Brazil

 

The objective is to estimate the vaccination coverage in the first year for children living in quilombola communities and 14 rural settlements in the Central region of Brazil and to analyze the factors associated with incomplete vaccination.

The topic is interesting and the approach is appropriate.

 

Comments and suggestions:

Abstract: The results should be better presented in this section.

 

Introduction:

How is immunization inequity defined?

- Reference is made to the decrease in vaccination during the COVID-19 pandemic; however, this retrospective study evaluates pre-pandemic years, so the unit of analysis is not the same. Assess the relevance of this information.

- It is important to understand the condition of these minorities to have official information about their classification in the socioeconomic level of the country.

 

Materials and methods

- Specify the objective of the mother study

- Specify with which computer program the map was made.

- It is not specified in this section if the study was retrospective. This should be clear.

 

Results

- The numbering of the tables is incorrect. Correct.

- There is enough space in the table to place the name of the vaccine and avoid the calls at the bottom of the table.

- At the end of the results there is a reference error, apparently the product of a misconfiguration with a reference manager

- The information described in the paragraphs is the same as that already found in the tables. This can be summed up.

 

References:

Review the correspondence between the references in the text and their correlative numbering in this section.

Author Response

Reviewer 2
Comment: The manuscript entitled: Low vaccine coverage and factors associated with incomplete childhood immunization in racial/ethnic minorities and rural 3 groups, Central Brazil. The objective is to estimate the vaccination coverage in the first year for children living in quilombola communities and 14 rural settlements in the Central region of Brazil and to analyze the factors associated with incomplete vaccination. The topic is interesting, and the approach is appropriate.

Response: We appreciate the reviewer's careful evaluation. Comments were
duly responded to and adjustments were made to the manuscript based on the
reviewer's notes. Thanks for the opportunity to improve our manuscript.

Comment: Abstract: The results should be better presented in this section.

Response: We thank you for your careful reading, and despite the restriction on the number of words, we have added the following information in the abstract results section: “Overall vaccination coverage was 52.8% (95% CI: 45.5% – 59.9% ) and ranged from 70.4% for the yellow fever vaccine to 78.3% for the rotavirus vaccine, without significant differences between the groups of quilombolas and settlers. Note that the chance of having incomplete general vaccination coverage was greater among children who did not receive a visit from a health professional.”

Introduction:
How is immunization inequity defined?
Comment: Reference is made to the decrease in vaccination during the COVID-
19 pandemic; however, this retrospective study evaluates pre-pandemic years,
so the unit of analysis is not the same. Assess the relevance of this information.
- It is important to understand the condition of these minorities to have official
information about their classification in the socioeconomic level of the country.

Response: We appreciate the comments and have removed the reference and information that addresses the COVID-19 pandemic. Furthermore, we believe that the results of this study point to an inequality in immunization for the minority groups studied.

Materials and methods
Comment: Specify the objective of the mother study

Response: We appreciate the suggestion and we agree with your placement. We add in the text: “The SanRural Project aims to promote knowledge about the conditions of sanitation and environmental health of settled and traditional communities, such as riverside communities and remnants of quilombos.”

Comment: Specify with which computer program the map was made.

Response: We thank you for the suggestion, we agree with your placement and we apologize for not having posted it earlier. We add the information: “Made with ArcGIS, version 3.24.3.”

Comment: It is not specified in this section if the study was retrospective. This
should be clear.

Response: We thank you again for the suggestion and we agree with your
placement. We add in writing.

Results
Comment: The numbering of the tables is incorrect. Correct.
Response: We appreciate your careful reading and apologize for the error. We have made the correction.

Comment: There is enough space in the table to place the name of the vaccine
and avoid the calls at the bottom of the table.

Response: Thanks for the suggestion. We removed the footnotes from Table 2. However, we left the meanings of the acronyms for a better understanding of the reader/editor without having to return to the text.

Comment: At the end of the results there is a reference error, apparently the
product of a misconfiguration with a reference manager

Response: We appreciate the suggestion and apologize for the error. We have made the correction.

Comment: The information described in the paragraphs is the same as that
already found in the tables. This can be summed up.

Response: We appreciate your careful reading and pertinent suggestions. We have modified some details in the text.

References:
Comment: Review the correspondence between the references in the text and
their correlative numbering in this section.

Response: We appreciate the suggestion and we corrected the correspondence between the references in the text and their correlative numbering in this section.

Reviewer 3 Report

I’d put the Population Characteristics in material and method

 

1) The main question is to estimate vaccination coverage in the first year for children living in quilombola communities and rural settlements in the central region of brazil and analyze the factors associated with incomplete vaccination

2-3) I no consider the topic original because the results (incomplete vaccination) and the associated factors are predictable. It address only a specific estimation of the vaccination in a specific area

4) the methodology is appropriate. There is only a formal request. I think that the “population characteristics” should be included in materal and method because the characteristics are no objective of the study

5) the conclusions are consistent and they address the objective

6) the references are appropriate

 

Author Response

Reviewer 3
Comment: I’d put the Population Characteristics in material and method.

Response: Thanks for your remark. We believe it is part of the results
(information obtained after data collection). Thus, we ask permission to keep in the results section.

Comment: The main question is to estimate vaccination coverage in the first year for children living in quilombola communities and rural settlements in the
central region of brazil and analyze the factors associated with incomplete
vaccination. I no consider the topic original because the results (incomplete
vaccination) and the associated factors are predictable. It address only a specific
estimation of the vaccination in a specific area.

Response: We appreciate the evaluation of our manuscript and notes. We
believe that the penultimate paragraph presents the importance of the study and the theme of vaccination coverage in groups of ethnic minorities and settlement residents. Once again, we appreciate your review.

Comment: the methodology is appropriate. There is only a formal request. I think that the “population characteristics” should be included in material and method because the characteristics are no objective of the study

Response: We appreciate your suggestions and your careful reading. We believeit is part of the results (information obtained after data collection). Thus, we ask permission to keep in the results section.

Comment: 5) the conclusions are consistent and they address the objective.
Response: We appreciate the evaluation of our manuscript and notes.

Comment: the references are appropriate.
Response: We appreciate the evaluation of our manuscript and notes.

Reviewer 4 Report

The authors should provide more details about the study protocol. They could improve their paper by helping the readers to understand the way they designed the sample selection. For example, was the visit time selected at random? what's the size of the targeted population (houses and children)? what is the age distribution, including for the vaccine coverage ? why the inequalities factors such as the mother education status was not documented in the study? are the 227 study children representative of the population of children in the target regions

The authors should also discuss the potential bias sources, such as participation bias and evaluate how this can impact the results.

 

Minor remarks:

line 129: replace Chart by Table

The number of stars used within this Table are different from the legend given bellow, that should be added at the end of line 130

The outcome variable is not defined in the manuscript

line 177: provide %femal and the distribution of the appropriate vaccination age

line 180: define this given amount ($) as the moderate/low income threshold and remove subsequently the use of "$" to replace it with a categorial variable (high vs. low income, for example), in Table 2 (previousely Table1)

 

 

 

Author Response

Reviewer 4
The authors should provide more details about the study protocol. They could
improve their paper by helping the readers to understand the way they designed the sample selection. For example, was the visit time selected at random? what's the size of the targeted population (houses and children)? what is the age distribution, including for the vaccine coverage ? why the inequalities factors such as the mother education status was not documented in the study? are the 227 study children representative of the population of children in the target regions

Comment: The authors should also discuss the potential bias sources, such as
participation bias and evaluate how this can impact the results.

Response: We appreciate your suggestions and your careful reading. We
complement and provide more details of the sample selection.

Minor remarks:
Comment: line 129: replace Chart by Table

Response: We appreciate the suggestion, and we apologize for the wrong chart formatting. We have modified the chart formatting according to chart guidelines. We ask for permission to keep it as a Chart.

Comment: The number of stars used within this Table are different from the
legend given bellow, that should be added at the end of line 130

Response: We appreciate your careful reading and suggestion. We rearrange the chart.

Comment: The outcome variable is not defined in the manuscript.

Response: We appreciate the evaluation of our manuscript and notes. The
variables were defined in the “Variables” subsection of the “Materials and
Methods” section: “The outcome variable was the incomplete general
vaccination coverage related to the vaccination situation (no/yes), according to
the applied doses described in the basic vaccination schedule, evaluated at 11
months and 29 days. “
“The independent variables were the sex of the child (male or female), the type
of community (settlement or quilombola), housing zone (rural or
urban/periurban), mesoregion of Goiás (Central Goiano, East Goiano, Northwest Goiano, North Goiano or South Goiano), mother's age (≤28 or ≥29 years), number of people in the house (≤5 or ≥6), has internet (yes or no), income (≤US$277.91 or ≥ US$277.92), health professional visit in the last year (yes or no) and, community health care unit available (yes or no). Quantitative variables (mother's age, number of people in the house and income) were categorized based on the mean of the variable (less than or equal to the mean versus greater than or equal to the mean).”

Comment: line 177: provide %femal and the distribution of the appropriate
vaccination age

Response: We appreciate the suggestions and we add in writing. The
distribution of the appropriate vaccination age are were present in table 2.

Comment: line 180: define this given amount ($) as the moderate/low income
threshold and remove subsequently the use of "$" to replace it with a categorial variable (high vs. low income, for example), in Table 2 (previousely Table1)

Response: We appreciate the suggestion, but we opted to make the cut considering the average found in the continuous variable. The changes were made in tables 1 and 2. As they are homogeneous groups, both with low socioeconomic status compared to the Brazilian population in general, it would be impracticable to stratify into high or low income.

Round 2

Reviewer 1 Report

Excellent revisions.  Congratulations on this interesting work.

Author Response

Thank you for your observations and contributions to the refinement of the study.

Reviewer 2 Report

My observations have been answered with arguments and changes that are sufficient to satisfy my concerns and observations.

Author Response

Obrigado por suas observações e contribuições para o refinamento do estudo.

Reviewer 4 Report

This version of the authors manuscript has been muchly improved

Author Response

Thank you for your observations and contributions to the refinement of the study.