Agnosys Health

Product Experience

The Problem Space

Agnosys health is focused on solving access issues within healthcare at a global scale. There are currently more than 4 Million Canadians without access to a family physician. While for patients that have physicians, it takes weeks to months to access them. On average, it takes 27 weeks to access specialist care in Canada.

The overarching issues are that most people in Canada and across the world cannot access the healthcare system when they need it and this has multiple implications.

1. Declining patient health - as patients are not able to access care as soon as they need it most patients wait and get worse before visiting emergency centres
2. Strain on the public care system - the rising number of emergency visits is causing a strain on ED facilities and healthcare workers
3. Rising overall cost of care - governments are spending more to combat declining health situations, replace healthcare workers and expand care facilities 

Problem Background 

The main cause of the access shortage in healthcare is the gap between demand and supply of healthcare providers relative to people who need care. The chart below illustrates the proportion of physicians to patients in Canada and the OECD

While Canada is on the lower end, the average number of about 3.5 physicians to 1000 people illustrates the global challenge. When the concentration of resources around urban areas is factored in, this ratio becomes even worse for marginalised communities.

Logically, there are 2 approaches to solving this issue of access.

The first approach could be to rapidly increase the number of physicians, however given multiple constraints, this has been a global challenge. Illustration of physician growth rate below 

The alternative approach approach is to increase the efficiency of care to improve access and this is Agnosy health’s focus

Research Insights

Last year, Canada’s healthcare wait times reached a staggering 27.4 weeks — 195% higher than the 9.3 weeks Canadians waited in 1993 (Waiting Your Turn: Wait Times for Health Care in Canada, 2022). The result is that patients are getting sicker, and physicians are more overwhelmed than ever.

“Obviously, the additional load that COVID-19 has brought on them with lots of backlogs and surgeries and diagnostics. So yes, it’s a very serious and severe situation now for workers and patients across Canada,” said Health Minister Jean Yves Duclos. 

After interviewing eight participants, we discovered that patients are struggling with more than just long wait times. 

87.5% mentioned feeling hesitant or frustrated about having to travel to a clinic 

60% of participants are unhappy with current online resources 

87.5% would prefer to check in with doctors before going to in-person care 

Physicians interviewed confirmed that 50% - 70% of interactions with patients did not require clinical intervention such as a prescription. Also, physicians (small businesses such as single physician practices) interviewed all stated they would use a service that helps them manage their patients when they are off work - off work hours, vacations.

Both patients and physicians expressed an interest in more online care, as long as it was moderated, authenticated, and overall trustworthy. Many interviewed patients said that they usually like to ask their friends or family members for medical questions first, but others didn’t have such a resource in their life. 

62.5% of participants said they would pay a small fee to communicate with a doctor online 

75% said they were interested in a community aspect to share stories and compare symptoms  

50% said they wanted doctors that could efficiently direct them to specialists or appropriate resources after diagnosis 

Solution Explanation 

Based on our research, it was clear that patients need a way to get quick, reliable answers to their non-emergency medical questions, and doctors need a way to manage large amounts of patients. 

We settled on the following features: 

24/7 Automated Care

Sometimes patients have a medical problem that can be solved quickly with over-the-counter treatment. If patients could consult an AI assistant at home, they can discover if they need to speak with a doctor or just treat at home. The AI is trained to identify concerning symptoms and escalate accordingly. 

Community Discussion Groups

By creating a community discussion board, patients in similar circumstances can compare experience and share tips and tricks. In order to prevent the spread of misinformation or cyber-bullying, all discussion boards will have at least one moderator to keep discussion on track. There are both private and public chat boards, and a doctor must be the one to create the group. 

Patient Database and Secure Messaging

Doctors can add new patients through email. Patients can then add in their medical data and update it as needed. Physicians will have the ability to sort through the patients, see and change upcoming appointments, and talk with their patients directly. This system will help doctors take on more patients and provide care more efficiently.  

‍Lofi & Hifi Mockups

Lo-fi Mockups

Usability Testing and Iterations

Things went smoothly during our first round of usability testing, but there was one glaring problem. Users were struggling to find the community feature. Despite there being two ways to access it (global nav and thumbnail on the landing page), users were getting lost. Our main CTA (interacting with the AI assistant) was so enticing (100% click success rate), users didn’t want to click on anything else.

For the V1 high-fidelity prototype, the article-like presentation for the page community was changed into a button-like card. We hoped the clickable nature of the card would catch the eye and encourage the user to scroll down and explore. 

Usability testing showed that this was definitely an improvement. 

Finally, we swapped the thumbnail image out for something more eye-catching.

We also revamped our homepage to look cleaner. The top blue bar looked too much like a button and took attention away from our main CTA. We added additional context so users could better understand what Agnosys offers. 

Implementation Details

Technical implementation

Where is it hosted? 

The Backend application is hosted on DigitalOcean while the frontend is hosted on Vercel.

What is your tech stack?
We use Django and Django Rest Framework with PostgreSQL for backend development and React, Next.js, and JavaScript/TypeScript for frontend development of our web app.

High level journey of a request?
When a user makes a request to the frontend application at https://communities.agnosyshealth.com/, the frontend server checks whether the user is authorised to access the requested page.

If the user is, the requested page is sent, else the user is redirected to the login page. Once the requested page is loaded on the user's browser, additional data is fetched from the backend app and displayed to the user.

Technical challenges

What was the hardest part of development? For me, the hardest part was implementing the Health Assistant (AI) and Decision Support functionality, we were unfamiliar with prompt engineering and had to delve into it within a short time. Also, implementing decision support which involves guiding the patient to the next flow (schedule appointment with a physician or call emergency) using the AI response was a bit tricky. The AI was not always consistent which means a lot of edge cases to handle.

Does your app have any scaling issues?

No

What are some key takeaways?
The development time was really short, due to this a short time was put into designing the architecture for this application.

There was no time to write unit or integration tests for the application, it was only tested manually. Some functionalities might have subtle bugs which will make them behave in an unintended way.

Future Steps 

This MVP will be further developed into a full fledged product to be launched in the first quarter of 2024.

In order to go from MVP to fully developed product, the following need to be executed:

1. Adherence to compliance standards (mainly Data compliance - HIPAA)
2. Development of an approach to validating physicians and healthcare professionals on the platform - relevant for true moderation
3. Further training of the AI model to improve relevance of results in case specific scenarios

Additional Assets