1/ There is widespread misunderstanding about the epidemiological models that prompted government decisions to issue stay-at-home orders and mandate business closures due to COVID-19.

2/ Teams at several universities, including Harvard and Columbia, provided input that led the White House Coronavirus Task Force to believe 1.5 to 2.2 million Americans might die from COVID-19 this year if nothing was done to slow the spread of the disease.

3/ Those teams also estimated that social distancing and other mitigation measures could reduce the U.S. death toll to a range from 100,000 to 240,000.

4/ At the March 31 White House briefing where this was announced, the Task Force also showed a chart of potential DAILY deaths from April through June. That chart came from a model produced by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington.

5/ Unfortunately, the White House Task Force didn’t point out that the IHME model was forecasting a MUCH LOWER range of total deaths than the other models. The IHME model is posted online, and I have been studying it daily since March 31.

6/ On March 31, the IHME model forecast a potential range of 37,000 to 152,000 deaths in the U.S., and it “projected” an outcome of 84,000 deaths by July 1. This does NOT mean IHME was predicting that 84,000 people would die.

7/ Probability risk analysis is an effort to estimate the degree of confidence we can have that an outcome WON’T EXCEED a certain number. The “projected” outcome is typically at the 50% confidence level.

8/ In approximate terms, this means the IHME model was forecasting 95% confidence that NO MORE THAN 152,000 people would die, 50% confidence that NO MORE THAN 84,000 people would die, and just 5% confidence that NO MORE THAN 37,000 people would die.

9/ On March 31, the IHME model was based on mitigation measures already mandated in about half the states. It did NOT assume full mitigation in states without such measures in place.

10/ The number of total U.S. COVID-19 deaths doubled every 2 days from March 20 through 26. When IHME prepared the March 31 risk analysis, the total was still doubling every 3 days. If that rate had continued, total deaths would exceed 37k by April 10 and 120k by April 15.

11/ About half the U.S. governors issued stay at home orders in late March, and a dozen more did so in the first few days of April. Today we’re approaching 17,000 COVID-19 deaths, and this total took 5 days to double (from 8700 on April 4).

12/ The growth rate is now slowing steadily, and it will probably take at least 10 days to double today’s total. Yesterday IHME updated its risk analysis based on new information.

13/ The IHM model now forecasts a potential range from 31,000 to 127,000 deaths and projects (presumably at the 50% confidence level) that total deaths WON’T EXCEED 60,000 by July 1. https://covid19.healthdata.org/united-states-of-america

14/ Now IHME, UW, and the Gates Foundation (which funds IHME) are being bombarded with criticism on Twitter. Some attacks are going viral. Many critics claim IHME predicted 100,000 to 240,000 COVID-19 deaths. (That estimate came from others.)

15/ Even when critics correctly state that IHME originally “projected” 84,000 deaths, it’s clear they don’t understand what this means. Such a projection indicates 50% confidence that total deaths WOULDN'T EXCEED 84,000 based on what was known on March 31.

16/ Another FALSE claim is that the IHME model assumed full mitigation, including lockdown orders, in all states from the start. You can filter the online model by state and see which mitigation measures are included in each state and when they started.

17/ The April 8 reduction in IHME's overall forecast is partly due to a dozen governors issuing stay at home orders shortly after the March 31 risk analysis was prepared.

18/ IHME also attempted to forecast the need for additional hospital resources, such as total beds, ICU beds, and ventilators in each state. They greatly overestimated usage of those resources.

19/ The forecast of severe shortages was partly attributable to incorrect or insufficient data from many states regarding existing hospital beds and ventilators. Even so, IHME forecast a need for 27k ventilators when blaring headlines claimed we might need hundreds of thousands.

20/ With such a new disease, so little data to go on, reports of massive impact on Italian hospitals in March, and unreliable information from China, I find it amazing that IHME managed to produce a reasonable projection of fatalities.

21/ Trying to forecast hospital resources was a bridge too far this time, but hopefully states will improve their hospital data collection before the next new disease hits.