Resolve Mapping Conflicts

Find and resolve cases where the same product (identified by ma_number/product_id) is mapped to different RxNorm concepts in ema-to-rxnorm.tsv vs latvia-to-rxnorm.tsv.

When to Use

When asked to find or fix conflicts between EMA and Latvia RxNorm mappings, or to align mappings across both sources.

Prerequisites

Refer to the find-concepts skill for searching RxNorm concepts, and the map-drugs skill for general mapping principles.

Step 1: Find Conflicts

python3 skills/resolve-conflicts/scripts/find_conflicts.py [limit]

• •With no arguments: prints all conflicting EMA folders
• •With a number argument: prints only that many folders (useful for batch work)

Conflicts are grouped by EMA folder since all ma_numbers in a folder share the same mapping.tsv and should be resolved together. Output shows for each folder:

• •Product name and conflicting ma_numbers
• •The EMA folder path and all Latvia folder paths containing mapping.tsv files to edit

Step 2: Resolve in Parallel

Launch one subagent per conflict group using the Task tool so they run in parallel with isolated context. Each subagent resolves one EMA folder group independently.

The prompt for each subagent should include:

• •The EMA folder path and all Latvia folder paths from the script output
• •Instructions to follow the resolution steps below
• •The map-drugs, find-concepts, map-ema-drugs, and map-latvia-drugs skill files as context (read them and include their contents in the prompt)

Example:

Task(subagent_type="general-purpose", prompt="""
Resolve mapping conflicts for <product_name>.

EMA folder: <ema_folder_path>
Latvia folders:
- <latvia_folder_path_1>
- <latvia_folder_path_2>

<include skill contents for map-drugs, find-concepts, map-ema-drugs, map-latvia-drugs>

Follow the resolution steps:
1. Read the EMA mapping.tsv and all Latvia mapping.tsv files
2. Read source data (parsed_data*.tsv for EMA, info.txt for Latvia)
3. Determine the correct mapping for each conflicting ma_number
4. Update the less accurate mapping.tsv file(s)
""")

Launch all subagents in a single message so they run concurrently.

Resolution Steps (for subagents)

Investigate

1. •Read the EMA mapping.tsv to see all mappings in that folder
2. •Read each Latvia mapping.tsv to see all mappings
3. •Read source data — parsed_data*.tsv (EMA) and info.txt (Latvia) to understand the actual product (strength, form, pack size)
4. •Compare the mappings for each conflicting ma_number — determine which concept is more accurate, or whether both are wrong

Common Conflict Patterns

Different strengths or dose forms chosen Read the source data carefully. The correct concept should match the actual strength and pharmaceutical form of the product. Use find-concepts to search if needed.

One side is BROAD, the other is EXACT Prefer the EXACT mapping if it accurately represents the product. If neither is exact, both should be BROAD with the same concept.

Branded vs unbranded concept When the brand name is the same in EU and US (e.g., Tobi, Erivedge, Lantus), always prefer the branded RxNorm concept — it is more precise. Only use unbranded concepts when the EU and US brands differ. See the map-drugs skill for full branding rules.

Both mapped to valid but different concepts Investigate which concept better represents the product. Consider:

• •Does the concept match the strength exactly?
• •Does the dose form match (e.g., Pen Injector vs Injectable Solution)?
• •Is one correctly branded and the other unbranded? Prefer branded when the brand matches across EU/US. Both mappings should converge on the same concept_id.

Update

Edit the mapping.tsv in the folder that has the less accurate mapping:

• •Update concept_id, concept_name, concept_code, mapping_type to match the correct mapping
• •Update last_updated_date to today's date
• •Add a comment noting the alignment if helpful

Important: Only edit the mapping.tsv in the product-level folder (the source of truth). The combined ema-to-rxnorm.tsv and latvia-to-rxnorm.tsv files are regenerated from these folders.

Step 3: Regenerate Combined Files

After all subagents complete, regenerate the combined TSV files from the updated individual mapping.tsv files:

python3 skills/resolve-conflicts/scripts/regenerate_combined.py

This runs both combine_tsv_files.py (EMA) and combine_latvia_mappings.py (Latvia) to rebuild ema-to-rxnorm.tsv and latvia-to-rxnorm.tsv.

Step 4: Verify

Re-run the conflict finder to confirm the conflict count decreased:

python3 skills/resolve-conflicts/scripts/find_conflicts.py