To share with the audience the difficulties and the ways to succeed in achieving a translated subset leading up to the adoption for use by healthcare providers in the electronic medical record (EMR) of the Erasme university hospital in Brussels.
Background: Semantic interoperability of health care data, using SNOMED CT, is a cornerstone of the Belgian eHealth plan 2015-2018. A project to translate a selection of SNOMED CT concept into the national languages (Dutch and French) started in 2010.
Methods: Belgian terminologists and medical experts reviewed a large subset (consisting of 79.287 Clinical Findings & 25.249 Procedures concepts) using formal and documented rules . This resulted in selection and translation into Dutch and French of 43% Clinical Finding concepts and 70% Procedures concepts, considered relevant for the documentation of the medical record. The subset was tested on hospitals records, showing about 85% of adequate coverage . From 2014 this subset was used to create an extension for the Erasme University Hospital in Brussels. Most of the French translated concepts were reviewed, corrected, standardized according to IHTSDO editorial rules. New concepts were added through requests of different medical specialties. This Hospital extension was deployed in the EMR in October 2015, using flat lists consisting of preferred terms without synonyms. Requests were handled through a request tool in the EMR, directly informing the terminologist about missing, obsolete and misspelled descriptions of concepts.
Results: In October 2015, the French extension was deployed in 70% of all hospital specialties for inpatients over a period of 6 months. The monthly rate of change requests per number of discharge letters peaked at 30% in the beginning and decreased to less than 10% after 1 year. Most of the requests concerned missing (91%) and misspelled (8%) descriptions. Updates were done weekly. Usage of available concepts was 6891/35600 (20%), 4376/18700 (23%) and 811/8700 (9%) for clinical findings, procedures and body structure, respectively. Usage tended to increase over time. Currently, about 1 million concepts were recorded. The hospital disposes of a terminology server which will allow to manage synonymy, to query subsets and to update the local extension adequately.
Conclusion: SNOMED CT translation is a delicate and time consuming work. When an extension is deployed live, tools must be provided to the end-user to communicate missing concepts. An extreme reactivity of terminologists is mandatory for the adoption of the structured terminology.