|
Although traditional coding processes may include the use of encoders or other computerized tools, they still require coders to read, or at least visually scan, all of the clinical documents from a treatment episode to identify health care provider-attested information. This information is used to justify the assignment of disease classification and medical procedure codes to the case for claim submission to a payer. Coders must then individually input a subset of the codes for which they have found justification to ensure completeness and accuracy of the claim. This repetitive, manual process not only absorbs much time and resources, but it also introduces the opportunity for human error and inconsistencies, resulting in rejected claims or underpayment from incomplete coding.
The next level of coding evolution is computer-assisted coding (CAC), defined by the American Health Information Management Association (AHIMA) as computer software that automatically generates a set of medical codes for review, validation and use based on clinical documentation provided by health care practitioners. The first health care environment to leverage CAC was the freestanding outpatient diagnostic site, which used CAC primarily for radiology, pathology and emergency department billing. Faced with substantially fewer and less complex documents and coding rules as compared to inpatient clinical areas, outpatient sites were able to use the technology -- even in its early stages -- to increase billing accuracy and turnaround. In essence, the technology transformed the role of coders from performing time-consuming, manual input to critical review and analysis of automatically generated data.
Now with significant advances in natural language processing (NLP) technology, rising adoption of electronic health records and industry migration to the ICD-10 coding system, CAC is poised to have a significant impact in the inpatient environment as well.
Artificial intelligence
NLP, which is the backbone of CAC, is software technology that uses artificial intelligence techniques to identify concepts in free text and associate codes from controlled vocabularies to the relevant phrases in the text, thereby enabling the extraction of computer-processable data elements or standard terms from free text to feed databases and facilitate data exchange.
In health care, an NLP-enabled CAC application automatically identifies medical concepts or terms and assigns disease classification and medical procedure codes to them after analyzing the context in which these words are used within the text document. For example, in the case of "viral pneumonia" and "aspiration pneumonia," an NLP-based CAC product knows that the word "viral" is used as an adjective while "aspiration" is used to describe a type of pneumonia, and identifies codes accordingly. Next, it forwards the codes it has identified to coders for review and validation based on clinical documentation.
Depending on its structure, a CAC solution may use either a statistics- or rules-based approach, or both, to assign codes to phrases in the text. A statistics-based NLP engine predicts which medical code may correspond to a particular term based on past statistical probability, while a rules-based NLP system uses algorithms or programmed rules based on linguistic analysis of the text.
Until recent years, the types of medical documents from which NLP could reliably extract meaning were rather limited. Outpatient care involves narrowly focused treatment and, generally, only a few types of documents and simpler coding rules for reimbursement. In contrast, inpatient treatment involves extensive documentation of many different types and even more challenging coding rules for reimbursement because patients are frequently treated for multiple, interrelated problems during a single admission. Due to advances in NLP technology, and its use in a growing number of specialized clinical domains, CAC is now poised to meet the complex demands of the inpatient setting.
Drivers of CAC adoption
Since a prerequisite of NLP-based CAC is that data must be in an electronic format, the American Recovery and Reinvestment Act (ARRA), with its $19.2 billion in incentives to encourage nationwide adoption of electronic health records (EHRs) by 2014, should help drive CAC adoption.
Additionally, the health care industry's migration to the ICD-10 coding system should accelerate CAC adoption. Already faced with a Jan. 1, 2012 compliance date for the HIPAA 5010 transaction standards, hospitals will be challenged to meet the Oct. 1, 2013 deadline for migrating to ICD-10. One of the biggest challenges related to ICD-10 adoption is its approximately 155,000 procedure and diagnosis codes compared to approximately 16,000 codes in ICD-9. With this tremendous increase in the quantity of codes, it will become even more critical to enhance the productivity of coders with automated assistants such as NLP-based CAC systems.
Although the use of CAC for inpatient coding is still limited to early adopters, health care delivery organizations are increasingly listing the technology as a prerequisite in their health information management requests for proposals (RFPs).
Challenges
Hospitals must overcome the hurdle of gaining coder buy-in. Recognizing the decrease in manual labor resulting from CAC deployment, coders may fear a reduction or total elimination of coder positions. However, this concern can be addressed by emphasizing the increased workload that ICD-10 will bring and that the organization is implementing the technology to help coders perform their job more effectively and efficiently in new, more complex environment. Facilities should also explain that CAC requires coders to apply their coding expertise to edit or validate the output of the NLP engine -- and do higher level data analysis -- rather than spend their time finding information in documents and entering codes manually at the beginning of the process. In other words, CAC helps transform the coder's role into a more strategic, knowledge-based contribution, rather than diminishing or eliminating it.
Another challenge facing institutions is selecting the right solution. Vendors of existing solutions currently must integrate with a third-party encoder. The next generation of technology, expected to be commercially available by the end of the year, will feature a CAC system with a built-in encoder that can also be integrated with an inpatient compliance application. As a result, institutions will be able to reduce interfacing and maintenance costs, and further improve the efficiency of compliance processes.
Business benefits
One way to understand the value of CAC is to compare its evolution to that of speech recognition technology as applied to medical transcription. Similar to speech recognition, CAC functionality was at first limited. However, technological improvements over several years led to significantly increased accuracy and broader applicability.
Both technologies align with physician workflows. Just as speech recognition does not change how physicians dictate, CAC does not alter how they document encounters. Instead, both technologies streamline back-end processes. For instance, speech recognition converts dictation into text documents more rapidly than a transcriptionist, allowing the transcription staff to shift their focus to editing for accuracy. Similarly, CAC finds the relevant information in a set of documents, highlights it and suggests relevant codes to the human coder and enables him or her to move directly to the last step of the process: reviewing for coding accuracy. The result is increased coder productivity, which directly impacts hospital revenue in two ways.
|