NEW DELHI (AP) – India’s excess deaths during the coronavirus pandemic could be a staggering 10 times the official COVID-19 toll, likely making it modern India’s worst human tragedy, according to the most comprehensive research yet on the ravages of the virus in the South Asian country.

Most experts believe India’s official toll of more than 414,000 dead is a vast undercount, but the government has dismissed those concerns as exaggerated and misleading.

The report released Tuesday estimated excess deaths—the gap between those recorded and those that would have been expected—to be 3.4 million to 4.7 million since the pandemic began through last month. It said an accurate figure may “prove elusive,” but the true death toll “is likely to be an order of magnitude greater than the official count.”

The report was published by Arvind Subramanian, the Indian government’s former chief economic adviser, and two other researchers at the Center for Global Development, a nonprofit think tank based in Washington, and Harvard University.

It said the count could have missed deaths that occurred in overwhelmed hospitals or while health care was disrupted, particularly during the devastating virus surge earlier this year.

“True deaths are likely to be in the several millions not hundreds of thousands, making this arguably India’s worst human tragedy since Partition and independence,” the report said.

The Partition of the British-ruled Indian subcontinent into independent India and Pakistan in 1947 led to the killing of up to 1 million people as gangs of Hindus and Muslims slaughtered each other.

The report on India’s virus toll used three calculation methods: data from the civil registration system that records births and deaths across seven states, blood tests showing the prevalence of the virus in India alongside global COVID-19 fatality rates, and an economic survey of nearly 900,000 people done three times a year.

Researchers cautioned that each method had weaknesses, such as the economic survey omitting the causes of death.

Instead, researchers looked at deaths from all causes and compared that data to mortality in previous years—a method widely considered an accurate metric.

Researchers also cautioned that virus prevalence and COVID-19 deaths in the seven states they studied may not translate to all of India, since the virus could have spread more in urban versus rural states and since health care quality varies greatly around India.

Other nations also are believed to have undercounted deaths in the pandemic. But India is thought to have a greater gap due to having the world’s second-highest population of 1.4 billion and because not all deaths were recorded even before the pandemic.

The health ministry did not immediately respond to an Associated Press request for comment on the report.

Dr. Jacob John, who studies viruses at the Christian Medical College at Vellore in southern India and was not part of the research, reviewed the report for the AP and said it underscores the devastating impact COVID-19 had on the country’s underprepared health system.

“This analysis reiterates the observations of other fearless investigative journalists that have highlighted the massive undercounting of deaths,” he said.

The report also estimated that nearly 2 million Indians died during the first surge in infections last year and said not “grasping the scale of the tragedy in real time” may have “bred collective complacency that led to the horrors” of the surge earlier this year.

Over the last few months, some Indian states have increased their COVID-19 death toll after finding thousands of previously unreported cases, raising concerns that many more fatalities were not officially recorded.

Several Indian journalists have also published higher numbers from some states using government data. Scientists say this new information is helping them better understand how COVID-19 spread in India.

Murad Banaji, who studies mathematics at Middlesex University and has been looking at India’s COVID-19 mortality figures, said the recent data has confirmed some of the suspicions about undercounting. Dr. Banaji said the new data also shows the virus wasn’t restricted to urban centers, as contemporary reports had indicated, and that India’s villages were also badly affected.

“A question we should ask is if some of those deaths were avoidable,” he said.