Behaivior Covid19 Early Detection System

The pandemic COVID-19 continues to spread worldwide. As such, there is a dire and urgent need for development of rapid, accurate wearable diagnostics to identify and isolate pre-symptomatic COVID-19 cases and track/prevent the spread of the virus.   

Behaivior proposes to utilize our infrastructure for time-series based modeling of physiological signals for early indication of COVID19 that can augment typical, self-service, survey-based diagnostic/ risk-stratification methodologies.    

Our solution fills a gap in the current market and need for COVID19 testing by way of two novel value propositions:

• Pre-screening at home : determining diagnostic testing needs 

• Continuous monitoring and alerting : Ongoing streaming of physiological data for monitoring and alerting for COVID19 risk, serving as a basis for obtaining a prescription from a PCP or drive/walk-through.

Scaled globally, our solution could help expedite a return to safer workplaces and help increase the dangerously limited testing capacity regarding Covid19 detection and intervention. 

Until a vaccine is widely available, regular testing will become a part of life. The US needs 35 million tests per day, encompassing ~10% of the US population, with more frequent testing and evaluation of front-line workers.  Today not nearly enough Americans are being tested for the COVID19 largely due to a shortage of diagnostic testing capabilities and the individuals’ resistance to leave their home due to potential exposure to the virus. Pre-screening technology therefore enables confirmatory testing for the general population concurrent and self-quarantining.  

Authorized at-home COVID-19 tests are available commercially but cost over $100, ruling them out for routine at-home testing particularly for marginalized populations.  An at-home, continuous monitoring option, which is essentially free, is available in wearable devices with typical physiological monitors including those computed from photoplethysmography.   Behaivior proposes a free wearable-data-based pre-screening technique that can serve as a precursor to an insurance reimbursable confirmatory test.  

We anticipate our solution will also help mitigate the challenges of testing low-risk individuals with no symptoms and likely negative test outcomes. Scaled globally, this technology could help expedite a return to safer workplaces and help increase testing capacity.

Behaivior proposes to utilize our infrastructure for time-series based modeling of physiological signals to offer an early indicator of COVID19 that can augment typical, self-service, survey-based diagnostic / risk-stratification methodologies. This tool will utilize trained machine learning models built on time-series data from a wide range of sensors provided on commercial off-the-shelf smartwatches. This technology can detect signs of febrile illness and specifically due to COVID19 that precede full-blown symptoms, and subsequently direct the user to resources for drive-through or walk-through testing or a PCP that can provide a prescription for the same. As such, self-reports by the smartwatch wearer would serve as labeled training data during training endeavors, and automated signaling based on algorithmic alerts would prompt the wearer to engage in therapeutic activities in our anticipated production embodiment. Our methodology for crowd sourcing signals and their corresponding labels as well as our existing infrastructure for developing time-series data-driven AI for classification of categorical states (i.e.m disease, emotional states, etc) will permit us to rapidly develop an minimum viable product (MVP) to diagnose COVID19 from wearables data.

A major issue of this pandemic is that infected individuals unknowingly spread the virus throughout the population. Behaivior’s solution can effectively minimize the spread through early detection, empowering individuals to self-isolate and then receive formal testing as needed. This is targeted towards people unable to be tested at home due to cost (i.e., $109-$150 per test, multiple tests required for confirmation due to the false negative rate with current tests); who need a health care provider’s note/approval to even access a test (required for insurance carriers and testing sites); workers who are at risk but are unable to get tested on a regular basis; and/or those who want to return to the workforce but can’t due to lack of testing - an easily adoptable solution (Pew: 1 in 5 Americans owns a smart watch) providing individuals with testing abilities at their fingertip. In particular, this tool will be powerful for essential workers facing risk of infection on a daily basis and to individuals exposed to those considered high-risk. While a vaccine may take 12 months of development and potentially longer for wide distribution, Behaivior’s Early Detection System can be developed and implemented more quickly, thereby reducing spread.

The health sector is focusing on testing patients who are symptomatic and people who think they may have been exposed. We anticipate that our advanced algorithms will extract COVID-like signs and symptoms from the raw data insights even before individuals may perceive them. Therefore, our solution aligns with MIT’s vision in regard to: a) being proactive (as opposed to reactive) in treating disease outbreaks; b)  having a solution that enables continuous monitoring and reporting aligned with the goal of tracking the spread of the outbreak in real-time to affect decision making of the general public. 

Administering coronavirus tests requires time and supplies that are already scarce yet aggressive testing has proven to be the best way to track and isolate the disease, stopping its spread. Further, the lack of testing continues to be a source of deep frustration nationally, with worried patients unable to find out whether they have the ordinary flu, the coronavirus or something else entirely. Our solution aims to fill this void by allowing end-users to evaluate their COVID19 risk from home and make an informed decision regarding their need for traditional testing. We anticipate this helping reduce the burden of testing low-risk individuals with no symptoms and likely negative test outcomes.

Behaivior is proposing to utilize our infrastructure for time-series based modeling of physiological signals to provide an early indicator of COVID19 that can augment typical, self-service, survey-based diagnostic / risk-stratification methodologies. This tool will utilize trained machine learning models built on time-series data from a wide range of sensors provided on commercial off-the-shelf smartwatches to detect signs of febrile illness and specifically to identify COVID19 that precede full-blown symptoms, and subsequently direct the user to resources that may alleviate the symptomatic conditions. As such, self-reports by the smartwatch wearer would serve as labeled training data during training endeavors, and automated signaling based on algorithmic alerts would prompt the wearer to engage in therapeutic activities in our anticipated production embodiment.       

Our methodology and existing app-infrastructure for crowd-sourcing signals and their corresponding labels as well as our existing infrastructure for developing time-series data-driven AI for classification of categorical states (i.e., disease, emotional states, etc) will allow us to rapidly develop a minimum viable product to diagnose COVID19 from wearables data. We have a full tech stack established.

Behaivior has developed proprietary iphone/apple watch applications for crowd sourcing physiological signals from a broad audience, with technology to label these signals based on user responses. Modification of our existing workflows will facilitate users of our applications, who consent to contributing their physiological data (i.e., from Apple Watch as well as iPhone activity, location, etc), to additionally inform us of their current status of febrile illness or a confirmed diagnosis of COVID19, so as to collectively create a first of its kind physiological waveform data set with self-reported labels, which may be leveraged for the purpose of training machine learning identifiers corresponding to febrile illness and more particularly, COVID19. Our methodology for crowd sourcing signals and their corresponding labels as well as our existing infrastructure for developing time-series data-driven AI for classification of categorical states (i.e., disease, emotional states, etc) will allow us to rapidly develop a minimum viable product (MVP) to diagnose COVID19 from wearables data.     We have validated our system’s many facets through IRB approved clinical studies funded by the National Institutes of Health where our focus was helping individuals struggling with opioid addiction. From our surveyed research participants: 2/3 (i.e., 66.66%) appreciated the physiological feedback they received from the wearable, allowing for health tracking and progress toward goals. Additionally, 50% named sleep feedback as something they like best about the wearable, and 41.66% liked the survey/self-report aspect of the feedback to monitor their cravings and cues such as stress.

Immediate outputs for target population:

Our initial contribution during the performance period of this award will be to release a data-gathering platform for time-series wearables’ data that fuels a first-of-its-kind public dataset of wearable time-series data along with ground-truth information regarding demographics and self-reported state of health of individuals. This dataset is foundational for data-science, and can model categorical endpoints of disease and viral illness as it relates to COVID19.

Longer-term outcomes for target population:

Our collected dataset and monitored population of wearables also may serve as a resource for modeling a plurality of meaningful leading clinical indicators in the future, including viral illness, febrile illness, etc.. 

Logical links between activities, outputs, and outcomes: 

In the US, 56.7 million will use a wearable device at least once a month in 2019. Further, ~23% of wearable users have an Apple Watch. At a 10% adoption rate, we expect to reach out to 1.3M adopters.  We anticipate reaching this milestone of adoption at the end of 12 months with rigorous digital marketing campaigning. Our system  has applicability for any country and the iPhone app can be downloaded from any country.

1. This solution is in development and is not yet available

2. In one year we will be serving 1.3M amount of people

3. In five years we will be serving 6.5M people assuming linear extrapolation on growth goals for year 1.   

The transformational impact of our COVID19 pre-screening technique and our vitals-data crowdsourcing model is one that can affect millions of lives because:

1. Crowdsourced vital data may be useful for unsupervised learning tasks that are beneficial for a plurality of analytical tasks, applicable to febrile or viral illness, in general;

2. Pre-screening for signs of illness may extend beyond COVID19 in terms of identifying viral or other illnesses.

In sum, we plan on generalizing our modeling approach and our dataset to model end-points that even go beyond COVID19. We also propose to offer open-access datasets for public consumption.  

Our impact goals within the next year are to reduce Covid19 death and infection rates, and within the next five years to expand the system to be utilized for other viral diseases.   

Within the next year our greatest barrier will be funding to support the development of this solution. Within the next five years we anticipate funding will also be the greatest barrier for us to accomplish our goals. 

Critical mass in terms of data collection is critical to developing meaningful models. Consequently, digital marketing of our solution to a large enough audience will be essential to develop useful and usable models with minimal false positive or negative alerts. 

Currently we are addressing the financial barrier via applying to grant funding opportunities. We are also exploring opportunities for investment. 

Digital marketing will be key to acquiring a significant user base of participating data contributors to develop meaningful and large-scale monitoring for COVID19.    

As of 6/8/2020

Full time: 1

Part-time: 3

Interns: 5

Mentors/Advisors: We have multiple advisors and mentors who join us as needed on relevant projects

Behaivior is comprised of a diverse team of experts with professional backgrounds including business, design, software development, machine learning, addiction science, wearables and consumer technology in addition to others. The team features individuals with roles including founders and leadership, specialists, advisors, and interns.

Ellie Gordon, Founder and Chief Executive Officer of Behaivior, is a behavior change technologist, designer, and biotech entrepreneur. For more than 13 years, her work has primarily focused on improving people’s lives through technology. She has successfully created, organized, led, and managed teams for everything from hackathons and codefests, to other science based startups and initiatives, to community development and research projects. She did a TEDx Pittsburgh talk as a behavioral technologist speaking about the design elements of what Behaivior is developing.
 
    Behaivior's CTO, Prahlad G Menon, PhD, has extensive experience with regard to commercialization of medical device technologies. He has spun off technologies from his academic research program as an Assistant / Associate Professor of Electrical & Computer Engineering (Carnegie Mellon) and Biomedical Engineering (University of Pittsburgh), which resulted in QuantMD LLC and CerebroScope (Science Plus Please LLC) - each entities focused on different biomedical device areas based on granted patents. QuantMD LLC was focused on clinical efforts relating to image based diagnostics for cardiovascular disease, and image guided surgery for cardiovascular interventions.
Doug Wolf is a project advisor and executive advisor to Behaivior. Doug will provide monitoring and guidance in the areas of financial performance, research & analytics, business development & strategy, and marketing.

N/A currently. If funded, we would like to partner with organizations who can help provide this tool to those who need it and who can facilitate promotion.

Real-time HRA for payers - subscription service

Hospitals for post-discharge patient monitoring - a telehealth extension   

We propose to offer our solution for covid19 pre-screening and real-time risk monitoring for free so as to serve the purpose of data collection at a large-scale and to develop a proof-of-concept pre-screening model. This model can augment current clinical triage workflows prior to drive-through and walk-through testing using a molecular or serological test.    

In the long term, we expect payers / insurers to be interested in our solution as a form of continuous-mode health-risk-assessment (HRA). HRAs are generally completed at new member intake and have poor completion rates via current modalities of data collection.   Our solution can improve compliance with HRA data collection efforts while augmenting HRAs to a continuous monitoring paradigm.        

Payers will subscribe on a per-member-per-month basis to our services for continuous monitoring and patient-specific alert reporting as it relates to a plurality of clinical endpoints, ranging from simple febrile illness to risk of suicide, addiction relapse, etc.. 

We anticipate that insurance payers will subscribe on a per-member-per-month basis to our services for continuous monitoring and patient-specific alert reporting as it relates to a plurality of clinical endpoints, ranging from simple febrile illness to risk of suicide, addiction relapse, etc..  This subscription service for the Medicaid population alone, in a single state (e.g., Delaware) includes 150k members and therefore forms a revenue stream of $1.5M at 10 per member per month.  We anticipate to achieve equilibrium within months after acquiring a first payer network as a client.        

Our team is at a critical stage in which we need an expanded resource network to successfully enter the final development and deployment stage. MIT’s network can take our massively scalable, science and technology-based tool and provide the needed funding as well as employ an objectives-based mentoring process with the goal of maximizing equity-value creation. We look forward to acquiring crucial funding (particularly given the urgent impact of Covid19) in addition to refining our product and team through the access and mentorship provided.

1. Talent Recruitment: We would like to expand our team to find additional talented team members who are passionate about helping others. 

2. Solution Technology: Given the opportunity to expand the capabilities of the system through the development and employment of additional foci (i.e., additional physiological and behavioral metrics), assistance with expertise and person-power to expand this solution would be of great value to our team. 

3. Funding and revenue model: We have conducted some initial customer discovery research. Due to time constraints and the changing landscape, further guidance regarding our funding and business model plans would be beneficial. 

4. Marketing, media, and exposure: As a small health tech startup we have a limited budget for marketing; any assistance in that realm through promotion or media opportunities would be greatly appreciated.

We would like to partner with healthcare providers to provide this solution to the individuals they serve. We would also like to collaborate with any MIT faculty and initiatives or Solve Members who have related expertise, and would like to join us in expanding our tool through research and development to better serve those who are experiencing barriers services access due to the Covid-19 pandemic and/or who are not in a position to access testing. 

Our team is working hard daily to help stem the tide of the highly contagious Covid19, which has affected millions of people globally, and has caused hundreds of thousands of deaths. After the death of Behaivior founder Ellie’s grandmother in April due to Covid19, we felt compelled to find a way to help stop the spread of Covid19 inspired by our existing infrastructure. Our goal with this tool is to help reduce the spread and death rate of the Covid19 coronavirus. We are elevating humanity by creating a novel detection and intervention system. We are adapting evidence-based solutions using human-centered design principles and user experience testing to a digital form for real time interventions that are personalized, empowering people at risk of infection. 

Our team will use the Elevate Prize to advance our solution and support our team who are putting their health and safety at risk to save the lives of and care for those who are affected by covid19 through funding a collaborative development and outreach effort. This would include hiring a community outreach coordinator to connect our tool to those who need it most, a developer to implement additional wearables data and processing into the application, and a collaborator with local support services to provide our system to those who are economically and/or socially disadvantaged based on identities including race and ethnicity.

Using our foundation and infrastructure for data science, artificial intelligence, and machine learning, we will utilize our infrastructure for time-series based modeling of physiological signals in order to offer up an early indicator of COVID19 that can augment typical, self-service, survey-based diagnostic / risk-stratification methodologies. This tool will utilize trained machine learning models built on time-series data from a wide range of sensors provided on commercial off-the-shelf smartwatches to detect signs of febrile illness and specifically due to COVID19 that precede full-blown symptoms, and subsequently direct the user to resources that may alleviate the symptomatic conditions. Our infrastructure with an addiction focus was supported by funding from the National Institutes of Health through the National Institute on Drug Abuse we collaborated with AI and machine learning experts at Carnegie Mellon University, the University of Washington, and the University of Pittsburgh, including Dr. Anind Dey, Dr. Grace Bey, and our CTO, Dr. Prahlad Menon using data we collected from IRB approved clinical research studies. 

Our team will use the AI for Humanity Prize to advance our solution that is dedicated to improving the lives of individuals and our global community through neuroscience research and information technology. Funding this collaborative development and outreach effort would enable hiring a community outreach coordinator to connect our tool to those who need it most along with a developer to implement additional wearables data and processing.

The People’s Prize, which celebrates and crowdfunds innovative achievements for the UN Sustainable Development Goals, will help advance our solution to help mitigate the Covid19 coronavirus pandemic. The UN broad goals of Good Health and Wellbeing and Reduced Inequalities are supported by our innovative achievements of our solution, which limits the spread of Covid19, and empowers people to rejoin or participate in the workforce more safely, taking charge of their own well-being regardless of income and access. Depending on the award level, we will use the People’s Prize funds for a collaborative development and outreach effort that would include hiring a community outreach coordinator to connect our tool for broad reach, a developer to incorporate additional wearables data and processing, a collaborator with local addiction support services to provide our system to those who are economically and socially disadvantaged including due to their racial and ethnic identities.