Is AI in Revenue Cycle’s Future? Experts Say the Answer is a Clear “Yes, But”
With all the hype surrounding artificial intelligence (AI) and machine learning in healthcare, it’s inevitable that there is now hype surrounding the imminent arrival (and in a very few quarters, actual arrival) of AI and machine learning tools in the revenue cycle management space. But, say experts, not so fast: most hospitals’ and health systems’ RCM infrastructures aren’t yet ready.
First, the experts say, we need to step back a step and look at the landscape around all of this. As everyone knows, during the spring of 2020, hospital, medical group, and health system revenues crashed after the Centers for Medicare & Medicaid Services (CMS) were compelled to effectively order a halt to virtually all elective procedures in hospitals, clinics, and surgery centers nationwide, as the COVID-19 pandemic exploded across the United States.
Provider leaders completely understood the order, and universally complied with it. But, as has long been noted, for many hospitals and other patient care organizations whose revenue base continues to come overwhelmingly from fee-for-service payment from Medicare and private health insurers (and to a lesser extent, from Medicaid), the income from elective surgical procedures continues to make the difference between a 1- or 2-percent revenue margin and a negative one; and as a result, revenues crashed last year healthcare system-wide. Indeed, Indeed, Kaufman Hall’s August 2020 “National Hospital Flash Report,” based on July data from over 800 hospitals, found that hospital operating margins had plunged 96 percent in the first seven months of 2020 compared to in the first seven months of 2019, before rebounding later in the year.
The roller-coaster of revenues history during 2020 and into 2021, experts agree, has shown hospital leaders how incredibly important it will be to absolutely optimize their revenue cycle management processes, during a period of financial instability and narrowed revenue margins.
Indeed, automation adoption is accelerating in revenue cycle management, according to an Aug. 19 press release announcing the results of a survey conducted by the South San Francisco-based AKASA. “The survey found 78 percent of health systems are currently using or are in the process of implementing automation in their revenue cycle operations—a 12-percentage point increase compared to results of last year's survey,” the press release noted. “The findings signify automation in healthcare is no longer an emerging trend but is mission-critical for driving efficiency and cost-effectiveness in revenue cycle operations.” AKASA CEO and co-founder Malinka Walaliyadde cited the opportunity for provider organizations “to expand their ambitions and scope for automation,” by implementing solutions “that can be deployed rapidly with minimal disruption.”
Still, the “eternal verities” in RCM remain very important, as revealed in a July 15 report from AKASA, based on a survey of RCM executives. That report found that “[T]he top five measures of revenue cycle operations success, in order of most to least important, include: net days in accounts receivable (in general); aged accounts receivable (billed >90 days); initial denials rate; discharged not final billed; [and] final write-offs rate.” And, with regard to that set of issues, Amy Raymond, head of revenue cycle operations at AKASA, says that a core issue that continues to dog RCM leaders is that, “As a revenue cycle leader, it feels as though you’re always behind. Staffing is the first issue. And AR [accounts receivable] is being worked 30-60 days after the moment of service. And as if we needed anything to make things even more challenging, the pandemic came,” she adds. “So the need is, we’re drowning here, we’re buried. Rev cycle has always had to decide which accounts to work, and now, those strategic decisions have been made even harder. It was a perfect storm with the pandemic, with reduced revenues, and changing roles for payers. The need to try to increase revenues again is intense; there’s just a real push to work smarter,” she adds. “You only have so many people and so many hands, and so many tasks to do.”
Dustin Cragun, the research director in revenue cycle management at the Pleasant Grove, Utah-based KLAS Research, says that things remain challenging in terms of shifting into more advanced solutions. “Providers are just looking for a result; they just want an outcome,” Cragun says. “And I think to some extent, they don’t even care how they achieve that outcome, whether traditional RPA [robotic process automation], or even screen-scraping, which work great for some things.”
In that regard, Cragun says, patient care organizations are tending to fall into two groups right now. “The first includes providers that have truly embraced digital transformation and are using it to drive competitive advantage for their entire culture. And that doesn’t just mean the tools, it speaks to the ways in which they use the tools.” The second group, he says, “tends to be focused on reducing costs, and start out by thinking they need to reduce headcount, particularly in staff-expensive areas like RCM. I don’t think there’s a right or wrong way to think about this,” he adds. “Every system is on its own journey. But it does bifurcate how the tools are adopted. The former group is more focused on the algorithmic processes, while the latter is focused on overall automation.”
What’s more, Cragun says, “The bulk of the technology is still RPA-based. I’m not yet seeing fast adoption of AI in this area.” Indeed, he says, though interest in AI is “incredibly high, “RPA is providing exception reports or isolating denials, and helping people to fix the claim and resubmit it,” and for many organizations, for now, that is what’s achievable. The question that those who would implement AI solutions need to ask themselves is, “Are you actually able to replace human intervention and resubmitting a claim without human intervention? That’s a pretty big chasm,” between what RPA solutions can do right now and what fully mature AI solutions might be able to do at some point.
In the trenches, ongoing challenges
CFOs and others working in the trenches are determined to move forward, even as they find themselves addressing present-state issues. That certainly is the case at the four-hospital, 850-bed Nebraska Methodist Health System, based in Omaha, says Jeff Francis, the organization’s vice president and CFO. The pedestrian yet persistent issue of denials management remains a day-to-day challenge, he confirms. So AI-based tools are extremely important. Prior to implementing an AI system (Methodist is a customer of AKASA) he reports, “We were finding that it was taking an average of seven minutes for a staff person to check the status of a claim; now, with our AI solution, they can do it in one minute; that makes us far more efficient.” In addition, the web-based solution they use made shifting to remote-based work during the pandemic far smoother a shift.
And, where the rubber really, really meets the road, Francis confirms, is around staffing. “With the example of checking status on claims, we had open positions, we were having to run overtime. We have fewer open positions now because of the efficiency of the tool. We’re definitely in a market where attracting highly skilled workers continues to be a challenge.”
As for the leaders of RCM teams in patient care organizations, Francis says that “If they’re still just trying to deal with claims statusing and claims denials, they’ll use RPA, because that’s the best tool for that job. So if you’re going to act on a claim—if you’re at a point where you’re making a decision as to what to do with that claim, you’ll be getting into the more advanced technology to help make the best decision for that claim. And the best decision might be not to resubmit it but rather to freshen it up, clean it up; and you might need an algorithm to help you make that decision; and that would require AI.”
Peering into the future
Bill Falconer, managing director for strategy and operations at the Impact Advisors consulting firm, based in the Chicago suburb of Naperville, Ill., who has spent more than 25 years in the RCM space in healthcare, working for hospitals, medical groups, and consulting firms, notes that “Something like a third to a half of organizations have at least dipped their toes into some form of RPA. When you start talking about true AI, native-language, we’re starting to see very progressive people going after these things,” says the Birmingham, Ala.-based Falconer, though he adds that fewer than 5 percent of revenue cycle teams have so far plunged directly into any kinds of AI solutions; but that number will soon accelerate, he predicts.
“Besides the fact that AI can be multi-factorial and get into the predictive space, probably more than 85 percent of the data we have in healthcare is free-form and unstructured, and that’s always been this murky space we couldn’t do anything with,” Falconer continues. “AI has the potential to make that free-form data useful and actionable. That’s what’s exciting to me. There are clear use cases. Everybody naturally goes to predictive denials” as a use case, he says, “and that’s valuable. But also, there’s a role for predictive admissions and transitions of care; in the value-based care space, the specificity and granularity of data are so important, in terms of doing a better job of taking care of the chronic and polychronic populations. So in the next two to four years, we’ll see meaningful change, and the folks involved in value-based care will move into the practical application of live implementation and expansion in that area.
How’s your foundation doing?
Falconer emphasizes that he sees that “There are two typical paths that I see organizations taking to get to the point where they are right now. There are those organizations disciplined and focused on their current operations; they have a really optimized EMR and integration across their enterprise; they have very robust analytical capabilities; and they’ll be successful over time in RPA and comfortable in moving forward into AI over time. The organizations that are potentially at risk are the folks who haven’t taken care of the table-stakes, price-of-admission things. They’ve dabbled into RPA, but they’ve started to run into challenges with their existing foundations. For some of them, they’ll have to do some rework” to repair and strengthen their infrastructure foundations before moving ahead. As he puts it, “You can’t have analytics be an IT function only and think of RPA as operational, as RCM, and then try to get to AI. People have to think of it as a multi-dimensional approach. It also needs to be a care delivery challenge, and a patient experience challenge. If you build the bricks all together, you’ll have success. You’ll struggle if all the different elements are disparate, with different owners.” Still, looking at the tremendous challenges of moving forward to optimize charge capture and keep their organizations financially stable, Falconer says that everything that’s happened in the past 21 months should be seen as a wakeup call: “For those who can look forward and see this as a general point of opportunity, I think it actually creates a lot of opportunity. I’m inherently an optimist. I think we will see some clear winners emerge from this year-and-a-half-plus of material disruption.”
Meanwhile, Nebraska Methodist’s Francis emphasizes that becoming hyper-efficient will become the norm expected of RCM teams in patient care organizations. It will all be about “the challenges trying to reduce the administrative burden, and needing to invest as many dollars as possible in patient care rather than the back office,” he concludes.
