Two Types of Computer-assisted Coding

Vol. 18 •Issue 1 • Page 19
Two Types of Computer-assisted Coding

Structured Input vs. Natural Language Processing: A look at how these two types of CAC software can work for you.

In a packed room in center city Philadelphia, HIM professionals scrawl notes and gaze at brightly colored PowerPoint slides. The unlucky ones find themselves standing in the back, leaning to jot down interesting tidbits and using knees as writing surfaces while teetering on one leg or settling back on haunches.

At the center of all this attention lies computer-assisted coding (CAC), a technology that’s been around for a while and put into practice by eager early adopters and test sites. Now, more HIM professionals are taking notice, as evidenced by the large crowd that flocked to the American Health Information Management Association’s (AHIMA) CAC workshop at the national conference last fall. HIM directors are keen on learning if the technology actually works and seeing if CAC can help ease the shortage of experienced coders. Coders at the education session listened intently to find out how, if implemented in their facilities, CAC will affect their jobs.

Two main types of CAC software exist: structured input and natural language processing (NLP). Here, with the help of some experts in the CAC field, ADVANCE aims to take a look at the two types to see how they work and how each could transform the way coding is done at your facility.

The Structured Approach

In structured input, physicians take center stage. Often used at the point of care, structured input systems present the physician with a data entry screen featuring point-and-click fields, pull-down menus, structured templates or macros. The system prompts the physician for more information if necessary, so physicians can’t simply opt out of providing detailed information. The words and phrases selected by the physician are linked to codes, which are automatically generated when the physician is through filling out the required fields. The codes are presented and then validated by a coder.

Physicians must change their behaviors for structured input CAC to work properly, and this may be looked upon as a disadvantage, as some physicians may be resistant to change. One big benefit in the change in behavior, however, is that transcription costs are eliminated for procedures where the CAC system is used.

Not all physicians would view the change of behavior as a bad thing, either. If a physician does a repetitive procedure, such as a routine colonoscopy, many times throughout the day, the physician may find it easier to click away rather than to sit down and dictate what they’ve done, over and over.

Sean Benson, co-founder, vice president, consulting, ProVation Medical Inc., Minneapolis, admitted, as a vendor of a structured-input system, that it’s not easy to get physicians to change their documentation behavior. In facilities where ProVation is used, however, it has a 90 percent usage rate. Benson also pointed out that having the physicians do the documentation results in a completed document with the appropriate level of detail. In manual coding, if the physician doesn’t enter the right amount of detail, coders may down code or may have to seek out physicians and hound them for more detail. With a structured input system, physicians are prompted to enter the right amount of detail to garner the most accurate codes. “Like they say, if it’s not documented, it didn’t happen!” Benson said.

Structured input typically fits best in procedurally driven specialties. ProVation’s MD software, for example, is typically used in gastroenterology, pulmonology, cardiology, orthopedics, pain management and urology.

A Peek at NLP

NLP, unlike structured input CAC, doesn’t require physicians to change behavior at all. In fact, physicians don’t really even have to know the CAC system exists. NLP imitates the way that people read and comprehend by using artificial intelligence. The technology scans the text documents of the record and singles out the important terms, converting them into codes. Usually, a transcribed report is the free text used when extrapolating the codes. A doctor dictates as usual, and the report is sent to the NLP engine, which then pulls out the relevant terms and comes up with a list of suggested codes. A coder validates that the codes are correct and makes necessary changes, and the codes are then sent to billing. Workflow changes very little.

NLP tools rely on the documentation in the record, which means that if a facility isn’t fully electronic, other documents that may be pertinent to coding may not be able to be used by the NLP engine. NLP vendor Artificial Medical Intelligence realized this problem and the fact that many facilities are operating in a hybrid environment. The vendor’s EMScribe™ DX system can pull in for viewing purposes only any scanned-in, handwritten progress notes or picture files, allowing the human coder to view the documentation and add codes based on that documentation, if necessary. “Many departments are not fully electronic, so if they’re capable of scanning in progress notes, we can put it in a format where it becomes part of the patient’s chart so the coder at least doesn’t have to do their proverbial paper chase and run around and get documents and play with sticky notes and write up their own handwritten templates,” said Stuart Covit, executive vice president, Artificial Medical Intelligence, Eatontown, NJ.

NLP is confined to electronic text documents, and many vendors are specialty-specific and offer niche systems. A-Life Medical, a San Diego-based NLP vendor, for example, began more than 10 years ago in the outpatient setting with the emergency medicine specialty and has since branched out into radiology and pathology.

Artificial Medical Intelligence is not specialty specific, and as long as the document is electronic and can be converted into some form of text, it can be coded. Outpatient as well as inpatient charts can be coded, and Artificial Medical Intelligence is ready to go live with an automated present on admission (POA) module for inpatient coding.

Impact on Coders

Whether in an inpatient or an outpatient setting, CAC will change how coders work on a daily basis. Coders will be responsible for validating the codes no matter what system, NLP or structured input, is used. The repetitive procedures, such as routine colonoscopies, will be relatively quick for coders to approve, while more complex procedures might require more time.

The CAC systems will also free up coders to work on more complex cases that absolutely require manual coding. Mark Morsch, vice president of NLP and new technology with A-Life Medical, sees the CAC as advancing the field for coders. “In many ways, coders can reduce some of their mundane tasks related to data entry or working on paper and focus on some of the more challenging aspects of the job, which include following up with denials, working with auditing, working on more complex procedures,” Morsch said.

What the Future Holds

More and more coders may see their roles changing in the near future. The vendors ADVANCE spoke with said that interest in CAC is on the rise. Artificial Medical Intelligence anticipates that the number of facilities using its technology will quadruple in 2008. A-Life has seen an increase in the emergency medicine CAC as well as a continued growth in the radiology specialty. A-Life has also introduced CAC for interventional radiology and certain vascular surgical procedures, and plans to extend into other specialty areas and inpatient coding in the future. ProVation just released a new documentation product for cardiology. “Our goal is to automate as much as we can, keeping in mind that there are some functions that really shouldn’t be automated and need that human interaction. It’s a real balance, but one that we never lose sight of,” Benson added.

There are hopes that the technology will catch on and gain widespread use. Coders shouldn’t fear that it will take jobs, but should instead look to CAC as a way of helping them do their jobs faster and more efficiently. HIM directors face challenges like the work force shortage and increasingly complex rules and regulations. All are good reasons to look to CAC as a solution, but the verdict is still out on whether or not the industry will embrace the technology across the board. “It’s still early yet — but it’s certainly evolving,” Covit said.


1.”Delving into Computer-assisted Coding.” AHIMA Practice Brief. HIM Body of Knowledge, FORE Library.

2.”Testimony of the American Health Information Management Association to the Standards and Security Subcommittee of the National Committee on Vital and Health Statistics.” Dr. Valerie Watzlaf and Dr. Mary Stanfill. July 26, 2005. From:

Lynn Jusinski is an assistant editor with ADVANCE.

ICD-10 and CAC Software

It’s coming. Sure, no one really knows when. But it is, and computer-assisted coding (CAC) just might be a savior when it does.

ICD-10’s been 2-3 years away for the last 10 years, according to Mark Morsch, vice president of natural language processing (NLP) and new technology with A-Life Medical, San Diego. ICD-10 contains roughly 10 times the number of codes of ICD-9, and Morsch admitted that making the switch will be a challenge across the industry. As a vendor, he is, however, hopeful. “CAC applications may become in some ways essential in an ICD-10 environment because of the number of codes and the fine granularity that separates out the different codes,” Morsch said.

Moving from so few codes to so many codes might be a motivator for moving to CAC. Changing the way that an NLP system, for example, reads a document and codes may be simpler than trying to change the way an entire staff of coders thinks. CAC might make the transition to ICD-10 a bit more palatable for HIM departments.

But how will vendors who already have CAC systems running ICD-9 in place deal with the change? Dee Lang, RHIT, president and owner, Dee-L and Associates LLC, Greensburg, PA, believes “it will not be a significant burden for some of the CAC vendors to switch to ICD-10, but this depends, in large part, on the underlying development of the technology. The bottom line is that CAC technology should help the HIM professionals make a smoother transition from ICD-9 to ICD-10.”

Many vendors have already taken steps toward the move to ICD-10. ProVation Medical Inc., Minneapolis, a structured input vendor, has a coding engine designed to handle the switch. “It’s a mapping engine, so it’s able to map any two sets of disparate data, regardless of what type of data it is,” said Sean Benson, co-founder, vice president, consulting. “Probably the biggest change for us will be reviewing our medical content — to make sure that it is detailed enough to support ICD-10. Once the content has been tweaked to be more specific, the actual mapping of that new content to ICD-10 codes will be relatively straightforward.”

At Artificial Medical Intelligence, Eatontown, NJ, the current dictionary on the company’s NLP CAC system isn’t outputting for ICD-10 yet, according to Stuart Covit, executive vice president. “We can ramp that up very quickly because our dictionary is modular. It’s basically swapping out our dictionary,” Covit explained.

In the end, the change to ICD-10 might prove a helpful boost in sales for CAC vendors across the board. “Medical NLP solutions will have an even more important role when ICD-10 is implemented because of the challenge in handling the complete ICD-10,” Morsch said.

By Lynn Jusinski

For Coders, by Coders

Allison Errickson, CPC-H, director of coding compliance with ProVation Medical, Minneapolis, didn’t really know what to expect when she sent her rŽsumŽ off to ProVation 8 years ago. She heard about the company through a colleague and was intrigued.

ProVation, a structured input computer-assisted coding (CAC) vendor, was developing its CAC system, and Errickson applied to ProVation to lend her coding expertise to the system. Formerly an auditor for a large health system, she’d seen encoders before. “However, this was the first front-end coding solution I’d seen,” Errickson said.

Once on board, Errickson worked alongside physicians at ProVation to build and structure the medical content to help the system choose the correct codes. Aided by a team of certified coders, she then mapped the CPT/HCPCS and ICD codes to the relevant medical content. Errickson also worked with the software development crew to incorporate the coding rules into ProVation’s coding engine.

Developing CAC systems requires a human, and specifically a coder’s, touch. Often, vendors hire teams of coders to help with the complex rules and to make sure the CAC engine runs smoothly and accurately. “Coders are vital to this process because they are the ones who know the coding rules. Unfortunately, coding rules are rarely black and white,” Errickson said.

With the ProVation product finished and on the market, Errickson is extremely proud she had a part in it, and customers are happy to know that human coders played a role in making the software work. “Our customers understand that we have a team of certified coders dedicated to researching these guidelines and ensuring documentation is built to support coding rules,” Errickson said.

By Lynn Jusinski

About The Author