Three questions with Jason Burke: The next chapter of revenue cycle automation

I sat down with the 3M Health Information Systems VP of Revenue Cycle Solutions Jason Burke to talk about the upcoming AHIMA Conference and the future of automation in the revenue cycle. 

Organizations are facing many challenges when trying to maintain a sustainable business model. What are some of the biggest concerns facing health information management (HIM) professionals today and what solutions can help them tackle these concerns?

The biggest problem that I hear a lot is lack of resources, specifically a lack of coders. There aren’t many extra coders out in the market ready to hire and this creates the need for many organizations to outsource coding and clinical documentation integrity (CDI) work. Lack of resources also means tight budgets which forces facilities to turn to outsourcing outside the U.S. to reduce costs.

One solution for organizations facing these pressures is to implement automation tools to help reduce coder burnout and increase efficiency. Generative artificial intelligence (AI) and deep learning models are starting to contribute to documentation improvement and can help coders focus on more complex coding tasks. There's also a huge potential for these tools to reduce the administrative responsibility currently challenging clinicians.

The need to show massive ROI to justify spend on these solutions can be a hurdle and it creates a lot of pressure to choose the right solutions. CFOs and revenue cycle department leaders are looking for a way to do more with the coders they do have, and the right tools can bridge that gap.

Speaking of AI, there is a lot of promise and hype around AI technology in the news, but automation isn’t necessarily a “set it and forget it” solution. Take medical coding: Why is automating coding using AI so complex? 

Coding sounds pretty straightforward because there are rules coders follow to get to the correct code. However, when you get past simple records like radiology and you get into complex outpatient cases and then very complex inpatient cases, you can have documents that are one thousand plus pages. These complicated patient records can benefit from AI technology: The models can pull out relevant data based on other similar records it has been trained on, but only if the model trained correctly.  

AI models require a massive amount of data to correlate the record to the codes that were actually coded. Those records have historically been coded by human coders which introduces some variation into the data because not every coder codes the exact same way. So, when trying to create a “golden corpus” data set you may expect perfection, but there will always be inaccuracies due to this coder variation. AI models can’t as easily understand the variation, which can produce errors down the line.

You have to address this variability with rules that can adjust and adapt, which can make AI models challenging to maintain at scale. Many companies that are trying to do AI are able to do some simple things like radiology or same day surgeries, but they're all finding that as you get more complex cases using the model is unsustainable. 

Looking ahead to the AHIMA Conference in October, we are exploring the next chapter of revenue cycle automation at our booth. What can health care organizations do today to better prepare for further automation in revenue cycle workflows?

Automated coding is going to happen faster if documentation is structured and standardized. The more structured the information, the easier it is for the model to consume it and to correlate data across the record. That can be challenging, though, as each physician has their own pattern the model has to adapt to, but as electronic health records (EHRs) evolve and as solutions improve, the AI is getting better.

When considering coding automation, it is important for hospitals to pick a partner they are comfortable with. Facilities must set up a lot of interfaces and pass a lot of data back and forth, so it’s better not to do that with multiple companies. Additionally, many organizations that have tried to start with an all-in, let’s automate everything approach to AI have had to walk back and rethink the feasibility. It is more sustainable to start small, with radiology for example, and try to get 10 to 20 percent autonomous and build up from there. 

I encourage all AHIMA attendees to stop by our booth to chat with our experts to learn more about the next chapter of automation.

Jason Burke is the vice president of revenue cycle solutions at 3M Health Information Systems.  

Deanna Berkowitz, marketing communications specialist at 3M Health Information Systems.