Author: Grant Messick
The earliest Western catalog of the stars was created by the Greek astronomer Hipparchus, building on earlier work going back to the Babylonians. Automation for revenue cycle operations is still a relatively early and evolving field and solution developers face particular challenges when attempting to map the universe of denials in order to apply automation. To many, automating denials may feel more like an aspirational concept than a reality. It may feel like an effort similar to mapping the stars – one that takes generations and some of mankind’s greatest thinkers. However, if accomplished, the benefits of an automated denials solution could be substantial: timely, accurate and effective appeals; maximized overturn rate; lower cost to collect; and freedom for denials staff to work at the top of their license, focusing on the hardest and most complex denials.
Denials is a particularly challenging area of revenue cycle operations to automate because it involves processes with seemingly infinite scenarios. Payers change their rules and their use of codes on a regular basis. Staff investigate each denial and then take one of several different approaches to fight, fix, and/or overturn the denial. Many denials require rigorous appeal processes and coordination across multiple departments. Some may question if automation in this space is even possible. At AKASA, we’ve developed proprietary technology and approaches called Unified Automation™ that apply machine learning to address the dynamic nature of denials. We often get asked, “how is machine learning different from traditional automation solutions such as robotic process automation (RPA),” In the case of denials, the difference is a game-changer.
Trying to build a rules-based engine to manage denials with RPA is like trying to map the stars. Between the complexity, the quantity, and the vast number of combinations, trying to tackle such a task in such a manual way may be impossible. This level of complexity is a perfect opportunity to apply a machine learning process that can continuously watch and learn from human work to ‘train’ an algorithm to automate the responses to specific denial codes. In the short term, the algorithm can select certain codes and build a high confidence level around the next steps required for a certain scenario. Over time, the algorithm will continuously evolve and add new codes and new scenarios to its repertoire – taking on more tasks and more complex functions as it learns. These machine-learning approaches have the ability to significantly accelerate our ability to automate denials. Just imagine what more Hipparchus may have achieved with the benefit of a modern telescope.
For example, a finite set of about 12 denial codes could be classified as eligibility denials, typically stemming from one of approximately 10 different eligibility issues upstream, and typically requiring one of approximately 6 different responses. The manual process of mapping the rules in order to develop RPA in this scenario likely isn’t worth the investment of time and manpower (12 x 10 x 6 = 720 rules). However, this is exactly the kind of problem best solved by machine learning. Certainly, some clinical denials will likely always require involvement from a medical professional who can justify the care that was delivered, but others may be addressed with machine-learning based solutions like Unified Automation™.
This may be the breath of fresh air that we’ve been looking for to effectively automate denials. Recent digital transformation efforts and other investment in denials management have yielded mixed and often underwhelming results. While denials still consistently rank among finance executives’ top priorities, organizations across the industry have struggled to achieve meaningful improvement. To make a dent in denials, organizations must deploy a substantial combination of expertise, analytics, stakeholder coordination, technology, meetings, and leadership — all of which takes additional time and investment. Many health systems are standing up dedicated denials teams and committees, as well. In spite of all of that, industry best practice overturn rates remain relatively stagnant around 60%, meaning we are still losing our appeals 40% of the time, and in some cases, denials quantities are so high that they cannot be worked before the denials age out and have to be written off as a complete loss.
Denials have, and will continue to be, a pervasive hurdle that revenue cycle departments must address. Fortunately, machine learning approaches are enabling the development of a new generation of automation solutions that can do the heavy lifting for us, minimizing the human touches required, maximizing overturn rates, and allowing denials professionals to get laser-focused on the small subset of work that truly requires human intervention. New approaches, such as Unified Automation™, allow us to efficiently connect “the stars” in a multitude of constellations — effectively mapping the universe of denials for automation that truly delivers on the aspirations of digital transformation.