Provider to Patient Ratios by 2020
In my latest book, The New Mobile Age, we spend some time talking about the growing demand for care (we’re getting older and older people need more care) as compared to the flat supply of healthcare providers. The outlook is bleak – we’re running out of young people to take care of old people. It is a global problem.
I can only think of one solution — to introduce more automation into the care delivery process. I’ve written about the need to adopt the use of technology to create one-to-many care delivery models in several recent blog posts looking at technology and the provider, voice technology, and connected health for our aging population. It was also the subject of a recent TEDx Talk.
In this context, I’ve become very interested in chatbot technology and am wondering if it’s ready for prime time. Could we create a near-term future where your first interaction with the healthcare system is via an automated chatbot? Are today’s chatbots up to the challenge? Are we able to seamlessly escalate to a human being, if necessary, so you don’t feel trapped in chatbot hell? Will individuals feel cared for and will we be able to spread our healthcare provider resources across larger populations of patients?
To achieve this, an individual’s conversation with a chatbot has to approach feeling human. This is not a new concept. In 1950, Alan Turing developed a test where a human being interacted with a computer and, in order to pass the test, the computer needed to seem human to the person. Over the years, computers have gotten better and better. Alexa and Siri are ubiquitous now, but do they pass the test? Only in a very narrow way.
James Vlahos is a journalist who writes about this topic and, in fact, created a ‘DadBot’ to immortalize his father’s life story before he passed away. You can read more about this in a very poignant story James wrote for Wired last year. He will be a keynote speaker at the Connected Health Conference on October 18 and I’m excited to hear what he has to say. In February, he wrote another intriguing article for the magazine, about Amazon’s Alexa prize, which offered a significant sum of money to create a bot that could carry on a conversation for 20 minutes. There was no winner, but I highly recommend reading the article, which illustrates the subtle ways our minds work and that, so far, no computer is able to think in such a nuanced, sophisticated way.
But, the functionality required in bots used for frontline healthcare interactions may not need to be as sophisticated as Amazon was shooting for. Several years ago, we collaborated with a computer scientist at Northeastern University (Tim Bickmore) to investigate whether a software bot could motivate people to be more physically active. We called her ‘Karen the virtual coach’ and she had both visual (a cartoon-like persona) and auditory (an early speech-to-voice engine) components. The paper is worth reading, but the bottom line is that we showed that folks who interacted with Karen three times per week were significantly more likely to achieve their exercise goals than those who did not. Interestingly enough, we found that participants either loved or hated Karen. At the end of the study, a few participants asked for her phone number as they contemplated inviting her on a date!
Our study was published in 2012, and in the last few years, a whole burgeoning industry of chatbots for health has sprung up. It is still early going and, in my experience, none of these is quite ready for prime time. Following is a brief review of several prominent companies working in this space. I’m sure I missed some, so if your favorite health chatbot did not make the list, leave a comment or send me an email so I can be educated.
Conversa Health focuses on three areas: care management to deliver efficient chronic care management, decrease remissions and optimize pre- and post-surgical care; marketing/patient experience to increase patient acquisition, improve patient satisfaction, patient retention, and generate incremental patient visits; and improving patient responses.
XebraPro is a decision support tool for physicians, not to be confused with Xebra, which is a medical imaging software platform. XebraPro, from Physician Cognition, has two versions, one for differential diagnosis and one called XebraED, an education tool. Unlike Conversa and some of the others, XebraPro is a tool to improve physician accuracy and efficiency.
Buoy, an online a symptom checker, guides you through a series of questions and then recommends a course of action. Their business model steers patients to a provider organization. Using their site or app does not feel exactly like having a ‘chat’ but rather filling out a questionnaire.
Medumo focuses on care navigation and instructional support to help patients. The primary delivery mechanism is text messaging or email, and the primary value proposition to the provider is getting patients to their appointments. Patients who don’t keep appointments generate a significant administrative burden and Medumo has set out to solve for that. Mostly their communication is outbound to the patient, so it does not feel like a chat, per se.
Babylon believes it is possible to put an accessible and affordable health service in the hands of every person on earth, using a combination of artificial intelligence and natural language processing. They made instant headlines by winning a contract with the National Health Service of Great Britain. Their interface is very similar to Buoy, guiding an individual through a series of branched logic about their symptoms to a recommended course of action.
Lark combines artificial intelligence with behavior change design to create a scalable, personalized care management platform. Lark claims to be clinically validated to deliver positive health outcomes across each major chronic disease state, with its main virtual coaching platforms used for diabetes and hypertension. The interface is a bit wooden in my experience, and there’s no attempt to disguise this to be anything but a dialogue with a software bot. The bot asks questions and you choose various answers. It seems overly simplistic to me.
Babylon Chatbot Interface
Memora is an automated discharge chatbot that can also manage follow up paperwork, enabling a care team can focus on the patient instead of the details of discharge planning. Their virtual assistant is named Felix, and the idea is for Felix to have an identity. The goal is threefold: to increase a care team’s capacity; rapidly identify high-risk patients, and deliver a world-class patient experience. It also streamlines time in front of a screen for the patient, promises 24/7 responsive team members, and an immediate start.
Tess by X2AI
X2AI seems like the most up-to-date version of AI of all of the technologies I have reviewed so far. With the goal of supplementing human therapists, researchers are instructed to teach empathy to artificially intelligent messaging tools. Their assistant is called Tess and she can schedule an appointment with a therapist, have a dialogue with a patient or connect the patient to one of the on-call therapists around the clock, for a small fee of $5 a month. Another useful feature is the nightly text encouragement from a therapist and suggestions that accompany each text. The goal of X2AI is to create a patient experience that includes interactions with Tess and some with people, with the software seamlessly triaging the interactions.
There is so much rapidly-moving innovation in this space — from companies with vastly different personas and consumer/patient interactivity — that we have to keep close tabs on it. From XebraPro, a diagnostic aid for the physician, to Buoy and Babylon, symptom checkers, to X2AI, a virtual mental health counselor. They all share one thing in common, which is a backbone of artificial intelligence.
At some point, the first part of any healthcare journey will be interacting with software like this. In some cases, that’s all that will be required. In others, the interaction will move to a telemedicine solution with a provider and, in still other cases, the interaction will move to a face-to-face experience with a provider.
How far into the future is this?