Hospitals turn to artificial intelligence to help with an age-old problem: Doctors’ poor bedside man

One of the most common complaints lodged against physicians aims at their communication skills — or lack thereof.

​

Doctors have a demanding job. They’re tasked with diagnosing symptoms, tending to back-to-back patients, filling out medical records, and the like. The hours are long and the stakes are high — mistakes can prove deadly. And somewhere in there, people skills can get lost. Studies suggest it’s one of the primary issues people have when they leave the doctor’s office.

It’s an age-old problem that Alex Young, a Bristol, England-based orthopedic surgeon, set out to solve in 2018.

“Complaints can happen anywhere, but in health care, they’re amplified tenfold because you’re doing things like breaking bad news or explaining a diagnosis to a patient who may not have medical understanding,” Young said.

Three years later, as the pandemic pushes medical facilities to innovate, his start-up, Virti, is supplying “virtual patients” to teach hospitals across Europe and the United States how to better talk to people.

“What we wanted to do with the virtual patient was create a scalable, data-driven way for people to practice their soft skills and communication,” Young said.

Think of virtual patients as AI-powered animations trained to interact with doctors and test them on practicing empathy and other interpersonal skills.

The software can live on a smartphone or computer. For greater immersion, the company can supply physicians with virtual reality headsets. Once the training session is over, doctors are scored based on their speed, the questions they asked the AI and whether they got the fake patient’s diagnosis correct.

Talking to the tech is supposed to improve how they engage with patients in the real world, Virti says.

Already, the Health Education Center at the University of Texas Medical Branch in Galveston picked up the firm’s tech to tutor future clinicians on communication and reasoning skills.

Cedars-Sinai Medical Center in Los Angeles started out as a customer before moving to invest. In 2019, Virti raised $2 million in a seed round originating from the health-care clinic. The investment helped the start-up launch a sales team and enhance its analytics software. These days, the software is used to train Cedar-Sinai medics on coronavirus-related processes. The United Kingdom’s publicly funded health-care system, NHS, also used the tech to teach staff how to properly use personal protective equipment and how to engage with patients and their families.

Virti’s cloud-based training program uses speech recognition, AI and computer-generated characters to simulate realistic interactions with patients. For instance, if a clinician asks the animated human to describe its symptoms, the AI will generate a relevant response.

Hospitals can customize how Virti's avatars look before doctor's train with them.​

Doctors can probe with further questions, and as they interact, the software analyzes their tone, cadence and quality of answers. VR headsets unlock other metrics, such as eye contact tracking.

Virti baked in speech, body language and mannerisms that mimic those held by real people, the company says. Still, computer-generated humans come with limitations. It’s hard to replicate human attitudes or the way individuals respond to pain and illness. Human patients don’t always give doctors the information they need to make an accurate diagnosis, and sometimes, real-life patients aren’t able to say much at all.

The medical training platform celebrated a soft launch last year when many nonessential medical clinics were on pause globally in the face of the rapidly spreading coronavirus. The AI software arrived just in time for Jack Boulter, an Exeter, England-based musculoskeletal podiatrist, to toy with it under lockdown.

After Boulter’s job at a private practice temporarily shuttered, the bone injury specialist signed up for Virti’s free trial. He preferred the role-playing on his smartphone but also went back and forth with the virtual patient on his Mac. He later extended his access to the service and demonstrated it with colleagues via Zoom.

“I was looking for kind of some professional development stuff to keep me occupied,” Boulter said. “After suddenly having a few months off, it’s easy to get a little bit rusty.”

Virtual patients are meant to combat other dilemmas weighing down the health-care industry.

The software was designed to train against implicit bias, which occurs when staff make unconscious assumptions about a person who walks in. The judgments may be well-intended but can also taint diagnoses.

The company addresses this by enabling hospitals to customize the patient's skin color, age, height, sex or gender. Digital faces are based on photos of various actors and other professionals, Virti says. There are up to 60 variations currently deployed.

In health care, communication training often relies on actors who come in and pretend to have an ailment. Peers or higher-ups evaluate clinicians on communication tactics deployed with the actor. Much of that teaching was interrupted by the pandemic.

Those in-person scenarios can “feel a bit forced” anyway, Boulter said, adding that fake patients “feel much more authentic than somebody making it up as they go along.”

Source