Seven in ten could read the label. One in three could take the right dose

Description and use

Concord creates accurate medication cards with pictograms and multilingual text, verified by two independent models to reduce dosage errors.

Problem

Patients often read medication labels correctly but still take the wrong dose, leading to medication errors.

Outcome

Concord produces accurate, pictogram‑based medication cards in 69 languages for 1,830 WHO‑classified medicines, with 98% of names matching the original and no miscounted doses.

Workflow

• 1. Generate a pictogram for a single dose using an image model.
• 2. Two independent models (from different companies) examine the image and describe its content in their own words.
• 3. If both descriptions agree on the content, the image is accepted; otherwise it is discarded and regenerated.
• 4. Repeat step 1–3 for the required number of doses to create a complete card.
• 5. Generate the medication name and dosage text in the patient’s language using a language model.
• 6. A second language model, never exposed to English, reads back the text to verify correctness.
• 7. Assemble the final card with the verified pictogram and multilingual text.
• 8. Deliver the card to the patient as a visual aid.
• 9. (Future) Conduct a comprehension study with patients and pharmacists.

Evidence

Among patients with low literacy, 71% could say the instruction back. Only 35% could count out the right number of pills

Source

In a randomized trial with 245 caregivers, picture instruction sheets cut the share of parents who measured a child's medicine wrong from 47.8% to 5.4%

Source

I tried it for a hackathon. I asked for one spoonful of medicine and got eight spoons. Across 168 pictures of whole doses, the model drew the wrong number of pills 83 times.

Source

I ran that on every prescribable medicine with a WHO classification code: 1,830 medicines in 69 languages, 126,270 names. 98% came back as the same medicine.

Source