A proposal for a LOINC automatic mapping support tool

Abstract

Clinical coding and classification systems are the lingua franca to overcome idiosyncrasies of the local information systems. Among those required into the Italian EHR, LOINC aims to uniquely identify both clinical document types indexed into EHR registries and tests into laboratory reports. Its use requires that local test codes are mapped to the codes of the standard so that equivalent concepts are aligned and can be easily understood and reused by other systems. This mapping process is cost and time-consuming as involves for many hours domain experts to find the right association. Therefore, this paper aims to make a proposal for an automatic mapping tool to support domain experts in finding the correct LOINC code to map to, specifically customized on the Italian language, combining textual similarity techniques with new NLP models and taking advantage of the so-called “wisdom of the crowd”.