Exploring the use of ontologies and automated reasoning to manage selection of reportable condition lab tests from LOINC

Epidemiologists publish criteria for laboratory tests that must be reported to public health agencies in order to initiate public health control measures. There are efforts to publish value sets of standard laboratory test names using Logical Observation and Identifier Names and Codes (LOINC®) codes to enable automated systems to use the codes to identify reportable events. Unfortunately, the set of lab tests (and thus codes) vary by state, are difficult to manually curate, and may be missing desired or include undesired tests. Previously, we developed an ontology that classified the terminology used to describe LOINC®-coded tests for Chlamydia. To test the extensibility of this model, we extended the ontology to handle tests for tuberculosis. The requirements for tuberculosis laboratory test reporting in Utah and New York City gathered for the CDC’s Reportable Conditions Knowledge Management System (RCKMS) project were reviewed. This provided the basis for manual queries to LOINC® to garner all possible tests for tuberculosis, and examination of these tests revealed new terms to add to the ontology. For each test, we created a new ontology term with a logical definition, and used the HermiT reasoner to automatically classify the tests into an ontology of tests. We used the new ontology to query for epidemiological selection, and compared to manually selected result sets. The LOINC® database provides structure that is useful to develop an application ontology to support epidemiologists with the task of managing sets of codes that meet reporting criteria. The automated classification strategy we propose is reproducible and extendable to address new diseases and problems found as the ontology is improved.