When the new coronavirus began appearing in the U.S. this year, Memorial Health System in Illinois knew it needed to prepare to screen large swaths of worried patients. So the Springfield-based system set up drive-through screening sites to help triage those with respiratory symptoms. What it didn’t expect was the level of interest, amassing more than 4,000 interactions with patients in one week.
“We had a lot of what I’ll call the ‘worried well,’ ” said Jay Roszhart, president of Memorial Health System’s ambulatory group. “A lot of individuals who were just wanting information.”
Enter AnneSarah—a chatbot that provides users with a COVID-19 risk assessment and directs them to next steps, such as a telemedicine visit or calling the system’s respiratory-care clinic hotline.
Chatbots, essentially personified computer programs that instant-message human users in plain English, had been “sporadically” adopted by hospitals and health systems prior to COVID-19, as part of a trend toward more consumer-focused healthcare, said Tom Kiesau, leader of the digital transformation unit at the Chartis Group, a healthcare advisory firm.
But the COVID-19 outbreak has driven a “very, very rapid shift in the use of chatbots as a tool to help triage” patients, he said.
Even the Centers for Disease Control and Prevention hosts a symptom-checker chatbot for the coronavirus on its website.
Memorial Health System in March launched its AnneSarah chatbot, which people access through the system’s patient app or website. Over the first three days, AnneSarah had more than 1,000 conversations, likely cutting the number of “worried well” who might otherwise have visited a drive-through site and pointing possible COVID-19 cases to the right resources.
“It really has saved us from needing to bring a lot more staff in to answer the phones, to make callbacks and to provide reassurance to the public,” Roszhart said.
Memorial Health System developed AnneSarah through a partnership with LifeLink, a company that develops chatbots for healthcare organizations. LifeLink said its chatbot platform starts at $5,000 per month, and scales up based on factors like patient interactions and the number of LifeLink tools added.
There’s a relatively “low barrier to entry” for deploying chatbots when it comes to screening for COVID-19 risk, since organizations can apply information from the CDC’s guidelines to assess patients, Kiesau said. Most chatbots ask questions including a user’s symptoms, travel history and whether they’ve had contact with someone diagnosed with COVID-19.
The more challenging component is connecting patients to relevant resources at the organization, after determining their level of risk.
“How do you handle somebody who’s a presumptive positive, versus somebody who’s maybe exposed, versus somebody who’s early symptomatic?” Kiesau said. “Every single one of those endpoints has a discrete set of process management challenges.”
Scientists are also learning more every day about the virus and how it spreads.
That means the way chatbots make decisions may need to be updated regularly to triage effectively.
“This is such a new disease process, and the virus has such a variety of symptoms,” said Dr. Mahalia Desruisseaux, an associate professor of internal medicine in Yale School of Medicine’s section for infectious diseases. “The algorithms need to be changed continuously for the platforms to work.”
Chatbots aren’t just for patients. At UCSF Health, employees are required to answer COVID-19 screening questions from a chatbot each day before going to work. Initially, UCSF Health set up volunteers at hospital entrances to manually screen workers before letting them into the building—but that took 40 seconds per employee. With the chatbot, from company Conversa Health, that’s gone down to 8 seconds. Employees complete the screening survey at home, and, if OK’d to work, are given an electronic entry pass to display at the door.
That’s more than a 75% reduction in time for “thousands of employees coming into work every day,” said Dr. Aaron Neinstein, director of clinical informatics at the UCSF Center for Digital Health Innovation. “That really adds up.”