Tutorial

Step 1: get the code

git clone https://github.com/electronic-patient-record/electronic-patient-record.git
cd electronic-patient-record/electronic-patient-record-matcher-with-rust

Step 2: create two records

Patient matching assesses whether two records refer to the same person. Create a file called a.json:

{"names":[{"family":"Rivers","given":["Morgan"]}],"birth_date":"1980-02-14"}

Create a file called b.json with a typing error in the given name:

{"names":[{"family":"Rivers","given":["Morgen"]}],"birth_date":"1980-02-14"}

Step 3: run the matcher

cargo run -- a.json b.json

The output contains a decision and a result for each field:

{ "score": 0.98, "decision": "auto_link", "fields": [
  { "field": "family_name", "similarity": 1.0, "note": "similarity 1.00" },
  { "field": "given_name", "similarity": 0.93, "note": "similarity 0.93" },
  { "field": "birth_date", "similarity": 1.0, "note": "similarity 1.00" },
  { "field": "sex", "similarity": null, "note": "sex unknown on one or both records" }
] }

The output above is shortened. sex is missing, so its similarity is null and it is excluded from the score.

Step 4: remove a field

Delete the "birth_date" field from b.json, keeping the JSON valid, and run it again. The names still match and the score remains high, but the decision changes to review, which requires a person to check it. conflicts gives the reason: birth date not available to confirm the link.

These thresholds have not been validated.

Step 5: run the tests

cargo test

Step 6: next steps

• Clinician interface: in electronic-patient-record-ui-with-svelte, run pnpm install then pnpm dev. It needs Node 26 and pnpm, and the API for live data.
• The API: it enforces login, so it needs an identity provider. The README has the commands, and Keycloak for development has the rest.
• The specification: start at the index.
• Contributing: see about.