LONDON — England's coronavirus death toll is being revised downward by more than 5,000 fatalities after experts belatedly concluded they were probably overcounting deaths.

This recently discovered “statistical anomaly” means that on Wednesday Britain’s official tally of deaths due to covid-19 was trimmed from 46,706 to 41,329 — a reduction of more than 10 percent.

A review revealed that a government agency had been counting people as having died of the virus regardless of when they tested positive — meaning even an asymptomatic carrier who was infected in March but was killed in a traffic accident in July would be considered a covid-19 death.

The miscount applied only to England, not Wales, Scotland or Northern Ireland. But the glitch is more than a rounding error.

The official government figures by Public Health England were a grim highlight for months at daily news conferences at 10 Downing Street by Prime Minister Boris Johnson, his top ministers and scientific advisers.

The numbers were widely distributed around the globe, featured in ubiquitous dashboard displays of pandemic daily counts of cases and deaths and used by the news media, public health services, the World Health Organization and academics tracking the pandemic, including Johns Hopkins Coronavirus Resource Center.

Ordinary people, too, still follow the numbers closely, feeling fear or relief in their rise or fall. The death count has also been central to a global scorecard on which nations are “winning or losing” in the pandemic.

Even with the adjustments, Britain has still suffered one of the highest overall death tolls in the world, behind the United States, Brazil and Mexico. The country remains the highest or among the highest, depending on how the counting is done, in Europe.

Public health experts have long warned about the pitfalls of “false precision,” as exact numbers of cases or deaths reported may be off, for all kinds of reasons, by a wide margin.

Tallying and comparing covid-19 cases and deaths is tricky. The number of cases depends on how many tests are carried out, especially among the asymptomatic. The number of deaths depends on how they are measured — by counting excess mortality over time, for example, or looking at death certificates to see if covid-19 is mentioned, or recording positive tests and then seeing if the tested patient dies.

As it turns out here, if a person once tested positive for the coronavirus and died even months later, that death would be counted by Public Health England (PHE) as a covid-19 death — until Wednesday.

That it took months into the epidemic to correct the flaw in reporting is unsettling, experts say, as Britain is known as a world leader in the epidemiology of infectious disease.

The “statistical anomaly” was first cited in a paper published July 16 by Yoon K. Loke of the University of East Anglia and Carl Heneghan of Oxford University. The pair noted that people living in England were increasingly concerned that PHE figures showed a relentless toll of more than 100 covid-19 deaths a day into July, long after the epidemic’s April peak and after the nation’s lockdown was loosened.

This was in stark contrast, they noted, to the more reassuring recovery in Wales, Scotland and Northern Ireland, where there were many days with no covid-19 deaths. As it turned out, Wales, Scotland and Northern Ireland were counting their deaths differently.

Taken to its logical extreme, by the official Public Health England definition, “no one with COVID in England is allowed to ever recover from their illness,” the researchers said. “A patient who has tested positive, but successfully treated and discharged from hospital, will still be counted as a COVID death even if they had a heart attack or were run over by a bus three months later,” researchers said.

A day after the paper appeared, British Health Secretary Matt Hancock called for an urgent review into how coronavirus deaths were recorded in England. The government paused reporting death tallies from Public Health England while the issue was being “resolved.”

From this week forward, Public Health England will provide two counts: Those who tested positive for the virus and died within 28 days, and those who tested positive and died within 60 days.

In the United States, the national death toll figures come from the Centers for Disease Control and Prevention, which in turn is dependent on data from states. Initially, the CDC included only deaths confirmed by a test but then expanded to include probable deaths, based on symptoms and possible exposure. Complicating matters, different states report figures in different ways, and many don’t include probable deaths.

David Spiegelhalter, a British statistician at the University of Cambridge, said that “there is no correct number for the number of people who have died of covid.” Global dashboards “tell a broad picture,” he said. “But everyone has different methods of reporting their cases, and in the end, for all of these, I’d want to look at excess mortality to make any final judgment.”

Excess mortality compares today’s deaths to a previous year, and would include in the excess not only those who died of covid-19 but all those who may have delayed heart or cancer treatment and died.

Spiegelhalter said that the episode “reinforces the idea that these numbers don’t just sit out there waiting to be dug up; we construct them on the basis of assumptions. We decide what a covid death is, and the problem then is there might be five different definitions.”

He said he “felt sympathy for the people who have to deal with this. They started from an impossible position of trying both to be accurate and produce a daily figure. These are incompatible aims.”

The Cambridge statistician stressed that the number presented in the daily news briefings “have never been what they were popularly portrayed as being.”

Spiegelhalter cautioned that the new method will swing the pendulum in the opposite direction. “What we will see now is this number will plummet by 80 to 90 percent. It will go to a fraction of what it was,” he said.

After news broke of the recount of deaths in England, some were relieved that the total number was a little lower, but others thought they had been had, that the government and its scientists were either incompetent or hyping the numbers to justify shutting down the economy during lockdown.

There is a dizzying number of ways that covid-19 deaths are tallied by officials. Depending on what reports you scroll through, the total death toll until July 31 in England alone can vary by over 12,000 fatalities.

Using the new method by Public Health England, covering a 28-day time limit, the number of deaths is 36,617. Using figures from Britain’s Office for National Statistics, which looks at death certificates that list covid-19 as a possible cause, the number of deaths is 49,183.