Digital health startup Buoy launches AI-powered, symptom-checking chatbot

By Heather Mack
Share

It has become an undeniable fact of life that people are going to Google their health symptoms. However, as most physicians are reminded every day by patients who come to the hospital or office armed with their smartphones or printouts of their search results, the internet often leads people to draw erroneous or even harmful conclusions about their health.

Fed up with the confusion and misinformation of patients he was encountering as a third-year medical resident in 2014, Harvard Medical School student Andrew Le thought there had to be a better way. Over the past three years, Le and his colleague Eddie Reyes worked with a team at the Innovation Laboratory at Harvard to develop a digital health tool that would function as a health-specific search engine while also giving patients pointers and compassion along the way.

“Right now, the status quo is Googling symptoms and getting scared. That is the front door of healthcare,” Le told MobiHealthNews in an interview. “So what we wanted to do was change the ramifications of that front door.”

The result is Buoy, an AI-powered resource, available either online or through an app, that sources from over 18,000 clinical papers, covering 5 million patients and spanning 1,700 conditions. Beyond a symptom checker, Buoy starts by asking age, sex, and symptoms, then measures against the proprietary and granular data to decide which questions to ask next. Over about two to three minutes, Buoy’s questions narrow down to get more and more specific before offering individuals a list of possible conditions, along with options for what to do next. Initially launched in August, Buoy is now coming out of its stealth mode period (during which it amassed over 5,000 users) to be fully available to anyone.

Le and his colleagues wanted to create Buoy to not only make searching for health information less terrifying, but to ensure people had a tool to make more contextual decisions about their health and improve treatment once they sought it.

“Most people Google their symptoms, but the information is unreliable,” said Le, who is CEO and cofounder of Buoy. “I was working in the ER as a medical resident, and I saw so many people Googling and essentially guessing from the results, and very often guessing wrong. It turned out that 78 percent of patients could have been taken care of out of the ER, and while that is better safe than sorry, that’s a lot of waste.”

Not only were some people suffering from the consequences of bad Googling in the form of wasted time and money, others were suffering from arriving at more benign conclusions about their health and foregoing medical treatment.

“Some people were overreacting and getting scared about something that turned out to be minor, but others were sleeping something off and not worrying about it, so that when they finally did seek medical help, they were in a worse-off state,” said Le.

One of those people was Le’s own father, who made the wrong decision from self-directed online research about his health and became very ill. That led Le to take a sabbatical from medical school to work on Buoy. Under the guidance of Harvard’s machine-learning expert Ryan Howard and Dr. Warner Slack, who conducted some of the first experiments between a human and a computer and is considered the “Father of Cybermedicine,” Le and his colleague Eddie Reyes began building the tool.
 
“When we first started, we knew we couldn’t build it ourselves, so we brought in luminaries Warner Slack and Ryan Adams and we figured out we basically had to build an algorithm based off statistics that were granular enough that, over time, as you see more patients, the program can learn and get better over time,” Le said. “For a while, it was a real chicken-and-egg problem.”

Le and his colleagues amassed enough data to teach the algorithm by reading those 18,000-plus papers by hand, then and creating an algorithm to crunch the numbers in real time, for every single person. Before bringing Buoy officially to market, the team put it through a battery of tests, including one that examined how Buoy interprets what a cough might mean compared to other leading web-based symptom checkers.  They examined 100 standardized cases involving 33 different diagnoses, ranging from a benign cough to life-threatening pulmonary embolism, as well as prevalence of rare diseases like histoplasmosis to a common cold. Buoy was right 92 percent of the time, compared to WebMD’s 56 percent, Healthline’s 53 percent and Mayo Clinic’s 38 percent.

“The difference with Buoy is it is learning about each person over time, and it also knows that some symptoms don’t mean anything. If you Google symptoms, by definition you are going to find an illness because nobody writes in the literature or medical records when symptoms don’t mean anything,” Le said. “But if this is a symptom that continues, Buoy can recognize when it goes from being something benign to something you need to go to the doctor for.”

Along with facilitating that kind of healthcare pathway, Buoy also sees its tool as new way to gather huge amounts of research that could otherwise be missed in traditional studies.

“For us, this is something that is a moonshot. Even thinking about the research possibilities just makes me glow, because as we are able to track people over time, we could be looking at a huge potential prospective study,” said Le.

While stealth mode users showed Le and his team that Buoy was being accessed for many common conditions like abdominal pain, coughs and ankle sprains, they have also seen many requests for mental health information.

“We cover the entire DS-V, and we saw a lot of people asking Buoy about things like depression and suicidal thoughts. That was very powerful for us and also very heartbreaking, but also inspiring because we saw a real opportunity to help someone who may not be open to seeking help from a real person, but felt comfortable asking Buoy,” said Le.

Those kind of insights have led to big plans for the future, ranging from partnerships with health systems and research institutions to organizations with resources for patients and their families. The company is already in the process of formalizing a few deals, including one with Boston hospital system Steward Health.  

“Buoy represents a groundbreaking idea in the emergence of on-demand health care,” Steward Health spokesperson Jeff Hall said in a statement.  “We are excited to be working with Buoy to identify ways in which this innovative technology will be used to benefit Steward Health Care’s existing and future patients, physicians, and frontline caregivers.”