BEIJING – Authorities in Hubei province reported good news Thursday: There were only 349 new coronavirus cases the previous day, the lowest tally in weeks.

The bad – and puzzling – news? Wuhan, the capital of Hubei, reported 615 new cases all by itself.

As Chinese leaders and state media strike a coordinated note this week on the government’s ability to contain the outbreak, inconsistencies and sudden changes in official data are leaving experts – and journalists – struggling to plot meaningful trends, or even place any confidence in the figures coming from government.

Hubei authorities have changed their criteria for counting cases twice in the past week. An earlier change that international researchers applauded led to a sudden spike in case numbers on Feb. 12. And the latest shift, the sixth time that national guidelines have been edited since Jan. 15, caused an overnight drop in new cases from 1,693 to 349.

Jonathan Read, an epidemiologist at the University of Lancaster, said case definitions sometimes do need to be edited as authorities come to grips with how a novel pathogen manifests itself.

“That said, it is very unhelpful for surveillance purposes to change how you define a case too often,” he said.

The latest inconsistency – under which one city appeared to have more cases than the total in the province – apparently arose because Hubei province deducted cases that have not been confirmed through genetic tests from a total reported case number, which includes all diagnoses made by physicians using other methods.

At best, the constant changes have frustrated scholars. At worst, they have raised suspicions.

“Sloppiness of cheating the case # is getting to level of ridiculous,” Eric Feigl-Ding, a visiting scientist at the Harvard T.H. Chan School of Public Health, said on Twitter.

There is no smoking gun suggesting that Chinese officials fabricate numbers – at least not since late January. But many researchers say that the official figures likely underestimate the true numbers because of limited testing capacity and the prevalence of cases with mild or no symptoms. That is why having case numbers collected with consistent methodology would help scholars chart the general contours, if not the precise values, of how the epidemic is unfolding.

When cases spiked on Feb. 12, which coincided with China’s ruling Communist Party naming several new officials to oversee Hubei and Wuhan provinces, Chinese political observers predicted that the move allowed the new regime to wipe the slate clean and be able to show progress. That prediction has been largely borne out.

The party boss of Wuhan, Wang Zhonglin, has ordered a citywide sweep to find all remaining cases of coronavirus infections so that the city would have “baseline” statistics to work with. At the same time, he issued a warning to local party cadres: If one more case were to be found in a household, he said, that district’s party secretary would be held accountable.

Politics aside, innate limitations in scientific modeling also pose challenges for the coronavirus count.

Covid-19’s fatality rate is around 2%, according to preliminary numbers. But that figure could be skewed in multiple ways, said Caitlin Rivers, a senior associate at the Johns Hopkins Center for Health Security. It’s often difficult to determine whether coronavirus was the central cause of death or a contributor. Along with cases doctors don’t yet know about, there are also recently infected patients whose survival remains uncertain. Fatality rates differ significantly across demographics.

While public health authorities are releasing primary medical data, scientists around the world are also pouring over secondary sources and publishing studies based on them. Rivers said these sources are helpful for experts, but cautioned that they can be confusing for a public that doesn’t know how to interpret the literature.

Knowing the severity of a disease is critical for modeling who might die and what demographics are most likely to burden a health system. For now, said John Allen Paulos, a professor of mathematics at Temple University, there’s an innate level of uncertainty around coronavirus figures.

The numbers are “nebulous and unclear,” he said. “We don’t know and for the time being we need to take whatever precautions might work.”

– – –

Berger reported from Washington.