Data Content, Structure, and Standards
This domain tests the raw material of health information: the data elements collected, the data sets that standardize them, the characteristics that make data trustworthy, and the terminologies and exchange standards that move data between systems.
Data elements, data sets, and sources
A data element is the smallest named unit of data with a defined meaning; a data dictionary specifies its format, allowed values, and source. The UHDDS governs inpatient discharge data - principal diagnosis, other diagnoses, procedures, discharge disposition, expected payer, and birth weight for newborns. OASIS covers home health, the MDS covers nursing facilities, DEEDS covers emergency departments, and HEDIS (from NCQA) derives quality measures largely from claims and administrative data. Primary sources are records created at the point of care; secondary sources such as indices, registries, claims, and the EHR data warehouse are derived and must be validated against the primary source.
Data quality characteristics
AHIMA's data quality model asks whether data are accurate, complete, consistent, timely, precise and granular, clearly defined, accessible, current, and relevant. Errors caught at the point of collection are cheapest to fix. Duplicate medical record numbers split a patient's history across two charts and create real safety risk, which is why master patient index maintenance matters.
Terminologies and exchange standards
- Reference terminologies: SNOMED CT for clinical findings and procedures, LOINC for laboratory and other observations, RxNorm for medications.
- ICD-10-CM/PCS and CPT/HCPCS are classifications built for reporting and payment, not for clinical granularity; mapping between a classification and a reference terminology requires human review.
- Interoperability: HL7 version 2 for messaging, CDA for clinical documents, and FHIR for resources and APIs. The US Core Data for Interoperability names the data classes that must be exchangeable, and the information blocking rules implement the 21st Century Cures Act.
Worked example
A laboratory reports a hemoglobin A1c result. The observation itself is represented with a LOINC code so any receiving system recognizes the test; the analyte result, unit, and reference range travel in a FHIR Observation resource; the patient's diabetes problem appears on the problem list as a SNOMED CT concept; and the claim uses ICD-10-CM. Four vocabularies, one clinical event - mapping, not substitution, connects them.
Traps that cost marks
Calling ICD-10-CM a clinical terminology; treating HEDIS as a documentation standard rather than a measure set; assuming FHIR is a document format like CDA; forgetting that secondary data must be validated; and ignoring the MPI as a data quality problem.