Modeling the potential trajectory and impact of COVID-19 has been in the news a lot lately, and models are being shared across the nation.
To understand the complexity of modeling, consider; “Do I need my rain gear for my bike rides this week?”
A simple question, but the modeling that goes into the answer that I get from my Dark Sky weather app is complex. This ability to predict the future is critical in our work to prepare our hospital systems and our state for the COVID-19 pandemic.
So, what do the models of COVID-19 tell us about the next 2 weeks?
Well, it depends on which type of model we use. One type is called a curve-fitting model. This model looks at the rise and fall of cases in China and Italy and plots where we in Minnesota are on those curves. It then looks ahead two weeks and using our population, hospital bed and ventilator numbers, extrapolates where we will be. It does not strive to explain and model all the details (it assumes that is impossible). It assumes we will follow their course. Using that model, we'll be well on our way to the predicted peak occurring on April 23 and our hospitals will be full.
In another method of modeling, however, we look at the specific factors that the curve is based on and make assumptions. We know the doubling rate in our community is now 3.8 days. We know that about one-third of people have very few or no symptoms. We know that 4% of patients need hospitalization, 1.5 % need the ICU and 1% need ventilators. Add in a host of other objective measures, and we have a "mechanistic" or epidemiologic model. By this model, in two weeks, our M Health Fairview system will have 92 more COVID patients recovering on the general wards. 35 more will be in the ICU, and 25 on ventilators. In this model, the peak will not occur until the end of May or early June.
So how do we reconcile the different models? We prepare for both. As with the weather, we know that tomorrow's forecast is more reliable than a month off. Each day brings more data and more scientific discoveries, and our models improve. We, together, are preparing for the storm, whether it be here in three weeks or two months.