Coronavirus cases in China may have been four times higher than officially reported numbers, according to a study published in the Lancet medical journal.

Infections would have been 232,000 in China as of Feb. 20 if the calculation of cases confirmed by “clinical diagnosis” had been applied throughout the outbreak, Lancet reported.

That compares to the 55,508 cases announced at the time by the National Health Commission, according to the report.

Measurement of infections can change significantly as a virus spreads, more information becomes available and testing evolves and increases.

The study’s authors said seven versions of the case definition were used by the National Health Commission in China from Jan. 15 to March 3 and changes should be considered when estimating growth rates.

A change of methodology that included cases diagnosed with CT imaging scans, alongside the previous method of nucleic acid testing kits, led to an addition of nearly 15,000 virus cases in a single day.

Previously, patients with pneumonia-like symptoms found only via CT scans weren’t confirmed as positive cases without the other test, which was short of supply back then.

“The case definition was initially narrow and was gradually broadened to allow detection of more cases as knowledge increased, particularly milder cases and those without epidemiological links to Wuhan, China,” according to the study published April 21 and led by Tim K Tsang.

While the study doesn’t project the current total of cases using the different methodology, it casts further doubt on the true scale of the outbreak in China, where data and death tolls have been repeatedly revised in the past three months. There have been 83,884 infections and 4,636 deaths in China, according to Johns Hopkins University data.

With the virus having infected more than 2.7 million people and killed 190,000 globally, countries are having issues working out accurate numbers of cases, especially with asymptomatic infections.

In China, serological surveys were initiated to detect how many people have been infected, while the government urged mass testing of both the virus and its antibodies nationwide.

Amid accusations of under-reporting the numbers and concealing the true extent of the outbreak, China last week added confirmed infections and 1,290 fatalities. The revision was aimed at including cases under-reported when people died without being treated or tested and overwhelmed health institutions failed to report them in time.

The Lancet report says among the 232,000 estimated cases nationwide by Feb. 20, 127,000 cases would have been from the city of Wuhan where the pathogen was first identified in December. This includes some 11,000 infections that met the fifth version of case definition with illness onset by the start of the year.

“Estimates of key epidemiological parameters using epidemic curves could be biased if they do not account for such changes in case definitions,” Tsang and other scientists wrote.

They added that when the case definition changed, there could be a backfill of cases that met the new definition around the change time.

Still, the findings suggest the lockdown measures in Hubei province turned out to be very effective in curbing growth as Wuhan implemented unprecedented control measures on Jan. 23. Before the lockdown, 92% of cases were undetected, the study said.