In recent weeks, residents outside Boston have died at home much more often than usual. In Detroit, authorities are responding to nearly four times the number of reports of dead bodies. And in New York, city officials are recording more than 200 home deaths per day — a nearly sixfold increase from recent years.

As of Tuesday afternoon, the United States had logged more than 592,000 cases of COVID-19 and more than 24,000 deaths, the most in the world, according to the Center for Systems Science and Engineering at Johns Hopkins University. But the official COVID-19 death count may, at least for now, be missing fatalities that are occurring outside of hospitals, data and interviews show. Cities are increasingly showing signs of Americans succumbing to the coronavirus in their own beds.

ProPublica requested death data from several major metropolitan areas. Its review provides an early look at the pandemic’s hidden toll.

Experts say it’s possible that some of the jump in at-home death stems from people infected by the virus who either didn’t seek treatment or did but were instructed to shelter in place, and that the undercount is exacerbated by lack of comprehensive testing. It’s also possible that the increase in at-home deaths reflects people dying from other ailments like heart attacks because they couldn’t get to a hospital or refused to go, fearful they’d contract COVID-19.

Mark Hayward, a sociology professor at the University of Texas-Austin who’s an expert on mortality statistics, said all of those deaths are part of the “overall burden of the pandemic.” He said an uptick in deaths, specifically in ProPublica’s findings for Massachusetts and Detroit, indicates an undercount is occurring.

You should think about the official coronavirus death counts, he said, “as just the tip of the iceberg.” The quality of the deaths data will improve as testing expands and fewer people die without getting tested, he added.

The reason having accurate death statistics is important is because they help signal the location of hot spots and prompt officials to deploy resources. Knowing someone died of COVID-19 also enables health officials to alert their contacts so they can quarantine themselves.

New York City was among the first to provide data on at-home deaths. Officials said last week that roughly 200 residents were dying each day outside of hospitals and nursing homes. That’s compared with about 35 per day on average between 2013 and 2017, according to city records.

ProPublica found similar patterns beyond America’s largest and most hard-hit city. Our review examined parts of states like Massachusetts, Michigan and Washington state, drawing upon information from vital-records departments, health agencies, 911 call centers and police departments. ProPublica then compared those findings with historical deaths provided by the federal Centers for Disease Control and Prevention.

In Middlesex, Massachusetts’s most populous county and home to Cambridge, Somerville and Lowell, officials reported 317 at-home deaths in March. That’s about a 20% increase from the same time period for the past three years, in which deaths ranged from 249 to 265. In late February, a Cambridge-based biotech firm hosted a conference in nearby Boston that was later reportedly linked to more than 100 infections; it’s unclear if anyone died.

Older people are particularly vulnerable to dying from COVID-19. In all of Massachusetts, deaths for people 65 and older increased by 3.6% in March from the same month, on average, during the previous three years. The comparison to 2019 was particularly dramatic, an additional 250 deaths across Massachusetts. At the same time, the data shows that increase can’t be accounted for by the official coronavirus tally alone: only 89 deaths statewide were attributed to the virus in March, according to state Health Department data.

The rise in deaths of the elderly is “a pretty dramatic change,” Hayward said, that’s “very consistent with COVID-related deaths.” (As of Monday, Massachusetts’ official COVID-related death count was more than 840.) The Massachusetts Health Department said that the data is preliminary and that there are year-to-year fluctuations.

In Detroit, authorities responded to more than 150 “dead person observed” calls in the first 10 days of April. It was around 40 during the same period for the past three years, according to city 911 call data. Almost all of the incidents in this year’s period occurred in areas where the median household income was less than $45,000, census data shows. Lower-income areas have been particularly hard hit by the coronavirus.

While specifics of each death weren’t available, a review of Detroit radio transmissions revealed instances of individuals found “dead on scene,” such as a March 31 call to a low-income apartment complex northwest of downtown. “Use your precautions here,” said the dispatcher after broadcasting a medical code but few other details. Detroit-area officials couldn’t immediately provide totals on at-home deaths.

In other parts of the U.S., 911 calls for medical assistance have dropped. In Seattle, an early epicenter of the pandemic, data shows that EMT and paramedic calls dropped by more than 25% in the first 10 days of April compared with the same time frame last year. It’s unclear how much, if any, of that drop is due to people being fearful of interacting with the health care system.

Guidance from the CDC says that coroners and medical professionals can list “probable” or “presumed” COVID-19 on death certificates in cases where the symptoms match, even if the person wasn’t tested. For the most part, official death tallies from the disease haven’t included people who died before they were confirmed positive.

Bureaucratic delays and restrictive laws are also contributing to an incomplete picture. Some state officials told ProPublica it would take weeks to provide complete death numbers because of thin staffing or antiquated computer systems. In California, a Health Department spokesman said a request for detailed figures on statewide deaths would take a month to process and would cost $325. (ProPublica plans to pay for it.) Hawaii has suspended the processing of public records.

When a city’s health care system is overwhelmed, the true scope of deaths due to the pandemic is easily missed “in the fog of war,” said Lorna Thorpe, director of the epidemiology division at New York University’s Langone Medical Center. The official numbers provided by health departments are imperfect, she said, and mostly used for “situational awareness” to show which areas are hardest hit.

“One of the reasons we count deaths is to allocate resources to where they need to go,” said Robert Anderson, the chief of the CDC’s mortality statistics branch. “It becomes a little more time sensitive when you’re dealing with something like a pandemic.”

He said it’s particularly important to catch potential hotspots before they explode. If officials can direct adequate resources to places early, “maybe we can keep things from going crazy.”

Given that the U.S. doesn’t have enough tests in the first place, the majority of the screenings are “probably marked for people who are alive,” and the shortage is particularly acute in rural areas with few confirmed cases, said Hayward, the University of Texas professor.

The CDC is logging COVID deaths by collecting death certificates from every state, territory and the District of Columbia, a meticulous process that involves time lags and is vulnerable to errors. Hayward, who is part of a CDC advisory council on vital statistics, said the quality and speed of the data coming in varies so much that it can feel like wrangling reports from more than 50 countries.

Even for routine illnesses like the flu, the national count based on death certificates is always an underestimate. The CDC uses a model that adjusts for the real toll of flu deaths every year, Anderson said, and his staff will create a new model for the coronavirus by comparing the total deaths recorded in the past few months against historic death rates.

At the beginning of the pandemic, “the undercount is going to be really high,” Hayward said. “I couldn’t give you a number. There is no good news on that front.”

Claire Perlman contributed reporting.

The U.S. Postal Service appears to be continuing its checkered response to the coronavirus. Workers across the country say they’re not being informed when colleagues have tested positive for COVID-19, despite a Postal Service policy to do so.

At the end of March, after workers complained, the Postal Service told its employees they would be notified if someone “in your workplace is confirmed to have COVID-19.”

But workers at 11 locations told ProPublica they found out about cases through colleagues or were only told by management days after word had already gotten out.

“People are scared because management has stopped communicating,” said David Denton, a mail handler at a Long Island processing plant. “We have a right to know from management exactly how many cases are confirmed in the building.” Denton said he’s heard of several positive cases at his location, but employees haven’t been told anything by management.

Three letter carriers in New Haven, Connecticut, said they only heard through co-workers — not management — that a colleague had the coronavirus. Their local union president, Vincent Mase, confirmed to ProPublica that one employee at their location had tested positive.

And a letter carrier in New Orleans who tested positive said his supervisor didn’t ask whom he had been in contact with until five days after the worker had raised it. An employee who had been in close contact with him said he didn’t receive any notification from management.

Some post offices are following through on the new policy. Managers at a post office in Gays Mills, Wisconsin, told employees when one of their colleagues tested positive for the coronavirus, asked those who were in close contact with him to quarantine and closed the site for cleaning.

USPS spokesman David Partenheimer said employees at postal facilities are being notified about cases in their facilities. He did not comment on workers saying that hasn’t happened.

Eleanor Murray, assistant professor of epidemiology at the Boston University School of Public Health, said that people who have been in close contact or shared objects with someone who has the coronavirus should be quarantined for 14 days.

“If someone in the workplace has tested positive, then there’s a good chance that at least a couple other people have been infected already even if they don’t feel sick,” she said. “Since we still don’t have great testing capacity, if you’ve been in direct contact with someone who has coronavirus symptoms but hasn’t been tested, then it’s a good idea to quarantine.”

The continued concerns come as the Postal Service nears potential financial collapse. The head of the agency told Congress last week that the agency will run out of money by September. The president rejected a bailout for the agency proposed by lawmakers in the $2 trillion stimulus.

The Postal Service announced its notification policy and other protections for workers after thousands of employees signed a petition for a safer workplace and ProPublica and other outlets reported letter carriers’ concerns over lack of protective equipment and being pressured to work despite having symptoms.

Senators recently wrote to Postmaster General Megan Brennan demanding answers about how the agency is protecting its workers. The Postal Service sent a five-page response outlining its efforts to implement Centers for Disease Control and Prevention guidelines.