Given the generative AI controversies in Hollywood (likeness theft), academia (cheating), and business (chatbots), you’re probably wondering how generative AI will impact the healthcare revenue cycle. While it’s still very new in our space, so far:
- Epic has incorporated ChatGPT in some health records processing
- Google Cloud now offers AI-enabled prior authorization processing
- telemedicine platform Doximity is rolling out a ChatGPT tool that can draft preauthorization and appeal letters
Very few examples of generative AI are actually working at a healthcare organization currently. A New-Orleans-based healthcare system has integrated ChatGPT into its email system to help clinicians answer emails. There’s also Chapel Hill’s UNC Health which has created a generative AI-based internal chatbot (using Microsoft's Azure OpenAI Service) to respond to questions, provide real-time recommendations and directions to locations.
Generative AI, a controversial improvement on traditional AI, has landed in the healthcare revenue cycle. Rest assured, you have time to explore the promises and pitfalls of this exciting and slightly terrifying new technology.
If you’re evaluating AI solutions for your revenue cycle, only by understanding AI nuances and its true potential as well as its risks will you be able to pinpoint your ideal solution. This article delivers a clear-eyed look at the improvements both traditional and generative AI in RCM solutions can make to your revenue cycle.
Generative AI v. traditional AI in RCM
The first thing a revenue cycle manager needs to know is that the generative AI getting all the buzz is different from the traditional AI that now powers the “AI-driven” revenue cycle solutions that most healthcare systems use today.
Most likely the AI fueling your eligibility checks, patient estimates, prior authorizations, and accounts recovery is considered traditional AI, more specifically, “machine learning,” a subset of AI. Understood as “problem-solving AI,” it accesses large datasets to identify patterns and execute specific tasks. It can make predictions or decisions without being explicitly programmed, but these decisions are based on human-defined rules (algorithms). Traditional AI can detect patterns sometimes missed by trained professionals.
Today’s traditional AI has been getting a lot done for the healthcare revenue cycle. Managers are using it in conjunction with automation to:
- Verify patient eligibility: the AI’s algorithms help systems rapidly verify patients' eligibility, benefits, and coverage, minimizing the risk of services being provided to ineligible patients. This streamlines the revenue cycle by reducing claim denials or delays due to eligibility issues, ensuring appropriate payment for services rendered.
- Accelerate and streamline prior authorizations: AI algorithms analyze patient data, compare it with predefined rules, and assess clinical necessity, resulting in more accurate and faster decisions. It helps automate workflows and ensure adherence to specific guidelines.
- Fill schedules: AI-powered systems can evaluate patient demand, physician availability, and resource utilization data to optimize appointment scheduling. By automating this process, clinics and hospitals can minimize wait times, reduce patient no-show rates, and enhance resource utilization, eventually increasing overall revenue.
- Enhance billing accuracy by automating the process of coding and billing. The AI-powered solution determines the relevant data to extract, ensuring accurate coding and appropriate billing, leading to fewer claim denials and faster reimbursement.
- Automate claims processing: by employing sophisticated algorithms to detect errors, inconsistencies, and fraudulent activities, it improves accuracy in claims processing. AI helps limit the need for additional RCM staff, reducing labor costs.
- Manage denials: by analyzing historical claims data and identifying patterns leading to denials, which reduces revenue leakage caused by claim rejections. The AI also implements corrective actions so that providers can identify areas for improvement in coding, documentation, and charge capture.
- Analyze the revenue cycle: AI algorithms coupled with automation, can crunch vast amounts of healthcare data to identify patterns and trends related to insurance claims, reimbursement rates, patient profiles, and physician performance. These insights enable healthcare organizations to make informed financial decisions by accurately forecasting revenue, identifying potential risks, and strategizing to optimize revenue generation.
How traditional AI in RCM reduces operational costs
Under intense pressure from the staffing shortage and clinician burnout, as well as rising costs and unrelenting inflation, some healthcare leaders are pinning their hopes on traditional AI for financial and operational relief. After all, many healthcare organizations aren’t even turning a profit.
Today, on average, hospital profit margins stand at 3 percent, according to a Moody’s mid-year report. Compare that to 2019’s margins of 7 percent. A recent Fitch Ratings annual report found half of hospitals are in the red, or not turning any profit at all. For long-term healthcare organization viability, costs must come down, somehow.
Across the board, the use of traditional AI in conjunction with automation is helping healthcare organizations not only cut labor costs but also lower prior authorization and claims denial rates, two moves that improve net revenue. Consulting firm McKinsey & Co. asserts that, by using automation and traditional AI, US healthcare overall could cut $200 to $360 billion in (mostly) administrative costs. Echoing these findings, an Institute for Robotic Process Information and Artificial Intelligence study quoted in a KPMG analysis asserts that traditional AI in conjunction with automation can create 25 to 50 percent savings in healthcare costs. Physician group, hospital, and healthcare system case studies confirming significant savings after using traditional-AI-plus-automation models appear on nearly every revenue cycle management company’s website.
AI-driven revenue cycle solutions also improve working conditions
Traditional AI is proving to do more than bring down costs, however. Physicians’ overwhelming administrative duties and other factors are fueling what the Association of American Medical Colleges foresees as a physician shortage of 37,800 and 124,000 physicians by 2034. Doctors bemoan the increasingly complex documentation that forces them to spend their off-hours on “pajama-time” records-keeping. They also pinpoint it as the prime driver of job dissatisfaction. This aspect of practicing medicine must change.
Revenue cycle staff share physicians’ frustrations. Utterly overburdened with prior authorizations, eligibility verifications, good faith estimates, denials, appeals, and more, these professionals can find less stressful work at similar pay in other industries. While specific turnover rates for RCM staff are not available, we do know that, according to the Study on Allied Health Workforce Retention which surveyed 1,000 support staff employees:
- 49 percent are considering leaving their current employer for a different healthcare role
- 39 percent are considering leaving their current position for a different industry
- 60 percent expect to leave their job in the next five years
Studies show that when AI- and automation-powered software supports staff, job satisfaction improves. When Salesforce surveyed 773 automation users recently, 89 percent reported higher job satisfaction after automation implementation and 84 percent were more satisfied with their company in general. Even better, 79 percent feel automation tools help them be more productive. Higher productivity, reduction in staff turnover, and improvement in company morale are just a few of the proven indirect benefits automation and AI can achieve. Of course, higher job satisfaction lowers staff turnover rates, a meaningful, money-saving consequence.
Generative AI…coming sometime to a healthcare RCM system near you
Generative AI (a.k.a. gen AI) in the revenue cycle is still in the development process. Still, knowing what’s coming prepares you to adopt a new, money-saving technology when the time is right.
No one wants to miss out on a technology shift that can disrupt entire industries and, more importantly, captivate consumers. We all remember how Blockbuster missed out on streaming video, Kodak on digital cameras, and Xerox on the graphical user interface (Steve Jobs walked right out the door with it!). And yet, jumping into a new technology too quickly risks bleeding cash with no gain.
So far, the most promising uses of generative AI in RCM for healthcare are:
While still in process, generative AI-powered note-reading and voice recognition dictation software has improved the speed and accuracy of physician’s notes. While a key step in attaining positive patient outcomes, notes overburden physicians. Until now, some health systems have attained limited success using optical character recognition (OCR) and natural language process (NLP) technology in improving the onerous, time-consuming task of physician notes.
OCR algorithms leveraging traditional AI have struggled with inaccuracies, however, particularly when processing messy handwritten notes or documents with complex layouts.
Generative AI methods are revolutionizing OCR systems, providing significant advancements. Through generative AI's context understanding capabilities, OCR systems can now comprehend words even when they are partially obscured or poorly written, enhancing accuracy and efficiency for doctors.
Similarly, software leveraging generative AI more accurately records, transcribes, and organizes notes from interactions between physicians and patients. It then draws from this information to handle administrative responsibilities seamlessly in the background, ultimately freeing physicians from the onerous, time-consuming note-taking, “pajama-time” tasks mentioned above.
Today’s engineers are also working on getting generative AI to assist in the detection of redundant patient records at the start of the patient journey. These tasks could be extended to automating eligibility assessment to jibe far more accurately to payer policies and agreements.
Solutions driven by automation and traditional AI in current form carry out only limited prior authorization tasks. By analyzing patient data, medical histories, and insurance information, generative AI can more promptly and accurately assess whether a patient meets the necessary criteria for a particular treatment or procedure. Further, generative AI can bring in more specificity, handling some of the time-consuming exceptions it now takes staff to complete. Because staff today is often so thin or inadequately trained, backup from technology is a last but critical solution for many physician groups and healthcare systems looking to remain viable.
Finally, developers are exploring gen AI’s potential to assist in accounts receivable by creating original, fact-based appeals to health insurers. Gen AI would use the healthcare organization’s past performance and appeals, along with the payer’s policy manuals, and contracted terms to write unique letters for every case. Think of it as ChatGPT for appeals, with all outreach triggered automatically.
Healthcare RCM leaders on generative AI
With the generative AI Kraken released in the healthcare arena, it’s up to you to decide whether you will be an early adopter or a more cautious observer for now. Most likely, you wonder what others in your field are doing.
Global consultancy Bain & Company surveyed healthcare system executives to gauge the current outlook on generative AI in the revenue cycle. Of the 254 respondents, 75 percent believe generative AI has reached a stage where it can have real impact on healthcare. Despite this outlook, just 6 percent have launched a generative AI strategy.
At this time, it is challenging to put an exact figure on the potential cost savings that generative AI could bring to physician groups, hospitals, and healthcare systems. You can watch this space via the Coalition for Health AI, a coalition of Johns Hopkins, Microsoft, Mayo Clinic and many more, organizing to create guidelines for responsible healthcare AI systems.
Too early? Risks involved in generative AI
Austin Brandt, AI early adopter and co-founder of healthcare utilization company Long Tail Health Solutions, cautions, “With any new technology, there's a hype curve, and we are in the hype phase now.” Hype without proven usefulness and accomplishment can become Bitcoin or Google Glass. No one wants that.
Even more nerve-wracking, AI leaders have pointed to the hazards involved in generative AI. They point to:
- accuracy and reliability: the ability of generative-AI’s ability to deliver precise and reliable outcomes is still under scrutiny. Erroneous results or misinterpreted data could potentially lead to financial losses and poor patient outcomes, putting the entire revenue cycle at risk.
- ethical and legal implications: Generative AI systems employ complex algorithms and models that learn and mimic patterns from large datasets. However, when the algorithms unknowingly replicate biased or discriminatory practices present in real-world healthcare billing, ethical and even legal complications can ensue.
- patient security and privacy: To generate accurate insights, AI systems require access to sensitive patient information, making them potential targets for cyberattacks or data breaches. Safeguarding patient privacy through robust data protection measures becomes paramount, ensuring stringent data access controls and encryption are in place to minimize security risks.
- limited or compromised clinical insight: generative AI, at its current capabilities, lacks the contextual understanding and clinical insight that physicians and experienced healthcare professionals possess. Revenue cycle management is intricately linked to the complexities of healthcare delivery and the nuances of patient care. Relying solely on AI-generated recommendations may overlook critical aspects that human experts can identify. The gap between AI-generated insights and the clinician's judgment should be acknowledged, and decisions be made in collaboration to avoid potential oversights or incorrect billing practices.
- implementation challenges: Healthcare organizations need to invest significant time and resources into implementing and integrating AI systems seamlessly. Training staff, adapting workflows, and addressing potential resistance to change pose additional hurdles.
Possibly the biggest AI solution risk: healthcare team bandwidth
Unfortunately, it’s too common that, after signing an annual contract for $100,000 worth of software, revenue cycle vice presidents and CFOs have neither the budget nor the staff to truly use it effectively. It’s shocking that 20 percent of IT and AI practitioners report 90 percent of their AI models go unused. Healthcare IT experts warn that 95 percent of hospital data goes unused.
To win the revenue gains they envision using either traditional or generative AI software, consulting company Bain & Company urges physician groups and healthcare systems to establish:
- unwavering team commitment to technology efforts, backed by a long-term vision and periodic goal measurement
- affected processes redesign to capture value fully
- a comprehensive approach to investing in technology
- a clear path to move from pilot to systemwide adoption
- sufficient talent and the strategy that ensures technology efforts succeed
- comprehensive indicators of future value to measure success
Without these pieces, pilot programs and software rollouts will fall short of stakeholders’ goals.
Exciting times for AI in RCM
Will generative AI drive healthcare towards a more efficient, cost-effective, and patient-centric future? Or will it crash and burn or, worse, end “not with a bang but a whimper” to quote poet T.S. Eliot?
With exciting headlines like “How AI is Reshaping Healthcare” and “No Longer Science Fiction: How AI and Robotics are Transforming Healthcare,” revenue cycle professionals tend to get excited about generative AI. Most also understand that real challenges – with the technology and with their own system limitations – lie ahead.
The bottom line is that vendors of “AI-driven” revenue cycle solutions for healthcare must not conflate the revenue cycle AI they’re using now (which really is machine learning) with what seems miraculous advances of generative AI, the likes of which is showcased by ChatGPT. At this writing, Gen AI isn’t performing the game-changing miracles in the healthcare revenue cycle that it is for that freshman struggling to write a paper on symbols of evil in Moby Dick. That literary feat truly shrinks the workload from 10 hours to under a minute. The healthcare revenue cycle is far more complicated than the five-paragraph essay.
Get critical revenue cycle tasks done without staff intervention with MD Clarity
Traditional and generative AI will only improve your net revenue and operations if they fit your needs. When your physician group, hospital or healthcare system needs cash flow, you can sweep in earned revenue by examining your current underpayments. Becker’s Hospital Review reports that providers lose one to three percent of their net revenue annually due to commercial payer underpayments. Other studies put that figure as high as 11 percent.
Have you compared expected to actual revenue recently? You might find a significant disparity. MD Clarity’s back-end product RevFind automatically scrutinizes every payment against contract terms, alerting you to underpayments.
With MD Clarity’s Clarity Flow, you can automate critical aspects of patient eligibility, payment estimations, and upfront collections. Enable patients to make up-front deposits directly from an online estimate that reaches them automatically after scheduling.
Get a demo to see how these products help limit your labor costs, reduce denial-triggering errors, and improve net revenue.