Problem Space 

With our solution, we aim to target the problem of physician burnout related to electronic health and medical records (EHR). In our visit to the ICU, we observed that when nurses are performing documentation and charting functions, they fill out many similar forms - where information is replicated across multiple fields. This is primarily due to the fact that the EHR is designed for insurance and billing purposes, rather than a clinician’s natural workflow. This unnatural and repetitive process adds to physician burnout, and presents a significant documentation burden. According to the AMA, for every hour spent with a patient, 2 hours are spent with the EHR. Thus, with our NLP-based solution, we aim to reduce staff frustration, increase efficiency of clinical documentation, and facilitate more time for patient care. 

Solution Space

For our solution, we aim to create a documentation assistance tool which utilizes NLP.

Features 

• Notes Analysis: NLP would analyze a healthcare professional’s unstructured clinical notes and connect directly to the EHR to fill in the appropriate fields

• Information Porting for Reduced Redundancy: tool ports the appropriate information to all of the fields, instead of clinicians having to repeatedly input replicated information

• Instant Feedback: tool provides feedback to users if the documentation doesn’t achieve clinical standards or is missing necessary follow up information

• Mixed-Initiative Learning: before the information is added, the user can review the information and provide feedback if necessary

Similar Solutions

Currently, there are similar technologies, such as digital scribes which take the audio of patient-healthcare professional interactions and convert them into clinical notes- namely DAX Copilot by Nuance. Such tools are convenient for the stages where a transcription is needed- but many professionals may prefer typing their own notes for accuracy and speed. Also, our tool more specifically targets documentation at a later stage in patient care - specifically EHR updating and charting. 

Solution Testing

The development of our solution consist of a few different phases (conceptualization, design, prototyping, testing/feedback). 

Conceptualization

In the conceptualization phase, we will primarily rely on input from nurses. Nurses face many data input tasks in their administration and monitoring capacities. We plan to ask them if they believe such a tool would provide assistance in their workflows and ask them which charting tasks take the longest so we can further brainstorm what to integrate into our solution to reduce nurses’ documentation burden. We will also ask them for feedback on our existing planned features. Finally, we can ask what potential barriers they foresee to implementing such a tool in a clinical setting. From this initial research, we can finalize the format and functionalities of our solution in order to best address the needs of clinical professionals. 

Prototyping

• Initial Prototyping: create design mockups to show the general layout of our tool as well as its basic functionalities

  • this is after refining our solution through discussion with nurses 

  • could conduct some UX surveys on these mockups

• Functional Prototype:  includes software development and backend support

• Final Testing and Feedback: ensure our planned features are working as expected and iteratively improve it based on feedback provided by healthcare professionals

  • multiple rounds of testing and feedback

Our Team

We are both interested in research applications that involve AI in healthcare.