Investigating Asthma Burden Attributable to Housing Conditions via High-resolution Geospatial Analyses

Abstract

Asthma is a major public health problem that contributes substantially to health care use and health disparities in the United States. Housing conditions are increasingly recognized as an important contributor to asthma burden and disparities, particularly through exposure to in-home asthma triggers associated with poor-quality housing environments. However, important research gaps remain in understanding how housing conditions relate to asthma burden and disparities at the population level and in identifying practical and cost-effective ways to support large-scale housing interventions. Several aspects of these research gaps were investigated through three studies in this dissertation. The first and second studies examined how housing conditions are associated with asthma burden at the population level using tenant-reported housing data and electronic health records. The first study investigated this association for adult asthma in Boston, Massachusetts, and found that higher rates of tenant reports of in-home asthma triggers were significantly associated with higher rates of adult asthma emergency department (ED) visits at the census block group level, after accounting for confounders and spatial autocorrelation. The second study further examined this association for childhood asthma in a different urban setting, Austin, Texas, and found that the positive association between housing conditions and asthma ED visits also existed for childhood asthma, even after accounting for sampling bias when using data from a single hospital. Together, these findings provide a more comprehensive picture of the population-level associations between housing conditions and asthma burden. The third study investigated the feasibility of using municipal data and electronic health records to predict poor housing conditions. By integrating tenant requests for housing code inspection, municipal property parcel data, and asthma-related ED visit information into machine-learning models, we showed that failed housing inspections could be predicted with acceptable performance, especially among larger properties. Tabular Prior-data Fitted Network, a tabular foundation model, outperformed Random Forest and Categorical Boosting and showed promising precision in identifying the highest-risk inspections, suggesting practical value for proactive targeting of housing interventions. In closing, these studies contribute new knowledge to the complex relationship between housing conditions and asthma burden by strengthening population-level evidence on the relationship between housing conditions and asthma, and by demonstrating the feasibility of using accessible data sources to support targeted interventions at scale.