Cleveland Clinic to implement generative AI for medical coding

CLEVELAND, Ohio — The Cleveland Clinic is turning to generative AI to review medical coding in a fraction of the time that it takes humans take to do the same task, the health system announced Tuesday.

An AI coding assistant, developed by the San Francisco-based company Akasa, will roll out across the Clinic’s U.S. locations over the next several weeks, the Clinic said.

The rollout will start at Clinic Fairview hospital, then go to the Weston, Florida location, main campus, Clinic Hillcrest and the rest of the health system’s U.S. locations.

The program will start at two or three Ohio-based Clinic hospitals within the next three or four months, said Clinic chief digital officer Rohit Chandra.

The Clinic will train Akasa’s generative AI-powered tool to review medical codes for patient care that are used to bill health insurance companies, work that is currently done by specialized staff. Generative AI is a type of artificial intelligence that uses models to produce text, images or other forms of data.

“Our main objective is to substantially streamline and simplify the work,” Chandra said. “If we train the machines well, they will be much more consistent than humans. Machines can have errors too, but if you train them well over time, they don’t make the same error over and over.”

Chandra said he didn’t know whether use of the generative AI tool will result in the loss of jobs for Clinic coding staff. The hospital system is exploring the use of AI for medical coding to make up for a shortage of medical coding specialists, he said.

This is the Clinic’s second foray into the use of AI. In February, at some U.S. outpatient locations, the Clinic began using AI technology to record conversations between caregivers and patients, and to generate written medical notes.

The goal is to increase personal interaction between patients and care providers, and to decrease paperwork, the Clinic said.

Specialists in medical coding are experts in the various numerical codes used to represent various elements of care given to patients. These specialists typically review more than 100 clinical documents per case, with information coming from a doctor’s office visit, emergency department visit, surgery, specialist consultation, inpatient hospital stay or other types of care.

Medical coding experts must select codes from more than 140,000 options to represent that care. These codes then are submitted to the patient’s insurance company to trigger reimbursement.

The complex process can take up to an hour for each instance in which a patient receives care in some form, the Clinic said.

But the Clinic’s new AI coding assistant can read a clinical document in less than two seconds and process more than 100 documents in under two minutes.

The generative AI coding assistant learns from real-world documentation practices and recognizes the nuances at individual health systems, the Clinic said.

“You can train a system on millions of records that we have, so that it understands the nuances of how we code,” Chandra said.

Human staff will generate the medical codes first, then the AI assistant will review the codes for accuracy and to ensure that the most appropriate codes have been used, Chandra said.

If the AI coding system flags questions, human medical coders will send the questions to the proper office to be answered. The Clinic may find that this allows insurance claims to be processed faster, Chandra said.

“That would obviously be nice for all parties concerned,” he said.

The Clinic declined to say how much it spent for the Akasa AI program.

The Clinic and Akasa also are testing a second AI tool that can perform the first round of coding itself, with the humans checking the program’s work afterwards, he said. The second tool should be rolled out later this year.

“We hope to get to a place where the AI can automate the bulk of it,” Chandra said. “We still have humans in the loop, but their function moves to a review function, as opposed to generating the codes.”

Julie Washington covers healthcare for cleveland.com. Read previous stories at this link.