| Logical Observation Identifiers Names and Codes (LOINCRTM) was developed in 1994 to provide a universal vocabulary for reporting laboratory and clinical observations. This dissertation was aimed at determining whether LOINC is meeting its goal when it is used in the real world.;Three institutions, Associated and Regional University Pathologist (ARUP), Intermountain Healthcare, and Regenstrief Institute, were invited to participate in this research. These institutions represented three of the seven institutions that provided their catalogue of laboratory test names for creating the first version of laboratory LOINC codes. After obtaining IRB approval, each institution provided 5 years (2003-2007) of laboratory data and their associated local codes and LOINC code mappings. Extensional definitions (EDs) were used to characterize the laboratory data reported by a specific LOINC code. EDs included frequency of testing, mean and standard deviation of the result values, coded variables, etc. To reduce privacy concerns, we distributed parsing and processing programs to each institution and the initial processing of the raw results occurred within the local systems, and only the deidentified EDs were sent to the primary investigator for combined analysis.;We used the EDs to evaluate the coverage, correctness, consistency and competence of LOINC. For coverage, we analyzed how many laboratory tests being routinely tested in daily operations could be assigned a correct LOINC code. For correctness, we verified the accuracy of LOINC mappings to local codes. For consistency and usefulness, we detected any inconsistencies in LOINC design and measured the degree of semantic interoperability that could be achieved using LOINC. Besides auditing LOINC code use, we also analyzed the result values that were associated with the LOINC results (i.e. characteristics like the type of result (number, coded value), units of measure, answer set (positive/negative) etc.). We also found that consistent use of result values was important in achieving semantic interoperability when exchanging laboratory data.;Our analysis produced the following results: 1. Completeness: LOINC can provide 99% coverage rate for the results in two typical health care institutions and 79% coverage for results from a reference laboratory. 2. Correctness: An error rate of 4.5% existed in mappings at the three institutions. 3. Consistency and usefulness: Several complicated or inconsistent designs for LOINC usage were found, which reduced the semantic interoperability of LOINC. |
