Three examples of AI applications in healthcare provide insight into how healthcare providers are using the technology today. The organizations are Auburn Community Hospital in Auburn, New York; Banner Health in Phoenix, Arizona; and Community Medical Centers, based in Fresno, California,
Addressing RCM staffing shortages: Auburn Community Hospital
Auburn Community Hospital, an independent 99-bed rural access hospital, uses robotic process automation, natural language processing and machine learning in revenue cycle management (RCM). The hospital started on this path nearly 10 years ago with the transition from the coding standard to ICD-10-CM from ICD-9-CM.
“We wanted to make sure our documentation was an accurate and complete reflection of the care provided,” said CIO Chris Ryan. “We used AI to harvest and ingest the data to prompt our providers and educate them on better documentation.”
During the industry’s move to ICD-10-CM, Auburn used AI in the form of computer-assisted coding to suggest appropriate medical codes based on relevant clinical data in the chart.
“When you have accurate documentation — and then AI reviews that documentation — you’re able to augment the role of the coder,” said Ryan. “The AI does the legwork, and it can do it so much faster than the human counterpart.”
Today, AI continues to promote accurate coding and documentation, and it also helps Auburn address RCM staffing shortages during times of business growth.
“AI has allowed us to add service lines without adding additional labor,” said Ryan. “We can do more with what we have. It helps us retain and attract more coders and be more efficient.”
Over the years, Auburn has seen a 50% decrease in discharged-not-final-billed cases, a more than 40% improvement in coder productivity, and a 4.6% increase in case mix index — all with the help of AI. The overall impact on its bottom line? A little over $1 million, more than 10 times its investment.
Improving RCM staff productivity: Banner Health
After serving as Banner Health’s continuous improvement program director since 2021, Jacci Schavone recently assumed a new role as revenue cycle automation program director, which was created after the health system realized the potential of RPA to transform the revenue cycle.
“As we put in improvements in workflows, there are clear opportunities where we could insert a bot as we make those workflows more streamlined and consistent,” said Schavone, who now reports to the senior director of revenue cycle analytics. “It happened organically from applying continuous improvement methodologies.”
For example, Banner Health automated much of its insurance coverage discovery using a combination of a service that finds each patient’s coverage and a bot that adds that coverage to the patient’s account in various financial systems.
A different bot handles insurance company requests for additional information.
“We get a lot of requests from insurance companies at so many points in the revenue cycle when the patient is still in the hospital and being treated,” said Schavone. “Insurance companies want clinical information to authorize the stay, and they also want certain information post-discharge to do medical necessity reviews. We’re looking at ways to ingest these requests and get that clinical information to payers so patients can receive a complete continuum of care.”
The health system also uses a bot to automatically generate appeal letters based on certain denial codes.
Schavone said using ML and NLP is the long-term goal as AI technology continues to evolve.
For example, Banner Health has also developed its own predictive model that determines whether a write-off is warranted based on certain denial codes and the low probability of payment.
“Machine learning and predictive analytics are great tools to ingest large amounts of data, find patterns and make decisions based on what’s already happened,” said Schavone. “In the very near future, AI is going to ingest that data and make recommendations for improvement. For example, consider our predictive model for write-offs. Soon, AI will synthesize that data and make improvement recommendations like our continuous improvement team does today.”
Combating payer denials: Community Medical Centers
At Community Medical Centers, a not-for-profit healthcare network based in Fresno, California, Eric Eckhart, director of patient financial services, said increasing claim volume continually added pressure on the revenue cycle management (RCM) team.
“I needed a tool in my toolbox to help my staff keep up,” said Eckhart. “My expenses are also highly monitored. Coming out of COVID, it has been rough. I needed something to give me an edge, and I wanted to try different things. AI is just a piece of that.”
The tool it chose is a product from its clearinghouse that reviews claims before they’re submitted and flags any that are likely to be denied based on Community Medical Centers’ own historical payment data and payer adjudication rules.
“It looks at all our claims and all our remittances back from payers and says over these thousands of claims for this particular payer for this particular denial, it’s denying 70% or 80% or 90% of the time. You need to pay attention to this,” said Eckhart.
Community Medical Centers currently uses AI to address two types of denials proactively: Lack of prior authorization and service not covered. For lack of prior authorization, staff simply obtain the authorization number and include it on the claim. Denials for service not covered require ongoing education for patient access staff in the ED, where turnover has been high.
“What we would typically find is that payers were using ‘service not covered’ for eligibility denials,” said Eckhart. “We have a lot of managed Medicaid in California along with traditional Medicaid, and there were a lot of upfront registration issues happening.”
Since deploying the service six months ago, the health system has seen a 22% decrease in prior authorization denials by commercial payers and an 18% decrease in denials for services not covered — without hiring additional RCM staff. Also, the health system saves 30-35 hours a week by not having to write as many back-end appeals.
Before deploying AI, Eckhart said, the RCM team took a retrospective approach to denials management, focusing mostly on high-dollar denials while missing “easy wins” that could have had a large impact. AI found what Eckhart calls a “funky, erroneous denial” pattern that spurred ongoing conversation with one large payer to resolve. “We might have been able to catch this trend through a denials analysis report, but the AI is another tool watching out for it,” said Eckhart.
In the future, healthcare organizations will increasingly use AI to generate documentation for administrative processes such as prior authorization letters or even clinical notes.
Eckhart envisions many future uses of AI, one of which could be to monitor remote RCM staff productivity.
“If we had AI learning [the] work patterns of employees — what’s common and what’s not — and then flagging and making us aware [of those patterns], that can provide some valuable insights and relieve some of my management team from having to pull manual reports, look through time logs and do some of the reviews,” he said.
The hospital is working with its computer-assisted coding vendor to build an application for its multi-specialty practices.
The experience of Community Medical Centers and others like it underscores how AI offers organizations opportunities to gain efficiency, revenue and cost savings by helping them to improve workflow and processes.
