As our provider clients get deeper into their ICD-10 remediation efforts, they are starting to realize what our payer clients came to grips with about six months ago: custom mappings (mappings that are not the standard General Equivalency Mappings/GEMs or reimbursement mappings) are an unavoidable necessity.

This is not necessarily true for transaction processing in ICD-10 (assuming that all systems are remediated to be ICD-10-literate), but it is certainly true for analytics and remediation activities requiring translating “artifacts” (lists of codes, medical policies, service line definitions, etc.) that contain ICD-9 codes to ICD-10.

There are a number of prospective usage patterns for mappings, but most ICD-10 practitioners would agree that they primarily will be used to a) consistently populate operational systems, databases and artifacts that use or contain ICD-9 codes with ICD-10 codes; and b) to support analytic and reporting needs for multiple purposes (clinical, financial, operational, etc.). Most ICD-10 practitioners also likely would agree that the existing mappings (GEMs, reimbursement) are inadequate to support these usage patterns.

The inadequacies are due to the clinical nature of the mappings. There is natural ambiguity in GEMs because there are often multiple clinically equivalent and accurate alternative mappings between the code sets (and in some cases, no alternatives). Therefore, using a clinically steeped mapping tool like GEMs introduces a number of challenges for remediation of operational systems:

  • If automated systems such as billing are remediated and there is more than one “right” answer or no answer for a certain translation, then additional logic will need to be developed to handle the ambiguity (or kick out the translation for manual processing).
  • If a clinical lens is applied to the resolution of mapping ambiguity, this may serve certain purposes, but it likely doesn’t address financial and operational analytic requirements.
  • If ambiguity is resolved by specific departments or process owners (each resolving ambiguity differently), there is a strong risk that there will be inconsistencies in how the same information is used across different operational processes, making data integration, analysis and root-cause troubleshooting of issues difficult. Additionally, there is significantly more effort required when departments perform mapping and code analysis separately and resolve ambiguities in a vacuum. This is true for initial remediation and for annual update processes as the Centers for Medicare & Medicaid Services (CMS) updates the GEMs each year.

We see the need for two types of mappings to support our clients’ operational and analytic needs: “best fit” mappings and “all possibilities” mappings.  “All possibilities” mappings identify all possible mappings for each code in the source code set in the target code set. “Best fit” mappings force a 1-to-1 mapping for all codes by identifying the single best code in the target code set for each code in the source code set. As an example, consider the ICD-9 code 260 for kwashiorkor and the GEMs corresponding ICD-10 codes E40 for kwashiorkor and E42 for marasmic kwashiorkor. In the ICD-9-to-ICD-10-CM GEMs mapping, 260 maps to E40 in a “1-to-many-approximate” relationship. In the ICD-10-to-ICD-9-CM GEMs mapping, E40 and E42 map to 260 in a “1-to-many-approximate” relationship. An ICD-9-to-ICD-10 “all possibilities” mapping would look bi-directionally and relate 260 to both E40 and E42. An ICD-9-to-ICD-10 “best fit” map would relate 260 to E40 only.

In general, ”all possibilities” mappings are useful when you are trying to translate an ICD-9-laden artifact and you don’t want to miss any ICD-10 codes that potentially could relate to the ICD-9 code. For example, translating a medical policy or a list of codes relevant to a clinical study would best leverage such a mapping.

“Best fit” mappings are useful when ICD-10 codes need to be inserted for ICD-9 codes in existing artifacts and a given ICD-10 code only can be inserted once in place of its ICD-9 counterpart. This mapping will identify the correct 1-to-many relationship to use. Consider having to replace ICD-9 codes in service line definitions in such a set of circumstances when an ICD-10 code only can exist in one service line: say you’re dealing with an ICD-10 code like F458 (Other Somatoform Disorders) that relates to 12 different ICD-9 codes, some of which are emergency codes and some that are not.

“Best fit” mappings are not easy to develop; with many clinically “correct” codes to choose from and the multitude of ICD-10 codes that relate to multiple ICD-9 codes, gaining consensus is not always easy. Nonetheless, these mappings will be required to remediate operational systems properly. “All possibility” mappings are much easier to develop, but “no maps” require resolution.

About the Author

John Wollman is the Executive Vice President of Healthcare for HighPoint Solutions, a Management and Information Technology consulting firm focused on Healthcare and Life Sciences.  John is responsible for HighPoint’s Healthcare industry group, catering to Payers and Providers.  John is a recognized expert in several healthcare business domains (Reform, HIPAA 5010, ICD-10, Platform Strategy) and technical domains (Master Data Management, Analytics).  Since graduating from Duke University, John has held executive level positions at consulting and technology companies over his 25 years in business.

Contact the Author

To comment on this article please go to

Share This Article