Medicare Wound Order Qualifier UI
Budget: $250 – $750 USD
I need a small application that can read both hand-written and typed clinical notes, recognise every wound documented, judge whether each one meets Medicare Local Coverage Determinations for surgical dressings, and then tell the user exactly which dressing types are reimbursable.
Here is what the system must do:
• Accept uploads in PDF or raw text. The source may be a scanned hand-written note or a generated EMR export; either way the content has to be extracted accurately.
• Run reliable OCR on handwriting and standard text extraction on typed notes, then feed everything through an NLP layer that can spot multiple wounds, locate their characteristics (location, stage, drainage, measurements, frequency of change, etc.) and keep them tied to the correct patient encounter.
• Cross-check every finding against the relevant Medicare LCDs and return a clear pass / fail for coverage, plus a short recommendation list of dressing codes and usage frequencies that qualify.
• Present the results in a clean, browser-based UI: the original note side-by-side with a structured summary and the recommended surgical dressing plan.
• Allow me to export that summary to PDF so it can be dropped straight into the EMR.
Acceptance will be based on three things: 1) test cases that mix two or more wounds in a single note are parsed correctly, 2) the LCD logic fires with 100 % accuracy on the latest published rules, and 3) the interface stays usable on a standard desktop browser without extra plugins.
If you have experience blending OCR (Tesseract, Google Vision, or similar) with NLP frameworks such as spaCy or TensorFlow, you’ll be able to move quickly on this. Let me know how you would tackle the handwriting recognition and guideline rules engine, and include a rough development timeline in your proposal.
Here is what the system must do:
• Accept uploads in PDF or raw text. The source may be a scanned hand-written note or a generated EMR export; either way the content has to be extracted accurately.
• Run reliable OCR on handwriting and standard text extraction on typed notes, then feed everything through an NLP layer that can spot multiple wounds, locate their characteristics (location, stage, drainage, measurements, frequency of change, etc.) and keep them tied to the correct patient encounter.
• Cross-check every finding against the relevant Medicare LCDs and return a clear pass / fail for coverage, plus a short recommendation list of dressing codes and usage frequencies that qualify.
• Present the results in a clean, browser-based UI: the original note side-by-side with a structured summary and the recommended surgical dressing plan.
• Allow me to export that summary to PDF so it can be dropped straight into the EMR.
Acceptance will be based on three things: 1) test cases that mix two or more wounds in a single note are parsed correctly, 2) the LCD logic fires with 100 % accuracy on the latest published rules, and 3) the interface stays usable on a standard desktop browser without extra plugins.
If you have experience blending OCR (Tesseract, Google Vision, or similar) with NLP frameworks such as spaCy or TensorFlow, you’ll be able to move quickly on this. Let me know how you would tackle the handwriting recognition and guideline rules engine, and include a rough development timeline in your proposal.