What happens when you wake up with a sore throat? Maybe you reach for a couple cough drops and go about your day or maybe you question whether or not you have strep throat. Should you see a doctor? Is it contagious?
A new app developed through a partnership between Treatment.com and medical experts at the University of Minnesota Medical School wants to help you answer that question. Called the Treatment app (for now), the program uses medical intel and artificial intelligence (AI) to show patients the likelihood that their symptoms are linked to certain illnesses.
“We’re building a global library of medicine,” said John Fraser, CEO of Treatment.com. “It used to be that when you wrote software, you would have if-then statements, but that doesn’t work in medicine because it’s too rigid. It doesn’t allow for thinking. Because the app has been designed by doctors, it’s been built to operate like a doctor.”
How does this rival WebMD? The Treatment app, unlike WebMD, walks patients through a self-assessment, considering the medications they are taking, their medical history, any risk factors (like smoking), epidemiology and more. Kevin Peterson, MD, MPH, a professor in the Department of Family Medicine and Community Health (FMCH) and director of the Center for Excellence in Primary Care, leads the Medical School’s efforts in helping develop the app.
“I’d compare WebMD to a textbook. It’s basically a summary of diseases, and you can’t just have a symptom and throw a textbook at a person,” Dr. Peterson said. “This app will make medicine more accessible to more people. But, I want to be clear—we’re not putting a doctor in a box. Doctors provide a lot more than just the facts. They provide compassion and a relationship. We’ll never replace a doctor, but we can have an autopilot where we help the pilot fly the plane. AI can help people better understand their symptoms and can help doctors provide more rapid, more comprehensive and safer evaluations. That’s what we see ourselves doing.”
Feeding the brain
For the app to work, it needs every scrap of medical knowledge available. The team currently works with seven fellows and four faculty from the departments of FMCH and Medicine. That’s where one of the medical research fellows—Rashedat Oshodi, MD—plays a role. She tackles one illness after another, inputting medical definitions, synonyms, risk factors and other data about each illness into the back-end of the app’s AI.
“All of this data helps the patients so that when they use the app, and the machine wants to give them possible diagnoses, it can ask if they smoke, if they drink, and it ties all that back to related illnesses,” Oshodi said. “If you have abdominal pain and fever, you’re more likely to have this compared to this.”
After Oshodi catalogs an illness into the system, Peterson and the team of faculty reviewers approve and publish the content. This same team reviews if the AI is reading the data correctly. Part of that testing involves giving the AI the OSCE (Objective Structured Clinical Examination) this spring.
“We’re targeting the OSCE, which is a clinical skills exam where actors have certain diseases and the medical students interview them to create a clinical diagnosis. Only this time, we’ll just have the exam taken by a mobile app,” Peterson said.
Building a lifelong
Once Peterson’s team finishes feeding and testing the AI, the app will be ready for download worldwide. They expect to launch in March or April 2020. As time passes, they plan to continue building the app to include even more features.
“We are going to add genomics and individualized treatments, so that you can put in your symptoms and all your information, get a set of diagnoses and when you select which diagnose you think it might be, then we’ll show the standard treatments for that kind of disease,” Fraser said.
This treatment information, Fraser says, may include referrals to doctors nearby or may provide a way for app users to order, for example, recommended cold and sinus medication from a local drugstore to treat their symptoms.
“The intent of this app is that it forms a potentially life-long relationship with you,” Fraser said. “We’re having doctors train it in a way where if the disease is important enough or a symptom is serious enough, then the AI should follow-up with the patient. That follow-up and looking at patient history—that more comprehensive relationship with the patient—from what I’m hearing from doctors, is critical. You have to have that to provide a good diagnostic infrastructure.”
Now, and in the future, the app will be maintained by a combination of medical experts from around the world. But, it’s the AI’s machine learning that will continue to make the app smarter.
“You can’t build an app like this unless you have a really good clinical foundation and it’s traceable back to some experts. That’s why the U of M Medical School is such a great partner,” Fraser said. “It’s really about building a platform where we take the medical expertise that’s out there already, put it into a common location where everybody can access it and then, that same community can make it smarter and smarter over time.”
That access is important to Peterson and his team, especially to Oshodi, who is a native of Nigeria where she practiced medicine for three years before moving to the Twin Cities in 2017.
“Thinking back to countries that I am from, sometimes, even though the medical doctors or hospitals are available, they are not affordable to some people,” she said. “If you do have the app, then you can decide if your symptoms are serious enough for you to see the doctor. Where I’m from, that’s a huge thing. People would rather not go seek medical help because they cannot afford it. Now, we’ve given them the option to decide.”
Peterson says half the world has little to no access to medical care and providing even a basic understanding to patients has tremendous capabilities. It’s a duty he feels emboldened to act on as a faculty member at a land-grant institution.
“Part of our future is the democratization of some of this testing and evaluation that’s been held within our medical environment. It should be available to people a little more easily and without such high barriers,” he said. “We really think everyone ought to have access to medical information that is relevant, appropriate and tailored to them, and we can do that now.”