Having proven their value in administrative uses, the Pennsylvania health system is now applying the cutting edge technology to clinical care.
With more than 80 AI bots deployed since 2020, Geisinger Health is ready to push the envelope and use the technology in patient care.
Narayana Murali, MD, CMO of medicine services at Geisinger, says health systems should take an intentional approach to AI bot adoption.
"It depends on how you view the solutions you are looking for," he says. "You want to understand the present state, and you should not start off with the solution on the front end. You try to look at the issues that need to be addressed or solved, then you start the process of the AI bot application."
The first issue is to understand the problem you want to solve, Murali explains.
"The second piece is not to focus on the solution—you need to understand the gaps or barriers in the workflow," he says. "Then you can start to design the necessary applications to address those gaps or barriers."
Murali gave several examples of areas where AI bots have been deployed at Geisinger.
- AI bots are being used for patient outreach such as scheduling laboratory testing or office visits.
- AI bots are promulgating policy changes mainly in the administrative sphere. They are not being used for policy changes related to patient safety or care quality.
- AI bots are being used for compliance such as program audits.
- AI bots are boosting productivity in patient communication such as helping clinicians manage electronic in-box messages.
- AI bots are being used in the emergency department such as the timely generation of notices of admission to reduce the number of claims denials.
"You need to understand AI bots from the standpoint of the entire workflow and where you need to address your challenges," Murali says. "Otherwise, you will come up with point solutions that are not capable of increasing efficiency."
AI bots have value from the perspective of the CMO, according to Murali.
"If I can achieve dramatic improvements in efficiency, it allows me to do things that reduce mundane work for physicians and frontline staff," he says. "AI bots can address pain points for the organization. You can reduce burnout among frontline staff by using AI bots."
Narayana Murali, MD, is CMO of medicine services at Geisinger Health. Photo courtesy of Geisinger Health.
Taking a careful approach to AI deployment
Health systems should take great care in AI bot adoption, according to James Blum, MD, chief health information officer at The University of Iowa.
"We have concerns about the nature of AI bots and the potential responses patients are going to get from them," he says. "One of the concerns is whether the responses from AI bots are going to be accurate and are going to provide the right information."
At this point, AI bots have the greatest potential in administrative functions, Blum explains.
"There is less concern in these cases, particularly if the tool is trying triage administrative issues," he says. "If there is a billing issue and the patient wants to set up a payment plan, that can be a good application. There is a place for AI bots in these areas, where the consequences are lower, and the risk is lower."
AI bots are more problematic in areas that affect patient outcomes, according to Blum.
"If you have anything that is potentially diagnostic or triage-based that impacts patient care, validating the quality of the response and making sure that is going to provide accurate assessment and triage of a patient is paramount," he says. "Many of these tools are doing symptom checking, then they are trying to figure out whether to send the patient to urgent care, the emergency department, or an office visit. We have real concerns about not having a human in the loop."
Customer experience is another area of concern regarding AI bots, Blum explains.
"You need to make sure that you have a patient population that wants to interact with an AI bot versus wanting to interact with a human," he says. "The question is whether you are providing a high-quality interaction that people desire or are you giving them lesser service at lower cost."
AI bots have the potential to support value-based care, according to Blum.
"We have looked at AI bot products in this domain, where the bots act as a coach for the patient," he says. "An AI bot can contact the patient to set up an annual physical and set that appointment. This communication can be conducted through text, and this presents an opportunity."
In the future, AI bots will be more appropriate in areas that impact patient care, Blum explains.
"In the coming years, there will be opportunities for AI bots to be far more triage-focused, to be doing diagnostics, and to eventually moving into ordering authority," he says.
Today, the focus of AI bots is on value-based arrangements such as following up with patients and administratively getting patients to the right place inside a health system to take care of administrative issues, according to Blum, who adds AI bots will eventually be appropriate for more advanced healthcare applications.
"With diagnostics, at some point the time will be right to provide triage services with AI bots—we will probably see that in the next five years," he says.
Christopher Cheney is the CMO editor at HealthLeaders.
KEY TAKEAWAYS
Geisinger Health started adopting AI bots in April 2020, and the health system has deployed more than 80 AI bots.
The health system is using AI bots in several areas, including patient outreach, compliance, and promulgating administrative policy changes.
The chief health information officer of The University of Iowa says health systems should be wary of using AI bots in areas that impact patient outcomes.