Why are US coronavirus deaths going down as covid-19 cases soar?
Coronavirus deaths are falling in the US even as cases skyrocket. In the UK, a lower proportion of people hospitalised with covid‑19 are dying. This has led to suggestions that the risk of dying if you are infected with the virus is falling, but the truth may be more complicated.
“At this point, I don’t think we have conclusive evidence that the death rate is going down,” says Tessa Bold at Stockholm University in Sweden.
Having plateaued at around 20,000 in May, the number of daily confirmed cases in the US began rising in June and has now exceeded 60,000. However, the number of deaths in the US reported as being due to covid-19 has fallen from more than 3000 a day in mid-April to well under 1000.
There are several possible explanations for this. For starters, it could be a result of better treatments, including use of the steroid dexamethasone.
Another reason why deaths aren’t tracking case numbers in the US could be the lag between people testing positive for the coronavirus and dying. Those who die usually do so around two weeks after developing symptoms and their deaths typically aren’t reported for another week. More widespread testing, no longer limited to those with serious symptoms, could mean that cases of coronavirus are being detected even earlier, increasing this lag.
It could also be that most new cases are in younger people, whose risk of dying from the virus is far lower. The median age of those testing positive in the US is falling, suggesting that while older people continue to shelter and avoid infection, younger people are being infected as they return to work and socialising.
“As this group begins to mingle with older relatives, we may see a spike in cases for the older,” says Richard Grewelle at Stanford University in California.
Plenty of mingling will have occurred over the Independence Day weekend, which could lead to a spike in deaths in late July, he says. “We’ll see if my prediction holds true.”
The situation in England points to a similar trend in the UK. An analysis of government data by Jason Oke at the University of Oxford and his colleagues suggests there has been a steady and steep decline in the proportion of people hospitalised with covid-19 dying. “It’s encouraging,” says Oke. “We are either getting better at treating this or it’s becoming less severe.”
But there might be other explanations, he cautions. It could just be an artefact of the data due to survivors staying in hospital longer. Another possibility is that hospitals are admitting less severe cases now they have the resources.
To know for sure if the odds of dying are falling, we really need to know how many of those who are infected succumb and if this is changing – that is, if the infection fatality rate (IFR) is declining.
Early estimates put the IFR across populations at between 0.6 and 1 per cent. Some thought this would turn out to be an overestimate, but recent estimates are similar. A statistical analysis by Grewelle and his colleague Giulio De Leo, for instance, suggests that the global IFR so far is 1 per cent.
Bold’s team has estimated IFRs for different countries around the world based on death rates in France, and also came up with relatively high numbers. For instance, Brazil, one of the world’s hardest-hit countries, should have an IFR of around 0.4 per cent given the ages of its inhabitants and their general health. Adjusting for the quality of healthcare, however, pushes the predicted IFR up to 0.8 per cent.
This matches research by Fernando Barros at the Catholic University of Pelotas in Brazil. He has tried to directly measure the nation’s IFR by doing antibody tests on more than 25,000 people. His team puts it at 1 per cent.
So far, though, there are no estimates of how IFRs are changing over time. “We have only one estimate, and not two or more points in time, so we are not in the position of studying trends,” says Barros.
Sign up to our free Health Check newsletter for a monthly round-up of all the health and fitness news you need to know
More on these topics: