Bayesian Hierarchical and Bivariate Regression Models to Assess the correlation between Crime Rates the Measles Vaccination Coverage in Nigeria

Introduction:

The motive of this study is to specify the correlation between crime rates, a social covariate, and the measles vaccine in order to improve future uses of bayesian hierarchical models by first analyzing the relation with their correlations with different bivariate regression models. I will achieve this goal by meeting the following objectives:

1) Acquiring geocoded data from Demographic and Health Surveys about Nigeria’s measles vaccination

2) Applying Baysian geospatial techniques to the the cluster level data, using public datasets on geospatial covariates, in order to map the vaccination coverage of the measles vaccine of the under under-5 population group

3) Incorporating a bivariate analysis to analyze the correlation between publicly available data on crime rates and the new information on vaccination coverage

4) Creating a new visual for vaccination coverage using crime rates within the bayesian model

This work will add on to previous research because it will use current methods of predicting vaccination coverage to create a correlational value between the measles vaccination coverage and crime rates in Nigeria.

Human Development Topic:

With two to three million people, mainly children, dying each year due to Vaccination Preventable Deaths, vaccination coverage is a prominent issue for many international institutions such as the World Health Organization (WHO) and the United Nations (UN) (Adedire et al., 2016). Specifically in 2018, 140,000 people, most of whom are children under 5-years old, suffered from a death due to measles (World Health Organization, 2019). Despite the significantly high number of deaths from measles, the measles vaccine has saved an estimated 23.2 million people from 2000 - 2018, demonstrating the effectiveness and necessity of receiving the measles vaccine (World Health Organization, 2019). The harms of this disease reaches society through an inequitable manner, as an “overwhelming majority (more than 95%) of measles deaths occur in countries with low per capita incomes and weak health infrastructures” (World Health Organization, 2019). This disease and its vaccination coverage warrants extensive research through its potential to kill millions of underprivileged children

The country of Nigeria, which is the focus of this study, is one of many nations which continue to struggle to maintain a suitable vaccination coverage. Within a study of the measles viral strain diversity in Nigeria from 2004-2006, the researchers, Kremer et al., found that the recorded measles strains had “multiple lineages of genotype B3” which validated that measles continued to be characterized as highly epidemic decades after the measles vaccine was introduced (pp 1727, 2010). Relating this to the lack of a vaccination coverage, the researchers cited that the “rapid restoration of a high genetic diversity was probably caused by low vaccination coverage” and other socioeconomic factors (Kremer et al., p 1727, 2010). This indicates that the inherent nature of the measles virus is that it will continue to cycle through lower- and middle-income countries as long as the vaccination coverage continues to be below the 90% national and 80% district-level suggested vaccination coverage of the World Health Organization (2019). As indicated in a study of the Atakumosa-west district, Osun State of Nigeria, vaccination coverage in those two areas range from 54% to 70% vaccination in children from 12-23 months old (Adedire et al., 2016). But increasing the tracking of vaccination coverage, by completing more studies such as this, could help increase vaccination coverage by pinpointing the areas that need vaccination coverage the most. For instance, since this group of vaccination teams in Nigeria began using GPS systems to track who they gave vaccines to for the polio virus, the number of wild poliovirus infections decreased (Barau, 2014). Moreover, in a study focused on measle control in Abia, Nigeria, researchers found that a decrease in vaccination coverage from 2007-2010 correlated with a significant increase in confirmed cases (Umeh & Ahaneku, 2013). So there is a clear validation to focus on Nigeria’s vaccination coverage and its influencers, as Nigeria’s districts vaccination coverage for measles are recently below the recommended vaccination coverage, increasing the likelihood of putting children at risk, and finding the spots without vaccination coverage could help organizations get the vaccine to them.

Focusing on measles vaccination coverage and its influencers is justified further by Amartya Sen’s discussion of freedom in his book, Development and Freedom. Relating this to Amartya Sen’s discussion of human development and freedom, if children, mainly babies, are denied the substantive freedom of life due to death, then human development loses potential, leading to a perpetual cycle of decreasing development (Sen 1999). Connecting this with earlier ideas, it suggests that if the vaccination coverage remains low, then the development will continue to decrease, which lead to the measles impacting lower- and middle-income countries losing potential growth which could then lead to them remaining in the lower- and middle-income status, thus repeating the cycle of decreased development.

The focus on the health of everyone is brought on by the United Nations Sustainable Development Goals (SDG). The most relatable SDG to this topic is to “Ensure healthy lives and promote well-being for all at all ages” under the third SDG (United Nations, 2019). This is because, as stated earlier, receiving the measles vaccines decreases the chances of dying from measles which ensure a healthier life than suffering from measles.

Human Development Process:

The study behind vaccination coverage, and sometimes illness rates, has been previously investigated for social concepts and institutions that would help researchers, epidemiologists, and other statisticians predict vaccination coverage, and sometimes illness rates, when they have little or no data. Within this large body of research, there are connections between education levels, perceived threat, and many other socio-economic variables that are used to accurately predict vaccination coverage, so it is critical to go over some of the overarching predictors in the current realm of research.

Education and economic levels continue to be a strong indicator of vaccination coverage both on a grand comparison of multiple countries and specifically analysing Nigeria. Concerning this topic, Wiysonge et al. conducted a multilevel logistic analysis on predictors of low immunisation of children in 24 sub-Saharan African countries and found that “children from poorest households, uneducated parents, mothers with no access to media, and mothers with low health seeking behaviours were more likely to be unimmunised” (p 4, 2012). Specifically concerning an economic focus researchers found that “children from the poorest households were 36% more likely to be unimmunised than counterparts from the richest households” (Wiysonge et al., p 1, 2012). Slightly contradicting this finding, researchers focusing on the correlation of poverty rates and tuberculosis notification rates in Cambodia found that the region of Cambodia with the highest household poverty rates in the country had the lowest sputum-positive tuberculosis case notification rates (Wong et al., 2013). Despite this interesting fact, the researchers in that study also found that “ same region also has the lowest vaccination coverage and the worst physical barrier of distance to health care facilities” which has been shown to increase the likelihood of becoming infected, so this suggests that people in the most poverty stricken area were probably unable to go to a hospital to identify with tuberculosis (Wong et al., p 28, 2013). Backing the findings of Wiysonge et al., the impact of education highly correlated with vaccination coverage is vigorously applicable to Nigeria. Focusing on the Northern region, researchs, Gunnala et al., found the vaccination coverage of random 12-23 month old participants from household cluster-level surveys from the 40 polio high risk, later drawing connections between vaccination coverage of different variables. In this study, lack of knowledge or education of the vaccination was the highest reported reason for non-vaccination (Gunnala et al., 2016).

Furthering the discussion on relatable covariates to vaccination, neighborhood associations also have a notable impact on vaccination coverage. Published in the Journal of Immunological Techniques in Infectious Diseases, Niyibizi et al. found that perception of safety in a community correlated with a positive view of vaccines in general, similarly indicating that views of unsafet would create more negative views of vaccines (2016). In addition, older age and less availability to vehicles per household had the same statistical significance as neighborhood security (Niyibizi et al., 2016). Niyibizi et al. thus began to focus on a conversation about how the mix of perception of safety, accessibility, and socioeconomic factors can potentially limit vaccination. Moreover, the perception of safety can also be linked to the economic status of a community, which has been shown to have significant link to vaccination coverage, as it has been highlighted that communities with higher income inequality levels had higher homicide rates (Menezes et al., 2013). With a link between income levels, safety, and income and vaccination, there is likely, through a transitive property sense, a link between crime rates and vaccination coverage.

Clearly, the number of variables are not limited to the presented key discoveries, but the sheer number of related factors indicates that the process of vaccination coverage and predicting vaccination coverage acts as a complex, adaptive social system. This is because there is a high number of variables. In addition, because of all the noted variables, predicting general, holistic vaccination coverage has become easier, as to be noted further within the discussion of methodologies. Despite this, as will be further noted later, predicting the specific changes in vaccination coverage is difficult to do in accuracy as finer resolutions of data or more difficult to capture.

Geospatial Methods

Finding a statistical method to predict vaccination coverage is critical to this research in order to accurately identify areas that need extra assistance with vaccination. For this reason, it is critical to identify, define, and evaluate the most used statistical methods of the research and likewise their associated data set. Within the studies that set out to predict the likelihood of a vaccination or sickness within a given area, methodologies split around Bayesian Hierarchical Model and Poisson Regression Model. A study conducted by Utazi et al. highlights the potential usefulness of a Bayesian Hierarchical Model in order to predict the measles vaccination coverage of the country of Nigeria, as part of its focus was the measles vaccination coverage of Nigeria. The method for this study can be broken down into two main parts which converge into one. First with data acquired from Demographic and Health Surveys (DHS), the researchers processed the data and aggregated it to a cluster-level which predicted the likely location of the surveys and then used age stratification to identify the cases for children under 5-years old (Utazi et al., 2018). Secondly incorporating spatial covariates, the researchers applied different sets of geospatial covariates to the cluster-level data to identify correlations and then ran a covariate selection process based on binomial generalized linear models (Utazi et al., 2018). Combining these two concepts through a Bayesian Hierarchical Model, the researchers can then create unequal distribution of vaccination coverage to identify which areas of Nigeria were more likely to be vaccinated based on the previous relationships with geospatial covariates and the cluster level data (Utazi et al., 2018).

On the other end of this decision, a study conducted by Wong et al. identifies the potential benefits of using a Poisson Regression Model. Best explained by the researchers, “the logarithm of the number of new sputum-positive TB cases notified in each OD was modelled as a function of household poverty with the logarithm of the OD population as an offset” (Wong et al., p 25, 2013). This means that poverty and population were treated as continuous variables to identify the likelihood of tuberculos notification. In addition to this, they conducted a univariate analysis to evaluate the relationship between household poverty rates and sputum positive tuberculosis case notification rates (Wong et al., p 25, 2013). Later, they used multivariate analyses to adjust for variables such as population density, distance to health facility, and HIV presence (Wong et al., p 25, 2013).

Considering these two methods, the possibly best choice for this research would be to use a Bayesian Hierarchical Model in additiation to bivariate models to assess the relationship between crime rates and vaccination coverage. Incorporating geospatial covariates previously shown to have correlations with vaccination coverage, this research can predict the total vaccination coverage and analyze how that correlates with crime rates by incorporating different bivariate models.

In the discussion of datasets, most research dealing with the analysis of pan-African countries acquired data from Demographic and Health Surveys. For instance, Utazi et al. and Wiysonge et al. both conducted surveys from data from DHS. Specifically relating this data to this study, notable research, such as Utazi et al., have used DHS data to predict the measles vaccination for the country. This is likely because of the reliable data from DHS. Because of this and the availability, this study will use data from DHS to identify cluster-level data on measles vaccination of under 5-years old people.

Discussion:

Up through this point of the research, notable links of different covariates have been identified to vaccination coverage. Some of the key points to focus on are the links between education and vaccination, crime and the socio-economic status of citizens and states, and the socio-economic status of vaccination. Continuously, it has been verified that education about the vaccine or education in general significantly correlates with vaccination coverage (Wiysonge et al., 2012). Likewise, simply living in a lower socio-economic area increases the likelihood of not being vaccinated (World Health Organization, 2019). As noted earlier, there is a link between income inequality and crime rates (Menezes et al., 2013), but not necessarily between crime rates and income level itself. To this point, an evident gap in the research present. Despite current knowledge on the impact geospatial covariates, education, and income on vaccination coverage, there is yet no clear link, on a national scale, between vaccination coverage and crime rates. In multiple cases, researchers could not go into some states in Nigeria and ask about vaccination coverage because of security risks (Barau, 2014). If this is the case, then the likelihood of citizens in perceiving the area as unsafe and also being limited to transport increases, creating a negative perception of vaccines (Niyibizi et al., 2016). Due to the following observations, this study will attempt to answer the question of how does the crime rate of Nigeria correlate with its vaccination coverage and if there is a correlation, does it prove to be more accurate than current covariates. The purpose of this research question is to identify areas of unvaccinated people which may have been overlooked in data due to security concerns or focus on other areas, such as education and income.

References

Adedire., E. B., Ajayi, I., Fawole, O. I., Ajumobi, O., Kasasa, S., Wasswa, P., & Nguku, P. (2016). Immunisation coverage and its determinants among children aged 12-23 months in Atakumosa-west district, Osun State Nigeria: a cross sectional study. BMC Public Health, 16(905). https://doi.org/10.1186/s12889-016-3531-x

Barau, I., Zubairu, M., Mwanza, M. N., & Seaman, V. Y. (2014). Improving polo vaccination coverage in Nigeria Through the use of geographic information system technology. Journal of Infectious Diseases, S102-110. doi: 10.1093/infdis/jiu010

Gunnala, R., Ogbuanu, I. U., Adegoke, O. J., Scobie, H. M., Uba, B. V.,m Wannemuehler, K.A., Ruiz, A., Elmousaad, H., Ohuabunwa, C. J., Mustafa, M., Nguku, P., Waziri, N. E., & Vertefeuille, J. F. (2016). Routine vaccination coverage in Northern Nigeria: results from 40 district-level cluster surveys, 2014-2015. PLOS ONE, https://doi.org/10.1371/journal.pone.0167835

Menezes, T., Silveira-Neto, R., Monteiro, C., & Ratton, L. J. (2013). Spatial correlation between homicide rates and inequality: evidence from urban neighborhoods. Economic Letters, 120(1), 97-79. https://doi.org/10.1016/j.econlet.2013.03.040

Niyibizi, I., Schamel, J., & Frew P. M. (2016). Neighborhood influences on seasonal influenza vaccination among older African Americans in Atlanta, Georgia. Journal of Immunological Techniques in Infectious Diseases, 5(2), 1-20. doi: 10.4172/2329-9541.1000139

Kremer, J. R., Nkwembe, E., Oyefolu, A. O. B., Smit, S. B., Pukuta, E., Omilabu, S. A., Tamfum, J. M., & Muller, C. P. (2010). Measles virus strains diversity, Nigeria and Democratic Republic of Congo. Emerging Infectious Diseases, 16(11), 1724-1730. https://dx.doi.org/10.3201/eid1611.100777

Sen, A. K. (1999). Development as Freedom. Anchor Books.

(2019, Dec. 5). Measles. World Health Organization. https://www.who.int/news-room/fact-sheets/detail/measles

(2019). Sustanable Development Goal 3. Sustainable development goals knowledge platform. https://sustainabledevelopment.un.org/sdg3

Umeh, C. A., & Ahaneku, H. P. (2013). The impact of declining vaccination coverage on measles control: a case study of Abia state Nigeria. Pan African Medical Journal, http://www.panafrican-med-journal.com/content/article/15/105/full/

Utazi, C.D., Thorley, J., Alegana, V. A., Ferrari, M.J., Takahashi, S., Metcalf, C. J. E., Lessler, J., & Tatem, A.J. (2018). High resolution age-structured mapping of childhood vaccination coverage in low and middle income countries. Vaccine, 36(12), 1583-1591. https://doi.org/10.1016/j.vaccine.2018.02.020

Wiysonge, C. S., Uthman, O. A., Ndumbe, P. M., & Hussey, G. D. (2012). Individual and contextual factors associated with low childhood immunisation coverage in sub-Saharan Africa: a multilevel analysis. Plos ONE, 7(5), 1-7. https://doi.org/10.1371/journal.pone.0037905

Wong, M. K, Yadav, R. P., Nishikiori, N., & Eang, M. T. (2013). The association between household poverty rates and tuberculosis case notifications rates in Cambodia, 2010. Western Pacific Survival and Response Journal, 4(1), 25 - 33. doi: 10.5365/WPSAR.2013.4.1.002