Experts say the true mortality rate of COVID-19 remains a mystery, with the World Health Organisation's (WHO) latest estimate almost certainly wrong.
On Wednesday, (NZ time), WHO director-general Tedros Adhanom said 3.4 percent of all confirmed infections of the coronavirus have resulted in death.
"By comparison, seasonal flu generally kills far fewer than 1 percent of those infected," he said.
That figure was reached by dividing the number of confirmed deaths due to COVID-19 - 3254 as of Thursday - by the number of confirmed infections (95,120).
It's significantly higher than previous estimates of between 1 and 2 percent, which were made when the outbreak was still contained within mainland China.
But medical experts say it's likely this figure is wrong.
Prof John Edmunds of the London School of Hygiene and Tropical Medicine said it could be much higher.
"In the case of COVID-19 the time between onset of disease and death is quite long - 2 to 3 weeks or more - so the number of cases that you should divide by is not the number of cases that we have seen to this point, but the number of cases that there were a few weeks ago.
"In a rapidly expanding epidemic, that number will be much less, so the true case-fatality-ratio will be higher."
In other words, those dying today would have showed up in the infection statistics about two weeks ago when there were about 75,000 confirmed cases.
Infectious disease specialist Prof Christl Donnelly of the University of Oxford called the WHO's figure "naive".
"When we make a robust statistical estimate of the case fatality ratio, we adjust for this diagnosis-to-death interval and the adjustment can substantially increase the naïve estimate.
"If the case mix of those being diagnosed stays the same, then it is likely that the 3.4 percent figure will increase further."
Doubts mortality will be that high
But it's more likely the mortality rate is well below 3.4 percent - and perhaps even on a par with the flu - they say.
"We do not report all the cases. In fact, we only usually report a small proportion of them," said Dr Edmunds. "If there are many more cases in reality then the case-fatality-ratio will be lower."
This appears to be the case in the US, where it's believed a lack of testing may have missed six weeks of transmission before it was formally detected, and Iran, which officially has about 3000 cases but unconfirmed reports it's running rampant, with even dozens of MPs falling sick.
Other experts said there would be many cases of people getting only mildly sick and not reporting to hospital - they wouldn't be showing up in official statistics, but can still spread it.
"Though there is disagreement about this, some studies have suggested that it is approximately 10 times too high," said Prof Mark Woolhouse, infectious disease specialist at the University of Edinburgh.
"There will be people with fairly mild illness we don't know about - so, this will reduce the percentage death rate," added Prof Paul Hunter of the University of East Anglia.
The virus also appears to have split into two main strains, Chinese scientists said this week. While the more aggressive strain was what ripped through epicentre Wuhan, since late January a less aggressive version of the disease has emerged dominant.
"The best estimates of case fatality ratio (CFR) would have to occur once an epidemic was over," said Tom Wingfield of the Liverpool School of Tropical Medicine.
"Estimating CFR in real-time during the COVID-19 epidemic is therefore fraught with difficulties including not taking into account those who currently have COVID-19 but for whom we do not yet know the outcome."
Past epidemic holds a clue
This is backed up by what happened with swine flu just over a decade ago. After the pandemic died down, WHO statistics suggested there had been about 18,500 fatalities.
Follow-up studies looking at mortality rates during the pandemic suggested around 500,000 people actually died, with up to a quarter of the entire world's population infected. The initial estimate only recorded cases confirmed at the time of death.
"Current quotes are estimates based on currently available data with assumptions or models where we do not have exact data," said Prof Hunter.
"Ultimately 1-2 percent is an estimate based on a number of assumptions which may or may not be true. But in my view 1-2 percent is still a reasonable estimate."