Vaccine Dashboard

In partnership with the University of Alabama at Birmingham we have launched a dashboard displaying relevant information near-real time for vaccine distribution and administration in the USA and the world. This includes information on progress of vaccination rollouts as well as information on adverse events reported following vaccination. This dashboard is an ongoing project where the DICE data team works on leveraging available data to output predictive models that bring new insights into our progression towards herd immunity and the social impact of current strategies.

View it live:

Adverse Events

In collaboration with data scientists and medical researchers from Pennsylvania State University, Mayo Clinic and the University of Texas MD Anderson Cancer Center, we have been working on analyzing reports on adverse events following vaccination. The Vaccine Adverse Events Reporting System (VAERS) is a post-licensure monitoring system that has been established to ensure the safety of vaccines. The VAERS collects and analyzes reports of adverse events that happen after vaccination. As part of our commitment to provide systems support for vaccine safety, we are conducting various data analytics studies on adverse events reported to the VAERS from December 15, 2020. The output of this ongoing research takes the form of a dashboard and research papers that improve our understanding of adverse events and leverage the generated insights to contribute to vaccine awareness efforts. A first paper has been submitted to JAMIA.

Deep Agent Based Modeling (DeepABM)

The current sanitary crisis has come at great cost to our economies. Sacrifices have already been made in order to curb the spread and safe lives. As hard choices still remain ahead of us, it is paramount to have a comprehensive understanding of how public health interventions, economic incentives and strategic lockdowns impact can be optimized to allow for economic recovery. The dynamic of these interactions, taking place while the epidemiological dynamics of the virus spread run in the background, is poorly understood. In macroeconomics, agent-based models (ABM) have been successfully used to understand these types of interactions among agents and institutions. Recent studies incorporate the COVID-19 spread into economic studies with the goal of modelling these interactions and eventually serve as gym environments for intervention optimization analysis. One bottle-neck of such models is the computational constraints. Realistic models with a sizable number of agents and interactions are unrealistic from the computational standpoint. Together with a group from MIT, we have been working on a framework called deepABM that uses the notion of convolution in a graph neural network architecture to model interactions between agents. By construction, these interactions can be parallelized and existing frameworks like Pytorch Geometric can be leveraged to efficiently implement realistic ABM models. This project is currently exploring such one implementation.


Javier Álvarez, Carlos Baquero, Elisa Cabana, Jaya Prakash Champati, Antonio Fernández Anta, Davide Frey, Augusto García-Agúndez, Chryssis Georgiou, Mathieu Goessens, Harold Hernández, Rosa Lillo, Raquel Menezes, Raúl Moreno, Nicolas Nicolaou, Oluwasegun Ojo, Antonio Ortega, Jesús Rufino, Efstathios Stavrakis, Govind Jeevan, Christin Glorioso. “Estimating Active Cases of COVID-19”, Aug. 2021. [PAPER]

Rajat Jain, Utkarsh Gupta, Sethuraman TV, Rohan Sukumaran, Christin Glorioso MD PhD. "Analysis of Tata-1mg data for Covid-19 second wave prediction in India", Aug. 2021. [PAPER]