Documentation has been a significant problem for healthcare. On the physician side, documentation decreases time available for patient care, limits the efficiency of practice, and causes quick burnout. On the patient’s end, stressed and overwhelmed doctors mean lower quality care. The current solution is to hire scribes, which is both costly and limited, to help lessen the workload. However, this is not a viable choice for many physicians. We wanted to design a cost-effective, scalable solution that would not only decrease the rate of physician burnout, but also increase physician efficiency and quality of patient care.

What it does

Scribr is a web application built to address the need for an efficient method of filling out electronic medical documentation. Physicians and patients will engage in conversation while Scribr is recording, and the application will convert their speech into text. This text will then be run through an algorithm to identify entries for fields in electronic medical record forms, and populate those fields with the entries. The end result is an online copy of the filled in form, ready for use soon after the conversation finishes.

How we built it

We split the problem along the traditional server and client lines for organization, and also for ease of access to the Google Cloud Platform through our backend server. We used Javascript through Node.js on the server side to handle communication with Google Cloud Services, our own medical data form, and a React.js middleman to facilitate the data exchange between them.

Challenges we ran into

Google Cloud Platform uses very particular input formats, and the correct format had to be determined via iterative experimentation. We also had some issues getting audio files transmitted properly between client servers.

Accomplishments that we're proud of

We were very proud to have been able to navigate through the Google Cloud Platform system for the first time, especially given its stringent authentication processes and foreign input formats.

What we learned

On the medical side: Through consultation with physicians and scribes here at MedHacks, we learned more about the role of scribes and what physicians would like from our application. These talks confirmed the need for more efficient documentation, and gave us helpful insight on areas to focus our efforts. As a result, we designed our application to significantly decrease the need for manual documentation. We also learned the components of a typical electronic medical record as we worked to integrate our application with one.

What's next for Scribr

We were limited in what we could accomplish during this hackathon by the time limit. Given more time to develop the application, we’d like to fully integrate natural language processing (NLP) algorithms into the application to be able to more effectively identify more EMR fields in more natural conversation. We also had several issues with accuracy in the speech-to-text application on the Google Cloud platform, so we'd like to use an improved application We’d also like to integrate our system to existing EMR systems to bring our technology to as many clinical settings as possible.

Share this project: