1. Infectious disease mapping
With relatively few exceptions, geographic mapping of infectious diseases relies on phenomenological descriptions of relationships between disease and its drivers and disregards a rich body of theory on transmission dynamics. The first theme of our research seeks to devise new approaches that leverage transmission dynamics theory to enhance the interpretability and utility of infectious disease mapping.
With relatively few exceptions, geographic mapping of infectious diseases relies on phenomenological descriptions of relationships between disease and its drivers and disregards a rich body of theory on transmission dynamics. The first theme of our research seeks to devise new approaches that leverage transmission dynamics theory to enhance the interpretability and utility of infectious disease mapping.
Zika: When Zika became a serious concern in early 2016, projecting the magnitude of the epidemic was a major priority for public health officials. Less than two weeks after the WHO declared the Zika epidemic an emergency, we published a preprint, and eventually a paper in Nature Microbiology, that made high-resolution projections of the number of people who could become infected. This was based on a method we developed to relate environmental variables to epidemic size based on theoretical relationships with the basic reproduction number, R0. This contrasted with other approaches, which applied distribution modeling to map much less interpretable metrics. Later work from us synthesized data from health surveillance and cohort studies to estimate that 132 million people (95% uncertainty: 111-170 million) may have been infected. This is similar in magnitude to our projection of 93 million (95% uncertainty: 81-117 million) at the beginning of the epidemic, which demonstrates the utility of this approach. These projections informed planning for Zika vaccine trials.
Yellow fever: We are developing new approaches for incorporating transmission dynamics into maps of yellow fever as part of the Vaccine Impact Modelling Consortium. So far, our approach has emphasized integration of multiple data types and has accounted for uncertainty around alternative assumptions about interpreting those data and other modeling choices. Ongoing work seeks to reduce uncertainty about those assumptions and to build models with more desirable features for capturing the relevant ecology and epidemiology of yellow fever. Work on this topic has been published here in Science Advances and here and here in eLife and has been used to inform investments in vaccination by Gavi.
Emerging diseases: We applied our framework for modeling yellow fever to several emerging zoonotic diseases prioritized by the WHO's R&D Blueprint, in support of efforts by the Coalition for Epidemic Preparedness Innovations (CEPI) to develop vaccines against these diseases. We recently published a preprint in which we applied this framework to project demand for vaccines during outbreak response for Lassa, MERS, Nipah, and Rift Valley fever. This work addresses a critical unknown in the vaccine development pipeline. In newly initiated work, we are leading modeling efforts for REDI-NET, a DoD-funded project seeking to advance capabilities for emerging infectious disease surveillance. Our models integrate historical epidemiological data with ongoing collection of pathogen detection data and seek to disentangle spatiotemporal patterns driven by separate processes related to transmission and surveillance.
2. Infectious disease forecasting
As has been demonstrated repeatedly throughout the COVID-19 pandemic, dynamic models that account for biological, social, and epidemiological factors have much to offer to public health decision making. The second theme of our research seeks to improve the scientific basis for infectious disease forecasting with dynamic, mechanistic models of transmission and other relevant processes.
As has been demonstrated repeatedly throughout the COVID-19 pandemic, dynamic models that account for biological, social, and epidemiological factors have much to offer to public health decision making. The second theme of our research seeks to improve the scientific basis for infectious disease forecasting with dynamic, mechanistic models of transmission and other relevant processes.
Zika: In addition to our real-time efforts during the Zika epidemic described under the first theme, we undertook an effort to retrospectively assess the performance of tools for forecasting that were available as the epidemic was unfolding, which we described in a paper in Nature Communications. We did so with support from an NSF RAPID grant and, later, in collaboration with UNICEF. Forecasting an emerging disease such as Zika is a major challenge, given that there are no historical data with which to train models. To address this challenge, we used Bayesian data assimilation to update a set of 16 alternative models iteratively over the course of the epidemic, finding that models were able to adjust and generate reasonable forecasts for the remainder of the epidemic after 8-12 weeks. Another finding from this work was that no single model from our set of 16 consistently performed best throughout the epidemic. Interestingly, we found that certain model features were most important during certain phases of the epidemic, suggesting that there may be potential in the future to learn from past epidemics in such a way that we can be better positioned to respond to future ones. Forecasting efforts during the COVID-19 pandemic corroborate our findings, suggesting that there is strong potential, and important demand, for further work on this topic.
Co-circulating pathogens: Our latest and most significant research project will build on our efforts to forecast Zika and COVID-19 with support from an R35 award from NIGMS through its MIRA program. In the wake of the Zika and COVID-19 pandemics, it has become apparent that, inevitably, these pathogens transition from epidemic to endemic circulation, alongside established pathogens with many ecological, immunological, and clinical similarities. Historically, however, the field of infectious disease modeling has been predominated by models of a single pathogen. To address the challenges that we face with these newly co-circulating pathogens—in one case, Zika, dengue, and chikungunya, and in another, SARS-CoV-2, influenza, and RSV, among others—we are developing new frameworks for modeling the dynamics of these pathogens.
3. Interventions for infectious disease prevention
Mathematical modeling has a key role to play in designing and interpreting studies to assess intervention efficacy and to make projections of the impact of interventions when deployed at population scales. The third theme of our research seeks to advance these capabilities for prevention of mosquito-borne and other diseases.
Mathematical modeling has a key role to play in designing and interpreting studies to assess intervention efficacy and to make projections of the impact of interventions when deployed at population scales. The third theme of our research seeks to advance these capabilities for prevention of mosquito-borne and other diseases.
Dengue vaccines: Our work has played a role in efforts to assess the potential impact of new and forthcoming dengue vaccines. This includes being one of eight groups to contribute to an effort organized by the WHO to project the public health impact of the Dengvaxia vaccine, described here in PLOS Medicine. Later, we published a paper that addressed questions about use of Dengvaxia following screening for prior infection, a requirement by WHO that followed its initial recommendation. This work then led to an opportunity to advise the CDC's Advisory Committee on Immunization Practices on their position on use of Dengvaxia in Puerto Rico, which we described here. With an eye towards broader questions around advancement of these vaccines, we published a paper that quantifies biases in estimates of dengue vaccine efficacy owing to complexities of dengue biology not accounted for in trial design. We published another paper that shows how uncertainty about breakthrough infections from dengue vaccine efficacy trials translates into considerable uncertainty in projections of a vaccine's impact
Spatial area repellents: We have been developing new statistical inference (here) and mathematical modeling (here) tools to confront challenges associated with the development of spatial area repellents (SARs) for mosquito-borne disease prevention. With support from a large award from Unitaid to support the Aegis project, we are extending these efforts to contribute to the analysis of data from epidemiological field trials of SARs for prevention of malaria and dengue. We are using a variety of modeling approaches to understand how SARs achieve their efficacy through different effects on mosquito behavior and life history. Additionally, we are translating estimates of how SARs achieve their efficacy into projections of what their impact would be in diverse contexts if scaled up and deployed by control programs in the event that they are recommended for public health use.
COVID-19: Early in the pandemic, we worked to inform the response to COVID-19. Recognizing the need to prioritize our efforts among the many topics that we could work on, we focused the majority of our activities on two themes. First, we adapted an existing agent-based model called FRED for modeling COVID-19, with a focus on the role of schools in driving community transmission. In advance of the fall 2020 semester, we presented work in a variety of fora, including at a MIDAS webinar on this topic, to the State of Indiana's pandemic response unit, and to the Ministry of Education in Bogotá, Colombia. As described in a paper in Epidemics, we stressed to these audiences the importance of high compliance with face masks and other preventive measures in schools, as well as the value of supporting families wishing to educate their children remotely in order to reduce densities within schools. Our model predicted that conditions in schools would play a major role in driving community transmission. Second, we collaborated with Army scientists to model strategies for preventing outbreaks in Army basic training, a vital function to ensure continued force strength. Our model offered new insights into transmission dynamics in this setting, including that drill sergeants and other staff pose a potentially greater risk for introducing the virus into this setting than the heavily tested and highly confined trainees that decision makers had directed most of their attention to. We are continuing to provide input to these decision makers as the pandemic evolves.
