Deduplication
Health Gorilla retrieves patient records from many organizations in several formats, and the same clinical facts often arrive more than once. Deduplication reduces matching records to one record that cites every source. Clinicians see each allergy once, not five times. Deduplication is always on for data Health Gorilla ingests and isn't configurable.
Deduplication is one of three processes that keep patient data consistent, and each answers a different question.
| Process | Question it answers | When it runs |
|---|---|---|
| Patient matching | Which patient does this record belong to? | At discovery, retrieval, and ingestion |
| Record deduplication | Do these two clinical records describe the same fact? | At ingestion, always on |
| Source categorization and deduplication | Does your own system already hold this record? | At view time, opt-in |
How Deduplication Works
Deduplication happens at two levels.
- Documents: When documents retrieved from a network have already been received for a patient, Health Gorilla keeps the earlier copies and doesn’t store the repeats. Repeats are recognized by their network identifiers or their content.
- Clinical records: Problems, allergies, medications, immunizations, and other clinical records are compared with existing records of the same type for the patient. C-CDA documents and FHIR resources are converted to a common format first. That way, problems can be matched whether they arrive in C-CDA documents or as FHIR
Conditionresources.
Records are duplicates when the details that identify the clinical fact agree, such as dates and RxNorm or SNOMED CT clinical codes. If compared values conflict, the records stay separate.
What Happens to Duplicates
Health Gorilla doesn’t create new records from duplicates. It keeps the record that’s already stored and reconciles duplicates into it:
- Empty fields are filled. On problems, allergies, medications, immunizations, and other medical-history records, empty fields on the kept record are filled from duplicates. Fields that already have values are never overwritten. On lab, radiology, and other result records, empty fields stay empty.
- Sources are added. Source references from duplicates are added to the kept record, which lets one entry cite several originating documents. Provenance resources in the FHIR API identify those documents.
- Duplicates are discarded.
Conflicting records are never merged. Similar entries that appear separately mean they differ in the underlying records. It isn’t a display problem.
Records Health Gorilla imports from your own system and
records from external networks are matched the same way. When their details agree, they’re duplicates, and the kept record cites both.
Records Entered by Hand
Records entered by hand, through the web app or the FHIR API, are compared only with other hand-entered records. New records that duplicate existing hand-entered records aren’t stored again. Hand-entered records are never reconciled with ingested records.
Testing Deduplication
Test deduplication against FHIR API responses, not just the Patient Chart. Some stored fields don’t appear in the Patient Chart, and entries that look identical on screen can differ underneath.
Hand-entered and ingested records stay separate. If you enter records by hand that match ingested records, both appear in the Patient Chart and in FHIR API responses.
Health Gorilla’s internal record identifiers play no part in deduplication. They reflect how data arrived and can change when data is reprocessed. Records are matched on clinical content instead.
For the full specification, including the fields compared for each record type, contact Health Gorilla.
Updated about 3 hours ago

