Hundreds of thousands of physicians, primary care physicians in specific, currently face severe burnout in the US. Physician burnout increases US healthcare billings by $4.6 billion every year. It contributes to more medical errors than unsafe medical conditions and is one of the leading causes for hundreds of thousands of medical error-related deaths. One of the primary factors of physician burnout is physician electronic health record (EHR) data entry during and after their shift. This can take up to 1-2 hours of a physician's time nearly every night, taking away from valuable time with family, friends, and other activities. Manual EHR data entry and, inherently, physician burnout negatively impact physicians, hospitals, and patients.

What it does

Our solution seamlessly automates the manual process of physicians typing and updating patients' EHRs. The physician only has to scan the traditional paper record and upload the pdf to Uzima takes the scanned pdf, the pdf gets sent over to the UiPath framework, puts it into OCR, grabs the data to send back to the user, and outputs a structured, actionable dataset that can be used to automatically fill EHR forms and other applications.

How we built it

Flask backend, bootstrap/JQuery frontend, UiPath integration via HTTP API/Orchestrator (in production. In prototype, manually accessed), accessed Google Cloud vision via api, and used UiPath to automatically fill in legacy EHR software forms to increase data portability and accessibility.

Challenges we ran into

Initially, we did not utilize UiPath to its maximum potential. Once we started using and implementing UiPath to our solution, we began to leverage the various features of UiPath.

What we learned

  • Learned UiPath and implementing OCR
  • Learned Bootstrap, Flask, and Jinja for full-stack web dev
  • Fully functioning prototype built from scratch
  • Learned standard industry tools like git

What's next for Uzima

  • Standardization of medical data
  • Move towards EHR interoperability
  • Further analysis/parsing of medical data
  • Actionable datasets for various applications
  • Academic research datasets

Built With

Share this project: