Assessing the Role of Different Environmental and Occupational Factors in the Spatial Distribution of Health Outcomes

Abstract

Complex epidemiological questions cannot be addressed with individual level analyses, but insights can be gained from aggregation and spatial methods. In this dissertation, I conducted three different epidemiological analyses in which spatial methods provided new insights to challenging questions. By using fine scale spatial methods and leveraging Bayesian statistics, I uncovered relationships between the physical, social, and infectious environment. In aim 1, I identified the relationship between cumulative P. falciparum burden and endemic Burkitt Lymphoma (eBL), the most common pediatric cancer which is fatal if untreated. I employed spatial estimates of annual P. falciparum burden to calculate lifetime P. falciparum infections for birth cohorts across three regions of Kenya, Uganda, and Tanzania to overcome the logistical and biological challenges of measuring an individual’s asymptomatic infections. After identifying a weak spatial correlation between the cumulative burden and eBL, I identified the relationship between endemic Burkitt lymphoma and lifetime P. falciparum exposure by region, age, and sex. I demonstrated that for every 100 lifetime P. falciparum infections, the risk of eBL increases 39% (95% CI: 6.1% – 81.0%) and this risk is greater among males and children between the ages 5-11. In aim 2, I used Michigan mortality records from January 2015 – November 2019 to predict the expected mortality from both acute respiratory infections and all other causes. I divided the pandemic period into distinct temporal periods from December 2019-November 2022 based on the strength of non-pharmaceutical interventions, vaccine availability, COVID-19 specific pharmaceutical treatments and shifts in dominant SARS-CoV-2 variants and compared the observed cause-specific mortality rates to the expected. Excess mortality from both ARI and non-ARI causes peaked during the first wave of the pandemic at 16.1 (13.44 -19.38) and 1.49 (1.43 – 1.55) excess deaths per expected deaths, respectively, and census tract level excess mortality rates are heterogenous across the state. Excess mortality, an aggregate metric with no analogous individual measurement, quantifies burden missed through traditional surveillance. In aim 3, I modeled the relationship between soil dust and drought, parameterized via the Standardized Precipitation Evapotranspiration Index with a 2-month lag (2-month SPEI). Exposure to soil dust has significant health burdens and is expected to increase under climate change. Using our modeled relationship between soil dust and 2-month SPEI, I estimated the future soil dust burden under climate change scenarios (Shared Socioeconomic Pathways; SSPs): SSP 2, a mild emissions scenario, and SSP 5, a more extreme emissions scenario. For every one unit increase in 2-month SPEI, indicating wetter soil, annual soil dust burden decreases 19.5% (95% CI: 19.4% – 19.6%). Under both climate change scenarios, overall dust burden is expected to increase in 2045-2055 and 2090-2100 and 243 (95% CI: 146-292) additional annual all-cause mortalities, 2913 (95% CI: 2334-3484) additional annual incidences of Alzheimer’s disease, and 356 (95% CI: 161-549) additional annual incidences of dementia. Using projected population projections at the county level, the expected increase in dust-mediated health outcomes is even greater. To summarize, the aims of this dissertation exemplify epidemiological questions that cannot be addressed with individual measures of exposures or outcomes, providing approaches to urgent problems in global and domestics infectious disease epidemiology. I hope this work inspires greater consideration for the application of spatial methods in situations where these methods are typically not utilized to examine the complex relationship between the environment and health.