Date of Award

2019

Document Type

Dissertation - MUSC Only

Degree Name

Doctor of Philosophy (PhD)

Department

Public Health Sciences

College

College of Graduate Studies

First Advisor

John E. Vena

Second Advisor

Andrew B. Lawson

Third Advisor

Erik R. Svendsen

Fourth Advisor

John L. Pearce

Fifth Advisor

Kathryn K. Cristaldi

Sixth Advisor

Allison E. Aiello

Abstract

Asthma is the leading chronic pediatric disease in the US and globally, making it a serious and costly public health issue. Emergency department (ED) visits for asthma also exhibit large disparities across many factors (e.g., race) that are not well understood. Furthermore, socioenvironmental determinants of health disparities have geographic components, and thus a clearer picture of the role that “place” has can better elucidate risk factors, synergistic effects, and improved interventions. We used the pediatric population (ages 5-19) of South Carolina (SC), a state with previously identified disparities in many health outcomes across numerous factors, as a case study. We employed a spatio-temporal Bayesian hierarchical modeling (BHM) framework to estimate associations between socioenvironmental factors, such as air pollutants and dimensions of socioeconomic status, and asthma ED visits. Methodologic innovations included an assignment algorithm for missing geographic identifiers of records, systematic variable selection, random effects for unmeasured confounding, geographically adaptive regression, spatio-temporal quantile regression, and a BHM case crossover model. Long-term (i.e., annual) and short-term (i.e., daily) models were used to detail the health effects of chronic and immediate socioenvironmental exposures, respectively. Key socioenvironmental factors associated with increased long-term asthma ED visit risk were demographics, distance to the nearest pharmacy, average household size, and complex interactions involving carbon monoxide (CO). These factors contributed to high risk clusters in southeastern SC, north of Columbia, the Lower Savannah River District, and particular urban areas. However, associations with ED visit risk varied spatially by census tract within the state, and additionally varied by percentile of risk. When considering environmental factors immediately preceding an ED visit, nitrogen oxides (NOx) and particulate matter sizes <10µm (PM10) were implicated across seasons. In the Fall, the highest asthma burden season, asthma ED visits were associated with PM10 and elemental carbon (EC). The PM10 association likely captures an association with Fall allergens, such as respirable ragweed antigens. Our spatio-temporal methods helped identify and locate health disparities for asthma in children, in addition to quantifying associations with potential long- and short-term drivers. Numerous place-based socioenvironmental exposures we identified should be considered by researchers and policymakers attempting to reduce health disparities. Specific Aims: Asthma is the leading chronic pediatric disease in the US, making it a serious and costly public health issue. Children visit the emergency department (ED) due to asthma for a variety of reasons that include neighborhood factors and patterns in healthcare. Asthma ED visits also exhibit large disparities across many social determinants of health (including local environment). Past research mainly highlighted increased disparities for urban locations, though more recent studies have detailed rural burdens of pediatric asthma ED visits, especially in the rural South. Much can still be learned and little research has contrasted disparities across different contextual factors, especially rural/urban residence. Furthermore, determinants of health disparities have geographic components, and thus a clearer picture of the role that “place” has in the asthma burden could better elucidate risk factors and their synergistic effects, leading to improved interventions to address disparities. Problematically, most previous studies in this field did not consider local geography, a scenario that can result in reduced precision of estimates due to assigning variation in the outcome to error terms rather than an effect of location (i.e., spatial confounding). Improving knowledge of asthma’s determinants has been highlighted as a research priority, and separating the effects of a variety of socioenvironmental risk factors that drive disparities is an ongoing challenge in epidemiologic research. Integrating spatial, environmental, and social epidemiology can support this mission, as improved assessment of risk factors over space and time can improve public health strategies for addressing disparities in this complex disease. The overarching goal of this research was to improve the understanding of socioenvironmental risk factors driving disparities in pediatric asthma ED visits. The primary hypothesis for this proposal was that disparities in such risk factors would associate with asthma- related health disparities in children and that these associations would be modified by characteristics of place. To address this, we employed innovative statistical methodologies to analyze highly resolved spatio-temporal data comprised of health outcomes, demographic and neighborhood information, and environmental measures from South Carolina, a diverse population that exhibits health disparities across numerous factors. Furthermore, we assessed differences in results and interpretations from two strategies to address missing census tract identifiers: geographic imputation and complete case analyses. In addition, we had access to novel highly resolved air pollution exposure data as well as advanced climatic estimates for covariates. The following specific aims and supporting hypotheses were proposed to investigate our primary hypothesis. Aim 1: Specific Aim 1: Quantify the drivers of long-term asthma disparities. Hypothesis 1.1: social and environmental neighborhood-level factors would be associated with risk for asthma ED visits that drove disparities. We used a Bayesian space-time “convolution” model of neighborhood-level asthma ED counts, incorporating covariates and random effects, and contrasted results for datasets employing geographic imputation versus complete case analyses. Results improved understanding of exposures and neighborhood-level risks associated with long- term asthma disparities. Aim 2: Specific Aim 2: Identify the spatio-temporal health disparity patterns in pediatric asthma. Hypothesis 2.1: Excess risk for pediatric asthma ED visits would cluster after known neighborhood risk factors were controlled. We used mapping, Bayesian exceedance probability clustering, quantile regression, and geographically adaptive regression to contrast results from geographic imputation against complete case analyses, in addition to a space-time uncorrelated random effect. Aim 3: Specific Aim 3: Quantify the drivers of short-term asthma disparities. Hypothesis 3.1: Acute, temporarily adverse environmental factors would modify the seasonal effects on individual ED visits after controlling for confounding. Within a case-crossover design extended to account for spatial confounding, we used logistic regression to model individual-level pediatric asthma ED visits. Results were important for detailing the modifying effects of seasonal variability in short-term exposures to environmental factors for asthma ED visits. The immediate impacts of this work identified, disaggregated, and refined understanding of the neighborhood factors that interacted and contributed to pediatric asthma ED visit disparities. The long-term impacts included advancing epidemiologic analysis of disparities in pediatric asthma, a complex and multifactorial disease that has proven challenging to study with existing methods. The research informed future case-control and cohort research designs of the importance for both collecting and analyzing highly precise spatio-temporal data to better detail health disparities. Furthermore, this research generated tangible location-specific evidence to direct interventions to reduce disparities in ED visits and their heavy human and financial costs.

Rights

All rights reserved. Copyright is held by the author.

Download

Share

COinS