Banner Health Gives Glimpse of Struggles, Triumphs in Deploying Revenue Cycle Automation

The executive director of revenue cycle management at Banner Health details lessons learned from deploying revenue cycle automation.

The healthcare industry is constantly evolving in ways that revenue cycle leaders can't control. Changes in billing requirements, payment models, and patient access can cause struggles for organizations with poor processes.

Successful organizations must enhance their revenue cycles and create the bandwidth to address these challenges. This means the use of technology and AI to streamline revenue cycle processes is essential.

Jamie Davis, executive director of revenue cycle management at Banner Health, recently spoke with HealthLeaders about Banner's journey in implementing the use of AI and automating its revenue cycle management. It wasn't an easy process, but it was necessary to protect the organization against revenue leakage.

Davis has over 20 years in healthcare and is the liaison for Banner Health's revenue cycle performance and strategic alignment with its regional c-suite. She oversees the revenue cycle portfolio, including vendor management, revenue cycle roadmap, continuous improvement, and strategy.

Banner Health currently has 22 day-to-day bots helping with its revenue cycle management. These bots complete tasks like adding insurance information and updating medical records. "All the things that our human resources shouldn't waste their time doing," Davis said.

These bots manage roughly 90 million records for Banner Health, and from mid-2020 to 2021, Banner has saved about 1.73 million man-hours by deploying them, Davis explained. Banner Health also has machine learning in its refund and variance space to help with credit and debt balances.

When it comes to taking on such a large automation endeavor, partnerships are needed Davis says.

"Our automation is done completely in partnership with our IT group. They have their own robotic process automation center of excellence, but they've also started dabbling in some process mining in our health plan data. And we are working towards automating things like low balance accounts receivable management and denials," Davis says. 

"The automation of the denials, low balance accounts receivable management, and the variances is really fun and innovative. That's where it's really rolling into that intelligent automation space since it's using machine learning that is predicting and reacting," Davis says.

Don’t go big at first

While Banner Health is fully rolling with a plethora of streamlined and strategic automation throughout its revenue cycle, it wasn't always smooth sailing. This, Davis says, is where you need to make sure to take your time and don't expect to go big at first.

"In full transparency, we tried to run first, and then we fell. We realized we needed to slow down a little bit, which was a great lesson learned," Davis said. "I think anyone who is trying to be innovative has those horror stories where something worked out really well in the boardroom and not so much in real life."

At the time Davis started developing its automation, Banner Health already had its IT team working with its medical records team to manage records and move them with bots. From there, it made sense to the team to expand its automation. 

"The theory is, we want our technology to pay for itself. We thought a great way to do that is to add machine learning to implant charge capture and it will increase that revenue. And we were very strategic about it—so we thought," Davis said. "We deployed process engineers and documented the processes. We found all the variants and began to write Python code and machine learning. And it worked until it didn't work. What happened was that the processes that we had documented varied from facility to facility."

Banner Health has 30 hospitals, and the appropriate workflows weren't aligning. 

"So, we automated a dysfunctional workflow, and it ended up being more cumbersome to utilize the machine learning. It was a good learning experience—we did the fail-fast theory."

Moving forward

After stepping back, realigning its strategic planning, and partnering with IT, Banner's deployment process started to turn around.

"We ended up creating hierarchal scoring for all of the automation that we wanted to consider. On one side, we have the benefits: for example, net revenues, compliance, or full-time employee re-allocation. We would then weigh those scores and compare them to the complexity of the build: for example, how many process variants does it have? How many systems are in there?" Davis shared.

After gathering those scores, Banner would use a classic grid to determine automation that was low-effort, low-return, and high-effort, high-return. This hierarchical approach made all the difference for them.

"Once we did that, we applied a continuous improvement team member to have oversight and to help be that subject matter expert in the revenue cycle to make sure we aren't recreating core processes. And from there, our automation just went gangbusters," Davis said.

Banner had about five bots at the time, and in almost no time they had 15 bots. Now they are at 22, Davis says.

"And we have governance over all of it as well. So, just because we can automate something, doesn't mean we should."

Managing all of this automation has become easier too, Davis said. The team now brings all of the automation to the table, just like they would do for any other project, and explain it to governance oversight so they can agree that it's strategically aligned.

It's also important to make sure the automation doesn't tax resources that are taxed elsewhere and to make sure there is monitoring in place. 

"We have dashboarding that monitors what those bots are doing, when they stopped doing it, and when they start doing it the wrong way. Because, you know, just like humans, a bot needs to learn, be educated, and be monitored," Davis said.

Davis credits its first failed attempt at deploying automation to the success it has today.

"I really think this playbook that we created has been really successful because now we have our feet under us from a bot perspective, and we are running really quickly."