A newly released federal report shows that long COVID was included on the death certificates of slightly more than 3,500 Americans in the first two and a half years of the pandemic. It’s a breakthrough for understanding the new condition. But is the number accurate?

The study, performed by the U.S. Centers for Disease Control and Prevention’s Vital Statistics System, examined death certificates that listed COVID as the cause of death with keywords referring to long COVID, including post-acute sequelae of COVID-19 (PASC), “post COVID,” and “long haul COVID.” The deaths occurred between Jan. 1, 2020, and June 30 of this year.

The majority of deaths with long COVID ties were reported among white people (nearly 79%). Black people accounted for 10% of such deaths, followed by Hispanic people (8%). All other racial groups comprised less than 2% of such deaths.

The 3,500 figure is “probably an undercount,” Dr. Georges Benjamin, executive director of the American Public Health Association, told Fortune. That’s because the method to determine such deaths is crude, and the fact that it’s early in the history of the disease. All long COVID deaths were coded for COVID because a cause-of-death code for long COVID hasn’t been implemented, according to the report.

“Long COVID deaths will go up as clinicians get better at identifying and recording COVID-related deaths (both acute and long COVID), and as the peak of long COVID cases moves through the system,” he said. “We have not yet seen the peak in cases of long COVID.”

Dr. Amesh Adalja, an infectious disease specialist and senior scholar at the Johns Hopkins Center for Health Security, said the study is difficult to interpret because it “relies on death certificate data for a condition that has nebulous diagnostic criteria.”

There is no agreed-upon definition of long COVID, generally defined as new symptoms that occur or persist after a COVID infection and last for weeks or months. The World Health Organization defines the condition as the continuation or development of new symptoms three months after the initial infection, with those symptoms lasting for two or more months, without other explanation. But other organizations like the CDC have definitions that vary.

Such data is “not reliable for something like this,” Adalja said, adding that it’s important to distinguish deaths due to long COVID from deaths among individuals with long COVID. 

Additionally, the risk of death in those with a severe COVID infection may remain high for a period of time after they recover, he added. But such a death may or may not be classified as a long COVID death, depending on one’s definition of the condition.

“There needs to be much more granularity to be able to understand a phenomenon like this,” he said.

In September 2020, the WHO approved a cause-of-death code for post-COVID conditions. It hasn’t been implemented in the U.S., the authors wrote, adding that their new study will pave the way for its implementation.

Almost half of COVID survivors—both children and adults—had lingering symptoms four months later, according to a landmark study published this month in an affiliate journal of The Lancet. Researchers performed an analysis of nearly 200 studies on prior COVID patients involving nearly 750,000 people around the globe.

Nearly 20% of American adults who’ve had COVID—an estimated 50 million—report having long COVID symptoms, according to data collected by the U.S. Census Bureau this summer.

Many experts contend that long COVID is best defined as a chronic-fatigue-syndrome-like condition that develops after COVID illness, similar to other post-viral syndromes that can occur after infection with herpes, Lyme disease, and even Ebola. Other post-COVID complications, like organ damage and post–intensive-care syndrome, should not be defined as long COVID, they say.