AI triage tools won't empower consumers without mastering convenient, personalized service

Chatbots, symptom checkers and other AI healthcare guidance services should take a page from Amazon and others on to engage users, experts say.
By Dave Muoio
Share

Healthcare has long had an access issue. While some patients might have the knowledge, experience and opportunity to pursue optimal care when they’re ill, countless others are unaware of the whens, whys and hows of seeking and receiving relief for their condition. 

“Access is really key. That can be access to an appointment to a provider [and] it’s also access to quality information,” Dr. Erin Jospe, chief medical officer of Kyruus, a maker of provider search and scheduling tools, said during an expert panel held this week at Boston’s MassForward: A Vision for 2013. “I’ve had a number of patients come in with printed tomes … from unreliable sources, or because Oprah told them they needed a test. Sometimes they do, but how can we ensure that people have access to the right information, that they have access to the right appointments and be truly creative in terms of thinking about what access can mean?”

The consequences of uninformed patients also extend beyond an individual’s poor clinical outcomes. Unnecessary healthcare encounters or preventable emergency admissions rack up bills that drain patients’ savings accounts, and place substantial strain on the health systems tasked to treat them. For health systems, payers and startups alike, consumer-friendly AI triage tools could mitigate these issues.

“Not all patients that are coming to the ED should be coming to the ED. These are not patients that are either high cost [or high need], but they’re not appropriate. They’re creating risk to other patients. They’re clogging up the system,” John Brownstein, chief innovation officer at Boston Children’s Hospital, said during the panel. “We want to be focused on the patients that we can really treat best in our institution and make sure that other patients are being routed to the right spots. That’s why a lot of our work has been about taking the know-how of the hospital and build it into software and devices. We know what we decide on a patient in the hospital — why not build that into a decision support software or a device that gives patients the right type of decision-making process and ultimately helps us?”

By providing users a handful of easy-to-answer questions on their condition and history, automated symptom checkers and triage chatbots can direct consumers toward an appointment booking service, a telehealth visit with a remote clinician or key information that the user can use to make informed decision regarding their personal health.

“There’s a lot of mis-self-triage, and a lot of that is most likely because when people are tackling the questions ‘What do I have?’ and ‘How do I fix it?’” Dr. Andrew Le, CEO and cofounder of healthcare chatbot maker Buoy Health, told MobiHealthNews. "So if you could give people a program that can help them distill the statistics around what’s most likely going on and do it in a way that’s scalable, easy to use [and] free, you can create a lot of value for society because the person can do this themselves, understand what may be going on and make an informed decision as if they had medical school distilled down into an easy-to-use program.”

Personalized, trustworthy and free

For consumers, one significant appeal of chatbots and symptom checkers is that the experience is, by design, tailored to the individual, Le said. As the increasing scale and specification of healthcare inches the industry away from patient-doctor relationships and toward impersonal service transactions, AI technology as a whole represents a way to keep things personal.

“Healthcare is still kind of a one-size-fits-all industry. If you look at any kind of patient portal or whatever, it’s basically a page full of links, a link farm. There’s nothing personalized about that,” he said. “The real kinds of personalization are coming at the patient-doctor level, when the primary care doctor knows me and there’s that kind of personal relationship. But even that’s becoming less vogue … I feel like we’re falling even further and further into a place where healthcare is not personalized. And this is where I see a huge opportunity for AI. If AI has democratized personalized guidance for other industries, it can do the same thing for healthcare — [especially] because the way people are interacting with health systems now is not great.”

The first step, however, is getting users in the door. Nearly every other consumer-facing industry has democratized access to information and recommendations, and they’ve done so while focusing on speed and usability. The result, noted Josep Carbó, global business development lead at symptom checker Mediktor, is that automated triage tools must now meet consumers’ expectations of convenience — and they must do so while balancing their inherent responsibility as a healthcare resource to provide high-quality guidance.

“One hundred sixty million Americans Google their symptoms every year. That means people prefer convenience than scientific rigor,” Carbó told MobiHealthNews in an email. “Symptom checkers [that are] scientifically validated can provide users both convenient and reliable access to services, just what they are used to in any other aspect of their life.”

Of course, there is also the ultimate factor in consumer adoption: the cost. If the end goal is to educate and support as many patients as possible, Le said that any price tag will limit both the reach of that service and the opportunities to refine it over time.

“If it’s not free, it’s not easy to use — you’re just compounding the problem of people not knowing what to do to fix their problem,” he said. “If you put a paywall on this, make it really difficult to access or only allow people to get to it through a certain paywall, it’s really not helping the consumers in a way that it ought to be.

“Guidance in today’s world is typically free, and that’s important, because if you’re using AI to make that guidance better, you need as many people to use it as possible. So if something is behind a paywall or hard to access in any way, it’s likely that product is going to be bad or suboptimal to begin with because it doesn’t have enough data to be good.”

Plenty to learn from industry leaders

Healthcare may be a hair behind the curve when it comes to consumer-friendly digital services, but that does provide the space a chance observe and learn from the trailblazers from other industries. For Le, there’s a long list of role models and cautionary tales that should serve as required reading for those looking to design and monetize their AI triage tools.

“If you look at something like Amazon, they really democratize the ability to understand what to buy. And it’s free to get all this great information around what products are good and what products are bad, and then Amazon is monetizing the product that ultimately gets sold,” Le said. “Yelp is another great example. They’re really democratizing the ability to understand where to eat at a really fine-tuned, personalized level — and there again, they monetize in a different way. Even companies like Spotify have a free version that is still driving their machine learning and providing really interesting AI guidance for what song to listen to.”

Assuming that the tool is free for the end user, this issue of an alternative monetization strategy is a major one for automated triage tool makers. Services that refer patients to a partner have an inherent conflict of interest, Le warned, and failing to address that risk can undermine consumer trust — or worse, negatively impact the quality of that user’s eventual care.

“So, for instance, if a company has telemedicine in house tied directly to their AI program or chatbot: are they or are they not economically incentivized to send [patients] to their doctor?” he said. “Telemedicine is great, but telemedicine is only fit for a certain number of situations and problems.

“I mean, look at what happened with Facebook. It was originally just a place for us to connect together, and then all of a sudden they layered on a business model that optimized for clicking on stuff and staying on the site. I would imagine that there was no ill intent, it was a natural evolution of the market force created by that particular business model,” he said. “So [is the navigator] truly agnostic to the services they might be navigating to? And if they are not, how are they going to deal with the conflicts of interest that are natural to their company?”

For his money, Le said that he hasn’t seen a healthcare guidance product (Buoy included) that stands up to Amazon, Facebook or any other tech giant’s mastery of personalization and user-focused design. However, he anticipates that whichever company does manage to find the appropriate balance of monetization, clinical validity and consumer service will have a similarly transformative impact on the healthcare industry.

“If you look at all these other companies that did a good job, it’s not like their algorithms are the thing that separated them in and of themselves,” Le said. “In other words, algorithms tend to relatively be similar. The differences between these companies and their competitors has been their ability to grab consumer insights from real people, which allowed them to leverage those same algorithms to drive the best recommendations, the best personalized guidance.

“So when I think about this space, I think that someone is going to do it the right way. Someone is going to win, and that someone is going to be the one who figures out the right business model so they don’t create distrust with the end user. They’re going to be the one who focuses most on adoption so they can get the most data to make the most personalized guidance. … That’s going to create a ton of positive value for society, and for that particular company.”

Consumerization of Healthcare

In April, we'll look at the consumerization of healthcare from a variety of angles, including how to treat patients as customers.

nike free run 5.0 infrared