| |

Socioeconomic Influences on Non-Medical School Vaccination Exemption Hotspots in Colorado: A Spatial & Statistical Examination

Introduction

Vaccination is one of the most cost-effective public health interventions that protect against various infectious diseases, including measles, mumps, rubella (MMR), and diphtheria, tetanus, and pertussis (DTaP). Despite high overall coverage, localized clusters of vaccine exemptions can compromise herd immunity, posing a significant risk to public health. In Colorado, the incidence of non-medical exemptions has been on the rise (Salmon et al., 2005), prompting concerns over potential outbreaks of vaccine-preventable diseases.

This study, conducted in 2024 using the most recent 2022 school vaccination data from the Colorado Department of Public Health and the Environment (CDPHE), focuses on identifying clusters and hotspots of counties within Colorado with high percentages of non-medical DTaP and MMR vaccine exemptions. These exemptions are commonly claimed due to personal or philosophical beliefs, despite the vaccines’ established safety and efficacy. Previous research has highlighted how state policies, community socioeconomic factors, and localized health education efforts significantly influence vaccination rates and exemption patterns. For example, stringent state immunization laws and policies promoting adherence to vaccination guidelines are associated with higher vaccination coverage and lower exemption rates (Shaw et al., 2018).

Furthermore, socioeconomic factors such as household income, medical insurance coverage, educational attainment, racial demographics, and rural living conditions are known to influence vaccination behavior. Studies have shown that areas characterized by higher affluence may exhibit lower vaccination rates due to vaccine hesitancy and greater access to exemptions, whereas socioeconomically disadvantaged areas might face barriers due to access issues (Hegde et al., 2019). This study will employ a correlation analysis to explore these socioeconomic variables in the context of Colorado, aiming to provide a detailed understanding of the factors driving non-medical vaccine exemptions.

By mapping these exemption hotspots and analyzing associated socioeconomic covariates, this research aims to offer insights into targeted public health strategies and interventions. These can potentially enhance vaccine uptake and reduce the pockets of susceptibility to outbreaks of diseases that are otherwise preventable through vaccination.

Methodology

This study utilized a multi-source data integration approach to examine the spatial distribution of non-medical DTaP and MMR vaccine exemptions across counties in Colorado. The objective was to identify spatial clusters and examine the correlations between vaccine exemptions and various socioeconomic factors. The methodology encompassed data collection, spatial analysis, and statistical testing.

Data Collection and Preparation

The primary dataset comprised school-level vaccination exemption records obtained from the Colorado Department of Public Health and the Environment (CDPHE). This dataset included counts of DTaP and MMR vaccine exemptions, which were aggregated to the county level using spatial join techniques in a PostgreSQL database with PostGIS extensions.

Additionally, socioeconomic and demographic data were sourced from the American Community Survey (ACS) five-year estimates for 2022 and the 2020 U.S. Census. These datasets provided detailed information on education levels, poverty status, median household income, race/ethnicity, and urban/rural residency status for each county.

All datasets were aligned and integrated using a common geographic identifier for county regions. The combined data were then imported into a GIS environment for further spatial processing and analysis.

Spatial Analysis

Spatial empirical Bayesian (SEB) rate smoothing was performed using Geoda software. This technique was applied to adjust for small population sizes in counties and to smooth out the vaccination exemption rates, providing a more stable estimate across the study area. The smoothed rates were calculated by taking the count of exemptions as the event variable and the total school population as the base variable. These rates were then multiplied by 100 to derive the smoothed percentage of exemptions per county.

Using ArcGIS Pro, several spatial analyses were conducted:

  • Rate Mapping: SEB rates for MMR and DTaP were visualized through choropleth mapping to identify geographical patterns and distributions.
  • Kriging: Empirical Bayes Kriging was used to interpolate exemption rates, creating continuous surface maps that highlight gradients and potential hotspots not immediately apparent in discrete data.
  • Hotspot Analysis: The Getis-Ord Gi* statistic was employed to identify statistically significant hotspots and cold spots of vaccine exemptions.
  • Cluster and Outlier Analysis: The Anselin Local Moran’s I was used to detect spatial clusters of high exemptions (high-high) and spatial outliers where high exemption rates were surrounded by low rates (high-low) and vice versa.

Statistical Analysis

Spearman correlation coefficients were calculated to assess the strength and direction of associations between the vaccine exemption rates and variables such as median household income, insurance coverage, educational attainment, racial demographics, and rural residency. This method evaluates the monotonic relationships without making assumptions about the frequency distribution of the variables.

  • Implementation: The analysis was conducted using the cor.test function from the base R package, applied iteratively across the selected socioeconomic variables.
  • Results Interpretation: Correlation coefficients and their significance were tabulated and visualized to provide clear insights into how each socioeconomic factor may influence vaccine exemption rates.

For significant correlations, scatter plots were generated to depict the relationships between vaccine exemptions and influential factors. Polynomial regression lines were added to these plots to illustrate the nature of the relationships, whether linear or nonlinear.

All statistical analyses and visualizations were conducted using R, leveraging packages such as ggplot2 for graphical output and sf for spatial data handling. The comprehensive use of these tools not only facilitated a robust analysis but also ensured that the findings were communicated effectively through high-quality graphical representations.

Visualization

The results of all spatial and statistical analyses were visualized using ArcGIS Pro for mapping and R software for creating scatter plots and correlation matrices. These visualizations facilitated an understanding of complex spatial relationships and correlations, aiding in the effective communication of findings to public health stakeholders and policymakers.

Results

Spearman’s Correlation

To determine if the variables violated the assumption of normality of Pearson’s correlation, normality was checked using the Shapiro-Wilk test.

Shapiro-Wilk Test Results for Normality Assessment
Variable Shapiro-Wilk Test Statistic P-Value Normality Assumption
MMR Exemption Percentage 0.6676 <0.001 No
DTaP Exemption Percentage 0.6915 <0.001 No
Median Income 0.5281 <0.001 No
Percent Insured 0.9659 0.0742 Yes
< 9th Grade Education 0.9264 <0.001 No
9-12th Grade Education 0.9608 0.0403 No
High School Graduate 0.9741 0.1961 Yes
Bachelor’s Degree or Higher 0.9297 0.0013 No
% White Residents 0.9060 <0.001 No
% Rural Residents 0.8213 <0.001 No

The results of the Shapiro-Wilk tests indicate statistically significant defiance from the normal distribution in the variables. A check of the histograms, density plot (blue curve), and normal curve (red curve) is further visual conformation of this. In these plots, mmr_ex_seb represents the MMR exemption percentage, and dtap_ex_seb represents the DTaP exemption percentage.

Due to the violation of the assumption of normality of regression and Pearson correlation, we will use the non-parametric Spearman correlation test to determine if there are statistically significant relationships between the dependent and independent variables. The results of the Spearman tests follow.

Correlation Results With Variables of Interest
Variable MMR Exemption Coef. MMR Exemption P-Val. DTaP Exemption Coef. DTaP Exemption P-Val.
Median Household Income -0.2567 0.0406 -0.2604 0.0377
Percent with Medical Insurance -0.1143 0.3683 -0.1111 0.3821
Less Than Ninth Grade -0.4041 <0.001 -0.3990 0.0011
Ninth to Twelth Grade -0.2227 0.0769 -0.2183 0.0831
High School Graduate -0.0887 0.4858 -0.0803 0.5282
Bachelor’s Degree or Higher 0.1436 0.2576 0.1353 0.2864
Percent White Alone 0.5451 <0.001 0.5587 <0.001
Percent Living in Rural Area 0.4097 <0.001 0.4182 <0.001

The significant relationships were plotted in scatterplots for visualization. The median household income, with a moderately strong negative relationship with both non-medical MMR and DTaP vaccination exemptions is shown first. Most of the observations are at the low end of the spectrum, with the median between $0 and $50,000 earned in the past twelve months for most counties.

The next significant relationships were between education levels and vaccination exemptions. The first, percentage with less than a ninth grade education, is shown below. Again, more than half of the observations are at the low end, less than 5%. But the relationship does appear to be a negative one, if nonlinear. The correlation coefficients were around -0.4 for both MMR and DTaP exemptions, indicating a moderate negative relationship, where an increase in those with less than a ninth grade education decreases the rate of DTaP and MMR non-medical vaccination exemptions.

The relationship between those with a ninth to twelfth grade education and vaccination exemptions, while not statistically significant at alpha=0.05, were each around p=0.08 and show similar relationships to the less than ninth grade results. Again, the majority of the observations were on the low end of the range, at less than 7.5%. Again, ,the relationships are negative, both at around -0.22, indicating that as the percentage of people who have a ninth-twelfth grade education increases, the rate of DTaP and MMR vaccination exemptions decreases.

The next statistically significant relationship was between the percentage of people who are white and the MMR and DTaP vaccination exemption rates. Both coefficients were around 0.55, indicating that as the percentage of people who identified as white increases, the percentage of MMR and DTaP vaccination exemptions increased.

The last significant relationship was between the percent of people living in rural areas and MMR and DTaP vaccination percentages, which showed a moderate relationship at around 0.41 for both vaccines. The distribution of data was interesting for the percent of people living in rural areas, as there were a large number of counties at 100%. This will make sense once we look at the cluster and hotspot maps.

Spatial Analyses

Rate Maps (Choropleth)

The first step of the spatial analysis was to create choropleth maps showing the areal distribution of the SEB vaccination exemption rates across the state. As shown in the two maps below, the Southwestern portion of Colorado appears to have higher rates of non-medical vaccination exemptions than the rest of the state, with a general trend for higher rates toward the West. The two maps are similar, except that the MMR rates are overall higher than those of the DTaP map.

Rate Maps (Heatmaps)

I also employed Empirical Bayes Kriging (EBK) to interpolate and smooth the non-medical DTaP and MMR vaccine exemption rates across Colorado counties. EBK is an advanced geostatistical technique that incorporates both the empirical Bayesian approach and kriging to optimize the prediction of spatial trends. This method adjusts the weight of the semivariogram model based on the distribution and density of the sample data, enhancing the accuracy and reliability of the spatial predictions.

For the SEB smoothed DTaP exemption rates, the results indicated an average CRPS (Continuous Ranked Probability Score) of 17.16, reflecting the predictive accuracy of the model within a reasonable range. Approximately 89.06% of the predictions fell within the 90% confidence interval, and 93.75% were within the 95% confidence interval, suggesting a high level of model reliability. The mean prediction error was -0.62, with a root-mean-square error of 38.48, and the root-mean-square standardized value approached unity (0.986), indicating that the model’s residuals are well standardized around the mean. The average standard error was approximately 34.49, providing further insights into the variability of the predictions.

Similarly, for the SEB smoothed MMR exemption rates, the analysis yielded an average CRPS of 17.59. This slightly higher CRPS for MMR rates suggests a comparable level of predictive accuracy as observed with the DTaP rates. About 90.63% of the predictions fell within the 90% confidence interval, and 93.75% were within the 95% confidence interval. The mean prediction error recorded was -0.76, with a root-mean-square error of 40.28. The mean standardized and root-mean-square standardized values were 0.00195 and 0.999, respectively, indicating an effective normalization of residuals. The average standard error was noted at 35.08.

These results highlight the effectiveness of EBK in modeling spatial variations in vaccine exemption rates, providing a robust framework for identifying areas with higher risks of under-vaccination. The upcoming maps will visually depict these spatial distributions, offering a geographic perspective on the regions that may require targeted public health interventions to enhance vaccine coverage and reduce exemption rates. The spatial distributions of the two vaccine exemption rates are highly similar, with the Western portion of the state exhibiting significantly higher rates. This is in agreement with the significant Spearman correlation between percent living in rural areas and the vaccination rates, as Western Colorado is highly rural.

In the spatial analysis, hotspot analysis using the Getis-Ord Gi* statistic identified regions with significantly high non-medical SEB vaccination exemptions. The analysis highlighted consistent hotspots for both DTaP and MMR exemption rates across Southwestern Colorado counties. Notably, Archuleta and Ouray counties were identified with higher confidence as hotspots in the MMR vaccination rate analysis compared to DTaP. This variation underscores subtle differences in local exemption patterns, emphasizing areas that may require focused public health interventions.

Following the hotspot analysis, the study further examined the spatial distribution of vaccination exemptions using the Cluster and Outlier Analysis (Anselin Local Moran’s I) in ArcGIS Pro. This method distinguishes between clusters of high values (high-high) and low values (low-low), as well as identifying outliers where high exemption rates occur near low rates (high-low) and vice versa. This analysis provides a nuanced view of the spatial patterns, highlighting not only where exemptions are concentrated but also where unusual patterns may indicate underlying socio-economic or demographic factors influencing vaccination behaviors.

The Cluster and Outlier Analysis identified significant spatial clusters and anomalies in vaccination exemption rates across Colorado counties. Gunnison and Hinsdale counties emerged as significant high-high clusters for both MMR and DTaP vaccination exemptions, indicating higher than average exemption rates. Additionally, Ouray county was also marked as a high-high cluster specifically for MMR exemptions. On the other hand, Mineral county was consistently identified as a Low-High outlier for both vaccines, suggesting lower exemption rates adjacent to counties with higher rates. Similarly, Rio Grande and La Plata counties were noted as Low-High outliers for DTaP and MMR exemption rates, respectively. These findings highlight specific areas with unusual exemption patterns, underscoring the need for targeted public health interventions in these counties.

Discussion

This study’s analysis of non-medical DTaP and MMR vaccine exemptions across Colorado has highlighted several key insights into the interplay between vaccine exemption rates and various socioeconomic factors. The spatial distribution of exemptions, characterized by significant clusters in certain counties, points to a complex landscape influenced by both sociopolitical and economic variables.

Sociopolitical Influences

Colorado’s political landscape, with a mix of rural conservative areas and urban liberal zones, plays a crucial role in vaccination behaviors. The identified clusters and hotspots in counties like Gunnison, Hinsdale, and Ouray, which are known for their distinct sociopolitical identities, suggest that local politics may significantly influence public health practices and perceptions. These regions may exhibit higher exemption rates due to varying levels of trust in government and differing views on individual rights and public health policies.

Socioeconomic Factors

In the correlation analysis of socioeconomic factors affecting non-medical vaccination exemptions in Colorado, the results indicate that socioeconomic conditions significantly impact vaccination behaviors:

  • Median Household Income: The analysis revealed that an increase in median household income is associated with a decrease in vaccination exemptions. This trend suggests that higher economic status may correlate with better access to healthcare and education about the benefits of vaccination, thereby reducing exemption rates.
  • Education Levels: Specifically, an increase in the proportion of the population with less than a ninth-grade education correlates with a decrease in vaccination exemptions. This could reflect targeted public health interventions or community outreach programs that effectively promote vaccination in lower-educated populations, possibly overcoming barriers to healthcare access or misinformation.
  • Demographic and Geographic Factors: Increases in the percentage of white residents and rural residency were found to increase vaccination exemptions. These trends might reflect cultural or ideological factors prevalent in predominantly white and rural communities, where there may be greater vaccine hesitancy or skepticism about public health mandates.

These findings underscore the complex interplay of income, education, and community characteristics in influencing public health outcomes, particularly in the context of vaccination uptake. The relationships highlight the need for nuanced public health strategies that consider both socioeconomic and demographic factors to effectively address and reduce vaccination exemptions.

Implications for Public Health Strategies

The spatial and statistical analysis conducted provides critical insights for targeted public health interventions. The identification of high-high clusters and low-high outliers underscores the need for localized health education campaigns tailored to the specific beliefs and socioeconomic conditions of these communities. Public health strategies could include:

  • Enhanced Outreach Programs: Focused on regions identified as high-risk areas, these programs could address vaccine hesitancy through community engagement initiatives that build trust and relay accurate vaccine information.
  • Policy Adjustments: Considering the political sensitivities around vaccination, policy-makers might need to navigate carefully to enhance vaccine uptake without infringing on personal freedoms, possibly through incentivization rather than compulsion.
  • Education and Access: Increasing access to healthcare services in underprivileged areas and enhancing educational efforts about the importance of vaccinations can help mitigate some of the socioeconomic barriers identified.

Conclusion

This study’s comprehensive examination of non-medical vaccination exemptions for DTaP and MMR in Colorado counties has provided valuable insights into the interplay between sociodemographic factors and public health practices. Through rigorous spatial and statistical analyses, we identified significant spatial clusters of exemptions and correlated these patterns with various socioeconomic variables.

Our findings indicate that higher median household incomes and higher levels of education (specifically less than a ninth-grade education) are associated with lower rates of vaccination exemptions. This suggests that economic stability and targeted educational interventions may effectively reduce hesitancy and improve vaccination rates. Conversely, an increase in the percentage of white and rural residents was associated with higher exemption rates, highlighting the influence of cultural and geographic factors on public health behaviors.

The spatial analyses using Empirical Bayes Kriging and hotspot detection provided a detailed map of the geographical distribution of exemptions, revealing areas where public health interventions could be most needed. By identifying counties such as Gunnison, Hinsdale, and Ouray as significant clusters, this research underscores the importance of tailored public health strategies that address local socioeconomic and cultural dynamics.

This study underscores the complexity of vaccine hesitancy and the multifaceted approach needed to address it. Public health strategies that consider the nuanced realities of different communities will be crucial in increasing vaccination coverage. Future research should focus on longitudinal studies to track the impact of interventions over time and explore the underlying reasons for exemptions through qualitative research to better tailor public health messaging and policies.

By enhancing our understanding of the factors that influence vaccination behaviors, we can better design and implement strategies that not only increase vaccination rates but also strengthen the resilience of public health systems against outbreaks of preventable diseases.

Limitations and Further Research

While this study provides a comprehensive overview of the factors influencing vaccine exemptions in Colorado, the findings are context-specific and may not be generalizable to other states or regions without similar sociopolitical and socioeconomic contexts. Future research could expand this approach to a multi-state analysis to explore the broader implications of these findings. Additionally, qualitative studies could provide deeper insights into the personal reasons behind vaccine hesitancy and exemption, complementing the quantitative data presented here.

References

  1. Shaw, J., Mader, E. M., Bennett, B. E., Vernyi-Kellogg, O. K., Yang, Y. T., & Morley, C. P. (2018). Immunization mandates, vaccination coverage, and exemption rates in the United States. Open Forum Infectious Diseases, 5(6).
  2. Hegde, S. T., Wagner, A. L., Clarke, P. J., Potter, R. C., Swanson, R. G., Boulton, M. L., & Dombkowski, K. J. (2019). Neighbourhood influences on the Fourth Dose of Diphtheria-Tetanus-Pertussis Vaccination. Public Health
  3. Williams, J. T. B., & O’Leary, S. T. (2019). Denver Religious Leaders’ Vaccine Attitudes, Practices, and Congregational Experiences. Journal of Religion and Health, 58, 1356-1367.
  4. Salmon, D. A., Moulton, L. H., Omer, S. B., DeHart, M. P., Stokley, S., & Halsey, N. A. (2005). Factors associated with refusal of childhood vaccines among parents of school-aged children: A case-control study. Archives of Pediatrics & Adolescent Medicine, 159(5), 470-476.
  5. Nicolich, K., Gerken, J., Mallahan, B., Ross, D. W., & Zapata, I. (2022). Preventable Disease, the Case of Colorado: School District Demographics and Childhood Immunizations. Vaccines, 10.
  6. Mohanty, S., Buttenheim, A., Joyce, C. M., Howa, A. C., Salmon, D., & Omer, S. (2018). Experiences With Medical Exemptions After a Change in Vaccine Exemption Policy in California. Pediatrics, 142.

Similar Posts

Leave a Reply