By Josh Pollatsek, MBA
Vol. 13 No. 3 P. 9
Walked into a lamppost? Code W22.02XA.
Walked into a lamppost again? Code W22.02XD.
The Wall Street Journal recently took something of a humorous look at the pending transition to ICD-10, an industry shift that will involve expansion from a 30-year-old, 13,000-diagnosis-code system to one encompassing more than 68,000 diagnosis codes. And some of these tens of thousands of codes could spark a bit of a laugh:
But what’s no laughing matter is the amount of time and preparation required to smoothly absorb this massive coding shift, an initiative designed to more accurately and precisely classify medical diagnoses and procedures to strengthen care guidelines and enhance payment processes.
For the industry’s approximately 180,000 coders experienced with the ICD-9 three- to five-character coding system, the transition to the ICD-10 three- to seven-character coding system represents a fundamental shift requiring significant education and retraining. The healthcare organizations that fail to effectively embrace readiness efforts long before the anticipated October 1, 2013, changeover may face substantial productivity and reimbursement losses.
According to the American Medical Association, “The ICD-10 code sets include greater detail, changes in terminology, and expanded concepts for injuries, laterality, and other related factors. The complexity of ICD-10 provides many benefits because of the increased level of detail conveyed in the codes. The complexity also underscores the need to be adequately trained on ICD-10 in order to fully understand reporting changes that will come with the new code sets.”
In addition, “Organizations that are not prepared could face major billing headaches and loss of compensation, since claims submitted after the October 2013 deadline not using the upgraded coding language will be rejected,” the American Health Information Management Association (AHIMA) recently cautioned. In fact, health organizations are heeding such warnings, as AHIMA’s latest survey revealed that 85% of respondents reported they are planning for or implementing ICD-10 coding, a significant upswing from the prior year’s 62% finding.
But how are these organizations doing it? Many are utilizing computer-assisted coding (CAC) solutions as a foundational strategy for efficiently and effectively addressing the ICD-10 transition well in advance of the deadline. Such technologies analyze rich clinical documentation—physicians’ dictation—to assess the services performed and diagnoses made to automatically assign the relevant billing codes. By no means do the solutions render human coders obsolete; they are meant to elevate the strategic nature of coders’ jobs, allowing them to focus their intelligence and experience on complex issues while the system automatically processes routine ones.
By applying CAC solutions to areas directly impacted by the transition, health organizations can take steps today to ensure they are prepared for and poised to succeed under the new ICD-10 coding system. And while the precise cutover deadline may appear somewhat indefinite in light of recent announcements from the Centers for Medicare & Medicaid Services, it is a certainty that health organizations will position themselves advantageously by using all available lead time to prepare to capture clear, concise information for use in improving patient care and business success.
Thorough preparation requires several areas of focus, including coder training and practice, clinical documentation analysis, institutional training and education, and systems and data review. But for this article, we’ll focus on one of the most important areas requiring change to function effectively within ICD-10: physician documentation.
ICD-10 requires a much higher level of specificity necessary, and more detailed clinical documentation is essential to accurately code to that higher level of specificity. Without additional clinical detail, coders will be forced to default to unspecified codes, those general codes available for use when clinical documentation doesn’t support the selection of more specific codes. This will limit physicians’ and organizations’ ability to take advantage of the greater clinical detail tracked under the new system.
Natural language processing solutions can be a key part in a healthcare facility’s approach to assessing this documentation deficiency as a first step in designing and providing physician training to close any gaps. By running clinical notes through a natural language processing engine, organizations can use technology to study and analyze language patterns—one record at a time while still encompassing the complete base of available records—for crucial areas of specificity to determine where physician documentation is satisfactory or lacking.
This approach allows organizations to glean specific and useful information from a fully valid sample of clinical documentation. This is in dramatic contrast to the labor-intensive and expensive assignment of human auditors/reviewers who are typically bounded by cost constraints to reviewing only a relatively small sample of records from which to make their assessments.
From the engine-driven analysis, detailed reports reveal exactly where physicians’ current coding practices satisfy ICD-10 coding demands and where deficiencies require a fundamental change in documentation approach to be able to take full advantage of the better care, analytics, and measurement that will become possible through compliance with the new coding system.
To delve more deeply into the subject of specificity, consider that the ICD-9 coding system offered approximately 13,000 diagnosis codes and 3,000 procedure codes. ICD-10 includes 68,000 diagnosis codes and 87,000 procedure codes. As these numbers indicate, there are many areas in the new system that call for dramatically increased specificity in documentation.
Several of the most influential areas, where attention today can make the greatest impact in preparedness for tomorrow, include laterality (left, right, bilateral, etc); encounter type (initial, subsequent, etc); diagnosis qualifiers (chronic, acute, mild, etc); and anatomy (not “pain in limb” but which part of the limb).
Natural language processing solutions can “read” current notes to study the language patterns of clinical documentation to discover the clues—or lack thereof—that yield the specificity required to select the most appropriate, accurate, and specific ICD-10 codes. The engine tracks instances in which insufficient documentation would require a default to an unspecified ICD-10 code. For example, when parsing a note for indication of laterality, the engine seeks not only instances of the words “right” and “left” but also many other telling laterality indicators, such as “bilateral,” “both,” and “plural,” as well as pluralized anatomical terms, such as “eyes,” “feet,” or “arms.”
Consider this example of current clinical documentation: Findings: AP and lateral X-ray ankle is reviewed. Bimalleolar fracture is evident radiographically; however, the medial malleolar area appears to be an antecedent injury. The alignment of the ankle is satisfactory. Casting material is present.
In this case, related to encounter type, a natural language processing engine would flag the word “casting” as secondary evidence indicating a subsequent encounter but would report this note as lacking a true indicator of healing status.
Consider a second example, this one reviewing documentation for anatomy detail:
Order Procedure: Foot 2 View Right
Reason for Exam: Sprain/845.02
Impression: Unremarkable right foot
Right foot 2 views.
History: Sprain. The right is unremarkable without evidence of fracture or dislocation.
As the engine processed this current note, it would identify a need to specify which ligament was sprained. Without such detail, a default to the unspecified code S93.601A would be necessary.
The sheer volume of content is causing consternation across the industry and will require clinicians and coders to think differently about what they do and how they work. But the overall equation is simple and presents a clear starting point: documentation. If the documentation is not sufficiently robust, coders will be forced to code to unspecified, and the benefits available via the detail of the ICD-10 system cannot be reaped, not to mention the potential impact on reimbursement. By applying natural language processing solutions to the analysis of physician documentation, practices and health organizations can determine now how their documentation would fare under the rigors of the ICD-10 system and can apply training to close those gaps based on the identification of real data deficiencies.
Physicians have been documenting to the specifications of the ICD-9 system for more than 30 years. A shift in approach combined with significant retraining is clearly required. With the right tools in place to identify the specific nature of the necessary reeducation and an attitude based on making changes to the areas specifically revealed by the system to require it, physicians and practices can master the ICD-10 conversion and take advantage of the benefits enabled by the new wealth of data captured.
In fact, thorough and early education and preparation should enable healthcare professionals to avoid F40.9 (phobic anxiety disorder, unspecified) and Z56.6 (other physical and mental strain related to work) or worse, Z56.1 (change of job).
— Josh Pollatsek, MBA, vice president of services for CodeRyte, has more than 20 years of experience in healthcare software development, marketing, and consulting.