Imagine being the first paramedic on the scene after a tanker truck has plowed into a city bus. Traffic is snarled, cars are honking and people are screaming. Who needs your attention most—the man on the concrete holding his bloodied knee or the woman on her back with closed eyes? What about the people inside the overturned bus? And what is that white vapor drifting from the truck’s tank?

In emergencies, prioritizing care for victims is called triage. The word comes from the French trier, “to select” or “to sift,” a usage that goes back to Dominique Larrey, the chief surgeon of Napoleon’s armies. A victim might be color-coded as red, meaning he needs help immediately; yellow, meaning he will need help soon; green, meaning he has minor injuries; or black, meaning he cannot be helped with available resources. But sorting human lives in this way is still more art than science.

This uncertainty persists because, although many rules exist to help rescuers, it is hard to evaluate whether those rules actually save lives. The decades-old Simple Triage and Rapid Treatment (START) system color-codes patients as described above. But like other such systems, said David C. Cone, M.D., associate professor of surgery (emergency medicine) and of epidemiology, “we have no idea if it works.”

Cone has spent his career thinking about disasters and mass-casualty triage. He studies emergency medical services and plans for chemical, biological and nuclear terrorist incidents. He has run disaster simulations at Tweed-New Haven Airport complete with volunteers smeared with fake blood. But triage research is inherently difficult. For one thing, said Cone, “we don’t even know what we want a mass-casualty triage system to do.” Is the best system the one that’s easiest to teach? Quickest to apply? Is the most important goal to get patients into ambulances as quickly as possible? Or to save the greatest number of lives? The complexities mount when one considers that every disaster is unique—it is almost impossible to compare triage systems in the real world. Now, though, Cone has begun working with a new tool to study start and other triage systems: a virtual-reality (VR) simulator.

While studying in Italy for a master’s degree in disaster management in 2004, Cone saw a VR simulator used to train Dutch firefighters and realized that the software could be adapted for triage research. Developed by the Dutch company E-Semble, the simulator looks like the highly realistic video game Grand Theft Auto. Learners at a laptop “walk around” a vivid scene, assessing and triaging victims. Dangers and distractions, like toxic spills or television reporters, can be added to the scenario. The learners are timed and their actions exported into a database that can then be analyzed.

Working with emergency medicine resident John Serra, M.D., and supported by the Centers for Disease Control and Prevention and the Laerdal Foundation, Cone plans to teach paramedic students two different triage systems several months apart, then compare how they did with each system in identical VR scenarios. “Once we get the software tuned, then we can design the larger studies,” said Cone. He plans eventually to use the tool to explore whether rules for triage are even necessary, or whether victims are better off being triaged by experienced rescuers who rely on clinical gestalt.

VR may one day allow researchers around the world to collaborate, exchange scenarios and compile “libraries” of standardized victims. Cone hopes the controlled VR environment will allow for real progress in triage research and ultimately save more lives during real disasters.