
Generative AI is poised to revolutionize healthcare’s revenue cycle management. While there’s much enthusiasm, the technology is still in its early stages, and challenges lie ahead. It holds promise in simplifying administrative tasks like generating appeal letters and improving front-end processes. However, it may struggle with nuanced patient interactions and complex denial issues. The road to widespread adoption will require overcoming regulatory hurdles and addressing data privacy concerns. In the next two to five years, generative AI is expected to significantly impact revenue cycle management, potentially transforming the healthcare industry’s operational efficiency.
Generative AI is making waves in the field of healthcare, but its most promising applications might just lie within revenue cycle management as healthcare providers seek ways to streamline their administrative tasks.
Healthcare is currently captivated by the potential of generative artificial intelligence (AI). This cutting-edge technology, exemplified by popular tools like ChatGPT, has demonstrated its prowess by passing medical exams, diagnosing complex conditions, and even contributing to the fight against COVID-19. Yet, its most significant impact on healthcare might materialize in the realm of revenue cycle management.
The revenue cycle management landscape is ripe for innovation. It encompasses a multitude of administrative tasks that must be meticulously executed for healthcare providers to receive reimbursements for their services and maintain smooth operations for patients.
Technology has long promised to simplify the intricate complexities of revenue cycle management, aiming to reduce administrative burdens, minimize costs, and enhance efficiency and productivity. Over time, healthcare organizations have invested in various technological solutions such as robotic process automation (RPA), natural language processing (NLP), and more recently, artificial intelligence (AI), with the goal of achieving these objectives.
Generative AI is the latest entrant in this technological revolution, offering the promise of improving processes and enhancing user and customer experiences. However, as is often the case, technology’s potential can sometimes outshine its practical implementation. After all, healthcare spending continues to rise rapidly, and administrative costs account for a significant portion of this expenditure.
The question remains: Is the enthusiasm surrounding generative AI for revenue cycle management justified, or is it simply hype? Can this new technology genuinely address the major drivers of healthcare spending and complexity?
FACT OR FICTION?
As per ChatGPT, one of the prominent generative AI tools, generative AI is defined as “a class of artificial intelligence techniques that involve training models to produce novel data samples or content, often in the form of text, images, music, or other media types.”
ChatGPT goes on to explain that generative AI leverages deep learning algorithms, particularly variants of neural networks, to create content that is not merely copied from existing examples but is generated based on the patterns and structures it has learned during training.
ChatGPT’s explanation not only sheds light on its inner workings but also exemplifies the capabilities of generative AI: the ability to predict coherent and usable text (or images) based on extensive datasets. These large language models (LLMs) powering generative AI can hold significant potential for revenue cycle management.
According to Varun Ganapathi, PhD, co-founder and chief technology officer at AKASA, “By leveraging an LLM, it lets you dramatically broaden the amount of training data that you have because you don’t need to label anything anymore. You have a ton of data from literally all of the English texts generated as part of the revenue cycle and elsewhere. There is a ton of revenue cycle applicability.”
However, it’s worth noting that the state of generative AI is still in its early stages, as observed by Austin Brandt, co-founder of Long Tail Health Solutions.
“With any new technology, there’s a hype curve, and we are in the hype phase now,” Brandt stated, referencing the Gartner Hype Cycle, which delineates the typical lifecycle stages of a technology from its inception to eventual obsolescence. These stages include the technology trigger, peak of inflated expectations, trough of disillusionment, slope of enlightenment, and plateau of productivity.
Expectations are soaring concerning generative AI, with media outlets and industry analysts showcasing its potential applications. This phase generates heightened interest in the technology, even though practical use cases remain somewhat limited.
Brandt explained that the healthcare sector is likely to traverse the trough of disillusionment as initial enthusiasm subsides and challenges related to generative AI’s application in revenue cycle management become apparent. However, at this juncture, “we will see what this actually means for revenue cycle management,” Brandt opined.
This echoes the journey revenue cycle leaders undertook with RPA a few years ago. While RPA held the promise of reducing waste and costs, many organizations encountered difficulties in achieving these goals at scale.
“At that time, it didn’t really do much for us, and now we’re at this point where there’s more sane adoption, more thoughtful adoption of the technology, and RPA is actually really good at certain things, but don’t touch it with other aspects of the revenue cycle because it’s not effective there,” Brandt remarked.
“We need to learn from past mistakes and try not to get caught up in the hype and force generative AI where it’s not intended,” he cautioned. “When you’re a hammer, everything looks like a nail.”
IDENTIFYING GENERATIVE AI USE CASES
Generative AI occupies a space between fact and fiction in healthcare, much like in other industries. However, some practical use cases are emerging within revenue cycle management. For instance, many healthcare organizations have utilized ChatGPT to generate appeal letters following claim denials by payers. Another evident area where ChatGPT has proven useful is in handling prior authorizations, as the LLM can sift through relevant medical and administrative information available on the internet.
Nevertheless, these are considered low-hanging fruits, according to Sunil Konda, chief product officer at revenue cycle management solution provider SYNERGEN Health. Generative AI, beyond the capabilities of ChatGPT, can address more substantial challenges, such as rectifying ineffective front-end processes.
“Having accurate patient data at the front end is crucial because inaccurate information at this stage often leads to claim denials or other complications, necessitating significant time and resources for resolution,” emphasized Konda. “Generative AI can contribute in areas like data validation and scrubbing within the front-end process.”
Claim denial rates are on the rise, with approximately 82 percent of these denials being potentially preventable. Most of these denials stem from issues related to registration, eligibility, medical necessity errors, authorization complications, and other front-end challenges. Addressing these issues consumes healthcare providers’ time, money, and resources as they must rework and resubmit claims.
Generative AI has the potential to prevent avoidable errors in data entry by meticulously analyzing thousands to millions of pages of payer contracts, policies, regulations, and other text-based documents to identify missing information or potential inaccuracies. Furthermore, it could be employed to optimize coding processes.
Another significant application in development is communication within the revenue cycle.
“As we move toward chat-based interactions, there will be evident applications and uses in payer and patient communications,” Brandt explained. “We’ve been transitioning to chat-based interactions for some time, and some vendors have already been successful in facilitating revenue cycle-related conversations through conversational user interfaces or chatbots.”
“Generative AI is a new technique for implementing a chatbot, and it might expedite development and broaden its capabilities, but it isn’t fundamentally introducing a new tool,” Brandt continued.
Similarly, generative AI-based chat tools can serve as valuable resources for training revenue cycle staff. These tools can provide faster responses to staff inquiries compared to the time it might take a professional to research and find the answer.
These are just a few of the use cases that revenue cycle management vendors and providers are beginning to explore. Ganapathi envisions a promising future for generative AI in the revenue cycle space, especially in scenarios where healthcare organizations need to transform unstructured data into structured data.
“That’s where these LLMs and generative AI will manifest most powerfully. It may not affect how the EHR communicates with the payer website or insurance company. Its true potential lies in scenarios where human-generated content, like English text, needs to be translated into a structured document,” Ganapathi asserted.
LIMITATIONS OF GENERATIVE AI IN REVENUE CYCLE
The healthcare industry faces the challenge of building use cases to determine where generative AI can revolutionize revenue cycle management and where its capabilities fall short. One gray area pertains to revenue cycle communication.
ChatGPT has demonstrated effectiveness in responding to patient queries, particularly in routine screenings and care. However, conversations regarding patient financial responsibility and financial assistance often necessitate a human touch. These interactions may hinge on understanding nuances in tone, a realm where AI chatbots may struggle. Human comprehension of tone plays a pivotal role in optimizing the patient experience, a dimension that AI chatbots cannot fully replicate, as elucidated by Konda.
Even some claim denials may prove too intricate for generative AI to resolve, despite its potential to enhance denials management. Certain denials may return with a denial code, but this information alone may not suffice to explain why a payer denied the claim. It could be attributed to a credentialing issue, according to Konda.
“Fixing the credentialing issue with the payer can result in the payment of all those claims. Generative AI may not be necessary or effective in addressing significant denials or understanding the underlying reasons for specific denial types. I don’t believe generative AI is currently equipped to grasp that level of intricacy,” Konda emphasized.
Generative AI is also unlikely to be the solution for process or workflow improvements. For instance, revenue cycle leaders might decide to reposition denials management staff to the front end of the revenue cycle to proactively address issues that lead to denials later in the process. This would necessitate the implementation of new systems, processes, and validation methods.
“Some of these areas require process modifications or the introduction of new technology to streamline data collection. In such instances, generative AI is likely to be less effective,” Konda pointed out.
In some cases, change management might prove more beneficial than AI. The healthcare sector grapples with inefficiencies arising from its complex system, compounded by substantial data quality issues. A survey conducted earlier revealed that approximately a third of healthcare providers believed that less than 76 percent of their data was accurate. Moreover, AI in healthcare has raised significant concerns related to algorithmic bias, data privacy, and security.
Generative AI relies heavily on the data it is trained on. If the training data contains biases or inaccuracies, the content generated by the AI may inadvertently reflect these biases. Biased or erroneous information could have severe consequences for both patients and revenue.
Biases or other forms of noise in the data can also increase the likelihood of “hallucinations” in which the generative AI tool generates text or images that are not based on real-world information or fail to faithfully represent the input data. Hallucinations can occur when the AI model generates information that it has not been explicitly trained on or when it misinterprets or extrapolates from the training data in an incorrect or fictional manner.
“Due to the sensitivity involved, it is imperative to establish policies governing the use of generative AI. Additionally, it’s crucial to educate teams on its appropriate usage,” Ganapathi cautioned. “Currently, the primary use case is for generating general information, and it’s important to remain vigilant as these large language models can sometimes produce erroneous or fabricated content. Therefore, it’s vital to ensure that the AI’s output can be easily verified.”
WHEN WILL GENERATIVE AI TRANSFORM REVENUE CYCLE?
The integration of generative AI into healthcare is a journey fraught with challenges, and there will be obstacles to overcome before revenue cycle teams can harness the technology to improve their operations.
“Healthcare is not known for being the fastest-moving industry. So, it’s unlikely that we’ll witness a transformative change overnight,” Ganapathi stated. “There may be grassroots initiatives underway, with doctors experimenting by posing questions to ChatGPT. However, for it to become the de facto method of operation, that will take time.”
The adoption of technology typically entails navigating through policies and regulations, especially concerning the security and privacy of the data fed into AI models. Healthcare, in particular, is among the most heavily regulated sectors when it comes to data security, primarily due to the sensitivity of patient data. Just in the first half of 2023, over 39 million individuals were affected by data breaches specific to healthcare.
“We cannot simply apply large language models without taking these considerations into account,” Ganapathi emphasized.
Technology companies must address data privacy concerns if they intend to deploy their technologies in healthcare. However, this is a customary part of the technology development and marketing process for companies targeting healthcare organizations, and it should not act as a barrier to the adoption of purpose-built, HIPAA-compliant generative AI tools for revenue cycle.
“These requirements are manageable,” Ganapathi explained, “and they can be addressed relatively soon.”
Technology leaders estimate that it will take between two to five years for generative AI to gain traction in revenue cycle management. On the shorter end of this spectrum, healthcare vendors and providers are likely to leverage generative AI to address low-hanging fruit, such as prior authorizations and appeal letters, according to Konda. Approximately three years down the line, generative AI is expected to play a more significant role in aspects of the revenue cycle, such as denials management, as confirmed by Brandt and Ganapathi.
“Inevitably, generative AI will bring about a dramatic transformation in revenue cycle management,” Ganapathi affirmed.
Finally, a new era of technology is on the horizon for revenue cycle management. Physicians are already exploring the capabilities of general generative AI models like ChatGPT to streamline various arduous healthcare administrative tasks. Larger healthcare organizations, such as UNC Health, are actively considering purpose-built generative AI models to target administrative use cases.
While it will take some time before generative AI becomes widely available and practically utilized to streamline revenue cycle processes, vendors and providers appear committed to exploring how generative AI can optimize a convoluted system.
“We all recognize that what we’re doing in revenue cycle management needs to evolve,” Brandt emphasized. “We need to leverage technology more and more as it becomes available and push the boundaries, because we can’t simply hire our way out of this situation. We can’t solely rely on staffing or process improvements to achieve higher levels of profitability or profitability at all. Technology must be an integral part of the solution.”