A Gem of a New Tool
By Judy Sturgeon, CCS, CCDS
For The Record
Vol. 27 No. 6 P. 6
The health care industry faces innumerable challenges as it makes the long-anticipated transition from ICD-9 to ICD-10. In order for any code—much less the most accurate one—to be assigned, physicians must document in greater detail in terms of both diagnoses and procedures. Both facility and physician coders need a better understanding of anatomy, pathophysiology, and medical terminology to perform even the most basic translations from chart documentation to the correct code.
Because ICD-9 will still be in effect for claims with dates of service prior to the implementation of ICD-10, payers, clearinghouses, and vendors still must be able to handle the old values. They also will have to simultaneously process any claims with dates of service after the implementation via the new code sets.
Teaching ICD-10 to coders and getting software engineers to process the new code sets is only the beginning of the obstacles that must be overcome during the conversion. For example, how will valid reports be created when ICD-9 data must be converted into ICD-10, and vice versa?
Consider the complexity of the data reporting issues created when the ICD-9 code sets are replaced by the new ICD-10 configurations. In ICD-9, 14,567 diagnosis codes and 3,878 procedure codes will be replaced by or substituted with 69,832 diagnosis codes and 71,920 procedure codes. The increase in quantity alone should be enough of a signal to health care leaders that there is no simple crosswalk. However, that's only one of many potential roadblocks to a smooth transition. For example, although many ICD-9 codes may be crunched into a single ICD-10 code, the reverse is more often true. Some ICD-9 codes will not have an ICD-10 counterpart. Conversely, there are ICD-10 codes that never existed in ICD-9.
Now consider how many standardized reports have been created to document payments, budgets, staffing levels, and dozens of other important topics. Then identify how many of them include ICD-9 codes as part of the reporting criteria. All of these reports must be changed. In fact, the task becomes exponentially more difficult than if the only need was simply preparing translations for new reports.
The Centers for Medicare & Medicaid Services (CMS) has created General Equivalence Mappings (GEMs), a table to convert ICD-9-CM data into ICD-10-CM and PCS (it also maps from 10 to 9). While GEMs are inappropriate for coding patient encounters, they can be used to translate data between the two code sets. They're especially helpful when translating large databases or reports.
However, the GEMs format is not particularly user friendly. It includes a source code (the code you have), the target code (the code you need to obtain), and flags that warn whether the desired code is even possible to achieve. The flags are confusing to a layperson; a series of 0s, 1s, and 2s codify whether the proposed match is approximate, identical, or impossible and whether there's a one-to-many or a many-to-one translation. There's even a "just give up" recommendation.
The following example from the diagnosis code (CM) GEMs illustrates that the ICD-9 diagnosis code 605 can be applied to any of the following five possible new codes:
• 605 N470 10000;
• 605 N471 10000;
• 605 N472 10000;
• 605 N475 10000; and
• 605 N478 10000.
Help Is Available
For those who cannot master the 623 kilobytes of numerical columns that comprise the ICD-10-CM GEMs zipped files, Andrew Boyd, an assistant professor of biomedical and health information sciences at the University of Illinois, Chicago, and a team of experts have created metrics and tools to make it easier to translate diagnosis codes with greater consistency.
When the researchers investigated the GEMs, which are the source of the online mapping portal for their tool, they expected the crosswalk to simplify data mapping. Instead, they arrived at a different conclusion. "When you map backward, you miss 30% of your historical ICD-9 codes. And when you map forward, you're only mapping forward to 25% of your ICD-10 codes," Boyd says.
To combat this flaw, the researchers took a different approach. "We began to use techniques that are very common in bioinformatics, the science of networks," Boyd says. "It's how you start to understand interconnected things in a way that you can act upon them. Using network methodology, you can look both ways at the same time."
From the GEMs with seemingly endless columns of codes and numerical flags, Boyd's team has devised two main pages where users can upload codes or entire files to be translated into the alternate code set. The results, which can be expressed as a network, an online table, or a text file, feature the codes requested, the text associated with those codes, their relationship, any corresponding translated code(s) with the related text definition(s), and the mapping category.
"You have to start somewhere," Boyd says, "so we started with ICD-9." This mapping category translates the numerical flags from the GEM into the following five comprehensible categories:
• Identity: One ICD-9 code translates to one ICD-10 counterpart.
• Class to Subclass: One ICD-9 code translates to multiple possible ICD-10 codes. For example, the same base code, with options for left, right, bilateral, or unspecified variants.
• Subclass to Class: Many ICD-9 codes translate to only one possible ICD-10 code.
• Convoluted: This translation involves merging or modified codes and terms, with no clear correlation or definition between code sets. In essence, the translation is approximate rather than exact.
• No Translation: There is no corresponding ICD-10 code.
Boyd discovered that while 36% of all ICD-9-CM codes have convoluted mappings, the percentage decreases substantially for the codes commonly used by specialties and subspecialties. In general, he suggests users set aside more time when creating reports to translate data.
Boyd, who cautions the translation metrics are only a tool, recommends users engage the help of expert clinicians and coders to build the most accurate translations for a report.
The mapping portals are available for free at the following links:
• For translating ICD-9-CM codes into corresponding ICD-10-CM codes, visit www.lussierlab.org/transition-to-ICD10CM.
• For translating ICD-10-CM codes into ICD-9-CM codes (reverse mapping), visit www.lussierlab.org/transition-to-ICD9CM.
For the adventurous, try translating the convoluted example of ICD-9 diagnosis code 658.41 Chorioamnionitis, delivered on this encounter, onset prior to delivery or unknown to its counterparts in ICD-10. This should provide users with a good idea of how the mapping portal functions and whether it meets with approval.
Boyd plans to update the information as often as changes occur to the GEMs, which the CMS plans to update over the course of the next three years. It's a labor of love, he says. "Part of my university job is to help other people. We're all going through the transition."
— Judy Sturgeon, CCS, CCDS, is the clinical coding/reimbursement compliance manager at Harris County Hospital District in Houston and a contributing editor at For The Record. While her initial education was in medical technology, she has been in hospital coding and compliance for 26 years.
RESOURCES AND USEFUL REFERENCE MATERIAL
• ICD-10 Transition Poses Problems for ICD-9 Translation: www.healthleadersmedia.com/content/TEC-314424/ICD10-Transition-Poses-Problems-for-ICD9-Translation
• ICD-9: www.cdc.gov/nchs/icd/icd9.htm
• History of the Development of the ICD: www.who.int/classifications/icd/en/HistoryOfICD.pdf
• Benchmarking With National ICD-9-CM Coded Data: http://library.ahima.org/xpedio/groups/public/documents/ahima/bok1_009215.hcsp?dDocName
• General Equivalence Mappings Frequently Asked Questions: www.cms.gov/Medicare/Coding/ICD10/downloads/gems-crosswalksbasicfaq.pdf
• 2014 ICD-10-CM and GEMs: www.cms.gov/Medicare/Coding/ICD10/2014-ICD-10-CM-and-GEMs.html
• 2014 ICD-10-PCS and GEMs: www.cms.gov/Medicare/Coding/ICD10/2014-ICD-10-PCS.html