ChatMD - Medical AI
ChatMD - A Virtual Medical Assistant
Internship Project | Avanade
Project Type: Product Design, Conversational Design & Research
Role: Product Design, User Research
ChatMD is a prototype project for a medical chatbot. This virtual assistant came to be after two weeks of user research around the current healthcare system as well as current artificial intelligence solutions. ChatMD is designed to assist patients with their medical needs and would allow users to search symptoms, schedule appointments, request prescriptions, contact their doctor or nurse, or to connect to external health wearables. ChatMD comes in a chatbot format that allows users to either type or speak to interact. It was important to us to give users the option to interact via conversational UI so that the virtual assistant interaction seemed more natural. ChatMD would be integrated in existing medical patient portals and would also function as a standalone mobile app. The primary challenge for our team when creating ChatMD was to create a medical virtual assistant that would make medical care more accessible, and the healthcare process more seamless.
For this project, I was able to do both user research and user experience design. The role was unique for me on this project because the UX design included designing for conversational AI as well as mobile. This posed a whole new set of challenges, as we had to think about how users might interact with the virtual assistant using voice, and make design decisions based on that research. Everyone on the team took on multiple roles, with each person conducting user research, interviewing, and testing as well as design. We worked together to consolidate our individual research and brainstormed together to come up with the user flows and chat dialogues for ChatMD. I had a large part in designing the chat flow and designing the internal interfaces that connected with ChatMD.
Initial Research & Interviews
The ChatMD project began with conducing some comparative analysis research and user interviews. First, everyone on the team did research on existing virtual medical assistants and how they are currently impacting the healthcare industry. We found many interesting results, but ultimately decided that a chatbot would be the most viable prototype that could be used by the greatest number of people. We found some similar medical chatbots but they offered limited functionality. Some offered robust symptom searches, but lacked the ability to schedule appointments or link up with current medical records. Others allowed these features but required payment and lacked other options.
After our comparative analysis research, we each went out into the world to interview potential end-users. We each wrote an interview script in order to get the most relevant information possible. I made sure the questions in mine were not leading questions, and that they had to do with technology usage, healthcare experiences, and remote communication with doctors. I practiced this survey with volunteers before going out to the public. Overall, my in-person interviews went well. I only had 5 respondents, but they were very diverse in age, gender, and background. Our team got around 25 participants total.
After we each conducted our interviews, we brought all of our key findings together. We found that many people have positive doctor's office experiences, but they find it challenging to figure out when or whether they should go to the doctor based on symptoms. People expressed an interest in a symptom checker they could trust that could help them decide if they need medical attention. Our respondents also repeatedly mentioned the need for a streamlined scheduling process. Some people said they had to wait weeks to get an appointment, or had to wait on hold for long amounts of time to schedule anything. The people that had EMS medical portals through their healthcare providers said that those were very helpful. When asked if they would be interested in using a chatbot to schedule appointments or search for symptoms, most respondents said yes, if it was reliable. From these key findings, we created user flows for scheduling and symptom checking functionality.
Part of making ChatMD a "reliable" chatbot included writing a chat dialogue that was conversational, friendly, and informative. We wanted the voice of our chatbot to sound more friendly and conversational than others we found in our initial research. We created chat dialogue flows for symptom checking, appointment scheduling, patient follow-up, prescription refill, and contacting the doctor.
Based on our user research, we decided that symptom checking was one of the top use cases for ChatMD. Because of this, we wanted to create an interface that worked within our chatbot to allow users to check their symptoms and get a list of possible causes, as well as a suggestion of whether and when to go to the doctor. We knew that we could not have our chatbot actually diagnose future patients for legal reasons, but we wanted to provide our end-users with the option to see what their symptoms might mean, and whether they were serious.
While designing ChatMD, we thought it would be most effective if it could be integrated with existing patient portals, as well as be a standalone mobile application. We designed some wireframes to showcase how ChatMD would integrate with its digital environment. In the medical patient portals, the chatbot would be collapsable, and would not cover up any of the page when expanded. In mobile form, the chat would be a separate entity from the home page, but would be easily navigable.
We decided to make prototypes using botsociety.io to create what the chatbot might look like. We made a prototype for each main use case and added potential end-user dialogue to showcase the core capabilities of ChatMD.
The first of these prototypes is the symptom checker. This was one of our users' key pain points, so it was important to us to fully flush out the symptom checking experience. Click to play the video below.
The next prototype is for the appointment scheduling process. This was another one of our users' key pain points, so creating a good chat dialogue for scheduling was important. We included an incorporated calendar widget to make scheduling easier for users. Click to play the video below.
The third chatbot dialogue prototype is for ordering prescriptions or getting them refilled. This was not as high a priority as symptom checking or scheduling, but was still a common pain point for those we interviewed. Getting prescriptions refilled can often take a couple days, so ChatMD would help to streamline that process.
The final prototype we created was for the use case of contacting a doctor. If a ChatMD user decided they needed to contact their doctor, we wanted them to be able to do that through the chatbot. Users have the option of calling, texting, or video chatting with their doctor.
After three weeks of researching, interviewing, designing, prototyping, and iterating, my team came up with our final prototypes for ChatMD. These final designs centered around the information and insights we got from our user research. We were able to create a product that would be useful for its end-users, and that could help improve the healthcare experience. The product, although a prototype, does have an intended purpose in the future, and versions of the conversational UI and chat flows are being developed out.
Overall, I think the ChatMD project went very well. We were able to create a working, high-fidelity prototype based on real user research in just a few short weeks. We were able to do plenty of research and interviewing up-front, which helped the rest of the design process go smoothly. If we had more time, it would have been nice to do more interviews and user research. We also experienced some limitations with the chatbot prototyping software we used, but it was a good tool to learn nonetheless. ChatMD ultimately fulfilled its purpose - which was to make medical care accessible. In the future, I would like to see this project, or one like it, developed and actualized. AI and virtual assistants have a great potential to help the healthcare industry serve its patients better.