Vaccination coverage in children under one year of age and associated socioeconomic factors: maps of spatial heterogeneity

ABSTRACT Objective: to analyze vaccination coverage spatial distribution in children under one year old and the socioeconomic factors associated with meeting the recommended goals in Minas Gerais. Methods: an ecological study, carried out in 853 municipalities in the state. Pentavalent, poliomyelitis, meningococcal conjugate, yellow fever, rotavirus, and 10-valent pneumococcal conjugate vaccination coverage were analyzed. Scan statistics and multiple logistic regression were performed to identify spatial clusters and factors associated with meeting coverage goals. Results: spatial analysis revealed clusters with risk of low coverage for all vaccines. Number of families with per capita income of up to 1/2 wage, Minas Gerais Social Responsibility Index and percentage of the poor or extremely poor population were associated with meeting the established goals. Conclusions: the results are useful for designing interventions regarding the structuring of vaccination services and the implementation of actions to increase vaccination coverage in clusters with less propensity to vaccinate.


INTRODUCTION
Vaccination coverage is a summary measure of performance used in the Brazilian National Immunization Programs (PNI) and can be monitored through administrative data or periodic vaccination coverage surveys (1)(2) .
The Global Vaccine Action Plan 2011-2020 proposed the achievement of coverage goals for all vaccines in the Brazilian national immunization schedule by 2020.However, less than two thirds of the countries reached the proposed goal, such as the third dose of diphtheria, pertussis, and tetanus vaccine, with 66% coverage (1) .
In Europe, countries have shown a decline in vaccination coverage since 2016, reaching almost 14 million children without the vaccination schedule for diphtheria, pertussis, and tetanus (DPT) and measles vaccines in 2019 (3) .In Montana, in the United States of America, less than two out of five children under two years of age had a complete schedule for childhood vaccines (4) .
In Brazil, since the 1990s, vaccination has shown satisfactory levels of coverage, guaranteeing access and greater equity in health (5) .However, as of 2016, vaccination coverage for children has declined by about 10 to 20 percentage points and, consequently, showing negative effects, such as the occurrence of epidemics, with the most recent of measles in Roraima and Amazonas (6)(7)(8) . .Recently, the COVID-19 pandemic has intensified health inequalities, with low vaccination coverage for poliomyelitis and measles in socially more vulnerable and unequal municipalities (11)(12) .
Studies carried out in Minas Gerais to analyze trends in vaccination coverage in children under two years of age, between 2014 and 2020, pointed to low vaccination coverage from 2015.Measles, mumps and rubella vaccine had coverage of less than 95% in all years analyzed.Pentavalent, Bacille Calmette-Guérin (BCG), poliomyelitis and rotavirus (LORV) vaccines were the ones that showed the greatest decreasing trend among the regions of the state (13)(14) .
Decreases in vaccination coverage are often related to a population's socioeconomic status and geographic conditions (1,(15)(16)(17) , characteristics related to structural conditions and supply and access to health services in each location (6,16,18) and more recently the COVID-19 pandemic (19)(20) .Systematic monitoring of immunization coverage is an indispensable activity to know the realities in which factors ranging from the quality of management of immunization programs to political and socioeconomic factors are inserted (15,17,21) .This monitoring allows knowing and identifying territories that need interventions in immunization services to increase vaccine coverage (11)(12)18) . Consdering the decline in vaccination coverage in the country and among the regions of Minas Gerais, the second most populous state in Brazil (9) , studies are needed to support the implementation of state health policies at the regional level to increase vaccination coverage.

OBJECTIVE
To analyze vaccination coverage spatial distribution in children under one year old and identify the socioeconomic factors associated with achieving the recommended coverage goals, in the state of Minas Gerais, in 2018.

Ethical aspects
This study, in which the guidelines of Resolution 466/2012 of the Brazilian National Health Council were followed, was approved by the Research Ethics Committee.Moreover, it is derived from a master's thesis entitled "Análise espacial da cobertura vacinal em menores de um ano, Minas Gerais, Brasil", presented to the Graduate Program in Nursing at the Universidade Federal de São João del-Rei, Centro-Oeste Dona Lindu Campus, in 2021, and is available at: https://ufsj.edu.br/pgenf/dissertacoes_defendidas.php

Study design, period and place
This is an ecological study carried out in the state of Minas Gerais in 2018.To elaborate the method, the STrengthening the Reporting of OBservational studies in Epidemiology recommendations were followed.

Study population; inclusion and exclusion criteria
The reference population for this study consisted of vaccinated children under one year old, living in the 853 municipalities of Minas Gerais.In 2017, a total of 260,959 children were registered in the Live Births Information System (23) , a fraction corresponding  to the denominator that makes up the basis for calculating the vaccination coverage indicator for 2018.
For spatial scanning analysis, the response variable was vaccination coverage, which is shown in the numerator the total doses that complete each vaccine's scheme, and in the denominator, the number of live births in the municipality multiplied by 100.It should be noted that the Brazilian PNI established a coverage goal of 90% for LORV vaccine, 95%, for pentavalent, poliomyelitis, 10VPC and MNC vaccines, and 100%, for YF (25) .
With the aim of identifying the socioeconomic factors associated with achieving the goals recommended by PNI for vaccines, explanatory variables were selected from the Brazilian Institute of Geography and Statistics (IBGE -Instituto Brasileiro de Geografia e Estatística) (22) and Fundação João Pinheiro (26) databases, which were included in logistic regression analyzes (Chart 1).

Analysis of results, and statistics
Initially, data were analyzed using Microsoft Excel, version 2016, in which it was possible to calculate vaccination coverage (25) .
To verify the existence of clusters based on the vaccination coverage indicator, spatial scanning analysis was used, using SaTScan 9.6, supported by the discrete Poisson model (27) .Scan statistic operates with scanning several circular search rays through the analyzed territory, i.e., the municipalities of Minas Gerais.To define the size of these analytical circles, the maximum size of the search radius is defined, which, for the present analysis, was considered the radius of 50% of the exposed population (goal population of the analyzed vaccines).Each cluster was statistically analyzed by the log-likelihood ratio test, and its statistical significance was assessed using Monte Carlo hypothesis tests (28) .
Calculations of estimates for relative risk (RR) were carried out for each of the clusters identified in spatial scanning analysis.From the respective ratio, it is possible to remove the population effect, which can trigger a distortion of the analytical findings.Thus, analysis considers a variable that indicates vaccination coverage in a given cluster (group of municipalities), associating it with a given population of the respective location.Thus, a cluster's RR is the quotient between this coverage observed in the cluster and the vaccination coverage in other municipalities in the state of Minas Gerais that do not belong to the identified cluster (29) .For the purposes of classification and interpretation of results in this study, when obtaining an RR>1, it can be considered that municipalities belonging to clusters have a greater chance of vaccinating their population, i.e., they are more likely to achieve high vaccination coverage compared to clusters with RR<1.Although RR is a measure calculated from a non-dichotomized variable, i.e., vaccination coverage, the respective interpretation (lower or higher chance of vaccination) was defined to provide a better understanding of results, given the large number of clusters and immunobiological agents analyzed in this research.Furthermore, it is important to highlight that RR with a value equal to one represents an unlikely association between the location and the chance of being vaccinated or not.
To prepare the choropleth maps with the results of the respective scan analysis, the cartographic base of Minas Gerais and its respective municipalities was used, obtained free of charge on the IBGE website and prepared using ArcGIS 10.8.
Considering the objective of identifying the factors associated with achieving vaccination coverage goals recommended by the PNI for the vaccines in this study, multiple logistic regressions were conducted.For this purpose, the dependent variable was considered based on the dichotomization of the municipalities that met or failed the vaccination coverage goal for the six vaccines analyzed.Thus, considering the municipalities as units of analysis for this research, they were classified as "0", if they failed to reach the vaccination coverage goal for each of the immunobiological agents analyzed, and "1", if this goal had been achieved by the respective location.Explanatory variables (Chart 1) were collected from different data sources to characterize the respective municipalities analyzed.
To select the final explanatory model, the lowest Akaike Information Criterion (AIC) value of the explanatory model was considered as a criterion, considering the stepwise technique of selection of variables to be included in the final statistical model.The AIC is an important metric to verify the statistical model quality, and the lower its value, the greater the quality and simplicity of the regression model.From this analytical perspective, it is important to highlight that the final explanatory model may not have all the variables presented in Chart 1, given the process of including and eliminating variables in the model in the search for the lowest possible AIC value (30) .
It should be noted that the Odds Ratio (OR) calculation considers "non-compliance with the recommended vaccination goal" as a reference variable, while the outcome was "achieving the recommended vaccination goal" (classification "0" for the municipality) and the outcome was "achieving the recommended vaccination goal" (classification "1") by the municipality.To analyze the final adjustment of the elaborated explanatory models, the Kolmogorov-Smirnov test was performed, a nonparametric test used to analyze whether the respective model's residuals follow a normal distribution.Another test performed was McFadden's pseudo R-squared, which measures the goodness of fit of the estimated model.Finally, the value below the Receiver Operating Characteristic Curve (ROC curve) was calculated, which analyzes variable sensitivity/specificity in the final model to predict the analyzed outcome, i.e., determines the final model's predictive power (30) .A multiple logistic regression analysis was carried out for each of the six vaccines, with the respective Confidence Interval (95% CI) and p-value of the explanatory variables being calculated.

RESULTS
Pentavalent, poliomyelitis, MNC, LORV, FAN, LORV, and 10VPC vaccination coverage was interpreted considering the health macro-regions (n=14) that make up the study analysis units (municipalities) of Minas Gerais.
The Center, Jequitinhonha and Triângulo do Sul macro-regions did not reach the recommended coverage goals for all vaccines analyzed.It was found that nine of the 14 macro-regions in Minas Gerais had adequate coverage for pentavalent, poliomyelitis and LORV vaccines, 11, for 10VPC vaccine, and eight, for MNC.The Triângulo do Norte macro-region was the only one to reach the YF vaccine goal (Table 1).
Spatial scan statistic detected the presence of statistically significant clusters for pentavalent, poliomyelitis, 10VPC, LORV, MNC and YF vaccination coverage.In the Center-North region, clusters with greater territorial extension (greater number of municipalities in the same cluster) were observed that had a lower chance of vaccinating their population.Considering the pentavalent, LORV and 10VPC vaccines, their clusters with the lowest chance of vaccination were identified with a propensity to form in eastern Minas Gerais, such as in the macro-regions Northeast, Jequitinhonha, East, Vale do Aço and East South.Poliomyelitis, MNC and YF vaccines were more available in central regions, encompassing macro-regions such as Center, North and Northwest, but without excluding the other regions mentioned above.
On the other hand, the Triângulo Norte and Triângulo Sul regions had clusters with a greater chance of vaccination for all immunobiological agents analyzed in the present study; however, it should be noted that these clusters had a considerably smaller territorial extension when compared to clusters with a lower chance of vaccination.It is also noteworthy that the South region was heterogeneous, as it presented clusters of lower and higher chances of vaccination together, demonstrating the complexity of the region.YF vaccine had the largest territorial dimension cluster with the greatest propensity to vaccinate the population, covering the Center-South region and four neighboring regions (Figure 2).To identify the socioeconomic factors associated with vaccination coverage goal achievement, multiple logistic regression was conducted; for this purpose, the municipalities of Minas Gerais were dichotomized between those that met and those that did not achieve the PNI goals.YF vaccine coverage was the one with the highest number of municipalities that did not reach the recommended vaccination goal (n=292), while 10VPC had the highest number of municipalities that reached the coverage goal (n=660). of

Vaccination coverage in children under one year of age and associated socioeconomic factors: maps of spatial heterogeneity
Pereira MAD, Arroyo LH, Serrano Gallardo MDP, Arcêncio RA, Gusmão JS, Amaral GG, et al.
Logistic regression analysis for socioeconomic factors to achieve the vaccination coverage goal in children under one year old, in Minas Gerais, identified six associated variables, namely: number of families with per capita income up to 1/2 minimum wage; percentage of poor or extremely poor populations in the single registry in relation to the municipality's total population; Minas Gerais Social Responsibility Index; ratio 20% richest/40% poorest; proportion of the population assisted by the Family Health Strategy; and proportion of hospital admissions for conditions sensitive to Primary Health Care.For all vaccines analyzed, it was evident that the number of families with per capita income of up to 1/2 the minimum wage, the percentage of poor or extremely poor populations in the single register in relation to the municipality's total population and the Minas Gerais Social Responsibility Index were associated with vaccination coverage goals.
The number of families with per capita income of up to 1/2 the minimum wage was associated with all six vaccines analyzed, representing a chance of up to 0.97 (ranging between 0.96 and 0.99, when considering the 95%CI) lower for municipalities to reach indicative coverage for each new family with this income.The increase in the proportion of hospitalizations due to conditions sensitive to Primary Health Care was a factor that reduced the chances of achieving vaccination coverage for 10VPC, while the ratio between the richest 20% and the poorest 40% was related to MNC and YF, reducing by 0.92 (ranging between 0.86 and 0.99 when considering the 95%CI) the chances of achieving vaccination goals with the increase in the respective indicator, i.e., with the increase in social inequality in the municipalities (Table 2).
Among the factors that are related to an increase in the chance of reaching the goals and coverage, the Minas Gerais Social Responsibility Index was the one that represented the highest OR values, meaning a greater impact for municipalities to meet LORV, MNC, pentavalent and YF vaccine goals.However, it is important to underline that the 95%CI of this variable was extensive, meaning that the sample used in the present study does not allow an accurate representation of the studied population mean.Added to this is the proportion of people assisted by the Family Health Strategy (associated with 10VPC and poliomyelitis) and poor or extremely poor populations in the single registry (associated with LORV, MNC and pentavalent), which also showed an increase in the chances of optimal coverage for the respective vaccines that showed a significant association.
Adjustment analysis of the logistic regression models performed showed adequate values for the Kolmogorov-Smirnov test, i.e., the respective models' residuals presented normality.When observing the results of McFadden's pseudo R-squared, it is verified that the models presented adequate values.
Finally, when considering the area under the ROC curve for the multiple logistic regression models for LORV (ROC=0.63),MNC (ROC=0.60),10VPC (ROC=0.60),pentavalent (ROC=0.62),poliomyelitis (ROC=0.54)and YF (ROC=0.63), it is identified that the models presented reasonable predictive power of the dependent variable, i.e., socioeconomic variables explain some conditions that lead or not the municipalities of Minas Gerais to reach the recommended PNI goal for vaccination coverage.

DISCUSSION
Spatial distribution of pentavalent, poliomyelitis, 10VPC, LORV, MNC vaccine coverage showed differences between themselves and between the health macro-regions of Minas Gerais.It is noteworthy that the vaccination coverage of all vaccines analyzed were below the recommended goals in the Center, Jequitinhonha and Triângulo do Sul macro-regions.Socioeconomic factors were associated with the achievement of these goals, with emphasis on the number of families with per capita income of up to 1/2 minimum wage, the percentage of poor or extremely poor population in the single registry in relation to the municipality's total population and the Minas Gerais Social Responsibility Index.with worse indicators of human development and social inequality (9,12,18,24,(33)(34) .In this study, the number of families with per capita income of up to 1/2 the minimum wage, the Minas Gerais Social Responsibility Index and the percentage of poor or extremely poor populations included in the single registry showed an association with achieving the vaccination coverage goals.The risk of non-compliance with vaccination coverage goals for children under one year old increases among families with per capita income of up to 1/2 the minimum wage.As families from less favored classes generally have less access to health services, it is likely that the spontaneous search for vaccination is low due to lack of infrastructure, greater distance and difficulty in accessing public services (43)(44) .In this regard, both in the national (45) and in the international scenarios (46)(47)(48) it becomes evident that low level of education and low socioeconomic level are associated with vaccination below the recommended.
A study carried out to investigate disparities in vaccination coverage related to socioeconomic status, urban/rural residence and the child's sex in 86 low-and middle-income countries identified that in 58 countries the highest levels of coverage were in urban areas and, in all countries, the poorest wealth quintile had the lowest immunization coverage (48) .In Brazil, research with children benefiting from Bolsa Família (Family Allowance), to assess vaccination coverage according to the family's socioeconomic level and maternal characteristics, found that belonging to the richest quintile (predominantly poor sample) and maternal education ≥ 9 years were associated with higher proportions of up-to-date vaccination (45) .
The percentage of poor or extremely poor populations included in the single registry in relation to the municipality's total population was positively associated with achieving vaccination coverage goals.Identifying the most vulnerable people and carrying out measures to control their social vulnerability can be an important factor in increasing vaccination coverage.A Brazilian population-based study that analyzed the impact of Programa Bolsa Família on child health found a positive association between receiving a benefit from the Program and greater childhood immunization coverage in low-income children (49) .However, a cohort carried out in São Luís and Ribeirão Preto, municipalities located in two regions with different socioeconomic conditions, identified that receiving the Programa Bolsa Família benefit had no influence on childhood vaccination, despite the high percentage of incomplete vaccines in São Luís (37.4%) compared to Ribeirão Preto (15.2%) (34) .This result may indicate that the Program's conditionality and the monitoring of the vaccination situation are not being carried out properly, since the percentages of vaccine incompleteness in beneficiary children were high.A national longitudinal study with beneficiaries of Programa Bolsa Família since 2018 also found a low percentage of children with adequate vaccination, both in the first and second year of life (45) .Social protection and social assistance are factors to be considered as policies to strengthen vaccination.However, more effective control of program conditionalities is needed, including those related to health (34) .
The Minas Gerais Social Responsibility Index had the greatest impact for municipalities to meet vaccination goals in children under one year old.A study carried out in 76 countries showed that a high Human Development Index (HDI) is a predictor for greater commitment and implementation of vaccination actions (50) .When reviewing the factors that influence childhood vaccination schedule compliance in different countries, especially related to socioeconomic conditions, authors observed that countries with lower HDI, such as Mozambique, Uganda and Kenya, have lower vaccine coverage for DPT than countries with HDI higher (51) .Recently, Brazilian studies have identified clusters of low vaccination coverage for poliomyelitis and measles associated with worse human development indicators, social inequality and less access to the Family Health Strategy, facts aggravated by the COVID-19 pandemic (11)(12) .
Achieving high and homogeneous vaccination coverage goals are essential for control and elimination of vaccine-preventable diseases, requiring global efforts and commitments to strengthen health systems and immunization services.Differences in childhood vaccination schedule compliance may be the result of different contexts of implementation of immunization programs, which pervade the health system characteristics, vaccination schedule complexity, records in the child's vaccination book, supply of immunobiological agents and, especially, due to socioeconomic conditions (12,18,45,52) .

Study limitations
Although this study sought to provide an overview of the correlates of achieving the recommended vaccination coverage goals, it is likely that there is regional variability within municipalities and also between other sets of clusters.The ecological character of this study is highlighted, in which the results presented here consider population clusters as the unit of analysis, making it impossible to interpret them at the individual level.
It can also be highlighted, as a limitation of this study, data quality and use of information produced in the PNI Information System, which can interfere with the actual calculations of vaccination coverage.A worrying reduction in the completeness of immunization records and vaccine coverage has been observed in Brazil, bringing the resurgence of some diseases hitherto overcome (9) .

Contributions to nursing, health, or public policies
This study confers an important originality, by addressing an emerging problem of great social impact related to socioeconomic conditions and the supply of services, and by considering the state of Minas Gerais as its scenario, the second most populous in the country and the fourth in territorial extension.The results can support the implementation of priority measures carried out by health professionals, specifically nurses responsible for immunization services, to avoid the resurgence at the epidemic level of vaccine-preventable diseases already controlled, particularly in the face of a COVID-19 pandemic scenario caused by SARS-CoV-2, which further aggravates the population's vaccination situation.

CONCLUSIONS
Spatial analysis revealed clusters with risk of low vaccination coverage for pentavalent, poliomyelitis, 10VPC, LORV, MNC vaccines in Minas Gerais.Socioeconomic factors were associated with achieving vaccination coverage goals.However, the reasonable values of the area under the ROC curve (ranging from 0.54 to 0.63) show that there are other variables or conditions that need to be better analyzed in order to understand more precisely which additional factors can influence vaccination coverage in children younger than one year.Other studies should be conducted to identify other determinants for vaccine coverage.
The identification of clusters with low coverage subsidizes priority measures regarding the implementation of state health policies at the regional level, in order to increase vaccination coverage in clusters with greater spatial risk and, consequently, greater transmission of vaccine-preventable diseases.

FUNDING
This work was carried out with financial support from the Coordination for the Improvement of Higher Education Personnel (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) (Code 001) and Minas Gerais Research Support Foundation (Fundação de Amparo à Pesquisa de Minas Gerais) (APQ-00638-21).

CONTRIBUTIONS
Pereira MAD, Arroyo LH, Gallardo MDPS, Arcêncio RA, Oliveira VC, Guimarães EAA contributed to study/research conception or design; data analysis and/or interpretation and final review with critical and intellectual participation in the manuscript.Gusmão, JS, Amaral GG contributed to data analysis and/or interpretation and final review with critical and intellectual participation in the manuscript.

Figure 1 -
Figure 1 -Macro-regions of the state of Minas Gerais, Brazil, 2022

Figure 2 -
Figure 2 -Areas of spatial clusters of vaccine coverage, referring to pentavalent, poliomyelitis, meningococcal conjugate, yellow fever, rotavirus and 10-valent pneumococcal conjugate vaccines, in children under one year old, Minas Gerais, Brazil, 2018

Vaccination coverage in children under one year of age and associated socioeconomic factors: maps of spatial heterogeneity
Pereira MAD, Arroyo LH, Serrano Gallardo MDP, Arcêncio RA, Gusmão JS, Amaral GG, et al.

Vaccination coverage in children under one year of age and associated socioeconomic factors: maps of spatial heterogeneity
Pereira MAD, Arroyo LH, Serrano Gallardo MDP, Arcêncio RA, Gusmão JS, Amaral GG, et al.

Table 1 -
Vaccination coverage in children under one year of age by health macro-region in Minas Gerais, Brazil, 2018 Note: *10VPC -10-valent pneumococcal conjugate; † LORV -rotavirus; ‡ MNC -meningococcal conjugate; § YF -yellow fever.Note: relative risk > 1 has a greater chance of vaccinating its population; relative risk < 1 lower chance of vaccinating the population.

Table 2 -
Result of logistic regression for socioeconomic factors to achieving the goal of vaccination coverage in children under one year old, Minas Gerais, Brazil, 2018

Vaccination coverage in children under one year of age and associated socioeconomic factors: maps of spatial heterogeneity
Pereira MAD, Arroyo LH, Serrano Gallardo MDP, Arcêncio RA, Gusmão JS, Amaral GG, et al.