The Malta and McConnelsville Fire Department in southeast Ohio serves an area so rural, the closest trauma center is at least an hour’s drive from their coverage area.
That makes the work of paramedics and first responders there all the more important.
Joshua Tilton knows this first-hand. He’s the department’s retired chief clinical officer, and has spent years working as a paramedic and EMS instructor.
Now, he’s defending his thesis for a doctorate in artificial intelligence.
Last year, he worked with the Malta and McConnelsville Fire Department to roll out an AI system in an effort to improve patient outcomes there.
The results were undeniable, he said.
“I know beyond a measure of a doubt that [the system] helped patients.”
How the AI tool works
The tool, called Artificial Intelligence Quality Assurance, collects information from emergency runs and analyzes it, highlighting ways individual paramedics and EMTs can improve.
From there, it creates short, tailored trainings for each practitioner, so they can provide better care for their patients in the future.
“That doesn't mean they're necessarily doing substandard care,” Tilton said. “It’s just, we can always improve, so it targets areas where they can be better, and then issues them continuing education within those identified sectors.”
The program’s impact
Very early on, Tilton said the program identified an area for the department to work on: chest pain.
“As a paramedic myself for 20-plus years, I didn't believe the data,” he said. Tilton pointed out that paramedics spend hours upon hours studying cardiology to qualify for the role.
But after running the modeling again and again with the same results, the team decided to buckle down. They did additional training to recognize less common heart attack symptoms, like back, shoulder and stomach pain.
“Then, we saw direct results of that,” Tilton said.
“Because we had had such an intensive focus on cardiac training, specifically atypical presentations because of the data, we saw a very, very large spike in treatment.”
Within six months of testing the AI tool, Tilton said patient treatment improved drastically.
“We were lucky enough to see care that was already amazing get better across the board,” he said. “And our citizens were able to be the recipients of this amazing care from our practitioners because of this.”
What’s next?
Tilton believes the tool could make a real difference in fire departments nationwide, but its biggest limitation is budgetary.
“I'm a paramedic. I’m an EMS instructor that's paying out of pocket to run the machine learning models as a doctoral student, and so the limitation there was really how much processing power I could afford myself,” he said.
“I understand that every penny that comes in from the taxpayers is judiciously analyzed and spent, but when we start to look at those closed machine learning models, it does unfortunately take money to operate.”
Unlike open models, closed machine learning models aren’t open to the public. That’s a critical feature for this program because it ensures private health information won’t be compromised.
Tilton believes the investment is worth it.
“I don't know how anyone can afford to not do something like this,” he said, “especially in this day and age when this is all readily available.”
