.jpg)
A Zoom-based application designed for healthcare providers to seamlessly transcribe and generate SOAP notes during virtual appointments.
Ux designer
Case Study
Internship
.jpg)
Healiom is an AI-driven healthcare logistics platform designed to enhance medical capacity and improve efficiency in healthcare systems, with a mission to enhance accessibility by connecting patients with providers.
As an intern with Healiom, I was tasked with creating a ZOOM application for healthcare providers to streamline the generation of SOAP* notes with AI assistance during virtual patient interactions.
*SOAP (Subjective, Objective, Assessment & Plan) - a standardized method of documenting patient information.
● Incorporation of AI suggestions into the UI.
● Medical transcription, summarization, and SOAP note capabilities.
Healthcare providers need to create SOAP notes for each patient encounter. These notes must be completed accurately and quickly to ensure proper patient care, provider payment, and malpractice protection.
Soap note building is tedious and time-consuming. Physicians nationwide express dissatisfaction with the process and seek a more efficient, streamlined solution.
The American Medical Association even highlighted several usability challenges with existing EHR (electronic health record) systems where SOAP notes are collected:
● Cluttered visual displays.
● Information stored or entered incorrectly.
● Lack of or unclear alerts.
● Difficulty locating functionalities within complex interfaces.
The current SOAP note generation process is lengthy and tedious, exacerbated by unintuitive user interfaces and physicians' time constraints.
The most widely used EHR systems for collecting SOAP notes are Epic, Cerner, and Athenahealth. These are the industry standards for taking SOAP notes, but where do these applications fall short, and what can we learn from them?
Upon exploring these platforms, similar to the research conducted by the American Medical Association, it was found that most:
● Had cluttered and outdated interfaces.
● Required many clicks to input a note.
● Several features are buried or difficult to locate.
For instance, a quick look at Epic's EHR system shows a visually overwhelming and cluttered interface. With the SOAP note taking process itself already being long and tedious, these UX/UI issues exacerbate the problem even more.
To better understand the problem space, healthcare providers were interviewed to gather insights into their experiences with creating SOAP notes.
Through these discussions with physicians, several core pain points were identified:
● SOAP notes are tedious. When patient volumes are high, both quality of notes and amount of time spent with patients can be effected.
● Rapid documentation in high-pressure medical setting can cause increased concerns for errors and burnout.
How can we design a Zoom platform to streamline the creation of SOAP notes for medical professionals, ensuring both efficiency and accuracy while prioritizing user experience?
With insights gathered, user personas were created to help empathize with future users of the Healiom Zoom App.
.svg)
Dr. Smith is a family medicine physician who often meets with patients over Zoom. She wants to have more time for direct patient care and spend less time on paperwork.
Finds current SOAP note creation processes time-consuming and tedious due to non-user-friendly UI's.
A product that can help speed up her SOAP note-taking process to allow her to spend more time with her patients.
.svg)
Dr. Patel is a neurologist who often meets patients over Zoom. He wants to ensure his patients receive the best care and ensure accuracy in his SOAP notes.
Finds it challenging to create detailed and accurate SOAP notes during fast-paced virtual consultations.
A solution that provides accurate documentation and reliable AI support to accelerate SOAP note creation.
User testing was conducted after high fidelity wireframes were created and prototyped. During product testing, users were asked to complete a series of tasks while verbalizing their thought process and feedback.
Such tasks included, "Accept the generated summary and then edit it" or "Delete the generated input you previously selected and then write your own."
The layout with the text box and AI suggestions underneath created excessive vertical height, leading to frequent scrolling.
The textbox and generated suggestions are combined into one. Upon hover, users can choose between accepting the generated suggestion or deleting it to type their own input, reducing scrolling and visual clutter.
1. A user noted that with this setup, if there are many tags, some might be overlooked.
2. Users wanted better clarification that these are AI-generated suggestions.
1.
Tags were rearranged into a vertical stack for improved readability. Medical codes were included for better organization and reference.
2.
A header was added to the AI suggestion box for clarification.
3.
A drop-down option was added to allow users to see all suggestions at once for easier viewing.
.png)
A tester suggested that some may inherently distrust AI-generated results. They asked how we can make the AI appear to be "thinking" more to generate results that seem more reliable.
To instill more trust in the generated results, the summary loading is broken up into more steps, showing the thought process the AI goes through in generating a summary, creating a more trustworthy experience for the user.
.png)
.png)
.png)
Ensure order and input accuracy and expedite entry with generated suggestions. Spend less time inputting data and more time with patients.
This project offered valuable learning experiences that I can apply to future projects.
1. Never underestimate the value of engaging with users during the design process.
I started this project with no prior knowledge of the medical industry, but talking with healthcare professionals allowed me to gain a better understanding of my target users and the problem space.
2. It is important to be flexible and understanding when working with developers.
Due to time constraints and feasibility concerns, some features may need to be reevaluated or redesigned to ensure they can be implemented effectively and efficiently.
