Hospitals turn to chatbots, AI for care
Chatbots and virtual personal assistants are helping to evaluate symptoms, manage medications and monitor chronic conditions.
Chatbots and virtual health assistants may be coming to a hospital near you.
As providers look to better engage patients and improve efficiency, many are turning to artificial intelligence to help them succeed.
AI-powered medical assistants can book appointments, remind patients to take their pills, monitor a patient’s health status and perform other time-intensive tasks. They can also assist with inventory, billing and claims management.
In Arizona, a doctor designed an influenza campaign around a chatbot that alerted patients when flu shots were in and invited them to get protection.
Rising smartphone use coupled with the increasing popularity of health apps and the Internet of Things, as well as growing adoption of telemedicine and other mobile health technologies, is driving market growth. According to Global Market Insights, the virtual health assistant market is expected to exceed $1.5 billion by 2024.
Major industry players include Nuance Communications, Next IT, Welltok, True Image Interactive and Medrespond, but there are a number of smaller, lesser-known firms that are having an impact in the space.
It’s still a “very, very emergent” market, Laura Craft, research director at IT consultancy Gartner, told Healthcare Dive.
“The real vision is for them to be avatars, perhaps used as the interface to have sentiment analysis,” said Craft, who sees the first tier of value in helping patients stay on meds or manage chronic conditions.
But while there are a few avatar-based health bots, in general there needs to be more maturing of the foundational technologies they rely on before they become mainstream in healthcare.
For example, while there have been major strides in natural speech recognition technology, sentiment analysis — the ability to intuit how a patient is feeling and ask questions that reflect on that — is still in very early stages. It could be highly applicable in spaces like behavioral health.
A long way from Eliza
Practically speaking, the market can be split into chatbots — which allow patients to have a conversation and offer some customized decision logic or act as a customer service interface — and more personalized virtual health assistants.
The idea of a robotic chat agent isn’t new. Eliza, the world’s first chatbot, was created by MIT scientists 50 years ago as a Rogerian psychotherapist that could reflect and provide feedback on what patients said. Current examples of virtual health assistants include:
- Florence acts as a virtual nurse helping patients with medication adherence and maintaining healthcare regimens. It also can help locate specialists and schedule appointments.
- Eva helps women track their menstrual cycles and pregnancies.
- Molly — an avatar-based virtual nurse assistant — connects patients with clinical advice to assess their condition and suggest appropriate followup.
- HealthTap allows people to chat with a doctor and send pictures and copies of lab results to see if a problem requires additional care.
- Your.MD asks users about symptoms and offers medically approved potential conditions and then makes referrals and schedules appointments.
- Ada is an AI-powered technology, asking people how they’re feeling and guiding them on next steps.
From an efficiency standpoint, chatbots offer a real opportunity to relieve staff of repetitive tasks. They also could help organizations deal with staff and budget constraints.
Virtual personal assistants, on the other hand, are clinically oriented to help maintain a patient’s health and well-being. “It has been shown that patients who engage with somebody are more likely to adhere to their care paths or protocols, so I think there is a huge opportunity from a clinical perspective,” Craft says.
Tell me how you feel
One company seeking to increase patient engagement is Sense.ly, maker of Molly. The virtual nurse platform uses speech recognition to engage patients in conversations about their health and then triage them to the appropriate level of care.
Here's how it works. A patient tells the virtual assistant that they are experiencing a headache and fever, and the app elicits information on the severity and duration of those symptoms. Based on the discussion, a back-end algorithm then analyzes whether the person should go to the emergency department, schedule a doctor appointment, speak to a nurse via a nurse phone line or provides self-help information, explains Adam Odessky, CEO and founder of the San Francisco-based startup.
Sense.ly also monitors patients with chronic conditions, engaging them to ask questions about their day-to-day health, and integrates with Bluetooth devices to take vitals like blood pressure and weight. The information that is collected is used to determine a patient’s risk of exacerbating their condition and potentially winding up in the hospital.
“The real goal of the virtual assistant is to essentially steer patients to the right level of care based on their condition and ultimately reduce costs and improve the outcomes and efficiency of the healthcare system,” Odessky tells Healthcare Dive.
Sense.ly, which has about 75,000 users on its system, is currently part of a National Health Service pilot study in the U.K. to evaluate the use of digital solutions in handling emergency helpline calls. According to Odessky, early results suggest about 20% of all interactions that people have with Molly are being steered to the nurse line, reducing demand on primary care physicians.
Barriers to adoption
Despite their potential, barriers remain to widespread adoption of chatbots and virtual personal assistants.
The elderly are more likely to resist using digital technologies, and the lack of a personalized approach in many of the current tools poses a challenge for business expansion, says Sumant Ugalmugale, manager of healthcare and medical devices at Global Market Insights.
It may be a generational barrier. Odessky says elderly patients tend to be more comfortable with voice recognition technology, while younger patients prefer text messaging.
Concerns about patient privacy and data security also affect market growth. Like other IT tools, chatbots and virtual assistants are susceptible to hacking, potentially putting personal information at risk.
“Keep in mind these are nascent, very emergent technologies … and one thing we’re dealing with in healthcare that is different in terms of maybe other technologies is our very, very low fault tolerance,” says Craft. That means the algorithms these technologies use and the way their capabilities are deployed need to be very precise because clinical decisions will be made based on their advice or risk scores.
That said, the potential for chatbots and virtual assistants to disrupt healthcare is huge. “Part of the vision is that a virtual personal health assistant will also say it’s time for your annual mammogram and can help schedule that for you, or it’s time for you to go get your blood workup and schedule a lab visit,” Craft adds. “There’s huge application, but we’re totally not there yet.”
That may be an understatement.
For now, chatbots fill a relatively niche role, but with providers adopting more telemedicine techniques, it's likely patients encounter them more frequently.