Associate professor of biostatistics Nicholas Reich and postdoc researcher Thomas McAndrew have teamed up for several weeks to poll epidemiologists and experts modeling infectious disease to put together a genuinely “educated guess” at how the COVID-19 pandemic is changing week to week.
Now Reich’s lab has made the work easier to view at a glance by compiling a page that shows how each of the big name models forecasts events over the coming month.
The range of projections for where the United States will be in mid-May varies considerably. The IHME projection, which has for better or worse become the favorite with the media and apparently with White House officials, calls for a number of deaths just under 60,000 at that point. On the same date, the MOBS lab model from Northeastern’s Laboratory for the Modeling of Biological and Sociotechnical Systems is somewhat less optimistic at 69,000 deaths. Both of these can be regarded as truly rosy compared with the output from the darkest timeline churned out by one of Columbia’s three models. The CU-80-Contact model calls for a heart-stomping 110,000 deaths by that date.
Of course, the Reich Labs chart doesn’t cover every option. For example, the truly statistically obsessed might find that Los Alamos National Laboratory’s model fits the bill. Los Alamos puts out a regular update that includes a truckload of downloadable data along with best case, worst case, best guess projections on a state-by-state basis. (Warning: Some of their “best guess” numbers aren’t pretty.)
When it comes to models, there’s a lot of variety in terms of how they’re done, how the aspects of the projection are weighted, and what they factor in about social distancing or other means of reducing the spread of COVID-19. Researchers have taken a wide variety of approaches in trying to find a way to describe the pandemic. But it’s not actually true that there’s a model to make everyone happy … because none of these results will make you happy. Unfortunately, the IHME model is close to the most optimistic values projected. Let’s hope even they turn out to be too high.