Identifier mapping, exact-name matching and pgvector similarity support medicine-to-BSO coding for Northern Ireland pharmacy workflows.
Tranz / Engineering in practiceStaff searched a large BSO medicine codebook to identify and enter the code for each prescription item. Repeating this work across prescriptions consumed time, with part-time staff used to help handle the workload. Endorsements and amendments added further administrative work.
We built a medicine-matching database and a three-stage algorithm: identifier mapping, exact-name matching, then vector similarity. The workflow supports coding from prescription information, including handwritten prescriptions, as reported by the delivery team.
Start with scanned pack information, resolve the dm+d identifier and use stored AMPP or VMPP mappings to find the linked BSO record.
If identifier mapping does not resolve the item, compare the prescription medicine name with stored AMPP and VMPP names linked to BSO codes.
If exact matching does not resolve it, use Sentence Transformers and pgvector to find similar medicine records.
Surface the mapped code or candidate match for the coding workflow. Similarity alone is not evidence of clinical equivalence.
The implementation links dm+d identifiers and medicine names with BSO code records. AMPP and VMPP describe actual and virtual medicinal product packs in the NHS dm+d model; AMP and VMP describe products without the pack level.
Prescription text may not be an exact match to a catalogue name. Vector matching uses Sentence Transformers to represent text and pgvector to compare it with stored records, providing a fallback for the handwritten-prescription workflow described by the team.
Pharmacies also need to understand whether reimbursement matches what they should receive. Alongside coding, the system identifies payment discrepancies for the pharmacy team to investigate, improving visibility into reimbursement and reducing manual checking.
BSO’s Northern Ireland guidance covers item coding, supporting endorsements and use of current codebook information. This case study focuses on medicine-to-code matching within that wider administrative workflow.
Concept layout · not a production screenshot.
Example content only.
Implementation and workflow outcomes reported by the Tranz delivery team.
Three-stage matching connects scanned identifiers, exact medicine names and vector candidates to stored BSO records.
Handwritten prescription support and automated lookups reduce staff effort spent searching and entering medicine codes.
Reconciliation identifies reimbursement discrepancies for investigation, helping pharmacies check whether they have been paid correctly.
BSO is the Business Services Organisation in Northern Ireland. Its coding guidance describes the codebook, current codes, quantities and supporting endorsements. This is a Northern Ireland reimbursement workflow, rather than a single coding process for the whole UK. The workflow connects scanned pack information with stored medicine mappings, then uses name and similarity fallbacks when needed.
Bring us the handoffs, document entry or verification steps that need attention. We can explore the integrations, review points and evidence needed for a useful implementation.
Discuss your pharmacy workflow