
Inside Trauma Research: Building a video game to improve triage decisions
Deepika Mohan, MD, MPH, has been working for more than a decade to develop Night Shift, a video game that simulates trauma triage decision-making. The goal of the game is to improve the ability of ED physicians at Level III, IV and V trauma centers to recognize severely injured patients who would benefit from being transferred to a higher level of care.

Dr. Mohan is a critical care physician at UPMC Presbyterian, a Level I trauma center in Pittsburgh. She received initial funding to develop Night Shift in 2015, and the first iteration of the game was introduced in 2016. She and her colleagues reported on the first trial in 2017, and the second iteration came out later that year.
Currently, Dr. Mohan and colleagues are conducting a clinical trial to test the video game among several hundred ED physicians nationwide. The trial, which is funded by the National Institute on Aging, will use Medicare claims data to determine the game’s impact on real-world triage results.
In the interim, the Night Shift research team has published a process evaluation designed to identify exactly why the game may or may not work as a behavioral intervention.
I spoke recently with Dr. Mohan to understand the inspiration behind the video game, the playing experience for physicians, and her surprising recent finding regarding the game’s mechanism of action. (This interview has been edited for length and clarity.)
How did you first come up with the idea for Night Shift?
The work really began out of the experiences I had as a fellow in Pittsburgh. I observed that some patients who were very badly hurt would come to us by ground after a delay at another hospital. And then there were other patients who were being flown to us, but when you looked at them you were like, There’s nothing wrong with this person other than the fact that they’re tachycardic because they’re screaming.
And I wondered, What is it that’s causing physicians to flag one patient as being badly hurt and who needs to fly immediately — while another patient who’s actually hurt worse is sitting at the hospital for hours waiting for transfer?
I was talking to my boss, Matthew Rosengart, one night after finishing call and I was like, “I just don’t get it. It doesn’t make any sense.” He was a great mentor. He basically said, “Well, you know you’ve been looking for a research project. Why don’t you try to understand what’s happening at these community hospitals that’s leading them down this pathway?”
Then another mentor encouraged me to connect with Carnegie Mellon, which is world renowned in social and decision sciences research. Bill Chase, Herbert Simon, Daniel Kahneman and Amos Tversky all came out of Carnegie Mellon, which has a great track record of doing really cool stuff.
And she said, “Why don’t you get in touch with them and see if anybody’s willing to work with you to try to understand this problem, because no one has really taken a behavioral science approach to the question of trauma triage.”
So I cold emailed the entire faculty of the Department of Social and Decision Sciences, and one person responded. It was Baruch Fischhoff and he agreed to meet with me. And it all kind of cascaded from there.
What is the Night Shift player experience?
The game is structured as a two-hour mystery video game where you take on the persona of a young doctor who moves to a small community after his grandfather disappears. And he takes a job at a community hospital where he encounters a series of patients, some of whom have obviously severe injuries, some of whom have less obviously severe injuries, and then some of whom have no injuries at all but have diagnostic challenges.
The severely injured cases that are obvious are designed to stimulate engagement, because you have to do things like intubate the patient and give blood. The music and the sound effects try to stimulate that kind of adrenaline rush.
And then the cases that are a little bit more obscure are supposed to teach you about the consequences of undertriage. For example, a patient comes in after a bicycle collision. He says his hands were tingling a little bit when he was in the field, but they’re better now. If you do your diagnostic imaging, you realize he has a cervical spine fracture. And then you have the option about what you want to do with him. You can transfer him to a trauma center, you can admit him or you can discharge him.
If he goes to the trauma center, then the next day you learn that he has had surgery and is doing really well, and your boss give you great feedback. But if you admit or discharge the patient, he comes back and now he has central cord syndrome and is paralyzed. The tingling of the hands was really this marker that he had a bad cord injury that you missed. And then your boss finds you and tells you that you made a mistake and the consequence is that this guy will likely be partially paralyzed because of your decision-making.
The idea is to use emotion as a means of making the decision principle memorable. We have found that most people who take the ATLS course are unable to remember the principles. Why is that? Because most adult education is designed in the same way that we try to teach kids. You’re in the classroom, you’re going to learn a bunch of things, and then you’re going to be tested on it. But what we really want those adults to do is to take that information and use it in the real world, and that’s a different process. So with this video game we are hoping to use emotion to kind of move information from working memory to long-term memory.
Your most recent paper describes a surprising finding. What was it?
Our initial work focused simply on whether or not the video game intervention changes behavior in the lab. Our current clinical trial is a real-world study to see how the game might impact actual triage decisions. And as an embedded project inside trial, we did a mechanism of action study to see whether or not the game is doing what we think it does in the laboratory.
The thing that’s different about this embedded study is that for the first time we tried to disaggregate the reasons why physicians might transfer or not transfer a patient. This work comes out of signal detection theory, which is from World War II.
Back then, a couple of military psychologists were trying to improve the use of radar technology, which was new at that time. They realized that there are two reasons why a radar operator might or might not respond to a perceived threat, and the first has to do with discrimination. Can you tell that what you’re hearing is a signal or a noise? It’s based on your experience. It’s based on your knowledge.
And then there’s your response, which is really about the relative cost of uncertainty. If you’re in London during the Blitz and you think there’s even a tiny probability that something on the radar might be a bomber and not, you know, a seagull, then you would be more likely to react.
However, if you’re at an airbase somewhere in the country, your considerations are different. You’re thinking about the fact that your guys will be going on a bombing run tomorrow, that tonight somebody’s inevitably going to get in an accident, that you’re going to waste fuel, etc. So you’re going to tolerate greater uncertainty than if you were in a place where the consequences of being wrong are much different.
And so we took this model and we use it for trauma triage with the goal of disaggregating the effect of intuition and knowledge from the effect of risk preferences, extrinsic constraints, and other things that move your threshold for transfer.
So in trauma, the two questions are, Can you tell the difference between a severely injured patient and a minimally injured patient? and What are the consequences of being right or wrong in either direction?
If you have only one ambulance in your county and you put someone in an ambulance who doesn’t actually need to go to a trauma center, that means the next time someone has a heart attack nobody will be available to go get that person. On the other hand, if you miss a severe injury, that person could end up with lifelong impairment.
So we wanted to know what the game was doing. Was it improving your recognition of severely injured patients, or was it shifting your threshold to make you more willing to transfer patients?
Originally, we designed the game to try to improve your perception and intuition about who’s badly hurt and who’s not. But through our mechanism of action study, we found that changing behavior via improved discrimination is not actually what the game does. Instead, we found that the game changes behavior by making you more willing to transfer patients.
How might this game impact trauma patient care in the real world?
I have a back of the envelope answer for you. There are 30,000 preventable deaths after injury in the United States every year, according to the National Academies of Sciences, Engineering and Medicine (NASEM). So if we were to reduce undertriage by 10%, we would potentially decrease 3,000 to 5,000 deaths per year.
It’s not a ton, but then I think the advantage is really in the morbidity associated with injury. For many people, severe injuries don’t cause death but they do start off a cascade where you lose your independence, particularly in older patients. Injury is a marker of frailty but it’s also a predictor of frailty.
Trauma centers exert their benefit in large part, not because of trauma surgeons, but because of the resources that surround the trauma center. It’s the physical therapists. It’s the occupational therapists. It’s the geriatricians. It’s the palliative care doctors. It’s the social workers.
So the hope is that if you can move more people into a place where they can get access to these resources, they will have better long-term outcomes.