MONOPOLI: a customizable model to provide forecasts of Covid-19 infections around the world using alternative non-pharmaceutical intervention policy scenarios, human movement data, and regional demographics

Arehart, Chris H.; Arehart, Jay H.; David, Michael Z.; D'Amico, Bernardino; Sozzi, Emanuele; Dukic, Vanja; Pomponi, Francesco

Abstract

During the global COVID-19 pandemic, policy makers, public health practitioners, medical experts, and laypersons have sought data that would enable evidence-based decisions about which interventions would be most effective at slowing the spread of SARS-CoV-2 disease. COVID-19 incidence curves have differed by country and region, and there has been great variability in the strategies adopted by governments around the world to respond to the pandemic. In the present study, measurements including 156 regions’ (105 countries, 50 United States, and Washington DC) confirmed case counts, demographics, socioeconomics, geography, government interventions, and changes in human mobility were included in a random forests modeling framework to predict the daily COVID-19 effective reproduction number (R(t)) through November of 2020. Variable selection methods were used to identify variables of high importance for transmission rates for this time-period before vaccines became available. Furthermore, the R(t) estimation is coupled with a susceptible-exposed-infectious-recovered (SEIR) epidemiologic model to obtain short- and long-term forecasts of the number of infections in each region over time.
Thus, the modeling and data visualization tool named MONOPOLI (Modeling Of NOnPharmaceutical Observed Long-term Interventions) offers real-time estimates and forecasts of R(t) under different nonpharmaceutical intervention (NPI) scenarios, while accounting for human mobility and demographic variables in each region. MONOPOLI can answer multi-intervention questions, both in
retrospect (hindcasting), as well as in the future (forecasting) contexts, including questions such as “what if country A were able to do this, at a specific time?” or “what if country B does this now?” The models for R(t) are dynamic to user input, and relative to a specified date, this method can illustrate the following: (1) What would happen under policy status quo from that date onward; (2) what would have happened in the past if a certain set of policies had been implemented; and (3) what is predicted to happen in the future under such policies. The United Kingdom is shown as an example to showcase the model’s capabilities, and detailed results are provided for all 156 countries in the Supplementary Material.

Citation

Arehart, C. H., Arehart, J. H., David, M. Z., D'Amico, B., Sozzi, E., Dukic, V., & Pomponi, F. (in press). MONOPOLI: a customizable model to provide forecasts of Covid-19 infections around the world using alternative non-pharmaceutical intervention policy scenarios, human movement data, and regional demographics. In B. Sriraman (Ed.), Handbook of Visual, Experimental and Computational Mathematics - Bridges through Data (1-29). Springer. https://doi.org/10.1007/978-3-030-93954-0_2-1

• Acceptance Date: Jul 8, 2023
• Online Publication Date: Feb 8, 2024
• Deposit Date: Jul 10, 2023
• Publicly Available Date: Feb 9, 2026
• Publisher: Springer
• Pages: 1-29
• Book Title: Handbook of Visual, Experimental and Computational Mathematics - Bridges through Data
• DOI: https://doi.org/10.1007/978-3-030-93954-0_2-1
• Keywords: COVID-19, policy, nonpharmaceutical interventions, pandemic, case reproduction number, machine learning
• Public URL: http://researchrepository.napier.ac.uk/Output/3143061
• Publisher URL: https://link.springer.com/referencework/10.1007/978-3-030-93954-0