Artificial intelligence is changing the world at breakneck speed. As a result, people often experience a consistency bias in thinking about machine intelligence. Because AI's infrastructure, availability, speed and sheer scale has enabled bolder algorithms to tackle more ambitious problems, humans can easily associate this technology with futuristic innovation — like self-driving cars and robots — and forget that tools they use or encounter in ordinary life — smartphones, online ads, navigation and ride-sharing — are driven by AI.
"Sometimes it's so practical in our everyday lives that we're using it in applications that we don't even think of as AI any longer," said Ashley Allen, Director of Strategy for Waystar.
In mid-September, more than a dozen executives, vice presidents and other senior-level healthcare professionals gathered in Chicago for a roundtable discussion on the relationship between AI and hospitals' revenue cycles as part of Becker's Hospital Review 4th Annual Health IT + Revenue Cycle Conference. The conversation was made possible by Waystar, which formed in 2017 through the combination of Navicure and ZirMed. Waystar reaches 450,000 providers, 22,000 healthcare organizations and 750 health systems and hospitals with its integrated RCM solutions.
Ms. Allen and Dan Ward, Vice President of Strategy for Waystar, led the hour-long roundtable with participants to discover how organizations are leveraging AI in the revenue cycle today, challenges they experience in doing so and AI applications they are clamoring for to strengthen the hospital revenue cycle.
"I want you to push the envelope here and think in your wildest dreams," Mr. Ward said. "If you could poach people from Amazon and Google, bring them into your system and offer them unlimited budgets and two years — what would you want to emerge? What would you really like to see in terms of your revenue cycle and AI applications?"
How hospitals are using AI today
Before the health system leaders answered that question and described their most ambitious AI-RCM aspirations, they began by level-setting and discussing their revenue cycle realities. When Ms. Allen asked participants how AI is used within their organizations today, participants paused to think. The consistency bias — associating AI with the most cutting-edge technologies versus those that are more familiar — was likely in effect. Ms. Allen followed up by asking who in the room leverages propensity-to-pay tools; several participants raised their hands.
"Propensity-to-pay platforms are artificial intelligence," she said. "Many scores rely heavily upon credit score information, but it is still integrated into some sort of algorithm as the foundation for creating that score and leveraging it back to you as an indicator of likelihood of payment.”
Machine learning capabilities mean AI has the statistical ability to recognize patterns in vast amounts of data. Given this power, Ms. Allen pointed out how propensity-to-pay platforms can grow more effective over time as they integrate information that can facilitate better understanding the financial health of a hospital's relationship with any given patient or guarantor.
AI and healthcare: Opportunities and challenges
For as much criticism as the healthcare industry withstands for opaque financial interactions and high administrative costs, Mr. Ward reasoned that the data-driven nature of healthcare puts the industry at a significant advantage for AI in the revenue cycle. Financial data is highly structured, and AI is in constant state of dealing with structured versus unstructured — or gray — data. Hospitals don't have to look far to find structured, uniform data that is highly organized.
"Every data scientist I talk to is incredibly envious of what we have in healthcare," Mr. Ward said. "We have CPT codes and ICD codes. We have discharge dispositions. We have DRGs. In other words, a significant amount of the activity is codified. Yes, there is unstructured data in terms of discharge notes, discharge summaries and the physician dictation, but the amount of data we have that is structured is overwhelming compared to other industries."
The amount of codified data within hospitals' reach is promising, but such a wide swath of information can leave hospitals — especially those early on in AI adoption — feeling overwhelmed. One participant recognized how crucial it is for organizations to deploy AI with clear intent, purpose and objectives. "[One challenge is] the overwhelming amount of information, and trying to focus on where you need to start, then following a path so you can continuously use AI," said the former CFO of a 400-physician medical group in the Northeast.
Although healthcare may be laden with the right kind of information for AI, the clash of technology and the employee experience is something participants cited as a concern. The AI revolution is poised to automate millions of jobs over the next decade and create new kinds of employment for those occupations made redundant.
Yet employment is not purely an economical function — it is also tied to our emotional wellbeing. A few participants said they are actively working to lessen anxiety and reassure members of their team worried about be replaced by technology. In doing so, they point out how AI is sifting away redundant tasks to leave employees with work that requires uniquely human capabilities, such as problem solving, critical thinking and creativity.
"There is fear of what they don't know," said the director of revenue cycle quality with one of the top 10 cancer hospitals in the country. "For automation, they think everyone will lose their job. With each new tool, [my job is] educating them on what it is, how it can make their role easier and how it makes their role more rewarding because the mundane widgets are gone, and they are now challenged with critical thinking. That usually challenges the team members you want to keep. For others who want to do things the way it's always been done, it may be ok if they leave the organization."
Discussion participants acknowledged hospital and health system workforces would benefit from more education about AI and how it may affect them in the short-term.
"My biggest thing I tell staff is, 'This will make your life so much easier,'" said the CFO of a nonprofit, multi-facility hospital campus in the Pacific Northwest that includes a 336-bed medical center. "You can be really transparent about how it will make them more resourceful and efficient versus AI being something that [will] eliminate their job."
Where and how hospitals want to put AI to use
After the participants discussed the opportunities and challenges of integrating AI into the hospital revenue cycle, they shared how they'd like to see the technology leveraged and the qualities they consider most necessary for AI to truly thrive in the revenue cycle — a distinct space where clinical, financial and consumer concerns merge.
Clinician involvement. Health systems' clinical and financial worlds are crashing together. As a result, attendees expressed need for clinical expertise in AI applications in the revenue cycle space. Mr. Ward recognized this by noting the sophisticated use of technology he has seen in health systems where a CMIO filters AI applications through a clinical lens. "Frequently, people who don't have a clinical understanding or understanding of what they're looking at will arrive at a statiscial conclusion that is contrary to operational reality," said Mr. Ward. "I think that's why where there is an active CMIO is where we see the most interesting applications of AI in the traditional clinical areas, but also as it percolates to the revenue cycle and the intersection of the revenue cycle and clinical areas."
Feedback loop with clinicians. While CMIO engagement and expertise sets great AI-revenue cycle pairs apart, roundtable participants also want AI to communicate and inform care teams on a continuous, ongoing basis. The AI-revenue cycle relationship should not be one-sided. "One thing that would be beneficial and helpful is a feedback loop with physicians," said the CFO from the hospital campus in the Pacific Northwest. "Give them real-time information, like 'Here are the things you can improve on in your documentation and ordering.' We find that quite a bit: If you would have documented 'X,' we could have billed for that. But you didn't, and we can't. We need to educate our physicians on building their skill set."
Price transparency. Although the promise of price transparency has received a lot of attention over the past five to eight years, several participants noted that information about procedures' pricetags remains complex, inaccessible and prone to variation. But what if AI were applied to price prediction? It would put providers in better positions to answer questions like this: "We're frustrated for a lot of our patients, who call and say, 'How much will I have to pay for a CT scan?' It sounds like an easy question, but it's not when you consider all of our payer contracts have different criteria," said the CFO of a 25-bed county-owned hospital in the Midwest. "So, you have to know with a payer if you have fixed fees or are they a fee per service? Is it just a simple CT scan or one with contrast? It's not a simple question. But to the patient, it is a simple question: What am I going to pay?"
Denial prevention. Several roundtable participants declared denial management to be a chief concern, particularly related to prior authorization. They expressed a desire to see AI deployed to help determine the need for a prior authorization based on a historical denials, CPT codes, payer rules and other criteria.
Conclusion
In closing, Mr. Ward and Ms. Allen noted the difficult nature of keeping a conversation about challenges, opportunities, preferences and needs for AI in the revenue cycle to 60 minutes. They encouraged participants to remember AI is not just a futuristic, cutting edge technology, but a tool Americans interact with every day. Think about how we encounter AI in daily life, they encouraged, and leave room for imagination on how its convenience, speed, adaptability and ease of use could carry over to healthcare.
"I just really want to see the rev cycle really embrace AI," Mr. Ward said. "We're talking about improving and saving lives, but somehow technology informing the movie we're going to watch on Friday evening is very often making better use of AI than helping people manage access to care and responsibly controlling the cost of having their lives saved."