Reconstruction of a Large-Scale Real-World HL7 v2.5 Database Using HL7 FHIR®

The 6NC-EHRs Database (6NC-EHRsDB) is based on clinical data output from electronic health records (EHRs) at six national centers. This study evaluated the reproducibility of the DB constructed in HL7 v2.5 to the HL7 FHIR® (FHIR). First, we clarified the correspondence between the 80 items collected in the 6NC-EHRsDB and each in FHIR. Next, FHIR resources (Condition, Medication Request, and Observation LabResult) extracted from EHRs were compared with data in 6NC-EHRsDB to evaluate the number of records, data item coverage, and the accuracy of data values were evaluated. 74 of the 80 items in 6NC-EHRsDB were successfully mapped to FHIR. The verification of coverage revealed that standard codes such as drug and disease names were not included in the data. In the records, 99.9% of Condition, 99.2% of Medication Request, and 99.9% of Observation LabResult were found to match after adjustment. Concerning the accuracy of data values, Medication Request had the lowest agreement rate of 69.1% for the prescription start date, and Observation LabResult had the lowest agreement rate of 58.9% for standard values. Based on the issues identified in this study, the reproducibility of DBs constructed with HL7 v2.5 using FHIR are expected to be improved. Our result will enhance data utilization and interoperability among medical information systems by eliminating deficiencies in implementing FHIR and enabling the assignment of accurate standard codes.