Boston Children’s Hospital investigating Alexa skills for doctors in the clinic

By Dave Muoio
Share

In early 2016, Boston Children’s Hospital was among the first healthcare organizations to leverage Amazon Alexa with KidsMD, a skill that helps parents understand their child’s symptoms, seek basic medication dosing guidance, or decide whether an in-person visit is necessary. With roughly 100,000 app interactions to date, more than 200 unique weekly, and plans for a second version of the skill underway, it’s safe to say that the hospital has seen some success with consumer-facing voice technology.

“If you’re a parent who’s concerned about if your child is experiencing weird symptoms, you want to know that you are getting an answer that comes from Boston Children’s Hospital or some other [credible] source,” Devin Nadar, partnerships manager at Boston Children’s Hospital’s Innovation and Digital Health Accelerator, said this week at the Voice of Healthcare Summit at Harvard Medical School, Boston. “We saw an opportunity with KidsMD to be that trusted source of pediatric information.”

Alexa and other similar voice technologies offer a clear opportunity to connect with patients directly, but that’s not to say that the hands-free applications won’t benefit the practitioner as well. In her talk, Nadar described how her group at Boston Children’s Hospital is investigation the role voice-first applications could play in the clinical setting.

Among these efforts is a yet-unpublished national survey of 111 US physicians’ perceptions of commercial voice assistants. Here, the hospital found that very few physicians had never heard of Amazon Alexa (4 percent), Google Home (13 percent) or Apple’s Siri (4 percent), and that a fair proportion both owned and used the three technologies (32 percent, 12 percent, 59 percent, respectively).

“We’ve spent a lot of time trying to get a provider perspective,” Nadar said. “Overall, what we’ve found is providers in the US are excited and interested in using voice. … I’ve heard a lot of people say that providers aren’t interested in using voice, they don’t use it at home. That’s not true.”

When asked how they viewed answers to health-related questions provided voice assistants, 60 percent said they would be “somewhat confident” or “very confident” in the reliability of their responses. However, 70 percent of these doctors said they would be more confident in the responses knowing that the content came directly from Boston Children’s Hospital, Nadar said.

Taking this to the next step, the survey also asked doctors if they would be willing to test a voice technology implemented within a clinical setting. Here, 47 percent said that they would be willing, 17 percent said they would not be willing, and 37 percent said that they were unsure. Following this, Nadar said that she conducted follow-up interviews with 21 doctors from the survey and her own hospital to demo one of the specific clinical Alexa skills Boston Children’s is piloting. After explaining step-by-step how the technology could be used in practice, all of these doctors said they would be willing to help test the implementation.

“The real big value out of this is real-time, hands-free solutions with the potential to reduce human error and aid in infection control,” Nadar said. “One of the things we’ve been focusing on with our pilots is just getting people comfortable with voice technology. I think there’s a lot of barriers, but there will be a day when voice is everywhere, so the sooner you can get physicians comfortable with using it, the sooner you can kind of feel that voice already has a role, I think the more successful we’ll be down the line in tackling really big problems and questions about integrating [voice] in the ER.”

To this end, the pediatric center has also been conducting clinical pilots to test and develop unique Alexa skills developed in-house. These skills were designed for three specific clinical use cases: a surgical checklist providing surgery teams with step-by-step, procedure-specific guidance; a voice-conducted organ verification process for transplant teams that automatically documents responses in a medical record; and an application for ICU workers to quickly reference which practitioners are assigned to a specific bed.

Alexa’s performance during these early pilots was hardly perfect, Nadar said, but the programs were generally well received, easy to implement within a unit, and a time saver for busy teams.

“These aren’t perfect. One of the things that’s interesting is 70 percent of the interactions we had with the ICU pilot resulted in exactly what the physician was looking for,” she said. “Not fabulous, but if you think about it, Alexa isn’t really great at recognizing clinical vocabulary, so we’re really impressed by this.”

Aside from the performance of the tech itself, Nadar said that the team was able to come away with a number of insights that could improve future rollouts. Chief among these was the importance of choosing the best location in a unit to deploy a voice-based device — for instance, nowhere that might have too much interfering noise — and the need to carefully walk practitioners through their first encounter with the application.

“We find helping your users understand what they can’t ask is very important. One of the things with voice is it seems super easy … but if you don’t ask something the right way, you get a little stuck,” Nadar explained. “With our ICU pilot, the first couple times people got stuck they didn’t want to use it again. So, if you lose your user the first time, they’re not coming back. What to keep in mind, and what we’ve been doing since then, is actually walking our users through it the first time … because once you learn it once, it’s really not that difficult and you can repeat the experience again.”

Nadar’s team also came away from their investigations with a number of other concerns voiced by practitioners and others involved with the pilots and designs, such as the challenges of security and privacy roadblocks, or worries that the applications’ benefits may not be worth the learning and development time. While Nadar said that the team was able to sidestep the former of these in “unique and creative ways” — such as by tracking patients by their bed number to compensate for Alexa’s HIPAA noncompliance — she said that addressing the latter would require design decisions that are respectful of a clinician’s workflow, continued testing, and efforts to ensure that the clinical voice applications being created are addressing real issues.

“We’ve done a lot of talking to a lot of different folks around the hospital to ask for feedback,” Nadar said. “One of the incredible benefits we have at Boston Children’s Hospital is we have folks coming up to us asking us to do this each and every day. So, we didn’t have a technology that we pushed onto a use case. The use cases and asks come to us, and we make a match between voice, the use case, and the critical pain point.”