AI coding at UMass Memorial tempers staffing shortage 

UMass Memorial Medical Center’s radiology department has been able to sidestep a lack of bill coding professionals by adopting AI in the medical center’s coding for radiology.  

“It’s really hard to get certified coders these days,” said Barbara Scully, associate vice president, physician billing with UMass Memorial, which is located in Worcester, Massachusetts.  

“It’s one of those positions that is in high demand,” Scully said. And because remote work has become more acceptable, and coders can work anywhere, hospitals are competing for those positions at more of a national level than they had been. 

Implementing an AI solution has allowed the coding staff to be more efficient, with automation systems assigning billing codes for more| simple procedures and human staffers taking on the complicated ones.  

Scully also said that any concerns the existing staff had about the possibility of being replaced by AI was offset by the fact that demand is so high for coders right now. AI allows human coders to focus on more complex, high-dollar procedures, and the computers to do the coding on automation-friendly procedures and care. 

With the adoption of AI in radiology, UMass’ automation percentage in that department has climbed to 86% from about 70% of its coding before implementation. 

The results have been positive.  

“I never thought we’d get to 86% (AI coded), so I’m thrilled,” Scully said. The system is now in the middle of adding AI to its pathology coding and will continue to add new types of coding to the system. 

“I think that the next thing is going to be endoscopy,” Scully said, adding that UMass is not likely to implement AI coding in more than one department at a time in order to not overburden the staff with excessive changes to the workflow. 

Saying goodbye to overtime 

The system’s approach has resulted in an elimination of both the need for overtime and of a backlog that showed up regularly after the weekend. 

“We have no backlog, which is huge,” Scully said. “The ORs were busy over the weekend, so the operation was behind the eight ball on Monday mornings all the time because we didn’t have a staff that worked on the weekend,” she said. 

Of course, automation-driven coding can work over the weekend. 

The system they adopted creates billing codes based on clinical notes, with some types of procedures doing especially well. 

“Almost 100% of our X-rays are coded by the system, Scully said. “Same with minimal mammograms [and coding rates are] very high for ultrasounds,” she said.   

Can’t do it all 

Where the automated coding is more difficult is in MRI’s and other complicated procedures with multiple variables, which is keeping the automated coding rates for those applications in the 70%-80% range. 

A concern going in, Scully said, was that quality would suffer.  

“So, we did a ton of our own audits, and our compliance department did as well, and it’s really worked out well,” she said.