Should we believe the coronavirus models?
COVID-19 PROJECTIONS ARE ALARMING, BUT ARE WE STAKING OUR FUTURE ON AN INEXACT SCIENCE?
The slide on page 13, Dr. Peter Donnelly warned those virtually tuned in to the technical briefing of Ontario’s grim COVID-19 projections late last week, would be the most disturbing of all. Indeed, the number was scary: Without physical distancing measures, Ontario would see 100,000 deaths over the course of the pandemic.
It wasn’t the only shocking figure: The impact of COVID-19 would have been an estimated 300,000 cases and 6,000 deaths by the end of this month alone with no interventions. An estimated 220,000 cases and 4,400 deaths have been spared by sealing off schools, banning large gatherings, shutting non-essential workplaces, closing outdoor rec facilities and, now, ticketing people $880 for walking their dogs through closed parks, according to the models. Another 1,350 deaths (from 1,600 to 250) could be prevented in the coming two weeks with further enforced measures, Donnelly, president and CEO of Public Health Ontario, said Friday.
The range of scenarios seems dizzying, the predictions seriously extreme. Already, critics are insisting the models bias high, leading many to wonder how much we should trust the models upon which so many life-upending decisions are being made.
It’s not possible to be exact about where we will end up … it’s difficult to know exactly where you stand … projections and modelling for a brand new viral disease are very inexact … this is not an exact science.
Donnelly used the word exact, or variations of it, six times throughout the briefing. Yet the imperfect science is informing Ontario’s strategy. And it underlines how assumptions used in modelling the pandemic may rest on “very flimsy foundations,” as Robert Dingwall, a professor of sociology at Britain’s Nottingham Trent University said this week in response to a study questioning the benefits of school closures in terms of scientific evidence.
The Ontario models made projections in terms of mortality based on the global experience thus far of COVID-19, as well as data gathered from Canada. The tables released Friday pegged the case fatality ratio (the percentage of confirmed infections that end in death) of 2.1 per cent overall for Ontario, from 0 per cent for people aged less than 40, to a truly scary 15.9 per cent for people 80 years and over. However, the case fatality ratio is based on known infections, and biases in both directions can inflate or underestimate it — notably, not counting mild cases can produce a falsely high one.
A study published last week in Lancet Infectious Diseases, based on 24 people who died in mainland China, and 165 recoveries outside China, came up with a case fatality ratio best estimate of 1.38 per cent. The estimate for the overall infection fatality ratio (the percentage of people who become infected and die, including those who show no symptoms) was 0.66 per cent.
Donnelly said the Ontario projections were based on “other things,” though he didn’t elaborate. Officials didn’t release the actual models, just the projections. It’s not clear what variables were plugged in.
Most of the projections are done by comparing the curves of different countries and applying the same trajectories here, said Murat Kristal, of York University’s Schulich School of Business, where scientists have developed an analytics dashboard to predict the five-day global spread of COVID-19.
Italy is one of the worst impacted countries in the world. On March 1, there were 1,701 confirmed cases in Italy. By April 6, there were 132,547, a nearly 78-fold increase. On April 1, the number of confirmed cases in Ontario was 2,292. “If we apply the same trajectory from Italy, the number of cases would be 180,000 in Ontario by April 30,” Kristal said.
The trajectory of confirmed cases in Ontario between March 1 and April 1 was 159-fold. “If the same trajectory is applied from April 1 to April 30, as a worst-case scenario, we can reach to 300,000 total confirmed cases.”
Social distancing doesn’t change the likelihood of getting the virus, just the timing. Never before seen in human beings, the virus is extraordinarily efficient at spreading between us, and it’s highly likely almost everyone will get it.
“The question is when, not if. And that’s a crucial question,” said Robert Smith, a disease modelling professor at the University of Ottawa.
The idea behind “flattening the curve” is to provide the chance for every seriously infected person to potentially benefit from intensive care and ventilation. But intubation doesn’t guarantee survival. Early reports suggest that, even with ventilator support, half of patients die between ICU day 1 and day 18.
The people at highest risk of requiring intubation are also at the highest risk of dying from competing causes, such as heart disease or chronic obstructive pulmonary disease. The 27 deaths in an Ontario nursing home are tragic. But the median survival of nursing home residents at end of life is five months, according to one 2010 study. While seriously disheartening, “It is quite likely that most, if not all of those fatalities would have happened regardless of the COVID-19 epidemic,” said one doctor, who we agreed not to identify “as any dissenters within the medical community are likely to be targeted,” the physician said.
“One has to consider that our health-care system is mortgaging the future prosperity of our youth for an impossible promise of immortality.”
In the early days of an epidemic, Donnelly said, it’s all about providing an important “early steer” to policy-makers about what they should be doing. As soon as the government’s “command table” saw the figure that suggested there could be an overall mortality of between 90,000 and 100,000 deaths, they moved quickly to shut schools.
Modelling is good if the data is perfect. But we never have perfect data, and critical information is missing. The data is changing almost daily. When looking at a time scale of months, it’s like driving through a blizzard.
According to the York analytics, it seems the growth rate in the number of confirmed cases in Canada appears to be slowing down. Epidemiologists are already bracing for the finger-pointing when the attack rates aren’t worst-case, because governments instituted social distancing.
But neither is an almost medieval “just close the door and stay in your house” public health approach feasible in the long term, given the impact on the economy — hundreds of billions of dollars to support the millions who have lost their jobs and businesses — and society at large, Stanford University epidemiologist Dr. John Ioannidis said on a recent Monk Debates podcast.
COVID-19 is a serious threat, he said, but “there’s a lot that we can get wrong if we don’t get it right.”