HINF 520 Week 2 Data Taxonomies and Classification Systems Example

E11.65, 4548-4 and RxCUI 6809: Building a Diabetes Value Set From Four Code Systems and the Local Laboratory Codes Two Acquired Hospitals Left Behind

[Student Name]

University of Phoenix

HINF/520: Data Management and Design in Health Administration

Week 2 Assignment

[Instructor Name]

[Date]

The health system, its local codes and value set are composites written for a model paper; code meanings were checked against National Library of Medicine tables, and research comes from the sources listed.

The previous paper showed that the composite health system counted diabetes differently depending on the data source. To build a single, reliable definition, the analytics team needed a value set: a defined list of codes identifying diabetes diagnoses, tests and medications across all its data. Building it required working with four standard code systems and with local laboratory codes that two hospitals acquired in 2021 had never replaced. This paper describes the taxonomies involved and how the value set was built.

Why Codes Matter

Computers cannot understand the phrase sugar diabetes, type 2, poorly controlled. They can match codes. Code systems give data shared meaning, so that a result from one laboratory and a result from another can be recognized as the same test, and a diagnosis in one hospital can be counted with a diagnosis in another. Without standard codes, every analysis depends on hand-built lists of spellings and abbreviations that break as soon as someone types something new.

Classification: ICD-10-CM

ICD-10-CM is a classification: it groups conditions into categories for counting, billing and statistics. Every condition must fit somewhere, and codes are organized hierarchically. The E11 category covers type 2 diabetes, and its codes add detail after the decimal: E11.9 marks the disease with no complications recorded, E11.65 marks it with hyperglycemia and E11.22 marks it with kidney damage caused by the diabetes. Classifications are designed for aggregation, not clinical detail, which is why a code may say little about how a patient is doing.

Terminology: SNOMED CT

SNOMED CT is a clinical terminology with hundreds of thousands of concepts linked by relationships, designed to record clinical meaning in detail. Problem lists in the system's record store SNOMED CT concepts behind the clinician-friendly terms they select, and the record maps them to ICD-10-CM for billing. Terminologies support decision support and detailed analysis; classifications support counting and payment.

Laboratory Codes: LOINC

LOINC gives universal codes and names to laboratory tests and clinical observations. A report on LOINC's first five years described how it was built to let results from different laboratories be recognized as the same test by naming the substance measured, the sample it is measured in, the kind of result and, where it matters, the laboratory method (McDonald et al., 2003). Hemoglobin A1c in blood has the code 4548-4, and the version measured by a particular laboratory method, high-performance liquid chromatography, has the code 17856-6. The value set includes both.

Drug Codes: RxNorm

RxNorm provides normalized names and identifiers for clinical drugs and links the many ways drugs are named across pharmacy systems and drug databases. Its developers described it as a way to connect different drug vocabularies through standard names for ingredients, strengths and dose forms (Nelson et al., 2011). The value set uses ingredient-level identifiers, such as 6809 for metformin and 274783 for insulin glargine, so that any product containing those ingredients is included regardless of brand or strength.

Building the Value Set

The diabetes value set combines four parts: ICD-10-CM codes in the E11 category and type 1 codes in E10, excluding gestational diabetes codes in the O24.4 range and prediabetes, R73.03; SNOMED CT concepts for diabetes on problem lists; LOINC codes 4548-4 and 17856-6 for A1c; and RxNorm ingredients for diabetes medications, with metformin handled separately because it is also used for prediabetes and other conditions. Each code was checked against the official tables published by the National Library of Medicine.

The Local Code Problem

The two acquired hospitals' laboratories still reported A1c under local codes: HGBA1C at one and A1C-POC and GLYCOHGB at the other. These results did not appear in any analysis using LOINC. The team mapped each local code to LOINC after checking the method and units with the laboratory directors. The point-of-care test, performed in clinics on a small device, was mapped to its own LOINC code rather than grouped with laboratory tests, because its accuracy differs. A test result that carries a local code is invisible to every report that speaks the standard language.

Testing the Mapping

After mapping, the team compared A1c counts before and after. Results for about 6,800 additional patients appeared in the analysis, most from the two acquired hospitals. A sample of 100 mapped results was checked against the original laboratory reports, with no errors.

ICD-11 on the Horizon

The World Health Organization's eleventh revision of the international disease classification took effect for international reporting in 2022. Its designers built it for digital use, with a structure linked to a foundation of concepts and the ability to combine codes to add detail (Harrison et al., 2021). The United States has not yet adopted it for billing and will continue to use ICD-10-CM for some time, but the team noted that any future transition will require remapping value sets.

Procedures and Services: CPT and HCPCS

The value set does not stop at diagnoses, tests and drugs. Some diabetes measures depend on services, such as retinal eye examinations, kidney function monitoring and diabetes self-management education, which are recorded with procedure codes. CPT codes describe physician and outpatient services, and HCPCS Level II codes cover supplies and services not in CPT. The team added service codes for eye examinations and education so that care gap reports could show which patients were missing them. Because some of these services happen outside the system and are reported only on claims from other providers, the team also receives payer claims files for patients in value-based contracts.

What Mapping Costs

Mapping is skilled work. The local laboratory code project took an analyst and two laboratory staff about 120 hours, most spent confirming methods and units rather than matching names. The team considered paying the acquired hospitals' laboratory vendor to replace local codes at the source, which would prevent new unmapped codes from appearing, and chose to do both: map the existing codes now and require LOINC for every new test built in the future.

Governing the Value Set

Code systems change: new ICD-10-CM codes take effect each October, LOINC and RxNorm release updates regularly and new diabetes medications come to market. The value set is stored in the data warehouse with a version number, reviewed each quarter by an analyst and an endocrinologist and approved by the clinical quality committee. Any change is logged, so reports can be traced to the version they used.

Beyond Diabetes

The same approach will extend to hypertension, asthma and depression. The team also plans to use value sets published by federal agencies for quality measures where they exist, rather than building its own, so that internal reports match what the system submits to payers and regulators.

Conclusion

A reliable diabetes definition required four code systems, each built for a different purpose, and the mapping of local codes that had kept thousands of results out of view. Classifications support counting, terminologies capture clinical meaning, LOINC identifies tests and RxNorm normalizes drugs. Combined in a governed value set, they give the health system a shared language for its data.

References

Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(Suppl 6), 206. https://doi.org/10.1186/s12911-021-01534-6

McDonald, C. J., Huff, S. M., Suico, J. G., Hill, G., Leavelle, D., Aller, R., Forrey, A., Mercer, K., DeMoor, G., Hook, J., Williams, W., Case, J., & Maloney, P. (2003). LOINC, a universal standard for identifying laboratory observations: A 5-year update. Clinical Chemistry, 49(4), 624-633. https://doi.org/10.1373/49.4.624

Nelson, S. J., Zeng, K., Kilbourne, J., Powell, T., & Moore, R. (2011). Normalized names for clinical drugs: RxNorm at 6 years. Journal of the American Medical Informatics Association, 18(4), 441-448. https://doi.org/10.1136/amiajnl-2011-000116