OCL content automatch discussion

Sep 30, 2024, @Michael Bontyes , @Andrew Kanter

Context

• Discuss how the automatch initiative could be helpful to other organizations than MSF/Madiro:
  https://github.com/MadiroGlobalHealth/clinical-content-tools

• Follow-up on session at OMRS24 in Nairobi:
  https://docs.google.com/document/d/1-dT0tiICLwr-njzCzX4F7lJyqMrd4eevcQbTy0lhni4/edit#heading=h.ynlgiqnz2612

General questions

• Everyone’s challenges/bottleneck?

• Current process?

• What would help the most?

• How to handle synonyms?

What’s Next?

• Update documentation with required columns in Excel template

• Mapping support?

• OpenMRS or OCL integration? How?

• Improve automatching algorithm - currently using Fuzzy search with Levenshtein distance

  • Adding other field in calculation such as data type and class

  • Adding context such as form or program

• Combine with AI like embedding-based search (meaning instead of characters)

• Viewing the top 10 results as an intermediary step would be useful like https://github.com/OHDSI/Usagi

  • Viewing the concept name, score, and mapping help to decide on which one is the right one

  • Example: AMPATH:1067, CIEL:1067, HL7-MedicationDispenseStatus:unknown, org.openmrs.module.emrapi:Unknown Cause of Death, PIH:unknown, PIH:1067, SNOMED CT:261665006