Chatbots made for Healthcare

mHealth apps are so 2014!

The high cost and mediocre outcome of our healthcare system can often be attributed to the lack of patient engagement after they leave clinics or hospitals. Study after study proved beyond doubt that we can achieve great savings and much-improved outcomes if clinicians and patients simply keep in touch after the patients go home. Effective patient engagement is truly the “trillion-dollar opportunity” in healthcare.

What’s wrong with mHealth apps?

In the past 8 years, many companies had invested heavily in mHealth applications to engage patients for everything from medication adherence, vital sign monitoring, self-reporting of symptoms, telemedicine consults, and behavior coaching etc. The results have been mixed. Some studies showed promise while others fell flat. Even for the ones that showed promise, the studies themselves often require a team of clinicians to follow up with patients to make sure that they use the technology solution on a regular basis. That by itself could defeat the “scalability” of such technologies.

The mediocre performance of mHealth apps is probably rooted in the lack of a truly engaging user experience. The need to install an app, and remember to open the app daily, is a chore, especially as the app reminds people their illness — people like games and entertainment in their mobile apps not sorrow and illness.

In fact, when you ask people why they do not take their prescribed medications, the top reason is “I just do not want to take medications”. Thinking about illness is hard.

Instead of asking patients to open yet another app, good interventions should bring engagement to the patients in communication channels patients already use. In that aspect, the old “non-scalable” nurse home calls are effective because patients already use their phones for phone calls. There is no more app to install and the intervention is “pushed” to the patients without them having to do anything.

Rise of the bots

Interestingly, after years of explosive growth, mobile apps have largely stopped growing. In the post-app world, visionary technology companies, such as Facebook, Microsoft, Amazon, and Apple, are making huge bets on a new user interface paradigm: automated agents in mobile messaging apps, known as chatbots. Can chatbots do better than mHealth apps in engaging patients?

By far, everyone’s favorite apps today are mobile messaging apps. It stands to reason that interacting with users directly inside of those messaging apps is much more effective than asking the user to exit and open another standalone app. Mobile app developers are giving up their standalone apps or skip app stores to write chatbots on mobile messaging platforms. Applications from e-commerce to productivity to enterprise to news to politics to entertainment are racing toward the “conversation UI”. Chatbot is a much better fit for patient engagement than standalone apps. Instead of hiring nurses to call patients at home, why not let the intelligent and empathetic robots do it?

Chat apps will come to be thought of as the new browsers; bots will be the new websites. This is the beginning of a new internet. — By Ted Livingston CEO of Kik.

Using a robotic chat agent to engage patients is not a new idea. In fact, the world’s very first chatbot (ELIZA from 50 years ago) was designed to be a Rogerian psychotherapist who can chat with human patients by reflecting on what the human said. In recent years, ProjectRED has famously developed a robotic nurse (called Louise) who can emphatically talk to patients to go over their medications and home care items at hospital discharge time. However, in general, without the recent advances in AI, chatbots can feel “robotic”, which is a flaw just as bad as those standalone mHealth apps from last year.

Artificial Intelligence is here

New advancements in AI have made today’s chatbots a lot more pleasant than before.

  • A large repository of public chat scripts enable chatbots to learn and mimic human conversations.
  • A whole new line of research on detecting emotions in natural language will enable robotic caregivers to engage patients empathetically.
  • New tools make it increasingly easy to incorporate cutting edge NLP, sentiment analysis, and concept extraction technologies into chat scripts.
  • New AI-as-a-service offerings make it easy to perform complex image recognition tasks, allowing chat users to send in photos, hand written notes, or even QR barcodes.

To see what a future healthcare chatbot might look like, checkout a mock chatbot from design group VIGET. It shows how a chatbot specialized in medication management might work.

For a more complete vision of a personalized health coach chatbot. Checkout Dr. Kvedar’s description of a futuristic health coach named Sam in his book The Internet of Healthy Things (chapter 1). The health coach is a “bot” on the iMessage app, and “he” knows everything about Dr. Kvedar. He guides the human towards healthy lifestyle proactively using insights derived from all kinds of contextual, wearable sensor, and social data.

Xiao Ice from China

Hmm, those sound nice. But are they science fiction? Can we program those chatbots today? To answer this question, look no further than Microsoft China’s famous chatbot called “Xiao Ice” (translation: Little Bing). Xiao Ice is used by over 40 million people, and 25% of them have told Xiao Ice “I love you”. Not surprisingly, many consider Xiao Ice their girlfriend. A recent article by Microsoft’s research director in China gave some very interesting in sights on how Xiao Ice responds to the emotion of the user. An example in the article stands out.

When the user sends a photo of swollen ankle to Xiao Ice, “she” recognized that the picture is a human ankle — a swollen ankle no less. She reasoned that a swollen ankle is probably painful, and then formulated an emotionally appropriate response: “Wow! Are you badly wounded?”. That level of intelligence and empathy are exactly what a healthcare chatbot needs.

Who’s doing it?

Xiao Ice from China is too far away, and not really specific to healthcare. What about VC-funded companies in the West? Are Chatbots really going to have a near term effect on the US healthcare system? The answer is probably yes.

  • Sense.ly is a virtual nurse that engages discharged patients to follow up with treatment plans and adherence. It is not a chatbot, but has many of the clinical and workflow elements we discussed. Sense.ly has raised around $4 million to commercialize its product.
  • Your.MD calls itself an artificial intelligence personal health assistant. It is chatbot that you can ask healthcare related questions. For example, you can ask about your symptoms. It is available on major messaging platforms, as well as its own standalone mobile app. It has raised over $7 million since 2015.
  • Babylon Health is an UK-based virtual consultation company. It utilizes AI chat functions to converse with the patient and determine what the patient wants. It has raised $25 million in 2016.
  • MedWhat is another artificial intelligence medical assistant that can answer questions, such as drug side effects, in a chat format. It is currently available as standalone apps. It has raised over $0.5 million.

We expect to see a lot more products and companies in this space in the coming months!


Dr. Michael Yuan is a co-founder of Ringful Health. He is writing a book on Chatbots to be published soon.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.