Artificial Intelligence

Autonomous coding, it won’t hurt a bit with CodaMetrix

September 30, 2021 8:54 pm

Medical coding processes for physician reimbursement are often a drain on resources, consuming a frustrating amount of time and money. CodaMetrix President and CEO Hamid Tabatabaie said his company’s technology helps alleviate those frustrations.

CodaMetrix — a spinoff of Massachusetts General Physician Organization, which makes them a subsidiary of the physician group with more than 3,500 physicians — was developed to make coding more efficient, reduce administrative burden on physicians, increase quality of results and significantly lower cost.

“They needed better tools than what commercially available solutions offered in order to abstract [information] from the medical records, and code claims to satisfy the reimbursement requirements,” Tabatabaie said.

Coding complexities

More than 75,000 codes describe diagnoses and about 8,000 codes describe procedures. When each encounter with a patient can have multiple procedures and diagnoses, code combination possibilities move into the billions. To add to the complexity of ensuring providers receive the correct and timely payment for services rendered, reimbursement requirements change from one payer to another, and there’s a fair amount of other red tape involved in submitting claims. Payers have different rules for submitting claims for the exact same type of patient encounters. One simple example is a scenario where a patient has to have multiple tests in one day. For the exact same scenario, one payer asks for the results to be “rolled” into one claim, while a different payer asks for them to be submitted as individual claims. Major providers have to deal with these types of exceptions hundreds of times each day.

“Thankfully, most doctors are specialized in a certain area, so what they do [and the corresponding coding required] is somewhat repetitive,” Tabatabaie said. “But, even the most repetitive function can have thousands of code combinations to choose from.”

“One of the sayings by one of our cofounders is, ‘Doctors didn’t go to medical school to learn how to code.’ But the way the electronic medical record works is that either the doctors spend a lot of time choosing these codes, which can literally add hours to their day,” and/or they hire certified coding staff, who are increasingly in short supply, and highly paid, which can get expensive, given the median salary for coders is $55,305.a

“The industry spends an enormous amount of money on coding — about $7 billion a year in coding costs, and that doesn’t account for the time doctors spend to code or facilitate coding,” Tabatabaie said.

In the past, handwritten notes made the process even more time-consuming.

Benefits of AI and automation

“But now that patient records are electronic and now that machine learning and AI are becoming part of our everyday language — you often hear something else has been automated with machine learning,” Tabatabaie said. “Our job is to use that technology to automate as much of the coding process as possible so, first and foremost, doctors have to do less of it. And secondly, the cost is brought down by increasing efficiency, so providers’ scarce coding resources can focus on the more complicated cases that machines can’t solve.

“There will always be things that machines can’t do. As time goes by, maybe the [issues AI cannot handle are] reduced. But there is plenty of work to do for coders to be able to address some of the more unusual and complicated cases.”

CodaMetrix uses its artificial intelligence technology to free up the time physicians and/or coding staff spend on coding and to reduce the costs of the administrative work handled by physicians and coding professionals.

Ubiquitous adoption of electronic health records (EHR) and electronification of health records have brought about a long list of positive results, but an unintended consequence has been that physicians spend a lot more time on documentation. Physician job satisfaction surveys reveal frustrations with the administrative work, including but not limited to coding, which limits the number of patients they can see, the time they can spend with patients, their time away from work and how much time they have with their families. Tabatabaie said this has led some doctors to retire early, causing larger healthcare organizations to try to recruit replacements.

“The exact thing we do, we automate the coding process,” he said. “But what we really do is solve major headaches.

“At Mass General, it’s gotten to a point that a great majority of surgeons don’t have to do coding anymore, and they literally stop the staff that they know are responsible for these solutions to thank them profusely,” said Tabatabaie.

“They repeat these stories at conventions and board meetings. They’re very, very supportive of what we do.”

Their partners and adopters are able to automate about 50% to 60% of their case volumes, saving 28% to 39% of coding costs, without needing new software or training.

“Our customers do exactly what they do every day before meeting us — use their existing tools for capturing patient information and recording interactions and diagnoses — usually their electronic health record,” he said. “CodaMetrix simply acts as a listening device and connects to their clients’ existing system(s). Our solution electronically looks at the records that need to be processed in a claim and we translate those to codes.”

When the system is sufficiently certain of the appropriateness of the translation — what in AI and machine learning is known as a confidence score — with high levels of confidence in the codes that are predicted, the system automates the process so human beings aren’t required to review the codes.

The information is then passed on to other existing revenue cycle systems that complete the billing process by putting the codes on the right forms, sending those forms to the right payer and collecting the reimbursement fees.

“You can think of us as a listening device that looks for the evidence to support the right codes, then skips the manual steps and hands over the right codes to the downstream steps in the process,” he said.

“It is a really simple system to adopt. Clients don’t have to install any software. They don’t have to train anybody. We are just a filter. The best tools are those that are non-invasive.”

As these non-invasive tools work more and see more, they’re able to do more.

“In machine learning and AI, the whole is always greater than the sum of its parts, meaning the more people use the system and the [more] data that goes through the system, the smarter the system gets,” Tabatabaie said. “It manages to be more accurate; it manages to automate more of a percentage of the total, and it manages to be generally more reliable and more useful. The way we use our technology is to really take advantage of that natural phenomenon.”

Looking ahead

CodaMetrix plans to invent new technologies that improve the overly complex models used now for reimbursement, finding ways for AI and machine learning to automate pre-authorization processes and serve as tools for population health and utilization management.

Tabatabaie pointed to the use of smartphones in auto insurance as a parallel to CodaMetrix’s long-term goals.

“Today, AI approves legitimate claims that are filed online using uploaded pictures of a damaged car, creating savings and convenience for consumers while insurers get better insights and improve their ability to manage risk better.”

“Our North Star is to invent technologies and deliver solutions that help payers and providers assess and auto-adjudicate reimbursement at the point of care,” he said. 

About CodaMetrix

CodaMetrix (CMX) enables physicians to rediscover the joy of practicing medicine by reducing administrative burden caused by the complexity of payer rules for coding, authorizations for service and other unique payer-by-payer requirements. Originally developed by the Massachusetts General Physicians Organization (MGPO), CodaMetrix uses Artificial Intelligence (AI) and Machine Learning (ML) algorithms to autonomously code a large portion of providers’ claims, which either eliminates the need for manual coding or significantly improves the process.

This published piece is provided solely for informational purposes. HFMA does not endorse the published material or warrant or guarantee its accuracy. The statements and opinions by participants are those of the participants and not those of HFMA. References to commercial manufacturers, vendors, products, or services that may appear do not constitute endorsements by HFMA.


a. “Certified professional coder salary in the United States,”, accessed Aug. 23, 2021.


googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text1' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text2' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text3' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text4' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text5' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text6' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-text7' ); } );
googletag.cmd.push( function () { googletag.display( 'hfma-gpt-leaderboard' ); } );