Thinking Citizen Blog: Mirror, mirror on the wall, which is the best coronavirus model of them all?

Thinking Citizen Blog: Thursday is Health, Health Care and Global Health Policy Day

Today’s Topic: Mirror, mirror on the wall, which is the best coronavirus model of them all? So have you been tracking the accuracy of the leading coronavirus forecasting models? Who has been? Who should be? What are the differences in critical assumptions that explain the outperformance of some versus others? Today, a few excerpts from an editorial on the “Murray model” (University of Washington Institute for Health Metrics and Evaluation). We’ll soon find out how accurate its predictions are. Experts — please chime in. Correct, elaborate, elucidate.

THE MURRAY MODEL — US deaths over the next four months (forecast as of 3/25/20)

1. 81,114 deaths

2. 95% confidence that the number will fall between 38,242 and 162,106.

3. “much lower than the 2.2 million that the President suggested as a worst case.”

METHODS

1. “Measures deaths in terms of population rather than confirmed cases since testing varies geographically.”

2. “It also extrapolates U.S. fatalities based on evidence from other hot spots and Wuhan in China after government lockdowns.”

THE MOST BASIC MATH — 27 day lag from lockdown, hospital beds, ICU beds in NY

1. “It took 27 days after strict social distancing was implemented in Wuhan before daily deaths peaked. New York, California and other states that took early action to close non-essential businesses are merely starting week three.”

2. New York will need an estimated 35,000 more hospital and 7,300 ICU beds next week when demand for care is projected to peak. That’s a lot, but blessedly fewer than the 55,000 to 110,000 hospital beds and 18,000 to 37,000 ICU beds that public health officials had said a week or so ago could be needed.”

3. “This shows how projections can vary and quickly change.”

NB: “Demand for health services rapidly increases in the last week of March and first 2 weeks of April and then slowly declines through the rest of April and May, with demand continuing well into June,” the Murray summary notes. “Daily deaths in the mean forecast exceed 2,300 by the second week of April. While peak demand will occur at the national level in the second week of April, this varies by state. The Murray model predicts population death rates will fall below 0.30 per million in most Northeastern states, Michigan, Indiana, Nevada and Louisiana by early May, but not until late June or early July for Wisconsin and Florida. The enormous geographical variation and modelling uncertainty no doubt informed the White House decision to extend social distancing guidelines nationwide for at least a month despite Mr. Trump’s previously stated desire to get the nation back to work.”

http://www.healthdata.org/sites/default/files/files/research_articles/2020/COVID-forecasting-03252020_4.pdf?mod=article_inline

Opinion | Coronavirus Models of Uncertainty

Why this Nobel laureate predicts a quicker coronavirus recovery: ‘We’re going to be fine’

Opinion | Worst-Case Coronavirus Science

Opinion | Yes, We Need Cost-Benefit Analysis