Without exact and reliable revenue cycle management, it will be very arduous to run a successful hospital or healthcare system. Right and reliable medical coding is also a gigantic part of that.
As more and more healthcare organizations see the potential for AI to raise revenue and reduce administrative burdens that trickle down to providers, medical coding professionals are concerned about AI accuracy and potential future job losses.
But the goal shouldn’t be autonomy, but rather augmentation — instead of removing humans from the equation, they can be equipped with modern, generative AI-powered assistive software, said Varun Ganapathi, co-founder and CTO at Akasa, a provider of generative AI-powered software. revenue cycle management technology. Ganapathi has a Ph.D. in computer science, with a focus on artificial intelligence, from Stanford University.
I spoke with Ganapathi to get his insights on how genAI can be leveraged in revenue cycle areas like medical coding, how the technology can be trained to provide rationale for its recommendations, what organizations need to consider when examining generative AI systems and their underlying technology, best practices for implementing genAI specific to coding and revenue cycle processes, and how genAI-based tools can support solve revenue cycle staffing challenges.
Q. How can genAI be used in revenue cycle areas such as medical coding? How would you warn users?
AND. GenAI is the future of revenue cycle. This intricate industry relies on highly manual and time-consuming tasks that include navigating intricate EHRs and payer portals, dealing with endless paperwork, and navigating ever-changing regulations and policies—all while trying to prioritize patient experience.
GenAI from powered by vast language models that deeply understand medical records and clinical data previously obscured by computers. With LLM, all of this is now accessible. LLM training in understanding the healthcare domain and its data opens up many possibilities, including a clearer path to truly solving some of the major pain points in the revenue cycle.
Historically, the coding has been too intricate for most legacy technologies to adequately solve. This is not the case for genAI, which can be trained specifically on healthcare data. This allows genAI to surface meaningful information from patient data and workflows, as opposed to a massive online database.
From there, genAI can work with existing coding professionals to generate suggested quotes and codes. Because genAI runs on LLM, it continues to learn. So if coding rules are updated in a given state or modern patient data is added, genAI can quickly adapt.
Users need to be cautious about autonomous coding. While the idea is invigorating, in practice autonomous coding carries a number of risks, including inexact coding suggestions. That’s why healthcare systems should always exploit a trusted genAI model that is fine-tuned to their data and keeps humans engaged in the process.
Q. You suggest that showing work is critical when using genAI in the revenue cycle. Why? And how can the technology be trained to provide reasoning for its recommendations and insights?
AND. Showing your homework is key in healthcare, especially when working with genAI. Hallucinations or genAI results that aren’t based on exact factual data are a real concern. Imagine a tool suggesting false codes for a patient. What starts as a routine visit can turn into huge medical bills and misdiagnosis.
Historically, much of AI has been a black box where we cannot see how the technology works or understand where the results are coming from. With the right genAI tool, coders can see what coding suggestions are being made and why. Where is the information coming from in the patient record?
By showing its work, genAI allows teams to validate that suggestion before it reaches the payer for exact reimbursement. It learns from specific data, so it learns how to capture the most effective codes for each individual organization and case mix index.
Q. What do organizations need to consider when investigating generative AI systems and their underlying technology?
AND. Organizations need to do some groundwork before implementing genAI. While genAI can learn on the fly and adapt to different workflows, it still requires some support getting up and running.
First, healthcare systems need to have as much information digitized as possible. Again, you want the genAI to be trained in the healthcare system, data, and workflows, and that can only happen if the information is digitized.
Then it is necessary each genAI tool works with the systems provided. Is it compatible with the organization’s EHR? With the vendor portals used? Can it scale across service lines, even intricate ones?
Lastly, but most importantly, organizations need to think about security. What is the genAI vendor’s data retention policy? Do they audit and encrypt all data? The same goes for organizations. Does the organization encrypt data, audit and only store what it needs to?
Q. What are the best practices for implementing genAI specific to coding and revenue cycle processes?
AND. It’s straightforward for organizations to get excited about genAI, and even easier to want to improve everything they can. Instead, look at the low-hanging fruit. What are the problem areas that aren’t overwhelmingly vast and intricate?
More importantly, which areas have the most data to train genAI? These could be great areas to pilot the technology and prove results.
For example, with coding, an organization can get more out of its team by using genAI, a tool that specializes in generating coding quotes and suggestions. This can even support with the need to include physicians in coding suggestions.
Q. How can genAI-based tools support solve revenue cycle staffing challenges such as the medical coder shortage?
AND. There is currently a major shortage of medical coders. Experienced coders are retiring and too few modern people are entering the workforce. Coding teams need to do more with less. But how? Historically, the answer has been technology.
Labor shortages lead to time pressures and a demand for coders to work faster than they otherwise would. This results in a lack of comprehensiveness as documents are skipped or minor details are overlooked. These details can lead to missing codes or incorrect codes, which can ultimately negatively impact quality metrics.
GenAI can support find codes that humans might otherwise miss. Some genAI models can review clinical records faster than human operators—and delve into files—resulting in greater accuracy and revenue at lower costs.
Some models suggest correct codes, leaving it up to the coders to audit or double-check their work. This not only allows senior coders to spend less time on tedious work, but also allows modern employees to work at the speed of an experienced coder.
Imagine genAI giving developers superpowers. Helping them work faster and better. Now imagine the potential across the entire revenue cycle.