A prescription is useful only when a person can take the next practical step: find a pharmacy that can help, near enough to reach.
The application reads prescription information and helps identify suitable nearby pharmacies against the medication requirements and the information available.
The problem
Handwriting and print are hard to read, medication names have many spellings, and calling pharmacies one by one is slow. A person with a prescription needs to know where to go and what to confirm before they travel.
What the system does
- Extraction of candidate medication text from an authorised prescription image or document.
- A confirmation step where a person fixes ambiguous or low-confidence text.
- Matching against pharmacy information and location, with results ordered by relevance and distance.
- A clear statement of what still needs direct confirmation, such as whether a pharmacy has an item today.
A representative journey
- Step 1: Provide a prescriptionThe person uploads or photographs an authorised sample. Nothing is shared with a pharmacy at this stage.
- Step 2: Extract candidatesMedication lines are read out of the document as candidates, each with a confidence.
- Step 3: Confirm unclear textA person confirms or corrects the ambiguous lines before they are used for matching.
- Step 4: MatchConfirmed medication names are normalised and matched against pharmacy information.
- Step 5: Show nearby resultsPharmacies are listed with distance and what is known about the match.
- Step 6: Confirm directlyThe result states what to check with the pharmacy before travelling.
Extraction · fully synthetic sample
1 line to confirm- Medicine A 500 mg, twice dailyHigh
- Medicine B 10 mg, nightlyHigh
- Medicine ?C 5 or 50 mgLow · unclear digit
A person confirms the unclear line before anything is matched.
Nearby results · fictional pharmacies
Pharmacy One 0.6 km
Carries 3 of 3 names; availability not confirmed
Pharmacy Two 1.4 km
Carries 2 of 3 names; availability not confirmed
Before you travel: confirm directly that the pharmacy has each item today.
Working prototypeUse a small working version of this on invented data, the way a person would.
Try the prototypeInspect the system
This anatomy is a proposed description of how such a system is put together. It is not an audit of a specific deployment.
Inspect the system
How Prescription-to-pharmacy discovery is put together
- What it is
- Secure intakeReceives an authorised prescription sample.
- What flows through it
- An image or document goes in. A protected upload comes out.
- The decision
- Personal data is handled minimally and kept apart from search logs.
Read the whole system as a list
- Secure intake. Receives an authorised prescription sample. Flow: An image or document goes in. A protected upload comes out. Decision: Personal data is handled minimally and kept apart from search logs.
- Extraction. Reads medication text from the document. Flow: A document goes in. Candidate lines with confidence come out. Decision: Low confidence is shown, not hidden.
- Confirmation. A person checks ambiguous text. Flow: Candidates go in. Confirmed medication names come out. Decision: Ambiguous lines are corrected by the person before matching.
- Matching and search. Matches confirmed names against pharmacy information. Flow: Medication names and pharmacy data go in. Ranked candidate pharmacies come out. Decision: Matching is based on confirmed data and explains why a pharmacy was suggested.
- Geospatial results. Orders and shows nearby pharmacies. Flow: A location and candidates go in. A distance-ordered result list comes out. Decision: Location is requested explicitly and used only for this search.
Key decisions
A person confirms the text
A wrongly read medication name is a bad starting point for any search. Ambiguous lines go to the person, not into the matcher.
Discovery, not dispensing
The application helps someone find a pharmacy. It does not diagnose, substitute a medicine or make medical decisions.
State freshness honestly
Availability changes during the day. Results say how current their information is and what to confirm directly.
Outcomes and limits
We publish no measured outcomes for this project, because we have none we can evidence. The description above is of what the system does, not of results it has produced.
- The OCR provider, medication normalisation, pharmacy data source, ranking criteria and update path are not specified here.
- No live stock integration, prescription accuracy or guaranteed availability is claimed.
- It does not diagnose, dispense or suggest substitutions. All examples are fully synthetic.
Guardrail
A person confirms the extracted text, and every result says what to confirm with the pharmacy. No diagnosis or substitution.
Technical detail
Technology named in the portfolio for this kind of system. Which parts a given engagement used is not stated.
- OCR / document extraction
- PostgreSQL + PostGIS
- Mobile and web client