Medical coding is the review of medical documentation for billing and database purposes. Coders, as professionals in the field are called, do so by assigning standard codes to items like the names of diseases, medical conditions, and even instruments used. 

Today, many coders are concerned if their positions will be replaced by computers. Let’s look to answer that concern, first by reviewing the profession’s history, then discussing what’s really going on between coders and computers.

Medical Coding A Brief History

Medical coding has been around centuries. The official coding of diseases began around the 17th century in England. Known as the London Bills of Mortality, certain data of diseases and ailments were collected and arranged into numerical codes. These were used to measure the most frequent causes of death. The problem was these bills did not match with their terminology. 

Florence Nightingale, who pioneered modern nursing, made a proposal that resulted in the development of the first model of systematic collection of hospital data used in modern medicine. Epidemiologist Dr. William Farr proposed in the late 1830s a uniform classification system of diseases and ailments. In 1893, French physician Jacques Bertillon introduced the Bertillon Classification of Causes of Death at the International Statistical Institute in Chicago. By the 1930s, these as well as other various ideas had evolved into the International List of Causes of Death. 

The list was adopted and used by the World Health Organization. It was expanded further and renamed as the International Classification of Diseases (ICD) in 1949. ICD-6 was the very first version of this new list.

ICD has undergone numerous updates as the knowledge of medicine expands. ICD-11 officially came into effect at the beginning of this year. Its most notable change is the introduction of entities, also called classes or nodes. While these normally represent a disease or a pathogen, they can also be an isolated symptom or (developmental) anomaly of the body. Other entities (and their codes) include reasons for contact with health services, social circumstances of the patient, and external causes of injury or death. 

Medical coders in the US currently use ICD-10. Rollover to 11 is expected to start in 2025 though it could be delayed until 2027.

Besides ICD, medical coders also rely on Current Procedure Terminology (CPT) to code outpatient services and procedures. 

The Healthcare Common Procedure Coding System (HCPCS) is also a mainstay for coders. It’s used to cover products, devices, and services not included in CPT. 

Other codes can come from the state and insurance companies.

Computers Will Not Be Replacing Medical Coders 

The title of this section pretty much answers the question. Much of the fear and concern stems from what happened to the medical transcription profession as covered in How Technology is Affecting and Improving Medical Transcription and What is the Future of Medical Transcription?

The popularization of electronic medical records (EMR) through the Health Information Technology for Economic and Clinical Health Act of 2009 (HITECH) increased those concerns. 

And yet the profession continues to grow. According to the US Bureau of Labor Statistics, demand for medical records and health information technicians like coders is expected to grow eight percent between 2019 and 2029. This is a higher rate of growth than the average across all fields. Much of this is attributed to the aging US population.

The Limitations of A.I. in Medical Coding

Computers are already used by medical coders. Patient records, which coders are given complete access to, are through EMR.

Computer Assisted Coding (CAC), which rose out of HITECH, is the other main computer program used by coders. It automates many coder functions like reviewing documents for items to code and making recommendations. It may also review coders’ work to find simple mistakes. 

Most coders are fine with CAC. It’s a useful tool. When coders speak of computers taking over their profession, they’re talking about artificial intelligence (AI). To many, this AI can: 

  • Take in any medical information whether it’s a call to the front desk, a provider’s SOAP note, or a patient record found on a medical computer
  • The ability to comprehend unstructured background material which leads it to figure out which code or codes are applicable. This includes figuring out incorrect or missing material, and requesting for it. 
  • Present the information in formats specified by requestor whether it’s insurance record or official government document. 
  • Accuracy high enough to need little to no review by human eyes. 
  • Able to present its reasons that can be understood by a human being. 
  • Be less expensive than an experienced medical coder.

Comprehension, Analysis, Interpretation 

There are currently no AIs capable of performing the above process. At the time of this writing, many medical coders’ duties are virtually impossible, too expensive, or both, to create algorithms for AIs. 

Much of a medical coder’s job is reviewing medical documentation and assigning codes to specific sections. This is easy when the proper keywords, descriptions, and terminology are present in the documentation. AI can be designed to handle this. 

Difficulty arises when different keywords are used. Or there are no keywords used but there’s a description of the issue. Or the information that may direct to the possible keywords are scattered across different sources.

The coder must now interpret what’s on the documentation to find the right codes. This is called a “gray area.” An operative report, which details a surgery, is one such example. The coder in the video had to figure out most of the codes based on incomplete and missing keywords as well as descriptions that didn’t match the official handbooks. She was able to do so because she could comprehend, analyze, and interpret the various parts of the report then put them together to find the right codes. 

Denial management is another example of “gray area.” An AI must sift through insurance information, codes, payer-specific guidelines, as well as any related minutiae that may have affected the denial. 

None of today’s AIs can deal with “gray areas” in coding like a human coder. 

Keeping Up with the Codes

Medical coders unsurprisingly must keep track of the codes that apply to their work. This not only includes those sourced from the ICD, CPT, and HCPCS mentioned above, but codes specific to the healthcare groups’ region (example: state) and the facility itself. 

An example of the last item are evaluation and management codes (E/M). These are developed where certain criteria are not developed by organizations like Medicare (example: what constitutes a “service” during a patient visit). They’re left instead to the healthcare organization to develop. An AI who replaces a human coder must keep a lookout for them as well as any changes in codes, which can arrive in an email, fax, or even a telephone call.   

Silo Separation

One of the advantages touted by EMRs is that they’ll allow patients’ records to be entered once into the system like through a medical tablet, then travel with them as they move about to different medical groups. 

This has not proven true. While HITECH and HIPAA encouraged the formation of EMR systems, they did not state they had to be compatible. This led to the creation of “silos” of patient information, each one distinct for each medical group and even EMR manufacturer. All are kept separate from one another. Patients switching to a new group with a different system may have to physically carry copies of their older records. These can range from pdfs to even paper copies (discussed below).

An AI replacing a human coder must contend with these silos. It must figure out differences between the two or more systems, update them to the one used by the current healthcare group, and merge everything so all the codes are correct. This is a daunting task even for people.

Coding Written Records

The silos discussed above can be connected eventually. But what happens when the patient’s records are all on paper? 

An AI medical coder unsurprisingly would be effective in a pure electronic media like EMR. But it needs to work with other information sources like facsimiles. Fax machines, as discussed in How Fax Machines Hinder EMR (and How to Solve it), are still very much in use in many medical groups and will be for years to come. The AI coder must be able to go through and interpret those scans with minimum to little human input to replace humans. These also include handwritten notes from providers and other medical support staff. 

Medical Coding and AI – Working in Harmony

Medical transcriptionist jobs diminished as computers were able to transcribe provider’s words into text. AI algorithms used in them are sophisticated enough to tell the proper usage of there, their, they’re, and dare in a sentence. Providers still have to edit their documents  to make sure there are no major errors. 

Medical coding is considerably more complex as pointed out above. Tools like CAC show most likely AI’s future in medical coding:

  • Handle simple and repetitious coding tasks. This frees human coders to deal with more complex cases like denial management.
  • Double check less experienced medical coders’ work so they don’t make obvious mistakes.

Human coders can work to improve such tools by:

  • Auditing any and all AI coding efforts on patient records. The algorithm learns from these efforts and hopefully doesn’t repeat them in the future. 
  • Testing new algorithms programmed into the AI. 

Closing Comments

Medical coding assigns various medical conditions and even instruments with codes for numerous purposes like billing. Coders, as professionals in the industry are called, are concerned that AI may eventually replace them. This is unlikely due to limitations ranging from incomplete documentation to facsimiles. 

If your medical company is looking to see how medical coders can use AI effectively in coding, contact a representative from Cybernet. Also follow Cybernet on Facebook, Twitter, and Linkedin to stay up to date on this and other relevant topics.