CovidGuru

Inspiration

Do you remember the first time you had your first child? Dealing with unknowns, seeking information on every new thing you faced and feeling a bit lost and nervous in the process. The book "What to Expect When You Are Expecting" has been a bestseller that helped generations of new parents prepare for the biggest challenge of their lives. Now, the entire world is faced with a once-in-a-century pandemic, and we are all learning about this novel virus together. Dealing with COVID-19 is like no other disease because there are so many unknowns. The journey can be nerve-racking when a person does not know what they may be facing next.

CovidGuru aims to be the equivalent of "What to Expect When You Are Expecting" for the Covid generation. We will build the knowledge base on-the-go, day by day with your personal contributions of your own health experiences and we will share the common wisdom back with you, just in time for you to brace and gear up for the next day. Moreover, this common anonymized knowledge of humanity will be made available to medical researchers to complement their clinical studies to provide further data on the phenomenology of the disease onset and progress.

We aim to learn more together one day at a time, as the patient, caregiver, and medical community with the help of meaningful and sophisticated analysis of daily patient-generated data from thousands of users.

What it does

Based on our own recent experience with COVID-19, we know the difficulties of each day spent with the disease: understanding what is common or normal, what signs to pay attention to, how to interpret any changes from one day to the other and more importantly being more informed about when to act and seek professional healthcare.

Our project, CovidGuru is an online platform where personal experience with COVID-19 will be shared and the information will be aggregated in a scientific, meaningful way. This will enable our collective experience to guide the way for others, including at home care and useful patient-reported information for the medical professionals as a credible source. However, this platform does not aim to provide medical advice, advocate any treatment regimen nor prescribe any medications.

CovidGuru platform gives the user a place to: Share their experience as they observe symptoms and fight the disease.
Learn from the experience of others as they go through the recovery journey. Monitor the changes in their vital information using available collective data as a benchmark, Recognize the signs/ changes in order to seek proper medical care in time, to avoid complications before it is too late.

Participants of the data platform would enter information on their health status tracking the daily progress of the disease. Participants can either be experiencing COVID-19 symptoms themselves, or be a caregiver at home, such as a spouse/ significant other, or the child/parent of the person with symptoms.

Through standardized data collection surveys, participants will enter their symptom frequency and intensity, any measurements from their available devices, emotional status and other relevant information such as fluid intake, diet and supplemental medications. Data on symptoms and mental health status would be collected through well-designed and validated patient-reported outcome (PRO) instruments for consistency.

This patient-centric information will be analyzed using machine learning techniques to identify meaningful patterns or signals in the data. A dashboard will present aggregated and anonymized data on a daily basis so that the participants can see the most common symptoms for each day of their course with the disease.

CovidGuru will shed light on the unknowns by tapping the collective experience and the wisdom of the community. As we individually navigate through the COVID-19 pandemic, the shared experience is collected for common good.

How I built it

User Interface/Experience: (1) The platform will have a mobile web interface. (2) Daily alerts/reminders for users to check-in their current statuses. (3) When a user is checking-in daily (or throughout the day) at time T, the system should present a list of predicted symptoms for him/her given recent symptoms reported plus other patients like him/her at time T-1. (4) Charts and dashboards to show historical progress of the user/patient.

Database: (1) we need to populate the initial knowledge base from public sources and maybe harvest data from social media sites before user #1, we do need some base information to provide him/her with his/her first set of recommendations, (2) original data entered by our own user base, (3) real-time infection dynamics data for the current ZIP code and the transmission dynamics in neighboring ZIP codes.

Data Analytics: We will use a combination of Machine-Learning and Predictive (Bayesian) Analytics techniques. As an illustrative scenario, for each new day in the morning, using the progression of symptoms and other parameters in all dimensions from other users, the trained Multi Layer Perceptron (MLP) or time delay neural network (TDNN) artificial neural network will offer new symptoms to watch for. People with similar predictions will be connected to confirm the outcome and communicate with each other. Trained network will be available in ONNX format to be used in other platforms such as Tensor, IOS, ML.NET and others. A collective intelligence network will be formed.

Data Privacy: Account data and gathered health data will be linked with a unique identifier but otherwise not be shared. Only aggregated information will be presented, and participants will never be identifiable. Data will be stored in an anonymized manner using the most compliant and secure practices.

IoT integration: Although the user can manually enter diagnostic data, bluetooth enabled consumer health diagnostic devices can also upload data.

Challenges I ran into

Gathering, compiling, and aligning health literature insights to create the baseline knowledge database before the addition of first user inputs (scientific problem). Getting people to sign up and share data actively (marketing problem). Weeding out the noise from the information (data curation problem). Narrowing down the approach to feasible first step for successful implementation (practical problem).

Accomplishments that I'm proud of

We are a team of highly accomplished, multinational professionals with diverse expertise in medicine, epidemiology, health informatics and data science, information technology, engineering, and marketing. We bring our combined knowledge, observations and passion to this project to improve the well-being of all people facing the threat of COVID-19 in order to find better ways to make patient-centric data and relevant education more accessible for all.

What I learned

Members of our team have suffered from COVID-19 both as caregivers and patients. During this difficult period, we learned from each other by sharing our knowledge and learnings. At the same time, we observed that for many people, unfortunately such information is not easily accessible. We believe that we can help empower each person dealing with COVID-19 with personalized information by building a self-learning data platform that gets better with more input every day as we heal and beat COVID-19 together.

What's next for CovidGuru

The platform can be improved to capture data automatically from home devices, such as thermometers, blood pressure monitors, pulse oximeters though either bluetooth connection or by reading and capturing the data through the camera pointed at the screen of these devices. It can be expanded to capture other types of data, such as a user’s voice recording repeating the same sentence every day in order to capture the changes in their breathing through digital sound analysis. We will regularly share the latest aggregate-level insights and best practices learned from our user community with the public using Twitter, Facebook, etc. Special Thanks to Uzay Kirbiyik, MD, and Omer Karay Akar for their contributions.

Built With

• dotnetcore