As presented in Input data model wiki page, the AI algorithm input data is represented as FHIR R4 models. This is why, to First we anonymize the openIMIS database and then we migrate it to FHIR.
Database anonymization
The following database tables and (not required) fields are anonymized (openIMIS-AI_Anonymization.sql):
tblFamilies: FamilyAddress
tblInsuree: CHFID, LastName, OtherNames, passport, Phone, CurrentAddress, GeoLocation
tblClaimAdmin: LastName, OtherNames, DOB, Phone, EmailId
tblOfficer: LastName, OtherNames, DOB, Phone, EmailId, permanentaddress, VEOLastName, VEOOtherNames, VEODOB, VEOPhone
tblUsers: LastName, OtherNames, DOB, Phone, EmailId
tblPayer: PayerName, PayerAddress, Phone, eMail, Fax
tblPhotos: CHFID, PhotoFileName
Extracted data
To develop the AI algorithm, we need to migrate the openIMIS data to FHIR. For this, we use the openIMIS to FHIR migration tool. For large data sets, ClaimResponse and Patient, we developed SQL scripts (GetClaimResponseJSON1Line.sql, GetPatientJSON1Line.sql) to generate the FHIR data directly from database.
...