CCA (CNCF): Mapping a Legacy Problem List to ICD-10-CM in a New EHR

The migration mixes terminologies with different purposes. SNOMED CT is a clinical terminology designed for detailed, concept-based documentation; ICD-10-CM is a classification designed for morbidity coding, billing, and statistics. A SNOMED concept does not always have a single equivalent ICD-10-CM code, and an ICD-10-CM code may represent a broader or narrower concept. The 'colon polyps' example shows semantic mismatch: an active polyp should not map to a personal-history code such as Z86.010; the active condition would more appropriately be K63.5 or a neoplasm code depending on documentation. The 18% historical entries marked active show a status/data-quality failure. The diabetes one-to-many map shows that map order, not clinical documentation, would determine the code, which is unacceptable. Free text adds synonymy, abbreviation, and negation problems. Automated mapping without human review can create incorrect claims, skew quality reporting, and propagate errors into the new EHR. The issues are semantic equivalence, granularity, status (active versus history), completeness, and mapping direction.

The plan should name a mapping owner and a multidisciplinary review group including HIM, clinicians, informatics, and billing. Prioritize high-volume, high-risk, and billing-sensitive concepts rather than trying to perfect all 9,400 entries. Use the official SNOMED CT to ICD-10-CM map and map advice where available, but require human review of every one-to-many and flagged concept. Build a test set of at least several hundred encounters with known correct ICD-10-CM codes and compare interface output. Freeze the map at a defined date, version it, and document every override and rationale. Create a feedback loop for clinicians to correct problem-list status and specificity. Do not allow free-text entries to map automatically to billing codes; route them to review or require structured selection. Define go-live rollback and monitoring metrics, such as unspecified-code rate, active/historical mismatch rate, and claim edit volume.

Historical entries need a status field that is clinically meaningful and visible to users. 'History of colon polyps' may legitimately map to Z86.010, but an active polyp should not. The clinic should separate active, resolved, and historical problems and require review before any historical entry is used for billing. For diabetes, the interface must not pick a code by map order. The clinic should require structured documentation of diabetes type and complications: type 1, type 2, or other; with or without complications; and the specific complication. E11.9 is appropriate only when type 2 diabetes without complications is documented. E11.65 requires documented hyperglycemia, and E11.69 requires another documented complication. The coder or clinician should review flagged cases, and the EHR should present the choices rather than defaulting. The governance group should monitor diabetes coding distributions after go-live to detect defaulting behavior.