Computer-Assisted Coding: Helpful or Hurtful?

Computer-Assisted Coding: Helpful or Hurtful?

Tune in to this monthly online coding column, facilitated by AHIMA’s coding experts, to learn about challenging areas and documentation opportunities for ICD-10-CM/PCS.

By Elena Miller, MPH, RHIA, CCS


When computer-assisted coding (CAC) was first introduced to the industry, it brought with it a feeling of uncertainty among many coding professionals. Would this software be replacing them, akin to the loss of transcription jobs in the wake of speech recognition software implementation? The rumor mill was churning, with some concerned that the computer could do it “better.” The idea was that CAC would increase the speed and accuracy of coders, essentially by taking over the heavy lifting of code assignment, while the coder took on a role where they would simply review and edit the codes on a second pass. With the time per record shortened, less coders would be needed—potentially solving the coder shortage. However, as we all know… good quality coders are as in demand as ever, and the industry still faces a coder shortage. The question remains: has CAC lived up to its promise of changing coding for the better?

Perhaps a quick review on the background of CAC would be helpful here. The Computer-Assisted Coding Toolkit in AHIMA’s HIM Body of Knowledge defines CAC as the process for extracting and translating transcribed free-text data or computer-generated discrete data into [information] for billing and coding purposes. Health Management Technology’s 2010 article titled “Computer-assisted coding: the secret weapon” states simply: “computer-assisted coding…automatically generates medical codes directly from clinical documentation.” That is exactly what the CAC does; generate codes from text contained within certain fields in a document. These codes are to be taken by the coder as suggestions that must be validated before finalizing the account. CAC is great at picking up text and suggesting codes. However, it does not—nor is it designed to—make intelligent decisions about the context of the record.

With that said, that doesn’t mean that the CAC is  ineffective. There are actually pockets of coding in which the accuracy of the CAC-suggested codes is extremely high, and accounts are finalized without any human intervention. But that is not the norm. The industry has definitely come to the realization that coders are still very much needed.

The question is: did the industry come to the realization too late? Let me explain. Initially, the primary focus of many when implementing CAC was to minimize the impact to production post-ICD-10 implementation. New inexperienced coders were brought into the industry trained to code using CAC. Remember—one goal of CAC was to reduce the amount of time that coders spent reading the records. By suggesting codes, CAC ensures the coders don’t miss anything while “skimming” the record. On the flipside, how does the coder validate the code without thoroughly reading the record?

Back in the day, most coders were trained using physical coding books. When the encoder went down, they were expected to pull out the code books and continue working. Then, there was a generation of coders that were only briefly trained on the books in school before diving right into the encoder to complete their coding course. Those coders weren’t as good at book coding during system downtimes. Finding a single code in a code book and coding a complete record using the book are two completely different things. Fast-forward to today, there is a whole generation of coders out there that have either only briefly been trained on coding without CAC or have never worked without CAC. How does this impact the development of the coder? Where the encoder can fully replace the book (i.e., access to the books are in the encoder), the CAC can NOT fully replace human interpretation of documentation. The coder must know how to interpret the documentation and apply coding guidelines.

Codes are highly scrutinized by everyone, from third-party payers to internal quality departments and physicians. Reimbursement and quality measures are dependent on the final diagnosis and procedure codes. Those codes need to be accurate and must be supported by documentation.

The question is not whether or not CAC is hurtful or helpful in general. There is no doubt that it can be very helpful to ensure all conditions are at least considered. However, with the reduced need to read the record, is CAC helpful or hurtful to coder development?


Elena Miller is the director of coding audit and education at a healthcare system.

Leave a comment


  1. Jim Taylor, MD, CRC AHIMA approved ICD-10 trainer

    I gave a presentation at AHIMA several years ago – “Shoot Out at the OK Corral” showing how CAC was the future. However, In CAC, the contender to pretender ratio isn’t very high. Some CAC engines are glorified PDF word searches that the coder then has to validate to findings in context. There are some out there that are excellent. My strong recommendation is to do a “bake off” where you give the same charts to the interested venders and then compare results.
    Things to consider – does the machine learning apply to your data set or is it all customers of that vender
    – time saved per note/chart
    – in the HCC world, the ability to set aside suspects
    Once again, I’m a firm believer that when this works, it is fantastic. When it doesn’t, it’s a very expensive misadventure.

  2. […] Computer-Assisted Coding: Helpful or Hurtful? | Journal of AHIMA Posted By AHIMA Staff on Dec 12, 2018 Tune in to this monthly online coding column, facilitated by AHIMA’s coding experts, to learn about challenging areas and documentation opportunities for ICD-10-CM/PCS. By Elena Miller, MPH, RHIA, CCS […]

  3. In 2016 employers started to demand CAC experience and competence as a condition for a job offer. Better stated, coders with CAC experience were using it to exclude coders without it, especially experienced coders, from work opportunities to advance themselves and to provide jobs to new and inexperienced coders. The same thing happened to me when the CCS credential was created back in about 2006. Suddenly job advertisements in the AHIMA Journal and other publications would require a CCS to be considered. Despite having a RHIA, I completed a CCS as soon as the money became available to avoid exclusion from job consideration. In 2016 I was employed by HCTech Partners as a remote inpatient contract coder for Yale New Haven Medical Centers until that client terminated after a year of employment in early January 2016. We were told that coding there was being offshored for financial reasons. A Catholic hospital system in the Ohio area hired me to begin as a contract coder through HCTech Partners but they were using CAC. Those already employed there seemed very inconvenienced with training me on CAC. To keep the employment opportunity I had to pass the training quickly and immediately meet production standards. This might have created a problem even without the additional problem of a too slow internet speed for that hospital system. Obtaining a fast enough internet speed required changing servers from AT&T to Comcast. By the time this change was made, the work opportunity was lost. So, this graduate with a B.S. in Medical Record Science in 1979 who passed the RRA exam by the end of that year and who had maintained the credential from then on, who started a first ICD-9-CM inpatient coding job with a Professional Standards Review Organization at a Trauma I medical center with a burn unit and an osteopathic hospital, a podiatry specialty hospital, and an affluent general hospital in 1980 and completed four years doing that before moving to a Working Supervisor role that included coding at Kaiser Hospital in San Francisco, CA and who then up until the year 2016 had completed many years of inpatient coding in acute care teaching hospitals to include some of the best in the USA – this credentialed RHIA, CCS was unemployed for a month or so until Berkshire Medical Center, MA made a hire to enable more experience with CAC. Adding to the misery, in January 2016 my son moved with his pregnant wife due in March and his three other children under the age of five for a job in China. Although he was a university graduate and had been accepted in the best Michigan university, he had been unable to find a suitable job in the USA which would silence the health care providers needed by his family. So, at the end of March 2016 I had the mail held for a month at the post office and left the house owned empty to stay in a hotel in Pittsfield and drive a rental car to the hospital, in the midst of an influenza epidemic requiring exclusion of most visitors, to gain experience with CAC and assist with inpatient coding while students from the local community college were trained as I sat in the remote manager’s vacated office while she sunned in a Carolina. When she returned to train on a new system basically my employment as a contractor through HCTech Partners ended barely in time to salvage credit due to a month away from bills. Yale New Haven Medical Centers hired me back in about June and to recall they had CAC but I was not allowed to use it or maybe had the choice to use it or not. After the employment opportunity with them ended due to lack of work in September 2016 I again was unemployed until April 2017 when Kaiser Permanente hired for an inpatient backlog project. They also had CAC but contractors were not allowed to use it. Not using CAC was fine with me. Experience using CAC for a month at BMS was enough to realize it was double work for an experienced coder and of little to no value. Now I have nearly completed another month of unemployment due to lack of hospital funding for contractors with no one else seeming to need an experienced coder after more than ten job submissions. Your article does not emphasize that all coders need to be able to use the books to code. Coding from the books is the ultimate authority. Since 1989 I have been using a 3M encoder. Coding with a 3M encoder is very different from coding from the books. Using a 3M encoder is an entirely different skill set. For example, if a delivery is preterm, you find the code by typing in “early”. Most of this knowledge is somehow automatic after practice and difficult to articulate. When checking for the correct code though the books must be used with the methodology of using the index correctly then reading the tabular with all the information that applies at the beginning of the chapter and three unit code. All auditors and documentation specialists must have mastery of this. To complicate the issue more possibly is the fact that coding books are included in the computer software. Other coders ridiculed me in 2007 at Alameda County Medical Center in Oakland, CA when coding books were hauled to and from work daily in shoulder bags and rolling briefcases instead of relying on the books in the computer. More than ten years later I still do not feel confident using the books in the computer. In addition I am frequently unemployed and need the books because software is not owned. Experience is the best teacher. New coders are the same as I was when beginning. They want to become experts and managers before they are ready and they have no patience. Over the years I have seen nurses take over documentation specialist and other coding roles. Health care practitioners are well suited to learning coding quickly. In about 1999 when I was in a RN program we were taught medical record review and coding. It seems to have a management position using coding more and more over the years you must be a nurse. Before I started to code all the science courses required for nurses and physicians had been completed with them. Please consider if that is the background needed to code instead of CAC. Having spent many of my adult years reading nearly illegible penmanship of physicians to enable coding to meet demanding production standards to the point of nervous breakdowns, what seems most important is changing documentation. Even though what must be read to code now is typewritten, much writing reviewed is not needed to code. A coder’s time and eyesight would be better used if they only needed to read what was necessary to code. Health care providers would be well advised to write only what they must to save their time and effort also. In the electronic records used now there are issues of copying and pasting, for example. When coding the same entries are read multiple times for whatever reasons. The same would be true for a nurse reviewing the record for patient care. Confusion from the duplicate documentation which often repeats past encounters or history which does not apply to the current encounter harms patient care and wastes time. The attitude of CAC trained coders is very hurtful to me. These coders are selfish and do not respect those who were experienced before they started using CAC. Even at the AHIMA Convention in Los Angeles in 2017 the arrogance of the CAC software vendors was intolerable as they presented innovations. They have marketed their product as a replacement to previously experienced coders in so many words. Whereas some may think this is encouraging to a new generation of needed coders, it only seems counterproductive and cruel to me.

Leave a Reply to Jim Taylor, MD, CRC AHIMA approved ICD-10 trainer Cancel reply

Your email address will not be published. Required fields are marked *