Leveraging over 15 years of support to the U.S. Department of Health and Human Services (HHS) for medical consequence modeling and our proprietary artificial intelligence (AI) models, IEM refined our AI model in late February and early March and are confident in its ability to provide accurate 7-day projections for operational and logistical planning. IEM has been providing our state-specific Coronavirus projections directly to state and local emergency managers to assist them as prepared for an increase in hospitalizations, planned for and locate drive-through testing facilities, and determine where increased levels of transmission may be occurring.
AI-based Model Background
IEM built our COVID-19 model much like a hurricane prediction model. We use all the actual data of confirmed cases and use AI to find the right parameter values that best fit the actual data, and do not assume the average number of secondary infections (R-value) stays the same over time. IEM’s AI model finds the best R-value over time to evaluate how it changes over the course of the outbreak. AI process includes approximately 11 million simulations to find the best parameter values for a disease model to fit the data and then to project the future number of cases. The future projections are like a hurricane spaghetti plot with thousands of possible paths, and we use the maximum likelihood of all these projections to make the best projection of future cases. While all models have limitations, we believe this information will help you with your planning as our projections have typically been within 20%, and are often within 10%, of actual confirmed cases.
The projections shown in these documents are based on data pulled daily from open sources. Check out our COVID-19 Projection Dashboard.