A few months after Hurricane Maria ravaged Puerto Rico in 2017, the governor announced the official death toll: 64.

To anyone who lived through the storm, as well as scientists nationwide, that number seemed impossibly low. After a flurry of studies, several research teams came up with their own estimates, which were some 15 to 65 times higher than the governor's estimate. Eventually, the official toll settled at 2,975—46 times the first number.

The National Hurricane Center tracks storm-related deaths in the continental United States. But previous studies and the example from Hurricane Maria show those counts may underestimate the total impacts. In addition, not all assessments are done in the same way, complicating comparisons between storms.

New research, published Wednesday in Science Advances, addresses those issues head-on. The study, which looks at 179 storms over the past 32 years, found that major storms contributed to more than 18,000 deaths in the month of and month following the storms—many more than in official tallies.

More than 90 percent of those who died came from poor or historically disadvantaged communities. To lead author Robbie Parks, an environmental epidemiologist at Columbia University, that highlights a critical point: "Cyclones are hitting randomly, but the effects are not random. They are dictated largely by social structures," he says. "The excess deaths that we estimated after tropical cyclones were disproportionately in the most socially vulnerable areas of the United States." The Centers for Disease Control and Prevention considers several factors for defining socially vulnerable communities including poverty, age and disability, access to transportation, to figure out which households might need support during and after disasters.

The massive disparity found in the study is not a surprise. Indiana University biostatistician Raul Cruz, who was not involved in the research, was one of the people who tried to come up with a more realistic estimate of deaths and their causes after Hurricane Maria. His team found substantial increases in deaths from heart disease, diabetes, and Alzheimer's disease in the months following the storm. Such conditions are manageable under normal conditions, he says. But they become dangerous or even deadly if not addressed, a hard or impossible task after a destructive storm.

For people from historically disadvantaged and poor communities, "when one of these storms comes and knocks you down, that can be what keeps you from the preventative treatment you need," Cruz says.

Those risks may have grown. More than 80 percent of the storm-associated deaths happened in the second, more recent, half of the researcher's study period, from 2004 to 2019. Hurricane Katrina, in 2005, was the deadliest, followed by Hurricanes Irma, in 2017, and Sandy, in 2012. The researchers did not include Hurricane Maria in their analysis because the dataset they worked from covered only the continental United States.

Climate change has already intensified hurricanes. Flood and wind risks have grown; Hurricane Harvey dumped 15 percent more rain over Texas than it would have in a world without fossil fuel-driven climate change. The possibility of two major storms like Katrina and Harvey making landfall within a few weeks, which is unheard of today, is increasingly likely by 2100.

The researchers didn't directly link climate change-intensified storms to deaths. But "anthropogenic climate change, that's certainly a factor," says Parks. People are also more often ending up in the pathways of storms; population growth is booming near coasts and even in flood zones. In North Carolina, 10 new houses have been built in floodplains for each one that went through a government-sponsored buyout because of flood risk.

Societal inequities play a major role as well. Broward County, Florida has some of the wealthiest and poorest communities in the state. After Hurricane Matthew in 2016, deaths were concentrated within the poor communities within the county, the research showed. "The rich have the planes and the rich have the second homes" to evacuate to, says Parks. "But what about people who just literally need, you know, waterproof equipment and power to be able to power the generators to power breathing apparatus or obtain a ride out of a place which is about to get hit by a big wave from a flash flood due to a tropical cyclone?"

More clearly identifying those at risk, Parks says, highlights how to better help vulnerable people during disasters. "Those are things which society can change," he says. "And it's really a choice."

Scott Zeger, a biostatistician at Johns Hopkins University, hopes agencies like the CDC start using the study's more comprehensive view of storm-related deaths. "Something like this ought to be the means by which we monitor this going forward," he says. Careful analyses showing which communities suffer deep losses after storms, Zeger says, could help policymakers figure out where to funnel support.