DARPA wants to modernize how first responders do triage during disasters

In Overmatched, we take a close look at the science and technology at the heart of the defense industry—the world of soldiers and spies.

IF A BUILDING COLLAPSES or a bomb goes off, there are often more people who need medical treatment than there are people who can help them. That mismatch is what defines a mass casualty incident. The military’s most famous R&D agency, DARPA, wants to figure out how to better handle those situations, so more people come out of them alive.

That’s the goal of what the agency is calling the DARPA Triage Challenge, a three-year program that kicks off November 6 and will bring together medical knowledge, autonomous vehicles, noninvasive sensors, and algorithms to prioritize and plan patient care when there are too many patients and not enough care—a process typically called triage. Teams, yet to be named, will compete to see if their systems can categorize injured people in large, complex situations and determine their need for treatment.

A sorting hat for disasters

Triage is no simple task, even for people who make it part of their profession, says Stacy Shackelford, the trauma medical director for the Defense Health Agency’s Colorado Springs region. Part of the agency’s mandate is to manage military hospitals and clinics. “Even in the trauma community, the idea of triage is somewhat of a mysterious topic,” she says. 

The word triage comes from the French, and it means, essentially, “sorting casualties.” When a host of humans get injured at the same time, first responders can’t give them all equal, simultaneous attention. So they sort them into categories: minimal, minorly injured; delayed, seriously injured but not in an immediately life-threatening way; immediate, severely injured in such a way that prompt treatment would likely be lifesaving; and expectant, dead or soon likely to be. “It really is a way to decide who needs lifesaving interventions and who can wait,” says Shackelford, “so that you can do the greatest good for the greatest number of people.”

The question of whom to treat when and how has always been important, but it’s come to the fore for the Defense Department as the nature of global tensions changes, and as disasters that primarily affect civilians do too. “A lot of the military threat currently revolves around what would happen if we went towards China or we went to war with Russia, and there’s these types of near-peer conflicts,” says Shackelford. The frightening implication is that there would be more injuries and deaths than in other recent conflicts. “Just the sheer number of possible casualties that could occur.” Look, too, at the war in Ukraine. 

The severity, frequency, and unpredictability of some nonmilitary disasters—floods, wildfires, and more—is also shifting as the climate changes. Meanwhile, mass shootings occur far too often; a damaged nuclear power plant could pose a radioactive risk; earthquakes topple buildings; poorly maintained buildings topple themselves. Even the pandemic, says Jeffrey Freeman, director of the National Center for Disaster Medicine and Public Health at the Uniformed Services University, has been a kind of slow-moving or rolling disaster. It’s not typically thought of as a mass casualty incident. But, says Freeman, “The effects are similar in some ways, in that you have large numbers of critically ill patients in need of care, but dissimilar in that those in need are not limited to a geographic area.” In either sort of scenario, he continues, “Triage is critical.”

Freeman’s organization is currently managing an assessment, mandated by Congress, of the National Medical Disaster System, which was set up in the 1980s to manage how the Department of Defense, military treatment facilities, Veterans Affairs medical centers, and civilian hospitals under the Department of Health and Human Services respond to large-scale catastrophes, including combat operations overseas. He sees the DARPA Triage Challenge as highly relevant to dealing with incidents that overwhelm the existing system—a good goal now and always. “Disasters or wars themselves are sort of unpredictable, seemingly infrequent events. They’re almost random in their occurrence,” he says. “The state of disaster or the state of catastrophe is actually consistent. There are always disasters occurring, there are always conflicts occurring.” 

He describes the global state of disaster as “continuous,” which makes the Triage Challenge, he says, “timeless.”

What’s more, the concept of triage, Shackelford says, hasn’t really evolved much in decades, which means the potential fruits of the DARPA Triage Challenge—if it pans out—could make a big difference in what the “greatest good, greatest number” approach can look like. With DARPA, though, research is always a gamble: The agency takes aim at tough scientific and technological goals, and often misses, a model called “high-risk, high-reward” research.

Jean-Paul Chretien, the Triage Challenge program manager at DARPA, does have some specific hopes for what will emerge from this risk—like the ability to identify victims who are more seriously injured than they seem. “It’s hard to tell by looking at them that they have these internal injuries,” he says. The typical biosignatures people check to determine a patient’s status are normal vital signs: pulse, blood pressure, respiration. “What we now know is that those are really lagging indicators of serious injury, because the body’s able to compensate,” Chretien says. But when it can’t anymore? “They really fall off a cliff,” he says. In other words, a patient’s pulse or blood pressure may seem OK, but a major injury may still be present, lurking beneath that seemingly good news. He hopes the Triage Challenge will uncover more timely physiological indicators of such injuries—indicators that can be detected before a patient is on the precipice.

Assessment from afar

The DARPA Triage Challenge could yield that result, as it tasks competitors—some of whom DARPA is paying to participate in the competition, and some of whom will fund themselves—with two separate goals. The first addresses the primary stage of triage (the sorting of people in the field) while the second deals with what to do once they’re in treatment. 

For the first stage, Triage Challenge competitors have to develop sensor systems that can assess victims at a distance, gathering data on physiological signatures of injury. Doing this from afar could keep responders from encountering hazards, like radioactivity or unstable buildings, during that process. The aim is to have the systems move autonomously by the end of the competition.

The signatures such systems seek may include, according to DARPA’s announcement of the project, things like “ability to move, severe hemorrhage, respiratory distress, and alertness.” Competitors could equip robots or drones with computer-vision or motion-tracking systems, instruments that use light to measure changes in blood volume, lasers that analyze breathing or heart activity, or speech recognition capabilities. Or all of the above. Algorithms the teams develop must then extract meaningful conclusions from the data collected—like who needs lifesaving treatment right now. 

The second focus of the DARPA Triage Challenge is the period after the most urgent casualties have received treatment—the secondary stage of triage. For this part, competitors will develop technology to dig deeper into patients’ statuses and watch for changes that are whispering for help. The real innovations for this stage will come from the algorithmic side: software that, for instance, parses the details of an electrocardiogram—perhaps using a noninvasive electrode in contact with the skin—looking at the whole waveform of the heart’s activity and not just the beep-beep of a beat, or software that does a similar stare into a pulse oximeter’s output to monitor the oxygen carried in red blood cells. 

For her part, Shackelford is interested in seeing teams incorporate a sense of time into triage—which sounds obvious but has been difficult in practice, in the chaos of a tragedy. Certain conditions are extremely chronologically limiting. Something fell on you and you can’t breathe? Responders have three minutes to fix that problem. Hemorrhaging? Five to 10 minutes to stop the bleeding, 30 minutes to get a blood transfusion, an hour for surgical intervention. “All of those factors really factor into what is going to help a person at any given time,” she says. And they also reveal what won’t help, and who can’t be helped anymore.

Simulating disasters

DARPA hasn’t announced the teams it plans to fund yet, and self-funded teams also haven’t revealed themselves. But whoever they are, over the coming three years, they will face a trio of competitions—one at the end of each year, each of which will address both the primary and secondary aspects of triage.

The primary triage stage competitions will be pretty active. “We’re going to mock up mass-casualty scenes,” says Chretien. There won’t be people with actual open wounds or third-degree burns, of course, but actors pretending to have been part of a disaster. Mannequins, too, will be strewn about. The teams will bring their sensor-laden drones and robots. “Those systems will have to, on their own, find the casualties,” he says. 

These competitions will feature three scenarios teams will cycle through, like a very stressful obstacle course. “We’ll score them based on how quickly they complete the test,” Chretien says, “how good they are at actually finding the casualties, and then how accurately they assess their medical status.” 

But it won’t be easy: The agency’s description of the scenarios says they might involve both tight spaces and big fields, full light and total darkness, “dust, fog, mist, smoke, talking, flashing light, hot spots, and gunshot and explosion sounds.” Victims may be buried under debris, or overlapping with each other, challenging sensors to detect and individuate them.

DARPA is also building a virtual world that mimics the on-the-ground scenarios, for a virtual version of the challenge. “This will be like a video-game-type environment but [with the] same idea,” he says. Teams that plan to do the concrete version can practice digitally, and Chretien also hopes that teams without all the hardware they need to patrol the physical world will still try their hands digitally. “It should be easier in terms of actually having the resources to participate,” he says. 

The secondary stage’s competitions will be a little less dramatic. “There’s no robotic system, no physical simulation going on there,” says Chretien. Teams will instead get real clinical trauma data, from patients hospitalized in the past, gathered from the Maryland Shock Trauma Center and the University of Pittsburgh. Their task is to use that anonymized patient data to determine each person’s status and whether and what interventions would have been called for when. 

At stake is $7 million in total prize money over three years, and for the first two years, only teams that DARPA didn’t already pay to participate are eligible to collect. 

Also at stake: a lot of lives. “What can we do, technologically, that can make us more efficient, more effective,” says Freeman, “with the limited amount of people that we have?” 

Read more PopSci+ stories.