Track My Exposure - COVID-19 Risk Engine

We are many months away from a vaccine to prevent coronavirus. As the pandemic drags on, individuals need to balance risk of COVID-1 exposure with the financial and mental drain that shelter-in-place has caused.

Track My Exposure is the first direct-to-consumer COVID-19 risk engine that gives users a way to see their exposure to coronavirus. As nations open up, we are sensitive to the risks of second waves of coronavirus in various communities. To mitigate these risks, we provide personalized up-to-date information in the form of a risk score that helps people make decisions on whether or not to meet up or attend events. 

Scaling this solution globally will put power back into the hands of individuals - they can make their own risk based decisions.

We remain months away from a vaccine to prevent coronavirus. As the pandemic drags on, public willingness to continue cumbersome but necessary countermeasures wanes. 

The U.S. economy isn’t built to handle extended sheltering-in-place. With 21 million currently unemployed in the US, and a global UN prediction of 6.7% of working hours wiped due to COVID-19, the effects of social isolation on both mental wellness and financial wellbeing intensify. “Quarantine fatigue” has been shown to cause cognitive impairment for risk analysis and decision making. We know from other “abstinence” policies that these measures are often more detrimental in the long run than helpful. As many governments open up prematurely, we are faced with a daily decision on how to minimize our COVID-19 risk. We need to know that risk is not binary and we need a way to accurately distinguish in between.

Globally, at least 7.8 million people have contracted coronavirus to date, and many more cases likely go undetected or untested. While our initial work will be focused in the US, we have the ability to scale this solution globally to impact all 5 billion in the world with access to cell phones and ultimately our tool. 

The tool Track My Exposure helps you calculate the risk of exposure to coronavirus to influence decisions to stay at home or continue with your activities. It is meant to be used during periods of time when the government is trying to ease stay at home restrictions for epidemics like coronavirus and individuals are now left to making decisions based on their own risk calculations. 

Users create an account through our online platform, answer a few questions about their county, any symptoms they may feel, and some behaviors they observe like mask-wearing. Based on the answers, our risk engine conducts a Bayesian analysis that creates an exposure risk score.

Users can use this information and share with people they may plan to meet with during the day to meet if the risk is low or to postpone if one or both parties have high risk scores.

Risk is calculated by a variety of factors including the county case information as a baseline, diagnostic information that helps define if their symptoms are more similar to COVID-19 or other respiratory illness like influenza or allergies, and common behavioral data like how much they are social distancing or wearing masks.

Our solution currently serves anyone in the United States who would like to have an easy tool to measure their COVID-19 risk and has access to the internet through their phone or laptop. Through our early studies including both qualitative user interviews and quantitative studies running ads to study our target audience, we have discovered the vertical of people most interested in using our application are 55 years or older or are at higher risk of fatality should they catch coronavirus. We are continuing to understand their use cases with ongoing user studies and continued study of product and marketing analytics. 

We will continue to engage with this audience to show up where they are digitally - namely facebook advertising will help us reach this group of people. We plan to also form community partnerships to be featured on local newspapers or government websites. 

For people who are at higher risk of fatality, managing COVID-19 risk is life or death. By using our risk tool, they can make more informed decisions and have a lower risk of contracting the virus as well as helping stop transmission to others.

COVID-19 has infected over 7.8 Million people worldwide and killed 480K at this point in time. People are scared and also eager to open back up after being in lockdowns for months. Risk does not have to be binary. We will prevent the spread of the virus by helping people understand the different risks of their daily activities, especially social meetings. This well help prevent transmission in this current pandemic until a vaccine is found and the same approach can be used to help curb future outbreaks or be used to manage public safety as a whole.

Traditionally infectious disease is contained using contact tracing and testing. Unfortunately, contact tracing only works reactively. Most models for prevention work at the community level. Our solution’s risk model aims to be more preventative at the individual level. Personalized risk decisions help people make decisions about who to see or where to go, or to avoid those situations altogether. 

Recently, we’ve also seen the rise of “daily health attestation” apps where employees log their health status every day before work. While this may work for more contained situations and ultimately make an assumption that we cannot help asymptomatic spread, this ignores the larger picture of the trajectory of existing cases in the community. 

We need both contact tracing and daily health attestation apps to help end COVID-19. We also need a preventative tool personalized for individuals and the daily social interactions that can take place. While contact tracing focuses on the already sick and health attestation apps work best for employers, our tool is designed for the people to help them manage both the business and personal meetings they may have. As of yet, we have not seen another app in the market and are excited to be the first direct to consumer COVID-19 risk engine.

We have developed a mobile-friendly, easy-to-use web app freely available at https://trackmyexposure.com that estimates your risk of having coronavirus after you respond to a few simple questions. The model keeps itself appraised of the latest case counts in every county in the U.S., and we update its logic as new pertinent findings become available in the scientific literature. Currently, we primarily use demographic and symptom data for the calculation, for example, your age range, county, symptoms, doctor’s opinion, and test results if available. One of the major challenges of estimating your probability of having coronavirus is that so many cases are asymptomatic. Demographic data, in conjunction with case frequencies among people similar to you, give us a window into your risk even if you may have an asymptomatic case. Your symptoms help us understand if you are more likely experiencing COVID-19, other infections like seasonal influenza or the common cold, or none of the above. We synthesize an overall probability based on all available evidence, and then map it to one of four scores: low (< 0.5%), moderate (0.5% to 12%), high (12% to 24%), or very high (> 24%). We are currently incorporating important behavioral factors such as self-isolation and mask-wearing into our model to emphasize the importance of the variables under a user’s control.

Our risk engine relies on Bayes’ law, a mathematical technique for estimating unknown variables (e.g., whether you have coronavirus) from known evidence (e.g., whether you are coughing) originally published by Reverend Thomas Bayes in his 1763 paper An essay towards solving a Problem in the Doctrine of Chances. Bayes’ law is arguably the foundation of modern statistics and has been deployed extensively and to great effect in applications as diverse as email spam filtering and genetic analysis. We leverage the latest medical research (https://trackmyexposure.com/sources.html) on COVID-19 and how to prevent it to define the parameters of our Bayesian model. The literature is rapidly evolving, and even since we began our project in April, the quality and scope of the available evidence has increased significantly, allowing us to improve the risk engine accordingly. The county case data we rely on (https://usafacts.org/visualizations/coronavirus-covid-19-spread-map/) is syndicated by USAFacts from state and local public health agencies.

Once we have a statistically significant number of users who have self-reported laboratory test results, we will further validate and refine our model, eventually using more advanced machine learning techniques that can train on users with test results. We will also collaborate with medical professionals to elicit their feedback on our diagnostic methodology. We employ industry best practices, such as continuous integration, cloud computing, automated unit testing, manual quality assurance, and user interviews to ensure that our software continues working as intended around the clock.

We have developed a COVID-19 risk engine that uses county case data, symptom diagnostics, and behavior data to predict a COVID-19 exposures core. This tool is provided through a web app that is easily accessible on desktop or mobile devices.

Individual users can use this exposure tracker every day to determine the COVID-19 risks that may occur. Risk can happen a number of different ways - based on county case data, symptoms or behaviors of people that they may meet, and helping postpone meetings on high risk situations will put the individuals at lower risk of contracting COVID-19. As states open up and many suffer from both the economic and mental aspects of stay-at-home orders, people are eager to get out back to the things that they used to do. We want to offer the option with added safety.

As individuals take more care of the risks that they have, communities benefit directly. The fewer transmissions that take place inside of a community, the lower rates of COVID-19. At the same time, this takes a more measured approach than shutting down business altogether so we don’t need another round of stay-at-home orders that would certainly destroy businesses.

As we learn more about coronavirus and it’s transmission variances, we’ve learned more about what we should do and what we should avoid. While CDC guidelines are helpful, it’s easy to fall back to heuristics. Due to the very nature of human evolutionary psychology, we are bad at computing risk. A risk engine during this time helps us more accurately think about our options. 

A few users of our app have remarked that they have postponed meetings with friends during May 2020. This is great anecdotal evidence and we will continue to quantify the number of meetings that are postponed. Overtime, we would like to be able to account for transmissions that didn’t happen because of the use of our tool.

As we are just starting out, we currently have 60+ users on our platform. We are working on scaling this rapidly with new features that promote organic sharing and continued investment in marketing and partnerships. We know that rapid scaling in the next few months is extremely important. We are targeting communities where local governments are taking the coronavirus outbreak seriously. Our goal is to scale up to 1,000 users in 2 months. 

In a year, if coronavirus is still a threat we are dealing with, we hope that we can create a lot of impact in the community with this solution by continuing to gain users and help them measure their risks through local government partnerships.

In five years, I hope we are no longer dealing with the coronavirus crisis and have pivoted this solution to combating other infectious diseases epidemics like seasonal influenza or sexually transmitted infections.

Our immediate impact goals are to help individuals manage their personal COVID-19 risks. We have a few different strategies to help us achieve this. 

Our first strategy is to go direct-to-consumer. This requires a strong marketing engine and partnerships with local governments to help get the word out. The first proof point we have to meet is that there is enough interest and desire for users to use the tool and come back. We will do this by reaching out via advertisements and content marketing like blogs and social media posts to find our first 1000 users. As a lean agile startup, we plan to listen attentively to user feedback and make adjustments to both our product and marketing to serve our users best. 

From there, we will continue our direct to consumer strategy by partnering with local governments to spread the awareness with a proven product. We are beginning the pipeline to build relationships now so that we will be ready for this phase in a couple of months. 

Our second strategy is to leverage the risk engine part of the solution by partnering with insurance or healthcare providers to scale the impact. The risk engine we have built uniquely calculates individual and county level risk in a way that could benefit insurance, health care providers, and epidemiologists. We don’t necessarily need a front-end web app to have impact and are exploring this avenue.

Our greatest barrier is the bridge between consumer trust and COVID solutions. We are currently at a 6% funnel rate of people who come to our site and use it, and by building a stronger backing we could build trust to increase that percentage.

The second barrier is that we cannot build desired features fast enough, with the time sensitivity of COVID-19. This is due to bandwidth constraints which we aim to solve, with funding, by bringing in additional contractors.

The third barrier is that because we are taking in case data from county codes, we are limited in the sources for applying the solution globally beyond the UK and Canada. While we can apply it for countries, our accessibility to case data isn’t specific enough to break down locations where cases may have peaked.

For the first barrier, we are looking at building stronger partnerships with local governments or entities. Through branding our site with their support, we can use that as leverage to build trust between the community and our solution.

For the technology barrier, we are actively looking for part time and full time help to accelerate our product development. We are well versed in agile software development practices and our team has a strong vision for the product so staffing is the primary hurdle to increasing the bandwidth. 

For the third barrier, we can overcome this barrier by communicating with the governing parties of other countries. That could allow us to access case data specific to them on a global scale, which can allow us to give more detailed information on where in a country a person may have a higher COVID-19 exposure risk than others.

We started working together in April 2020 and have not yet filed legal paperwork. We have explored for-profit and non-profit models. We are planning to file as a for-profit organization with a social mission and dedication to use our profits for continued venture growth. 

Full-time: 2 people

• Technical Led with software engineering and data science expertise.

• Design Lead with marketing and sales expertise.

Part-time: 1 person.

• Product lead with operations and behavioral science expertise.

Interns: 2 people. A marketing intern and a software intern.

We have been working on this solution for 2.5 months. We started in early April 2020 and have very quickly stood up a Beta product June 1st and marketing engine that has brought over 14,000 views in this short time. We are actively working on growing our team and building out the product quickly because of the time sensitivity. 

Our founding team of MIT, Stanford, UMass and Google alumni has a cumulative 20+ years of entrepreneurial experience. Kurt is a software engineer and data scientist from Stanford and Google. He has 8 years of software engineering/data science experience on high-velocity projects spanning data pipelines and analysis, advanced scientific applications, Linux operating systems, and behavior change apps. Sagger has a 10 year background in startups, B2C marketing, gamification, and branding apps. He’s created systems to generate content and drive users to our app, as well as being responsible for the design, analytics, and copywriting. Cathy comes from a strong product management and operations background. She’s well versed in the behavior sciences and has honed skills developing products that improve human behavior in ed tech, financial services, and DoD contracting.

In addition to the skills we bring into the venture, we bring both strong networks and teamwork and collaboration together. We’ve already attracted the attention of over 300 graduate and undergraduate interns, 2 of which we are mentoring this summer. We are hoping to further grow this team with paid contractors to help more rapidly expand our solution.

We do not currently have any partnerships but will begin exploring partnerships shortly. We are planning to partner with:

Local Government - We can help local governments curb the spread of coronavirus and hope to find communities to partner with to help spread the word of this tool. 

Healthcare Providers - We’d like to be able to be a tool that healthcare systems can recommend to their patients as a preventative method that is more personalized than following CDC guidelines. 

Our key beneficiaries are the individuals who are using the COVID-19 risk engine to make better informed decisions about their social activities. The type of intervention they would receive is through a web tool that can be accessed by computer or mobile phone. Users can log in daily to update their latest information and receive new personalized risk information. The Track My Exposure tool provides a more quantitative approach to COVID-19 risk so that individuals can make an informed decision about the activities and behaviors they do day to day that may increase contraction of COVID-19. 

We are providing this tool for free to the general public and are planning to offer sponsorship packages for ad or branding space on our website. Our sponsors will be provided with a socially responsible opportunity to show their brand to their target demographic.  

We are reaching our users through social media and blogs and our sponsors through LinkedIn and our network.

As a company with a social mission to curb transmission of coronavirus, we offer this tool for free to our users. This means to fund the development of our work, we are primarily relying on grants and sponsorships.

In the long run we have other strategies we can explore for revenue. An option where we’ve gotten some interest but haven’t explored further is the B2B model. Organizations like insurance companies may be interested in our risk engine to access via API. Other organizations may be interested to use our solution to keep their employees safe.

Solve allows us to solve some core avenues for growth for us. Firstly, we’d like to see the expertise of SOLVE advisors to help accelerate our growth and scale with the current time sensitive topic of COVID-19.

Secondly, we’re excited about the potential partnership opportunities which can help scale our solution with local governments and MIT research labs.

And finally, with joining the global network of Solvers, it would allow us to network and prep our solution for other avenues of growth.

Currently there is a distrust with COVID-19 between the public and apps. We are looking for support with the service distribution in order to build better trust between us and the consumers. With appropriate partnerships, we can bridge that gap in order to better approach government entities, private entities, and the public.

MIT Media Lab COVID-19 Response - Join people with a similar mission to see how our different solutions can augment each other.

MIT Center for Collective Intelligence - Help user research on the best methods to incentivize users to follow our computer generated score and bridge the gap between human and computer intelligence.

City of Cambridge or other local governments - to help provide feedback and deploy our solution.

MIT Senseable City Lab - to see how this technology can impact the design of cities post COVID-19.

Safe Paths - to collaborate on data aggregation and contact tracing.

Covid Act Now - to see if our predictive methodologies can be joined together.

We are currently using data science and statistics techniques like Bayesian Updates to calculate the the risk score. On our roadmap, we'd like to use AI and ML techniques to further the prediction quality of our risk engine. As we collect more information on individual behaviors, we can run analysis side by side with guidelines by the CDC to see which factors lower transmission the most.