Virtual assistants, chatbots and digital humans: Health practices discover surprising advantages in conversational A.I.

Practices are interacting with patients in the digital realm more than ever before. During the COVID-19 pandemic, providers began to leverage new telehealth capabilities and intensified usage of mobile apps and web portals. These virtual channels provide immediate care, while protecting patients and frontline staff from transmission risks associated with unwarranted in-person contact.

To guide those arriving on their digital doorstep, many practices are incorporating virtual assistants, chatbots and even “digital humans,” which combine artificial intelligence (A.I.) with a lifelike face and verbal inflection to express empathy and convey emotional context in their dialogue.

Conversational A.I., or the technology that enables users and systems to communicate by text or voice using conventional language, is being deployed for remote triage, tracking appointments and services and helping individuals manage life changes such as retirement, along with its corresponding healthcare implications. These hyper-realistic digital avatars listen to patients, talk them through the hospital discharge process, help them stay on track with post-operative rehabilitation exercises and medication, and manage their mental health.

Any non-diagnostic, patient-facing contact is a candidate to be augmented with conversational A.I. The operative terms in that statement are “non-diagnostic” and “augmented.” Currently, there is no appropriate role for A.I. to independently perform diagnosis due to associated risks and technological limitations. Healthcare diagnostics will continue to revolve around the relationships between human physicians and their patients for the foreseeable future. But by exploring the ways digital humans can augment those relationships, the healthcare sector is discovering exciting possibilities.

Escalating disclosure rates

Healthcare professionals know that patients may downplay symptoms or withhold information about their health in one-on-one discussions. Sharing personal information can be uncomfortable for many, and individuals may limit their disclosure to another person to avoid stigma. In a post-operative discharge interview with a hospital staff member, for example, a patient may be reluctant to ask questions when they feel they are taking up the other person’s time, or out of embarrassment due to being unfamiliar with terminology or procedures they feel they should already know.

Research has shown that digital humans don’t trigger those concerns because subjects know they are communicating with a nonhuman construction in a judgment-free exchange; but that is only part of the story.

In a study for the U.S. military, researchers at the University of Southern California set out to learn how Ellie, a digital human therapist, would affect returning soldiers’ disclosure of post-traumatic stress disorder (PTSD) symptoms. In the first part of the trial, researchers found that, given a choice of an in-person interview or screening by a faceless system, the veterans were more likely to confide in the chatbot than to a real person because they felt less inclined to hide their feelings out of fear of being judged.1

Those results corroborated a body of research showing the absence of social cues in an interaction promotes a greater sense of privacy, particularly when the information is sensitive. In the second phase, researchers added a computer-generated face and body to the chatbot to create Ellie. Researchers found that returning veterans were twice as likely to report PTSD symptoms to the digital human than the faceless system. Anecdotal feedback shows the soldiers experienced a rapport or emotional connection with Ellie, which made them feel more comfortable telling her more information.2

Boston Medical Center received similar responses after it introduced a digital avatar to conduct discharge interviews. The project found that patients took more time asking the chatbot questions than they would have with a human because they didn’t get the impression they were delaying a busy hospital administrator or being judged on their grasp of healthcare literacy.3

These studies illustrate the escalating rates of disclosure an organization can achieve by augmenting in-person encounters with alternatives powered by A.I. Digital humans and chatbots create a temporary sensation of anonymity that encourages admission of symptoms or lack of understanding. Rapport-building through the emotional connection of facial expressions gives digital humans their additional edge over faceless chatbots, and there is no system of face-to-face conversation to rival that combination.

Easing burdensome tasks

Other advantages of conversational A.I. relate to the automation of repetitive or burdensome tasks and reducing the number of physical interactions required between patients and providers. In times of crisis, such as the current pandemic, fear and uncertainty will drive people to seek expert advice to resolve their concerns.

A chatbot, virtual assistant or digital human can learn and respond to these frequently asked questions and educate inquirers about illnesses, risks and other related factors, 24 hours a day, seven days a week. Additionally, their multilingual capabilities allow digital humans to be understood by anyone, no matter their background or geographic location.

Likewise, digital humans can step in as the first point of contact that doesn’t require a face-to-face conversation. More than a convenience, these offerings serve a preference for many individuals, such as members of high-risk groups who may otherwise decline to seek medical attention from an emergency department or doctor’s office due to COVID-19 concerns.

Our firm developed a conversational digital human to demonstrate its applicability for this type of screening. The A.I. solution didn’t attempt a diagnosis; however, it built a baseline of risk factors, such as a person’s travel history and contacts, to recommend whether they seek further evaluation from a physician.4

Faces of the future

Systems that incorporate A.I. and a wide range of verbal and visual communication have the potential to deliver far more than a pleasant customer service experience in telehealth. It’s also important for digital humans to be co-designed with patients and their unique needs in mind. This ensures the proper deployment of digital humans so that they represent the diversity of society, regardless of background or even accessibility challenges.

At the front end of the healthcare relationship, a virtual contact may be the only way some patients will seek medical advice during a pandemic. The bonus of anonymity and the emotional connection a patient experiences with a digital human offer better disclosure of the information healthcare professionals rely upon to make an accurate diagnosis. After a correct diagnosis, referral to a specialist or treatment, a digital human can ensure patient engagement in post-discharge treatments.

When patients experience an emotional connection that builds trust, they openly engage with the practice’s A.I. system and fully disclose their symptoms and feelings. This would enable healthcare professionals to use that data to accurately diagnose conditions and design an appropriate treatment plan.

Dig deeper

• “Why give AI a face?”
• “Will people trust a digital human the same way they trust people?”

Notes:

1. Brigida AC. “A virtual therapist.” USC Viterbi School of Engineering. Oct. 18, 2013. Available from: bit.ly/3qfj7HI.
2. Ibid.
3. Cancino RS, et al. “Project RED Impacts Patient Experience.” Journal of Patient Experience, vol. 4, no. 4, Dec. 2017, pp. 185–190, doi:10.1177/2374373517714454.
4. UneeQ. “AI steps up to fight COVID-19 with new digital human health advisor.” April 7, 2020. Available from: bit.ly/37tbNiW.

Digital innovation in neurology: Retooling rather than rehiring

When the time came for more virtual care during the COVID-19 pandemic, the job of Fran Saperstein, chief operating officer, Center for Complex Neurology EDS & POTS, Phoenix, became that of a telehealth evangelist.

For patients who would establish care in person and then head out of state in the summer months, it was nothing new to be seen via telemedicine. For patients who lived across Arizona and faced hours of driving for an appointment in Phoenix, there was a clear advantage to avoiding that hassle. “When you have a chronic, debilitating condition, the drive takes it all out of you. … This saves you time, it saves you energy, it saves you gas money,” Saperstein said.

Early in the pandemic, the practice had a front office worker quit as the scheduling of new patients dwindled and the office was going to be closed as the first wave of COVID-19 surged in Arizona. That left Saperstein, a care coordinator and one other team member to answer phones, handling medications and payments. Additionally, the practice’s first option for a telehealth platform became difficult to use after initial success during the pandemic, which pushed them to find another solution.

Given these added administrative tasks, shifting away from in-office care meant needing a new infrastructure for virtual care delivery. The group already had a secure, third-party platform for patient communications that allowed patients to fill out registration forms and helped the group answer phones and manage text messaging from patients. When that platform added a telemedicine option to its software package, it made lots of sense.

“We tell them, ‘All you have to do is click the picture of the video camera and you’re in the appointment,’” Saperstein said, adding that patients’ familiarity with the platform aided in making the transition to a virtual visit easier.

Another step was the incorporation of a virtual assistant. In this instance, the virtual assistant was not an A.I. solution, but rather an outsourced worker with experience in insurance and patient scheduling.

“She made my life so much easier that we brought on another [virtual assistant],” Saperstein said. The second virtual assistant had experience as a pharmacist and was able to assist with prior authorizations. Eventually, the group brought on a third assistant — specializing in transcription service — as a scribe for the doctor.

These virtual assistants receive a flat hourly rate, and then the group provides the assistant access to the EHR and Microsoft Office 365 through a virtual private network (VPN) and the group’s VoIP-based phone system to ensure the work remains secure.

“We’re looking at ways to continue to improve their function so that they can be even more helpful to us,” Saperstein said. “It has been pretty amazing in changing the way that our practice runs.”

This simplification of administrative work and expansion of virtual care means patients get a better experience overall, which is important given the complexity and acuity of their health issues. “It’s so much easier for our patients. … They are happier to wait at home than they would be to wait in our office.”