AIM/MCRN Summer School on COVID-19: Day 4

On the fourth day of the summer school on “Dynamics and Data in the COVID-19 Pandemic”—organized by the American Institute of Mathematics (AIM) and the Mathematics and Climate Research Network (MCRN), participants heard from experts in the morning and practitioners in the afternoon. Experts, whose specialties ranged from mathematical epidemiology to biomathematics and statistics, included Codelia McGehee (University of Minnesota), Jack O’Brien (Bowdoin College), Pauline van den Driessche (University of Victoria), and Jianhong Wu (York University). Andrew Roberts and Nick Ma (both of Cerner Corporation) served as the practitioners.

What are the important questions and considerations?

 Participants offered many suggestions. Here are a few (in no particular order):

• Be aware of the different ways in which clinical trial data are reported
• Develop granular (meta)population models for specific communities, such as long-term care facilities, homeless people, prison populations, etc.
• Introduce spatial variation, group differences, and delays in ordinary differential equation models
• Design and evaluate strategies for de-escalation, including social distancing, stratified lockdowns, reopening the economy, school reopenings, etc. 
• Devise data collection schemes for global monitoring of infectious diseases (citizen science?) and use livestock for the pilot project
• Study the effect of multiple simultaneous epidemics via co-infection models
• Investigate scenarios pertaining to the timing of recovery transitions for long-term care facilities separately from the general population
• Different diseases may have shared symptoms. How do we distinguish them?
• How can we capture resilience—of an infrastructure, for example—in a mathematical model? What type of data do we need to assess resilience?
• Can we model the influence of human activities (e.g., land use changes) on the spread of zoonotic diseases?
• There seems to be a one-week cycle in daily infection rates; is there a possible correlation with weekend sociability (stochastic forcing)?
• What have we learned? Epidemics will continue to occur, so we must develop strategies for avoiding future outbreaks.

COVID-19 forecasting with Cerner Intelligence

• Curve fitting, informed by susceptible-infectious-recovered (SIR) models has had limited success
• Institute for Health Metrics and Evaluation model, granularity down to state level
• SIR, susceptible-exposed-infected-recovered (SEIR) models, parameter estimation, role of \(\beta\), and uncertainty
• Hospital IQ, based on curve fitting
• Penn Medicine’s COVID-19 Hospital Impact Model for Epidemics, SIR model, and individual hospital data
• Qventus538.

General approach

• Follow the medical rule: stabilize, then triage, then treatment
• Economics: uncertainty of new enrollments in Medicaid next year.
• Cerner’s Reopening Risk Index (low, medium, high, and very high), determined from daily case count at the local level
• Upscaling from death rate, downscaling from daily case rate.

For the evening’s homework assignment, participants worked with other simulators from the shared spreadsheet.