Curai Health

Curai is a startup with a mission to make primary care accessible to all by providing truly affordable telehealth. Its goal is to build an AI-assisted patient servicing platform that merges messaging capabilities with its own electronic health record system (EHR).  With the help of AI, healthcare providers would be able to serve many more patients per day, each with higher quality of care.  

Goal

Define, design, and build an AI-Assisted EHR and patient servicing platform.

Process

When I first joined Curai, the EHR was in a pre-MVP state with some basic UI to enable primary care providers to chat with the patients while training the AI algorithm. Most of the medical operations were handled in spreadsheets and docs, or a 3rd party messaging app. 

The following was my process to get the product from pre-MVP to MVP, and even North Star:

Visioning

As a fully remote company, I led virtual visioning workshops with the product manager across each group of stakeholders. The goal was to define the vision of the EHR, and what functionality and workflows were required to get us to that goal.

The vision for the Curai EHR is to:

“Create an AI tool that enables each medical staff to provide online care to 10,000 patients per year.”

The stakeholder groups were:

1. Primary care physicians
2. Medical Assistants
3. Clinical Assistants
4. Leadership team
5. Product team
6. Engineering + AI team
7. Patient product team

For each stakeholder group, we brainstormed around the six focus areas of the EHR, corresponding to each step of a medical provider’s user flow:

1. Reviewing patient records
2. Video & chat
3. Taking notes
4. Staff collaboration
5. Patient education, monitoring, & adherence
6. Partnerships

Feature Prioritization

After sorting through results from the visioning workshop, we could then prioritize the features to determine what is the north star, and what could be mvp, and how we could phase development given limited engineering resources.

Information Architecture: The Key to intuitive navigation

EHRs contain a massive amount of data. In addition, this EHR also needed to provide messaging, chat, and video/audio call functionality. The IA process is where we figure out how this information can be organized in a human-navigable way. 

User Flows: Understanding medical Ops

Since the medical team has particular protocols for working with patients, each other, and 3rd parties like insurance and labs, we needed to understand how they work by mapping out the user flows for each role—primary care doctor, medical assistant, and clinical associate. All of them use the tool in different capacities. The user flows were mapped out with the product manager who is also a primary care physician, along with the medical operations team.

Wireframes: Defining Layout and Functionality

The wireframe stage allowed us to explore different UI’s and interaction models. Ultimately, we decided on a 3 panel layout, with AI and other smart features integrated throughout the tool. That way, medical staff can simultaneously review patient info, communicate with the patient, and take notes all at the same time.

We designed a UI skeleton that would work for the north star, then created a pared down version for the MVP. With this approach, we preemptively designed a responsive UI that would allow the MVP to grow incrementally into the north star with each iteration, rather than having to redesign the UI each time. We also spec'd out the layout and conditional logic required for each of the user flows above. 

‍

Dashboard Features

• Intuitive navigation to navigate to different tasks and active patients
• Dashboard that shows patient queue and the key information medical providers need at a glance, such as lab results, active prescriptions, file attachments, etc.
• Smart sorting of patients to prioritize and delegate patients to staff they’ve spoken to previously.
• Dashboard with preview of patient info and incoming notifications.
• Functionality defined for different roles

Patient Profile Features

• 3 panel layout allows medical staff to simultaneously review patient records, message the patient, and take notes at the same time
• Keyboard shortcuts and syntax allow users to pull up information and features quickly without moving their hands off the keyboard.
• AI generated summary directly from the chat or video call, with medical staff capable.
• Find where in the chat or conversation a specific note refers to
• Take actions directly from the chat. For example to add something as a task
• AI-assisted autofill in the chatbox with suggested actions based on patient record and condition.

Visual Design: Ensuring Usability

After sorting out the layout and functionality in the wireframing stage, the visual and interaction design stage is where we specced out the details of each feature’s UI.  In the case of the EHR, which has an immense amount of information and different types of users, patients, and medical information. Visual design is key for usability. Designing each feature down to the pixel was key to ensure that the medical staff can quickly understand the patient context. A key strategy is to include easily recognizable iconography and styles to highlight what the provider needs to see.

For example, it was important for the medical staff to quickly identify if a patient was urgent, had a lab result, active prescriptions, etc. These were translated into icons that the doctor could easily identify. The staff also had to quickly identify the source of the patient information, whether it came from themselves, the patients, another staff member, or AI generated. They also wanted to quickly grasp which conditions the patients have currently versus previously. In the visual design, we explored how to highlight these important facts without overwhelming the UI. See below for examples.

Design Systems

Once the visual design direction established, the design team composed a design system in Figma and Storybook to ensure smooth implementation that is scalable for the long run.

‍