AI can help hospitals bill complete medical record

Photo: Hinterhaus Productions/Getty Images

Billing inaccuracies are estimated to cost hospitals billions annually, but physician-trained artificial intelligence can help hospitals recover millions in lost revenue by reducing denials, according to Dr. Michael Gao, CEO and cofounder of SmarterDx.

Gao said he is on a mission to transform the way hospitals bill using artificial intelligence. 

The AI works in the revenue cycle to pull together patient information from labs, medications, orders, physician notes and other sources to deduce treatment for the patient. This is used by hospital coders to find gaps in what is billed and avoid errors of omission.

Another approach is to use the physician's behavior to find out what the doctor was treating. If the physician prescribed a high blood pressure medication, coders can infer the patient has high blood pressure whether this was written down or not, Gao said.

"Our ability to use the fingerprint of the actual care for the patient ensures notes are accurate," he said.

Gao did his residency at Weill Cornell Medical College in New York City, which is affiliated with Presbyterian Hospital, where he oversaw applied AI for the health system from 2017 through 2020.

"When I worked at New York Presbyterian, I started thinking about using AI for operational purposes, like an air traffic controller," Gao said. "There's a clinical traffic controller to think about workflow."

In early 2010, some hospitals started using AI to match words in physicians' notes to medical coding, according to Gao. In its early uses, natural language processing was used to pick up phrases. 

"AI, over the past decade, used to be able to do the most basic stuff," Gao said. "As it's advanced, we're able to program computers to do more complicated things. That requires more judgment and consideration."

The AI software platform allows computers to transmit information to hospitals and insurers in a readable format, to "true-up" medical records to accurately represent what happened, he said.

The software is in five hospitals and academic medical centers nationwide. The average hospital has gained $800,000 in new revenue per 100 beds, Gao said.

"In the long run," Gao said, "we can all agree, a fully accurate representation of care provided is great for everybody, including the patient."

Gao is cautious about future AI use. When used in the back office, if AI makes a mistake, it can be corrected without patient harm, he said. On the clinical side, there's more work to be done to integrate AI into the healthcare system safely, he said.

Twitter: @SusanJMorse
Email the writer: SMorse@himss.org