Computer - Assisted Coding

Angela Brown
BINF 3301
Prof. Manger
November 15, 2009

Computer –assisted coding (CAC) is a computerized tool that automates a set of medical or surgical codes, based on clinical documentation from a healthcare provider, which is used for review and validation. With the assistance of these new automation tools, coding or HIM professionals can easily translate clinical data input into useful clinical data output. Increased amounts of clinical coding is done by machines, which saves time and human participation for more complex coding cases and data analysis tasks. Factors, such as advances in natural language processing, EHR adoption, compliance issues and mandates for labor – intensive administrative reporting processes reduction, influenced the demand of CAC. Traditionally, clinical documentation (whether paper or electronic) is analyzed by a coder, translated into the appropriate ICD – 9 CM or CPT/HCPCS codes with the help of coding books or encoders and entered into a database. These new coding automation tools assists HIM professionals in translating data by automated code assignment instead of manual review and translation alone.
As early as the 1950s, the technology of CAC – enabled tools, particularly Natural Language Processing (NLP), started with formal language theory. Throughout this time, technological progress was slow but technology has rapidly progressed and is constantly advancing at an exponential rate since the 1990s. Coding is a difficult task because it has a four- dimensional complexity. First, coding rules’ volume and intricacy makes selecting the right diagnosis/ procedure code and code modifiers difficult. In an article by Yuval Lirov (2009), the author gives example to this level of complexity by stating, “ For instance, a claim will get denied if you charged for two CPT codes but provided an ICD – 9 code that shows medical necessity for one CPT code only” (Lirov 2009). Secondly, the complexity is exacerbated by payer- specific modifications, which creates the need to code or process the same procedures differently, depending on the payer. Third, continuous coding education and re- training is necessary due to the constant coding and regulations changes over time. Finally, the generation of charges and claims follow ups are sometimes disconnected in time and space and many times, are performed by different people which adds to the costs of the claims processing cycle.
Currently, there are two formats of CAC technology; Natural Language Processing (NLP) and Structured Input. Much like language translation software that translates one language from another, NLP software can translate data from clinical language into the language of CPT/ ICD 9 CM codes. In NLP, artificial intelligence is used to extract pertinent data and clinical terms derived from text – based documents and they are then converted into a set of codes to be used or edited by a coding professional. This software is often referred to as the NLP “engine”. Coders, first, begin by reviewing suggested codes instead of coding from scratch. The NLP system allows the coder to review, edit, approve and finalize codes for each record. By using a recommended list of codes, coders will improve their productivity and elevate their role from researcher to quality data analyst.
Within NLP, there are two approaches to NLP software. The rules – based approach creates an understanding of clinical words and phrases and the medical codes that are used to report these words and phrases with a complex, extensive series of rules. This approach is also called the knowledge – based approach because CAC’s coding ability depends on the coder’s expertise to provide the complex rules to properly assign code numbers to words and phrases. The statistics – based approach creates an understanding of clinical words and phrases and the medical codes that are used to report these words and phrases with a large body (or corpus) of reports. This approach is also called the data driven approach because the coding ability of the software is based on a group of statistics. The software predicts the appropriate code that should be assigned to a given word or phrase based on the statistics that indicates the assigned code to these words/phrases in CAC’s corpus of data.
Currently, NLP is starting to become prevalent in specific clinical settings. It is most effective when transcription system interfaces is available and the system can be remotely accessed. NLP works best in outpatient settings. Outpatient’s clinical documentation is often electronic and there are limited medical terms listed. Also, NLP engines work well when electronic data and text files are received from an electronic template, diction, or speech recognition system. Although transcription is not necessary for CAC, it is suggested that transcription interfaces are to be implemented for optimal workflow.
Structured input, also known as codified input, is based on clinical term menus. This approach does not require human interaction code selection and is generated from a pick list. Structured input is most effective in physician offices settings. These applicable medical codes are embedded into the system, giving providers the option to select specific codes for each technique that is used during a procedure. When a menu item is selected, a narrative text phrase is generated and is integrated into the health record. The diagnosis and procedure codes, entered by the clinician, are identified automatically and is matched to the correct code. Once the codes are assigned, it is presented to the clinician for confirmation and is then, given to the coding professional for manual code review. Structured input works well in this setting because each clinician’s workflow may have a limited number of diagnoses and procedures.

CAC is also present in inpatient settings, as well in outpatient settings. Inpatient CAC can increase productivity and quality by encouraging necessary changes in coding workflow. The software can automate inpatient coding tasks, which saves time for coders and reduces the generation of DNFB reports. Chart processing time decreases with the use of inpatient CAC. In an article by Dee Lang (2008), the author adds by stating, “Imagine coders having all data elements and documentation required for coding in a single format and view. Think about how many documents an inpatient coder had to hunt through to code a record ….Also, consider the various sources (some electronic, some paper , some scanned images) where the documentation is found”( Lang 2008). The author believes that CAC solutions are able to pull these documents from any source and in any format into a single view.
Inpatient CAC has the ability to read and interpret a clinician’s spoken language and can determine when coding is necessary. The system, then, assigns diagnosis and procedure and then codes are submitted to the coder for validation and sequencing. Finally, the CAC solution automatically assigns a present on admission (POA) indicators, which is validated by the coder. CAC technology decreases the burden on coders and increases overall coder performance. Due to its advanced level of comprehension in diverse terminology, CAC technology can also assist and automate some outpatient coding. Inpatient CAC technology can rely on data fields that can distinguish the difference between inpatient and outpatient records so the appropriate coding guidelines can be applied.
In an article by Cheryl Servais, MPH, RHIA (2006), a case study was conducted to evaluate CAC software and its ability to increase the productivity of a sample group of remote coding staff. It was found that because of the vast number of codes presented to the coder, the software did not improve the group productivity, as expected. Also, inpatient CAC needs additional adjustments before it will completely improve coder productivity and accuracy. The author concludes by stating, “In truth, computer- assisted inpatient coding is the most sophisticated task that anyone could ask a CAC software module to perform. As opposed to studies that report on the successful performance of CAC for outpatient coding, these studies did not have a parallel success for inpatient coding” (Servais 2006).
NLP or structured input computer-assisted coding tools has revolutionized traditional coding workflows, in certain domains. CAC technology alters coding workflow with the use of either NLP or structured input coding tools. The coding engine will show improvement based on the coder’s feedback. These coding tools are implemented as a best practice via the code assist model. The code assist model uses software that conducts an initial screening with well defined terms and creates a preliminary set of “draft” codes. These codes are reviewed, edited and revised by a coder to generate a final set of codes.
There some advantages and disadvantages to using NLP and structured input CAC tools. Improvements to CAC system through the use of free text for recording documentation is one positive aspect to utilizing NLP based CAC tools. On the other hand, the reliance on electronic documents, software development efforts and potential coding mistakes could deter productivity levels within the system. For structured input CAC tools, improvements, decreased costs and the ancillary creation of documentation can lead to system success. However, how the structured input is used can cause system setbacks. To ensure coding integrity for CAC tools, reliability and validity should be audited routinely. This is useful in determining quality improvements results and efficiency gains.
In order to prepare HIM professionals for the coming coding revolution AHIMA has created the computer – assisted coding e –HIM work group. This work group was chartered to help healthcare facilities and organizations navigate through the system and understand the preparation process for their changing work environment. In a series of studies conducted by the e- HIM work group to assess accuracy rates of NLP -based CAC tools, it was found that accuracy rates ranges from 57% to 98%. However, researchers found difficulty in defining coding accuracy when assessing code output quality from NLP-based CAC tools. Also, researchers encountered problems the variability of the codes assigned by coders. It was found that CAC tools are much faster that human coders but are not necessarily more accurate.
During the first stages of CAC software implementation, coding professionals began to worry about the existence in coding departments. Coding professionals believed that CAC solutions were implemented to replace the human coder. In an article by Chris Dimick (2008), Mythily Srinivasan, coding manager, expressed her anxieties about the software by stating, “Like everybody else, I thought, ‘Oh my god, it is going to come and am I going to be replaced? (Dimick 2008). In the same article, the author explains the connection between CAC and the coder. When the software was installed, there was a lot of feedback from the coders. This allowed the coders to see that the coding knowledge base and judgment was vital to the coding process and the system was only implemented to serve as an aid.
Although CAC technology became increasingly sophisticated, computers will not replace the work of the clinical coder; it will only reduce manual code assignment time. Computers are not capable of performing many of the tasks clinical coders can do. CAC tools should only be used to assist coders rather than replace them .With human interaction, these tools are not advanced enough to operate on their own. Although CAC technology may be fast and efficient, machines are not capable of interpreting and analyzing data the way a coding professional can. CAC lessens the burden of tedious, repetitive coding tasks so knowledge –based coding becomes the new role of coders. With the use of this technology, “knowledge workers” can concentrate on tasks that require critical thinking skills, such as the interpretation and analysis of documentation or aggregate data.
The preparation and implementation of CAC tools will challenge coding professionals to advance their skills and competencies in the scrutiny and clarification of clinical data. The development of production quality NLP –based CAC applications are dependent on traditional CAC applications. A process –driven approach is required for its development and quality assurance. In order to move forward with CAC technology, coding professionals are expected to move from production coding to knowledge- based coding. With the use of CAC tools, coders save time from doing repetitive code assignments that could have been computed. For organizational participation, electronic clinical documentation is required for CAC tools. The health record should be divided into portions that can be used for code assignments and available or readily convertible electronic forms.
The financial benefits of CAC are expected to improve, as well. In a research study conducted WinterGreen Research, Inc., medical information analysis was done on computer assisted coding. It examined market shares, forecasts and strategies for the years 2008 to 2014. It was found that CAC markets are expected to increase from $50.9 million in 2008 to $2.7 billion in 2014. These markets were primarily based on U.S. markets because of the role insurance plays in the nation’s health care delivery system.
An organization must follow certain steps in order to acquire maximum benefits from a CAC technology tool. First, the organization must evaluate their current clinical documentation. Documentation evaluations must be done for each treatment settings with input from the HIM department, as well as from a diverse provider and user work force. Then, an assessment must be done to determine how a CAC tool can alter the current workflow. Next, the organization must balance productivity and accuracy by defining expectations to the staff. By maintain high productivity and accuracy rates, the facility can achieve return on investments. The next step to maximizing on CAC success is for the organization to set goals for CAC tool utilization, develop a plan and evaluate if the goals are achievable. Organizations, should then, present opportunities where coding staff can improve their knowledge. By equipping them with the necessary tools, they can capitalize on advantages that are offered by the software. Finally, facilities must carefully plan their success to survive the change. The necessary are changes identified and a plan is created to implement them.
Since the 1990s, Computer -Assisted Coding has revolutionized the world of coding. Although skepticism and reluctance was present during the early stages, CAC has lessened coder workloads and decreased coding cost, as well as increased coding productivity consistency and accuracy. CAC has improved and evolved tremendously over the years. Coding professionals now understand and appreciate CAC technologies. Although these systems may not be capable of completing coding tasks by themselves, CAC will be around for years to come.

References
Computer Assisted Coding (CAC) Of Medical Information Market Shares, Forecasts, and Strategies, 2008-2014 (2008). Markets and Research. Retrieved November 4, 2009. Available http://www.researchandmarkets.com/reportinfo.asp?report_id=649883&t=d&cat_id=

Dimick C (2008). Computer, Coding and Change. Journal of AHIMA. Retrieved November 4, 2009. Available http://library.ahima.org/xpedio/groups/public/documents/ahima/bok1_036238.hcsp?dDocName=bok1_036238

Lang, D (2007). Inpatient Computer- Assisted Coding: Imagine the Possibilities. For the Record. Retrieved November 4, 2009. Available http://www.fortherecordmag.com/archives/ftr_11262007p8.shtml

Livol Y (2007). Medical Billing Control and Compliance With Computer Aided Coding Software. Ezine articles. Retrieved November 2, 2009. Available http://ezinearticles.com/?Medical-Billing-Control-and-Compliance-With-Computer-Aided-Coding-Software&id=240570
The Impact of Computer- Assisted Coding. Medquist (2005). [Online Website]. Available http://www.medquist.com/Portals/0/whitepapers/Coding%20Productivity%20Whitepaper.pdf
Van Der Graf A, Weber J, Weiner M (2004). Delving into Computer –Assisted Coding. Journal of AHIMA. Retrieved November 4, 2009. Available http://library.ahima.org/xpedio/groups/public/documents/ahima/bok1_025099.hcsp?dDocName=bok1_025099

Van Der Graf A, Weber J, Weiner M (2004). Delving into Computer –Assisted Coding: Appendix A: Primer on NLP for Medical Coding. Journal of AHIMA. Retrieved November 7, 2009. Available http://library.ahima.org/xpedio/groups/public/documents/ahima/bok1_025037.hcsp?dDocName=bok1_025037 Van Der Graf A, Weber J, Weiner M (2004). Delving into Computer –Assisted Coding: Appendix C: Advantages and Disadvantages of CAC Technology. Journal of AHIMA. Retrieved November 7, 2009. Available http://library.ahima.org/xpedio/groups/public/documents/ahima/bok1_025038.hcsp?dDocName=bok1_025038