Digital data streams and machine learning for real-time modeling of vaccine-preventable infectious diseases

Over the last 30 years, a new field--known as computational epidemiology (comp epi)--has emerged at the intersection of digital data streams (e.g., news and social media, search query, and mobility data), machine learning (e.g., nonlinear optimization, natural language processing, and agent-based modeling), and public health crises. Due to the ongoing SARS-CoV-2 pandemic, as well as other pathogens that have resurged in the United States (e.g., mumps virus), comp epi has shifted part of its focus as a field to improving public health decision-making during outbreaks and epidemics of infectious disease. In this proposal, we present four foundational challenges within the context of infectious disease research and comp epi more broadly. While the first three of these challenges are more conventionally scientific in nature, the fourth involves scientific community-building: (1) estimating the time-varying transmissibility (i.e., the effective reproduction number, REff) of a given infectious disease; (2) real-time monitoring and measurement of transmissibility-relevant behaviors (e.g., human mobility, etc.); (3) forecasting of outbreaks and epidemics as a function of individual health behaviors; and (4) recruitment of new scholars to the yet-insular field of comp epi. To address these challenges, we propose the development of (1) a meta-analytical tool for ensemble estimation of REff across multiple research groups; (2) a surveillance system to monitor transmissibility-relevant behaviors, including an inference system to produce more representative measures for human mobility; (3) a generalizable agent-based model for epidemic forecasting that features behavioral parameters, as informed by the aforementioned surveillance and inference systems; and (4) a virtual laboratory for comp epi scholars to collaborate on infectious disease research across institutional silos. By addressing the first three challenges, we hope to help clinicians and public health policymakers make data-informed decisions during infectious disease crises while simultaneously providing opportunities for other public health researchers to augment their own efforts in transmissibility estimation and epidemic forecasting by harnessing expected products from our proposed research. Meanwhile, by addressing the fourth challenge, we hope to help new scholars form meaningful collaborations both with pioneers in comp epi and with each other, while simultaneously promoting growth of the field as we move forward.