2024
2024
Financial decision-makers are learning that AI will turn revenue cycle management upside down, but it might also save the day.
Financial decision-makers are learning that AI will turn revenue cycle management upside down, but it might also save the day.
- ‹ prev
- next ›
- 1
- 2
- 3
- 4
- 5
- 6
- 7
AI’s Irresistible Lure
AI promises to take that even further. Algorithms that “learn” from those processes and use incrementally more data from more sources will not only improve efficiency, but give RCM managers new insights, eventually even recommending outcomes.
“It's become obvious that there's no shortage of opportunities for AI, machine learning and automation to help pull paper out of people's day-to-day work,” says Dane Hudelson, Enterprise Director of Data & Analytics at Sanford Health.
“It's become obvious that there's no shortage of opportunities for AI, machine learning and automation to help pull paper out of people's day-to-day work.”
—Dane Hudelson, Enterprise Director of Data & Analytics at Sanford Health
But finding the right way to use AI isn’t easy. There is very little “low-hanging fruit” in RCM upon which to guarantee good ROI.
AI “is a tricky question, and one that requires you to manage cost, expectations and return,” says Seth Katz, MPH, RHIA, FAHIMA, VP of Revenue Cycle, HIM and Finance at University Health KC. “I don’t think plowing ahead just to say, ‘Hey, we implemented some AI in our revenue cycle’ is the win that it’s being made out to be.”
Sanford Health
Sanford Health is one of the largest rural health systems in the United States, serving more than 1.4 million patients through more than 400 locations and covering a multi-state region of some 250,000 square miles. Sanford Health and its senior care division, the Good Samaritan Society, have more than 44,000 employees.
More importantly, healthcare providers are faced with mounting financial pressure, labor shortages, and operating margins that leave little room for experimentation. New ideas have to prove themselves quickly.
At Sanford, Hudelson and his team first used AI to create a “digital employee” to handle a lot of the back-end details that tie up staff at the end of each month.
“With a little over 160 facilities, our first project was aligning with their operational groups to hopefully make life a little easier on some of the processes that they were having to do for each of their facilities for the month-end closing process,” he says. This “historically required a team of five individuals that would invest, give or take, 80 to 100 hours in some instances, even a little bit longer.”
At organizations like Moffitt Cancer Center, meanwhile, the burden of hiring enough coders to keep up with the relentless flow of medical records is overwhelming. AI offered a lifeline.
“We just couldn’t hire enough people to keep pace,” Bill Arneson, Director of Business Operational Transformation at Moffitt Cancer Center, says. By offloading repetitive coding tasks to AI, Moffitt found a way to address a workforce gap that threatened their financial performance.
But while AI may seem like the savior in a labor-starved environment, it’s not always a cure-all.
“Just because you automate coding doesn’t mean everything will run smoothly without any work on our part,” Katz says.
Katz’s warning is echoed by many in the group—AI can only do so much. When vendors pitch automation solutions, they often oversell the capabilities. The reality is that while AI fixes some problems, it creates others.
Unfulfilled Promises and Hidden Costs
AI’s most significant promise is that it can replace manual, repetitive work, freeing up revenue cycle teams to focus on higher-value tasks, like denials and prior authorizations. But the truth, as many healthcare organizations have discovered, is that automation can open up a Pandora’s box of new problems.
The initial cost of AI implementation, for example, can be staggering. Organizations face steep costs when implementing AI—costs that go beyond just the software.
“It’s not just about installing an AI tool,” says Clark Caserella, data scientist at Sanford Health. “You’ve got to pay for massive data storage, train your staff, and manage the technology day-to-day. It’s an ongoing investment.”
While AI promises low overhead in the long run, the up-front investment can be prohibitive, particularly for smaller or safety-net hospitals that operate on razor-thin margins.
University Health KC
University Health in Kansas City, Missouri is a large, non-profit health system that includes two acute care hospitals, University Health Truman Medical Center (a 238-bed hospital) and University Health Lakewood Medical Center (a 110-bed hospital with a focus on primary care, obstetrics, women's care, orthopedics, and geriatrics). University Health KC has more than 70 primary and specialty clinics and more than 4,500 employees.
“We sometimes wait nine months to a year just to get resource support after we’ve had a plan in place.”
—Shannan L. Bolton, VP of Revenue Cycle Optimization at Stanford Health Care
AI is often marketed as a plug-and-play solution, but the truth is in the details. New AI technologies aren’t well understood by healthcare leadership, making it difficult to pitch and even harder to implement.
Shannan L. Bolton, VP of Revenue Cycle Optimization at Stanford Health Care, noted that the process of reviewing AI technology, signing contracts, and rolling out the system can take months, if not years, particularly with the labyrinthine security checks and reference verifications that are required.
"We sometimes wait nine months to a year just to get resource support after we’ve had a plan in place," she says.
Beyond the cost, there’s the challenge of managing increasingly large amounts of data—or finding the right vendor partner to do that job. And because AI is an evolving technology, health system executives need to monitor those processes. That means not only keeping a “human in the loop” to review all AI output, but creating a governance team and protocols for continuous monitoring.
Moffitt Cancer Center
Moffitt is the third largest cancer center in the United States by patient volume. It's also one of only 47 National Cancer Institute-designated Comprehensive Cancer Centers. Moffitt treats patients from: All 67 Florida counties, all 50 states, and more than 130 countries.
“It needs its own steering committee, leadership and policies that govern how the tools are evaluated, in what order they are rolled out and metrics for success,” says Katz. “It can’t be ad hoc, where the clinical, financial and business operations sides all work independently and just buy and implement whatever they want willy-nilly.”
The key to ensuring data integrity, panelists said, is having access to the data being used. That may be a problem with vendors who control data access. It may require a different type of vendor agreement than the health system has used in the past, or safeguards to ensure that any errors that crop up are quickly addressed.
The Denial Problem
One of the most painful parts of the revenue cycle is claim denials, and AI is frequently held up as a solution to this long-standing problem.
Denials are a major source of lost revenue for hospitals, often triggered by inaccurate or incomplete documentation from physicians. Many organizations are betting on AI to solve this by analyzing historical data to predict, prevent, and manage denials before they happen. But here again, the promise of AI comes with caveats.
Sanford Health is focused on using AI to ensure timely and accurate physician documentation—a key driver in reducing denials. In theory, AI should be able to process this documentation in real-time, cutting down on delays and preventing errors. But what happens when the AI itself makes a mistake? Hospitals are forced to monitor AI outputs rigorously, and without proper oversight, errors can still lead to costly denials.
At University Health KC, Katz emphasizes that AI should be seen as an augmentation of human effort rather than a replacement. “It helps keep things efficient,” he says, “but it also opens up other problems.”
This is a sentiment echoed by many in the industry—AI can ease the burden of denials, but it’s no magic bullet.
It’s also important to note that payers have been using AI as well, some longer than providers, and they’ve been using the technology to, in some cases, fast-track the denials process. Some health systems have in fact launched AI programs with the express intent of catching up to payers.
“AI should be seen as an augmentation of human effort rather than a replacement.”
—Seth Katz, MPH, RHIA, FAHIMA, VP of Revenue Cycle, HIM and Finance at University Health KC.
Stanford Health
Stanford Health Care is a major health system based in Palo Alto, California, with more than 18,000 employees, including 2,283 physicians, 4,153 nurses and 1,615 residents and fellows. The health system has 604 licensed beds, including 119 licensed ICU beds, and handles 1.2 million outpatient visits annually, as well as 74,299 adult emergency room visits and 23,506 pediatric emergency room visits each year.”
Revenue cycle executives see the value of AI in denials management through two lenses. One, it will help them understand why payers deny claims and help them to create strategies that address each payer’s tendencies and protocols. Second, it will help them improve their own processes to proactively reduce denials. In both cases AI is seen as a tool to reduce the “human touches” that cause denials and fine-tune the administrative work that will allow providers and payers to have more meaningful interactions (and perhaps even like each other).
A Clash of Cultures: The AI Revolution Meets Healthcare’s Conservative Mindset
Even as the benefits of AI become more apparent, healthcare organizations are often slow to adopt the technology. Many health systems don’t have the data storage and computing resources, or the people with the right skills to handle those tasks.
At Sanford Health, Hudelson said the health system developed an enterprise data and analytics team in 2015 to address growing automation needs, including the inevitable advance of AI. That team now consists of some 70 people—including mathematics experts, Epic-certified report developers, statisticians, even a research nuclear physicist and some psychology folks.
Steven Kos, MSHCA, CHCIO, Senior Director of Revenue Cycle Application at Florida’s Baptist Health, says the rapid adoption of AI has caused health systems to develop and hire new skillsets, including revenue cycle informatics, revenue integrity and even patient advocacy or engagement. This puts healthcare in direct competition with other industries in recruiting—and it puts healthcare at a distinct disadvantage, as hospitals can’t match the salaries and perks offered by the likes of Amazon and Google.
Multiple Mastermind participants are deeply concerned about security and liability risks, which slow the adoption process to a crawl. Because AI is such an innovative technology, it has caught many health systems by surprise, in that staff don’t fully understand what risks it poses. Security teams must therefore spend months vetting AI solutions, and even then, the fear of a data breach can stall the process.
Baptist Health
Baptist Health (Jacksonville) is non-profit health system with six hospitals, 1,168 beds, and over 200 patient access points of care. The system also includes a cancer center, four satellite emergency departments, and more than 50 primary care offices in northeast Florida and southeast Georgia.
“The way vendors present AI, it sounds like you can flip a switch and start reaping the benefits. But it doesn’t work like that.”
—Jacqueline Samuel, MBA, PMP, Director of Revenue Cycle Quality, Strategy, and Analytics at Grady Health
At Grady Health System, the biggest challenge is not just finding the resources to implement AI but also convincing staff to embrace it. “How do we get staff from here to there in using more advanced tools?” asked Jacqueline Samuel, MBA, PMP, Director Revenue Cycle Quality, Strategy, and Analytics at Grady Health, highlighting the challenge of change management in healthcare.
When AI does get implemented, it often takes longer than expected to see the results. As Katz points out, “The way vendors present AI, it sounds like you can flip a switch and start reaping the benefits. But it doesn’t work like that.” AI requires training, ongoing monitoring, and—perhaps most importantly—a cultural shift in how hospitals approach technology.
The Road Ahead: Is AI Worth the Struggle?
The future of AI in revenue cycle management looks promising, but it’s also still uncertain.
Hospitals that are already using AI report significant improvements in productivity, particularly with tasks like coding, claims management, and denial prevention. However, as with any new technology, the challenges of implementation and ongoing management cannot be ignored.
AI has allowed health systems like Allegheny Health Network (AHN) to dramatically reduce the time it takes to complete certain processes. By tracking the time it takes to perform tasks manually and then automating them, AHN has made significant strides in improving efficiency, especially in claims processing.
Grady Health
Grady Health System is a large academic medical center in downtown Atlanta, Georgia that provides emergency, acute, and outpatient medical care. Grady Memorial Hospital has 953 licensed beds, and Grady Health System has 5,001–10,000 employees and 3,000 physicians representing over 80 medical specialties.
Allegheny Health Network
Allegheny Health Network has 14 hospitals and more than 200 primary- and specialty-care practices in more than 300 clinical locations and offices. AHN has approximately 2,600 physicians in every clinical specialty, 21,000 employees, and thousands of volunteers.
Community Health Systems
Community Health Systems is one of the largest healthcare delivery systems in the United States and operates 69 acute-care hospitals and has roughly 12,000 acute care hospital beds.
"We're all trying to remove touches from the claims process," says Brian Ice, Vice President of Clinical Revenue Cycle for the Allegheny Health Network. "We're all trying to come up with ways to make that process more efficient."
But even in success, there are frustrations. “When we first rolled out bots, we didn’t realize we’d need more user licenses,” Ice recalls, describing how even small logistical issues can derail a project.
Many of the Mastermind participants see AI giving RCM executives more opportunities to address patient financial issues. Indeed, the biggest benefit of AI going forward could lie in tying RCM to the patient journey.
Shannan Bolton, Stanford Health Care’s Vice President of Optimization and Performance Improvement, sees AI becoming a powerful; tool for educations and financial counseling, helping patients to both understand their financial responsibilities and the options available to them for paying their bills.
“That’s where we fall short with patients,” she said.
And Christina Slemp, MHA, MSHI, Vice President of Revenue Cycle for Tennessee-based Community Health Systems, added that AI can help reduce the stress for patients by giving them the information they need quickly, rather than waiting around for explanations.
Ultimately, AI in the revenue cycle isn’t a quick fix—it’s a long game. It requires patience, careful planning, and a willingness to embrace both the rewards and the risks. For healthcare organizations ready to commit to the journey, AI offers the potential to revolutionize the revenue cycle.
In the end, AI may save the day—but it won’t be without a few bumps along the way.
Looking for even more on what our HealthLeaders Mastermind participants achieved?
Looking for even more on what our HealthLeaders Mastermind participants achieved?
ewicklund@healthleadersmedia.com, Linkedin
