Category of Allergy Identification from Free-Text Medical Records for Data Interoperability.

ABSTRACT

The use of different data formats complicates the standardization and exchange of valuable medical data. Moreover, a big part of medical data is stored as unstructured medical records that are complicated to process. In this work we solve the task of unstructured allergy anamnesis categorization according to categories provided by FHIR. We applied two stage classification model with manually labeled records. On the first stage the model filters records with information about allergies and on the second stage it categorizes each record. The model showed high performance. The development of this approach will ensure secondary use of data and interoperability.