Readers who’ve been following along with me know I’ve been frustrated by the fuzziness of the coronavirus statistics. In particular, it is hard to nail down such crucial numbers as what a “confirmed case” is (e.g., is it a case of the virus or the potentially lethal disease caused by the virus?), and what counts as a “coronavirus death” (does it mean the virus or disease was merely present in the patient, that it played some contributory role in death, or that it was the proximate cause of death?).
I hope to have more on that soon. In the meantime, though, let me draw attention to a report by our Tobias Hoonhout that an influential model, on which the White House Coronavirus Task Force relies to project COVID-19 deaths and related statistics, has just been dramatically revised to lower its estimates. The University of Washington’s Institute for Health Metrics and Evaluation now projects 12 percent fewer deaths (81,766, down from 93,531). While estimates about daily deaths in the peak period have actually increased, it is believed the peak may come somewhat earlier and be shorter in duration.
What is truly gripping is the sudden reduction, from a projection made just three days earlier, of the estimate of how many hospital beds that will be needed at peak. On April 2, IHME predicted 262,092; on April 5, that was reduced to 140,823. The estimated needs for ICU beds and invasive ventilators were also slashed.
The fact that the projections have been reduced by so much so quickly is not the only notable thing. Former New York Times reporter Alex Berenson has been following the IHME modeling like a hawk on Twitter. As he tartly notes, it is not just that the IHME projections on which the government so heavily relies have proved unreliable; it is that they are wrong even about the here and now. The IHME projected on April 5 that hard hit New York would need about 24,000 hospital beds, including about 6,000 for ICU patients at that point; of course, April 5 has already happened, and New York announced that it had 16,479 people being hospitalized, including 4,376 ICU patients — i.e., the model was about a third off on the day it was published.
Berenson reasonably asks, if the model is significantly inaccurate in its application to current conditions that are more easily knowable, how much confidence should we have in its projections for a week from now, or a month from now?
Based on these models, the government is making policy, now and going forward. It is vital that we get more and better information about what is informing them. The decisions to substantially shut down the economy have been based on the assessment that surges in coronavirus infections will outstrip our health-care system’s capacity to treat patients, with the knock-on effect that otherwise treatable ailments will not receive treatment, needlessly increasing the number of serious illnesses and fatalities.
How reliable are these assessments? Do they justify the damage being done by the extreme measures that have been taken to counter the predicted scenarios? We need better information about the inputs.