Identifying Vulnerable Communities for Infectious Disease Outbreaks

PROJECT SUMMARY The COVID-19 pandemic's unequal toll on racial and ethnic minority groups in the United States underscored that vulnerable communities need unique attention from public health officials to address health disparities stemming from a cumulative history of injustices. Compared to white Americans, Black and Hispanic Americans as well as indigenous populations have increased odds of hospitalization and higher deaths rates due to COVID- 19. A rapid, focused public health response is necessary for future outbreak preparedness, especially among minority populations that are more vulnerable to disease. Artificial Intelligence (AI) has been used to predict potential disease outbreaks; however, machine learning (ML), a branch of AI, has yet to be broadly used in identifying vulnerable populations and underserved communities at risk for disease outbreaks and track heterogeneities in risks at the neighborhood level. Furthermore, while disease incidence is often calculated at a county or zip code level, understanding heterogeneities in risk among neighborhoods in community transmission of diseases requires a more granular geographic unit for analysis. To this end, epidemiologic, geospatial, and machine learning tools to rapidly and accurately identify vulnerable neighborhoods based on local needs will be imperative to achieve health equity during infectious disease outbreaks. In Aim 1, we will explore associations and trends between respiratory infectious disease incidence (ex. influenza, tuberculosis, pertussis, and COVID- 19), vaccination coverage (MMR, DTaP, HPV, and influenza), and socioeconomic disadvantage considering geography in Philadelphia. Area Deprivation Index and Social Vulnerability Index will be used to measure socioeconomic disadvantage. Poisson and linear regression models will be used to find associations between infectious disease incidence, low vaccination coverage, and social determinants of health. Bayesian spatial regression modeling will be used to assess the change in the proportion of vulnerable communities affected by infectious diseases and identify any gaps in vaccination coverage differentially by neighborhood-level factors. In Aim 2, we will train a geographic information system (GIS)-based ML model, fit to the aggregated geospatial disease, vaccination, and social determinants of health data from Aim 1, and test its predictive capability on Philadelphia COVID-19 case data. Our goal will be to assess the predictive capability of GIS-based ML models on identifying areas for public health intervention. This innovative research will help us predict neighborhoods at risk of future infectious disease outbreaks and aid in timely identification of vulnerable populations to guide public health resources, which would be very useful for emergency preparedness efforts for future infectious disease outbreaks. The accompanying training plan consists of both didactic and experiential learning opportunities, and will enable the applicant to develop the skills and experience necessary to become an independent investigator and applied epidemiologist in the field of infectious diseases.