Abstract
Background: The increasing need to transfer data from the performing laboratory to hospitals, clinics, utilization management reviewers, clinical researchers, public health workers and others that use laboratory data requires a level of standardization that is inherent in neither data traditionally produced by the laboratory nor reporting information system. Portability and merge-ability of the data is essential to conduct multi-site clinical trials and epidemiological studies. Comparing or combining lab results performed in different laboratories, which is integral to the Virtual Data Warehouse (VDW) of the Cancer Research Network (CRN), is challenging. Barriers include locally defined test codes. Most codes are easily recognized, such as K+ for potassium, others are less intuitive. It is also common to find normal ranges of different labs to be significantly different. This creates difficulties when trying to combine into one dataset for analysis as the values are not necessarily equivalent. This was historically addressed by using indirect criteria to compare data from different sites instead of absolute criteria. The INR for prothrombin time (PT) is an exception as the World Health Organization directed a massive standardization effort for PT in 1983. LOINC was designed to address these relational issues. The feasibility of using LOINC for the standardization of lab tests in the VDW is being explored. This report describes the strengths and limitations of using LOINC at Henry Ford Health System (HFHS).
Methods: Regenstrief Institute’s RELMA program was used on lab data to test LOINC implementation. The process of mapping test code to LOINC was manual and iterative, but the resulting file contains sufficient detail to create a one-to-one relationship between data sources.
Results: The introduction of LOINC to laboratory databases allows the known methodological nuances to become part of the reported lab result. Nearly 31,000 of the 41,000 LOINC observation codes pertain to laboratory testing.
Conclusions: It is advantageous to model the VDW databases to conform to LOINC in structure and nomenclature. Designing our data to universal standards will facilitate use in healthcare informatics. This process will require a person with considerable laboratory knowledge to accomplish accurate LOINC implementation. Coding complexities and redundancies require thorough investigation to determine appropriate LOINC codes are selected for data retrieval.
- Received September 11, 2008.
