ClinScribe

There are 3 big issues for the current clinical procedure in the UK:

1. Clinicians spend 1/3 of their workday on documentation, away from actual clinical duties.
2. Most notes are recorded post-consultation, increasing the surface area for human errors.
3. Writing notes during consultations decreases the presence with patients, harming the quality of the interaction.

During the UCL FHIR hackathon 2024, we created a solution to ease this burden, leveraging FIHR, a global standard allowing healthcare data interoperability. In response to these problems, we've developed a solution that automatically transcribes clinical consultation audio and generates artefacts such as clinical letters or FHIR queries to update the EHR system. Our solution leverages the Azure stack with an App Service orchestrating the data flow between the user and different components of the system.

1. Upon uploading a consultation recording, the Whisper Model from the Azure OpenAI Service is utilised for transcription generation.
2. This transcription is then transformed into FHIR-compatible JSON by the GPT4-Turbo model.
3. Finally, after the clinician confirms the content and format, the output is dispatched to a FHIR-compatible database. This makes our solution EHR-agnostic and ready to integrate with different EHR products.