Dr. Yaa Kumah, assistant professor of biomedical informatics at Vanderbilt University Medical Center, discussed the opportunities and hurdles for voice technology in provider organizations last week at Harvard Medical School.
Time is one of a doctor’s most precious resources. But with the increasing emphasis on documentation, doctors are often snowed under with paperwork. Scores of startups and provider innovation organizations are looking for ways that new technologies could help tackle this problem.
Recently, voice technology has come onto the digital health scene as a new tool with the potential to help lighten the workload.
“Now as opposed to having to learn how to tell the computer what you want it to do, you can literally tell the computer what you want it to do,” Dr. Yaa Kumah, assistant professor of biomedical informatics at Vanderbilt University Medical Center, said during The Voice of Healthcare Summit at Harvard Medical School last week. “That is just amazing and can allow so many more people who don’t have all the tools and resources to interact with a physical keyboard to still get the same benefits.”
When it comes to saving time, voice has its advantages, Kumah explained. For example, writing by hand produces 15 words per minute, typing is about 40 words per minute, and texting on a phone is 30 words per minute. However, speech trumps all of those mediums with a speed of 150 words per minute, she said.
“The argument is that if it takes any less time to say it than to find it, voice wins. It’s an argument that is an easy one to make,” she said. “If it’s easier for me to get the work done, doing it this way versus the other way, then why not do it this way? … You talk faster than you can type. You can get information you need to take care of your patients without having to struggle and worry about typos.”
Kumah has a team piloting a new voice user interface for EHRs at Vanderbilt Medical. EHRs, she said, are notoriously not user friendly and can take additional time.
“You want doctors to be as well-informed about you and what is going on with your care so when they walk into a room they can give you the most effective care,” she said. “Right now, unfortunately, the way our EHRs are laid out, we are pointing and clicking just to find a simple value like a weight. But why? It’s all there. This gets back to the issue of not great design from our EHRs. EHRs are built by people who are technologists, but not necessarily informed by people who are the end users. I know that when I see a person’s A1C it would make sense for me to be able to trend it very easily. But it takes five extra clicks for me to do that natural step. But the information is all there.”
In the new pilot, the voice user interface was designed to give the physician information about the patient in a similar fashion that an attending physician would ask information from a resident. Kumah said it’s important for the system to be designed in a workflow familiar to the providers.
Where voice is falling short of user expectations
While Kumah said the advantages of voice is clear, there are also unique challenges with the technology.
“One of the hardest things is that everybody knows how to talk and everybody has expectations for how it is going to work, and when it doesn’t work precisely as they expect them to, then there is this distrust in the system,” she said.
For example, she said the medical world has its own set of linguistic norms.
“Doctors don’t have a lot of time. They want to get to the facts, they want to know what they need to know in terms of next steps of taking care of their patients,” Kumah said. “We learned that very quickly with one of our first interactions with blood pressure. When you first asked about blood pressure, what you would get back is, ‘120 millimeters of mercury over 80 millimeters of mercury.’ The provider said, ‘This is taking too much time. Just say the blood pressure is 120/80.’”
While this might be an annoyance for a provider, units are important in medicine, Kumah said, so finding the correct balance is tricky and an ongoing process.
Another issue in voice is timing. When you speak, on average it takes the other person 200 milliseconds to respond, she said. But that doesn’t always happen in voice technology. It can take a few seconds for the technology to find the information and speak to the user.
“People have expectations about how things are supposed to sound, and when you violate that expectation you end up giving a very bad user experience,” Kumah said.
While the technology has a lot of promises, it’s still not perfect. Kumah said she and her team are still grappling with questions over timing.
“What we need to figure out is, how early days is this? When is it appropriate to put it out there for everyone, and when is it considered rushing out the gate to get something really exciting in front of people, and then they consider it not quite ready and then they lose trust in the process?”
Date: August 14, 2019