Forget GEMs: Assign ICD-10-CM Codes from the Record
by John Verhovshek, MA, CPC
ICD-10 is just a few weeks from “going live.” If your implementation plan involves using the Centers for Medicare & Medicaid Services’ (CMS) General Equivalence Mappings (GEMs) to “translate” your most common ICD-9 codes to ICD-10, you should (quickly) adopt a new plan.
CMS created the General Equivalence Mappings, or GEMs to assist in the conversion of I-9 to I-10 (other organizations have created similar mapping tools). GEMs allow you to plug in an I-9 code to determine a likely match in I-10 (or vice versa). But even CMS admits the GEMs have serious shortcomings.
Of nearly 70,000 diagnosis codes in I-10, approximately 5 percent match I-9 descriptors. In all other cases, ICD-9 and ICD-10 “… differ so widely that all attempts at translation offer only a series of compromises and subjective choices. This is necessarily so because there is no ‘mirror image’ of one code set in the other,” warns Lolita M. Jones, RHIA, CCS and Stanley Nachimson of Nachimson Advisors. Per their calculations:
- There are 445 instances where a single ICD-9 code can map to more than 50 ICD-10 codes
- There are 210 instances where a single ICD-9 can map to more than 100 ICD-10 codes
- There are 6,821 instances in the mappings for diseases where a single ICD-10 code can map back to more than one ICD-9 code
Jones and Nachimson continue, “If clinical equivalency is the most important factor there are some cases where your data may not mean the same thing in ICD-10 that it did in ICD-9 when a clinically equivalent match does not exist.”
CMS confirms the imperfect nature of the GEMS, stating, “there is not a one-to-one match between ICD-9-CM and ICD-10 … there are instances where there is no plausible translation from a code in one system to any code in the other system” [emphasis in original].
The lesson for physicians and other clinicians is clear: Forget GEMs and assign ICD-10-CM codes from the record. CMS concurs, recommending, “In coding individual claims, it will be more efficient to work from the medical record documentation and then select the appropriate code(s),” rather than attempt to translate information from one code set to the other.
As Jones and Nachimson rightly argue, I-10 “… is not a coding problem at all. It is actually a clinical documentation problem.” To succeed going forward, clinicians should concentrate on documenting the concepts necessary to support accurate code assignment directly in I-10 (such as laterality, episode of care, acute vs. chronic, etc.), rather than worry how the ICD-9 codes “translate” to ICD-10.