HI530 Unit 10 Implementation Recommendation Example

Capture Meaning, Derive Codes, Carry the Losses: A Unit 10 Implementation Recommendation for Clinical Terminologies at a Health System, With a Terminology Server and a Loss Register

[Student Name]

Purdue University Global

HI530: Clinical Terminologies, Vocabularies, and Information Standards

Unit 10 Implementation Recommendation

[Instructor Name]

[Date]

Tolland Ridge Health, its current state and its costs are composite; the standards and sources cited are real.

Decision Requested

Approve an eighteen-month program to standardize clinical terminology at Tolland Ridge Health, at a composite one-time cost of $410,000 and a recurring cost of $160,000 a year, including one additional terminology analyst. Without it, the health system will continue to report quality measures from codes that were never designed to carry clinical meaning, will keep retyping medication lists at transitions and will not be able to exchange coded data with partners who expect standard terminologies.

Current State

Three of the system's clinics record problem lists directly in ICD-10-CM, choosing billing codes from a search box, so lists are full of "unspecified" codes and symptom codes for conditions that were later confirmed. Laboratory results travel under about 1,100 local codes, of which a little over half are mapped to LOINC, some incorrectly, as last year's glucose incident showed. Medications are recorded in RxNorm in the hospital but partly as free text in two clinics. Maps from problem lists to classification codes exist only inside one vendor's system, and nobody outside coding can see what they lose.

Recommendation, Part 1: Capture

Record clinical meaning where it is created, in the reference terminology built for it: problem lists in SNOMED CT, through an interface terminology that lets clinicians search in plain language and stores the concept behind the phrase they pick; results in LOINC, with every local code mapped and the specimen and method confirmed; and drugs as RxNorm concepts that carry ingredient, strength and form. Reference terminologies were built to record detail that classifications set aside (Cimino, 1998), and capturing in them first means nothing has to be recovered later.

Recommendation, Part 2: Derivation

Derive ICD-10-CM codes for problem-list reporting from SNOMED CT concepts through the National Library of Medicine's published map, reviewed locally for Tolland Ridge's high-volume concepts (U.S. National Library of Medicine, 2025). Rule-dependent targets go to coders for confirmation rather than taking a default. Claims coding for encounters continues to follow official coding rules from the full documentation, not from the problem list alone.

Recommendation, Part 3: Mechanism

Host the terminologies, value sets and maps on a single FHIR terminology server that every system calls for lookups, validation and translation, rather than keeping copies inside each application. The FHIR terminology module defines standard operations for validating a code, expanding a value set and translating between code systems (HL7 International, 2019), so the record, the laboratory system, the reporting warehouse and the exchange engine all draw on one current version.

Recommendation, Part 4: Loss Register

Every map entry that is not equivalent, whether broader, rule-dependent or without a target, is recorded in a loss register with what is lost and which reports use the derived code. Each quality report built on derived codes carries a footnote stating the share of its population that came through lossy map entries. The register is reviewed quarterly by the terminology committee. A map is only safe to report from when its losses travel with the numbers, and the loss register is the mechanism that makes them travel.

What Changes for Clinicians

For most clinicians, the change is small and mostly invisible. A physician adding a problem types "type 2 diabetes with kidney disease" into the same search box as today; the difference is that the phrase now resolves to a SNOMED CT concept that captures both conditions and their link, instead of a billing code chosen from a list. Laboratory staff see no change in their screens, though the codes behind results become trustworthy across systems. Pharmacists and clinic nurses entering medications choose from a drug list rather than typing free text, which the hospital already does. Coders see the most change: they confirm derived codes flagged as rule-dependent before reports run, and they own the classification side of the loss register. The program asks the most of the people who maintain terminology, not of the people who deliver care, which is why the analyst position is part of the recurring cost rather than an optional extra.

Alternatives Considered

1. Code directly in ICD-10-CM at the point of care everywhere. Rejected: it forces clinicians to choose billing categories for clinical meaning, produces the unspecified-heavy lists the clinics already have and cannot express suspected conditions or risks.

2. Extract coded meaning from free text later with language processing. Deferred: promising for legacy notes, but its error rates would have to be measured and explained for every report, and it does nothing to improve exchange now.

3. Keep the current state. Rejected: the costs are already visible in inflated or missing quality numbers and retyped medication lists.

Phased Timeline

1. Months 1 to 6: install the terminology server; complete LOINC mapping with specimen and method review; begin the loss register with existing vendor maps.

2. Months 7 to 12: move the three clinics' problem lists to SNOMED CT capture through the interface terminology; finish RxNorm coverage in the two clinics.

3. Months 13 to 18: point the reporting warehouse and exchange engine at the terminology server; add loss footnotes to quality reports.

Costs

One-time: terminology server licensing and setup, $180,000; mapping and review labor, $150,000; clinic training and build, $80,000. Recurring: server hosting and support, $55,000 a year; one terminology analyst, $105,000 a year including benefits. These figures are estimates for this composite system and should be confirmed with vendor quotes.

Measures

1. Share of new problem-list entries captured as SNOMED CT concepts: target 95 percent by month 12.

2. Share of resulted laboratory tests carrying a reviewed LOINC code: target 98 percent by month 6.

3. Share of clinic medication entries coded in RxNorm: target 95 percent by month 12.

4. Share of quality reports carrying loss footnotes: target 100 percent by month 18.

5. Quarterly mapping audit error rate: below 2 percent.

Risks

Clinician resistance to new search screens is the largest risk; the interface terminology and short training sessions address it. Vendor support for the terminology server varies, so contracts should require standard operations. And the program depends on the terminology analyst role; without it, maps and the register will decay. Interoperability texts stress that terminology programs fail more often from neglected maintenance than from poor design (Benson & Grieve, 2021).

References

Benson, T., & Grieve, G. (2021). Principles of health interoperability: FHIR, HL7 and SNOMED CT (4th ed.). Springer.

Cimino, J. J. (1998). Desiderata for controlled medical vocabularies in the twenty-first century. Methods of Information in Medicine, 37(4-5), 394-403. https://doi.org/10.1055/s-0038-1634558

HL7 International. (2019). FHIR release 4: Terminology module. Health Level Seven International.

U.S. National Library of Medicine. (2025). SNOMED CT to ICD-10-CM map. National Institutes of Health.