Date of Award

4-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Program

Biomedical Sciences

Track

Microbiology, Immunology, and Biochemistry

Research Advisor

Minghui Li, PhD

Keywords

Artificial Intelligence; Bias Mitigation; COVID-19; Pandemic Management; Prediction Model; Testing Protocol

Abstract

The COVID-19 pandemic, along with prior outbreaks such as SARS and Ebola, has revealed critical gaps in global pandemic preparedness, emphasizing the need for integrated approaches to infectious disease management. This dissertation addresses three key areas of pandemic response: developing scalable screening strategies, constructing reliable predictive models, and mitigating demographic biases in artificial intelligence (AI) models used for healthcare decision-making.

The first study introduces a global screening strategy designed to optimize resource use and increase testing capacity, particularly in low- and middle-income countries (LMICs) where resources are limited. This study leverages pooled testing methods, where multiple samples are combined into a single test to conserve resources, and advanced mathematical models to ensure accuracy. The stratified, algorithm-guided, multi-level testing strategies developed in this research dynamically adjust testing frequency and pool sizes based on infection prevalence, allowing large-scale screenings to be conducted efficiently without compromising detection accuracy. Simulations showed that pooled testing can significantly reduce the number of tests needed in regions with low infection rates while maintaining reliable results. This method has the potential to be deployed not only for COVID-19 but also for other infectious diseases that require mass testing, especially in resource-constrained environments.

The second study focuses on the construction of a predictive model for real-time forecasting of COVID-19 infections and mortality. The model integrates real-time data from countries with different levels of epidemic control and is designed to adjust predictions continuously as new data becomes available. Factors such as population density, healthcare infrastructure, and public health interventions are incorporated into the model to improve its adaptability and accuracy. The model demonstrated its utility in predicting both the rise and fall of infection rates, as well as the effectiveness of various interventions. Countries that implemented early and strict public health measures saw reductions in predicted mortality, underscoring the importance of timely action in pandemic management. The predictive model provides governments and public health officials with valuable insights for resource allocation and intervention planning during ongoing and future pandemics.

The third study investigates the demographic biases inherent in AI models used to predict COVID-19 mortality, with a particular focus on racial and ethnic minority groups. Existing AI models often underperform for minority populations, exacerbating health disparities. Using population-based data from the Centers for Disease Control and Prevention (CDC), the study applied transfer learning techniques to adapt AI models trained on majority populations to better predict outcomes for underrepresented groups. The study focused on improving predictive accuracy for Non-Hispanic Black, Asian, and American Indian populations. Results demonstrated significant improvements in model fairness and performance, particularly in reducing mortality prediction bias for minority groups. The study highlights the importance of ensuring that AI models used in healthcare are equitable and do not perpetuate systemic biases that disproportionately affect vulnerable populations.

Together, these three studies provide a comprehensive approach to improving global pandemic preparedness. By addressing critical gaps in scalable testing, predictive modeling, and equitable healthcare delivery, this dissertation contributes to the development of more robust, adaptable, and fair strategies for managing future pandemics. The research offers valuable insights for policymakers, public health officials, and AI developers, ensuring that public health interventions are not only effective but also inclusive and equitable, particularly for underserved populations.

Declaration of Authorship

Declaration of Authorship is included in the supplemental files.

ORCID

https://orcid.org/0009-0009-0287-3423

DOI

10.21007/etd.cghs.2025.0684

Additional Files
2025-001-Gu-DOA.pdf (202 kB)
Download

Share

COinS