Chatbots that analyze users’ symptoms and offer advice are all the rage in Silicon Valley.
Ada Health, a European tech company that is bringing its symptom-checking software to the United States.
About four million people have downloaded and used Ada’s app, according to its website. But in a few weeks, a version of its symptom-checking software will also be available for folks to use for free via Sutter Health’s website, without needing to sign up via the app.
As luck would have it — or rather, bad luck — I recently had a bout of what seemed like food poisoning to put it to the test.
Here’s how it went:
I signed up with my first name, date of birth and some basic health information, like whether I’m a smoker. At that point, I started typing in my symptoms, primarily nausea and vomiting, as well as how long I’d had them. It wasn’t particularly serious, so I noted that it cleared up in less than a day.
From there, Ada led me through a series of questions to find out more about how I was feeling. I noted that I didn’t feel any abdominal bloating, dizziness, or cramping, among other symptoms. All in all, it took me about ten minutes to get through all of the prompts.
That led me to a web page that listed a breakdown of plausible explanations for my ailment. The most likely was viral gastroenteritis, which is a condition that I’m told I could manage from home. But some of the others, such as irritable bowel syndrome, came with a recommendation that I see a doctor. I was planning to stay at home and I was inclined to believe the most obvious explanation, so this gave me a little extra reassurance.
Had the chatbot sent me down the right path?
I called up Sutter Health’s chief of digital patient experience Albert Chan, to get his perspective on whether he agrees with the assessment, why he chose to work with Ada Health, and what he thinks these sort of medical AI tools will be used for in the coming years.
Chan believes there’s a gap in the market for technologies that can help patients figure out what to do when they get sick, which led him to reach out to Ada Health about a partnership. As Chan points out, it’s not always obvious for most people whether to stay home to rest and recover, see a primary care doctor or go straight to the emergency room.
“We want them to get the appropriate level of care,” said Chan.
Chan is also a doctor, so I briefly ran him through my symptoms and he felt that the AI had advised me well. I noted that three of the five potential diagnoses indicated that I should seek medical care, while the most likely (viral gastroenteritis) suggested that I could stay home.
Chan said these tools are meant to be “advisory,” rather than a replacement for a doctor or a formal diagnosis. If I got any worse, he said, I should consider seeing a doctor.
I called up Ada Health’s chief commercial officer Jeff Cutler, who noted that the tool is designed to learn about its users over time.
“It really depends on the person, as some might take action based on the most severe decision while others are reassured by the most likely one,” he said. That’s why Ada lists a range of possible explanations to help patients make their choice.
Cutler said that if I were using the Ada app, these recommendations would get more accurate as Ada learned more about me. (The Sutter Health version on the website is anonymous, but I can save the assessment and download the app from there.)
While it’s sexy to imagine that an AI will replace doctors — and many in Silicon Valley do find that vision enticing — that’s not going to happen anytime soon. There are so many technological and regulatory limitations before an AI can diagnose disease.
In my case, I sort of knew what I wanted to do: I was feeling OK and didn’t want the hassle of going to a clinic, so I took Ada’s least-frightening suggestion to heart. But I could imagine someone else in my shoes wanting to play it safe and booking an appointment regardless. Technology can help be a guide, but personal health decisions are often subjective based on the patient’s personality, family history and a range of other factors.
But here’s where I do think tools like Ada could prove useful:
- More honesty: Sometimes it’s less awkward to share embarrassing symptoms with a chatbot, which might mean patients are more upfront. At one point, Ada asked me if I had experienced any anal discharge, a question that would have prompted a nervous giggle if a doctor had been asking.
- Informing patients: In Chan’s view, doctors can use these tools to make the most of their time with a patient, as the most obvious questions will get asked upfront. “When they (the patient) gets to the doctor’s office, they’ve both already thought about it and have a leg up.”
- Replacing tools like WebMD: Many of the websites and apps that people use today for symptom-checking are far dumber, and will often lead the patient to believe they have something deadly like cancer, when their actual diagnosis is fairly minor. Tools like Ada are clearly a step up from that.
- Better navigation, in some cases: Over time, technologies like these will get smarter about directing the patient to the right place, even if they do typically recommending medical care over staying home in cases where it’s not clear either way.
Sutter’s integration with Ada goes later this month. If you have a chance to use it, let me know what you think @chrissyfarr on Twitter.