Dr. Gian Garcia, who recently received his PhD from U-M’s Industrial & Operations Engineering program, analyzes data to help shape human and machine processes and improve decision-making. His recent published work in the Journal of Neurosurgery delved into the world of concussion, particularly making the Sport Concussion Assessment Tool (SCAT) more efficient for concussion diagnosis. 

Dr. Gian Garcia (right).

The SCAT is a standardized concussion assessment tool, encompassing three different assessments used to evaluate athletes. Although thorough, SCAT can be a time-consuming task for clinicians. Because of that, Garcia set out to determine which subset of SCAT’s components could be used to accurately diagnose a concussion. 

Using CARE Consortium data and under the direction of Dr. Steven Broglio, director of the Michigan Concussion Center and Dr. Mariel Lavieri, associate professor of Industrial & Operations Engineering, Garcia combined mixed-integer programming (a methodology used for making optimal decisions) with traditional statistical modeling techniques. Through this process, they discovered that five symptoms could be labeled as “key indicators” of concussion:

• Headaches
• Dizziness
• Athlete says he/she doesn’t feel right
• Symptoms get worse with physical activity
• Symptoms get worse with mental activity

The concentration score and delayed recall from the Standard Assessment of Concussion (SAC), one of the key SCAT assessments, also appeared as concussion indicators in his model. 

“If someone runs through the parts of the SCAT we identified, and they don’t see anything indicative of a concussion, they might add the other parts of the assessment,” Garcia said. “But if you can identify a concussion early on with the smaller subset, that can save a lot of time in the long run.”

One assessment, the Balance Error Scoring System (BESS), didn’t show up in Garcia’s results. “That indicates to us that in the presence of other things like symptoms and SAC, BESS isn’t adding a lot of value. We don’t want to get rid of that aspect of the assessment, but if we find something more sensitive to concussion, that would be better served here,” he said.

One finding surprised Garcia: the subset symptoms accurately identified concussions better than the entire SCAT.

“Whenever you are taking these summaries of higher assessments, you might be inadvertently adding noise to your assessment,” he explained. “In our analysis, we were trying to get the most bang for our buck, but in some cases, by adding in these things that may be noise at this point, you can actually hurt your performance and assessment.”

This isn’t Garcia’s first time working with concussion-related data. He previously used statistical models to develop a framework aiding clinicians in concussion diagnosis. The framework utilizes an athlete’s testing performance to determine if they have an unlikely, possible, probable, or definite concussion.

He is also researching how athlete over and under-reporting of concussion-related symptoms influences the clinician’s return-to-play decision-making and the risk it brings to the athlete when they are not being honest.

Garcia has been using his engineering background to solve health-related crises since he was an undergraduate student. 

“I thought it was cool that I could use this mathematical modeling and apply it to high impact problems and do some social good,” he said. “When I was applying to graduate school, I knew I wanted to keep doing work in health care, but I wanted to try something in Industrial Engineering that no one else has done.”

He discovered concussions during his initial research and, in speaking with Broglio, discovered a sweet spot surrounding the types of challenges that clinicians face, how they can be addressed using methodological tools, and their potential for positive social impact. 

In August, Garcia will begin a one-year postdoctoral position at the Massachusetts General Hospital Institute for Technology Assessment at Harvard Medical School.  While there, he will use data-driven modeling and decision-making to analyze the impact of social and behavioral components on issues surrounding the opioid crisis. “This is a high-impact problem, and there are aspects that I can take into future research. I’m hoping I can use this postdoc experience to help strengthen the foundation I have in terms of research methodology and broaden my research experience,” Garcia said.

Once he completes his postdoc, Garcia will start as an assistant professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech University.