Remote COVID-19 Diagnostic Testing

To combat the spread of the novel coronavirus, COVID19orNo.com will be a reliable, instantaneous, remote COVID-19 diagnostic medical device available for free to any respondent (a user engaging with the device on behalf of a patient). The respondent will complete an online questionnaire covering the patient’s symptoms, risk factors, and other key information, such as location, lifestyle information, and any recent trips to a workplace or hospital. Machine learning will analyze the provided data and will present a percent likelihood that the patient has COVID-19.

COVID19orNo.com will combat two urgent problems: 1) address the scarcity of COVID-19 in-vitro diagnostic tests by providing healthcare facilities a cost-effective alternative that can be used an infinite number of times, and 2) track disease progression and predict outbreaks using respondent location, enabling policy-makers and healthcare workers to effectively respond to future emerging pandemics. 

 COVID19orNo.com will transform the healthcare industry.

The Coronavirus Disease of 2019 (COVID-19) is a highly transmittable respiratory disease that was declared a pandemic by the World Health Organization (WHO) in March 2020. By June 2020, COVID-19 has infected over 7.8 million individuals globally and killed over 430,000 people. As cases continue to soar, there are too few reliable in-vitro diagnostic tests readily available to test all who may potentially have the disease. 

A remote diagnostic solution like COVID19orNo.com would combat the current scarcity of resources such as in-vitro diagnostic test kits, PPE, and medical personnel. Literature suggests that use of machine learning (ML) to analyze patient responses to a clinical questionnaire can yield a reliable, remote COVID-19 diagnostic solution. As an online medical device rather than a physical one, COVID19orNo.com can be used an infinite number of times.

Another current issue is that of prediction: authorities and the public have been uncertain how the COVID-19 pandemic would unfold and will continue to affect the globe. In COVID19orNo.com, user location will be paired with other response data to track disease progression and predict outbreaks. This will enable effective pandemic response in the future by the healthcare industry and by policy-makers in general.

COVID19orNo.com will be an online COVID-19 diagnostic test that uses machine learning (ML) to evaluate patient-reported data and diagnose COVID-19 remotely. The product will be a software-only diagnostic medical device eligible for Emergency Use Authorization (EUA) by the U.S. Food and Drug Administration, potentially the first fully at-home diagnostic device for COVID-19.

To develop the ML, data must be collected. A clinical questionnaire study will be conducted of patients with known COVID-19 test results. Given that the ML output for each user will be a binary response (infected or not infected), we will test ML models that, when given a number of features, yield a probability of a user having contracted the disease, such as a logistic regression model, a random forest classifier, or an XGBoost classifier.  The model (or model ensemble) that best predicts subjects’ COVID-19 in-vitro diagnostic test result will be implemented in the product. UserWise personnel will work closely with a ML expert and will gather other expert input in order to make the diagnostic medical device both safe and effective.

The final product will ask the user to report on significant variables, and will provide the user a percent likelihood of COVID-19 positivity and relevant instructions.

COVID19orNo.com serves lay user respondents (e.g. patients or caregivers), medical personnel, and healthcare and government institutions. The product will replace COVID-19 diagnostic test kits, facilitating easy, reliable, at-home testing.

For respondents, this product will diagnose COVID-19 and ease their uncertainty about symptoms or other details they are concerned about. Users can complete the test an unlimited number of times from the comfort and safety of their homes, minimizing spread of SARS-CoV-2. The software will also educate the public by communicating information from CDC and WHO. Additionally, UserWise is experienced in designing simple, intuitive user interfaces for lay users.

For medical personnel, COVID19orNo.com will keep patients at home, reducing the large caseload of risky contact currently necessary to conduct in-person diagnostic testing. Remote testing can be conducted early and often with instant results, potentially identifying COVID-19 patients and enabling their monitoring and care earlier than currently possible.

For healthcare and government institutions, the use of our online product will mitigate current test kit shortages and amass useful regional data. Geographic location, percent likelihood of COVID-19 and other details can be used to model spread, arming policy-makers with the data to make informed decisions for their institution or for the public.

COVID19orNo.com is a disease spread tracking solution, collecting patient-reported data that can help predict emerging outbreaks. The product is a low-cost rapid diagnostic tool that reduces spread of SARS-CoV-2 by enabling full diagnostic testing at home, eliminating risk for mail carriers and lab personnel. Additionally, use of COVID19orNo.com will reduce the amount of contact for health workers and thus reduce infection rates. Policy decisions can be made based on the aggregated data.

This research and development can also inform approaches to future pandemics with the evaluation of the concept that ML can be used to diagnose medical conditions.

COVID19orNo.com will be an online COVID-19 diagnostic test that uses ML to evaluate patient-reported data and diagnose COVID-19 remotely. This platform will be a diagnostic medical device eligible for Emergency Use Authorization (EUA) by the U.S. Food and Drug Administration. To-date, there are no FDA-approved online COVID-19 diagnostic products.

Currently, COVID19orNo.com is a HIPAA-compliant website that tracks each user’s symptoms and provides them with applicable guidance from the U.S. CDC and WHO websites based on the symptoms they provide. Other online COVID-19 self-check systems exist, providing recommendations regarding what to do if the respondent has symptoms. However, none of these competitors is intended to be used for the diagnosis or treatment of the COVID-19 disease or other conditions.

COVID19orNo.com is built for simplicity and understandability. The current product is already unique in that it allows users to compare their symptoms and timeline of those symptoms with those of confirmed COVID-19 cases. It has an intuitive user interface designed by experts in human factors and website development, and it shows whether your symptoms are symptoms of COVID-19 or if they are unrelated. 

Additionally, we offer paper completion of the tool to underserved populations or those who do not have access to computing devices.

With collection of clinical data and ML model development, COVID19orNo.com has potential to be the first fully at-home software-only diagnostic medical device eligible for EUA by the FDA. Existing COVID-19 self-check systems do not incorporate machine learning to correlate patient-reported symptoms to positive COVID-19 results.

The ML behind COVID19orNo.com is developed using data collected in a questionnaire-based clinical study. Given that the ML output for each user will be a binary response (infected or not infected), we will test ML models that yield a probability of a user having contracted the disease, including logistic regression models, random forest classifiers, or XGBoost classifiers. These approaches all model the probability of a binary event, like disease contraction, given a number of features.

Upon testing various models, we will either ensemble them on top of one another, or simply pick the one with the highest test accuracy. Once model types are assessed, we can either choose to make one our primary model, or we can “stack” multiple models on top of one another and average their predicted responses to yield one final outcome. The website will be redesigned to ask about the variables needed for the ML to assess presence of COVID-19, and to provide the user a result of the percent likelihood that they are COVID-19 positive.

The user’s responses will be sent to the server containing the ML, and the ML will calculate the percent likelihood that the user is COVID-19 positive. The ML server will respond to the browser with this likelihood, to be displayed on the software. The user will then be shown their percent likelihood on the results page of the software, and their test results will be sent to their healthcare provider, if the subscribing healthcare facility chooses to activate this feature.

As an informational website, COVID19orNo.com has been live for about 4 months and has received more than 2300 responses across the globe, achieving great success in attracting respondents through campaigns on various social media platforms. We have also been offering customization of the existing informational product to city officials for inclusion on their city websites. Through this process, we have been evaluating user interest; we have received positive feedback from several cities and respondents about the informational site.

Our goal is to implement machine learning into our website to serve as an online diagnostic tool. Researchers at Columbia University demonstrate the effectiveness of machine learning and artificial intelligence are becoming more common as a method of the detection and diagnosis of disease. Additionally, a study of two hospitals in Wenzhou, Zhejiang, China demonstrates that AI with predictive analytics applied to historical COVID-19 patient data can effectively identify patients that develop acute respiratory distress syndrome, based on initial symptom presentation. 

Validation of the ML models developed from the questionnaire-based clinical study will demonstrate if diagnosis based on patient-reported information, rather than historical data, is feasible. 

Product Demonstration

COVID19orNo.com will be an automated, remote, at-home diagnostic software that will calculate the percent likelihood that the user is COVID-19 positive. The user will then be shown their percent likelihood on the results page of the software, and their test results will be sent to their healthcare provider, if the subscribing healthcare facility chooses to activate this feature. The test can be completed and the patient diagnosed in a matter of minutes.

These product details support our immediate goals to administer COVID-19 diagnostic testing remotely and promote understanding of COVID-19 symptoms, progression, and mitigation strategies. These immediate goals link to our long-term outcomes including reducing human contact and COVID-19 exposure typically required during testing, conserving limited resources like PPE and COVID-19 test kits, and combating the spread of COVID-19. To evaluate our predicted outcomes, we will monitor the number of people who use our product. Assuming there is a direct correlation between remote, at-home testing and reduced human contact, PPE usage, and spread of COVID-19, the results will demonstrate if remote testing with COVID-19orNo.com can effectively achieve our long-term outcomes.

The current informational product serves any user worldwide who chooses to use the site. Since deployment in March 2020, we have received over 2300 quiz responses.

Within six months of product launch, we hope to reach 5% of total reported COVID-19 tests in the U.S., or approximately 1.2 million people as of June 2020. In the next five years, COVID19orNo is projected to test 5% of the global population tested for COVID-19, or approximately 365 million people. Pending support from the National Science Foundation and Biomedical Advanced Research and Development Authority in the United States will accelerate these projections. Additionally, the product can be modified to apply to diagnosis or screening for other diseases; this project is also a proof-of-concept that ML can be used to diagnose disease and generally to study patterns behind patient reported symptoms and disease development. 

After we complete the clinical study, we will apply the ML to create an automated, remote and at-home diagnostic tool, our goal is to obtain Emergency Use Authorization through the U.S. FDA. With this authorization, we will shift to presenting our product as a software-only diagnostic medical device, with healthcare facilities as our main customers and partners. We also plan to run advertisements on social media such that patients whose healthcare facilities do not use our product may choose to be tested individually (and pay a small fee).

In the next year, the primary barriers are financial, market, and technical.

In order for our product to be successful, we must obtain funding from grants and awards to support development of the ML model and user interface. This financial support will also allow us to maintain the COVID19orNo.com platform as product adoption is just beginning.

It is difficult to gauge the extent to which the healthcare industry will adopt a remote, online diagnostic medical device for COVID-19. This project is both a proof-of-concept that ML applied to patient-reported data can diagnose COVID-19, as well as an experiment regarding market readiness for this technology. Its biggest assets are its low cost, low personnel and equipment requirements, reduced exposure to infected individuals, and instantaneous results.

The application of machine learning and artificial intelligence to diagnostic efforts is a relatively new field. Although data informs many diagnostics today, further research is necessary to determine if patient-reported data, rather than historical data, can serve as an accurate predictor of COVID-19 diagnosis. It is possible that our site will be sufficiently reliable to screen patients for testing with COVID-19 test kits, but not reliable enough to diagnose on its own.

Within 5 years, we hope that COVID-19 will no longer be a threat, yet predict that our product will be needed for the same or similar purposes. At that point, we will pursue international device approval, additionally assessing cultural and legal hurdles associated with expanding globally. 

Our software is currently being backed by our established company, UserWise, Inc. In order to sustain continued development, we are applying for grants and seeking partnerships to secure additional funding and technical expertise. We have applied for funding through the National Science Foundation and Biomedical Advanced Research and Development Authority. Our goal is to raise at least $500,000 by the end of the year.

UserWise is discussing the product concept with healthcare professionals and policy makers to assess market readiness and understand which features are most important to emphasize when promoting the product upon release.

To address our technical barriers, UserWise is consulting an ML expert and would like to partner with other institutions to ensure success and validate our results.

Regarding legal barriers in the next year, our team are experts on performing clinical studies and ensuring HIPAA compliance. As we will be working with human subjects, we have attained IRB approval to ensure safety of our patients. Additionally, the company has retained a qualified lawyer to consult and assist with the patent application process and will file a provisional patent application prior to submitting a non-provisional utility patent within one year of provisional patent submission.

Regarding legal and cultural barriers associated with transitioning to a global audience, our team will 1) use our regulatory expertise to comply with international requirements, and 2) use our expertise in understanding user populations to assess cultural impact.

UserWise is a consultancy of twelve full-time employees. We will assign 3 human factors engineers to this project and will consult 1 programmer part-time and 1 machine learning expert part-time. 

UserWise was established in 2014 to help medical device developers to design safer medical products and swiftly navigate FDA submissions. UserWise has become a leading expert in safety engineering globally. UserWise has expertise in partnering with manufacturers and working directly with the FDA to successfully develop novel medical devices and combination products. UserWise has a large portfolio, working with companies in all stages of development from startups to Fortune 500 companies.

In the light of COVID-19, UserWise has focused all efforts on combating this pandemic. Our team has an in-depth knowledge in conducting HIPAA-compliant, IRB-approved research studies. We have conducted several studies involving human participants and are adept in conducting interviews about medical products as well as collecting and rapidly analyzing data. Our team is also adept at website development, statistical analysis, machine learning and artificial intelligence. Please see relevant aspects of our core expertise:

• User research (e.g. user needs development, literature reviews to investigate known use errors)

• Website development (including data analytics, statistical analysis, and artificial intelligence development)

• Risk analysis (including development of design requirements to ensure safe use)

• Clinical testing preparation, recruitment, implementation, analysis, and reporting

• Regulatory submission authoring and strategy consulting (e.g. for Emergency Use Authorization, 510(k), de Novo, PMA, or Usability Engineering File for outside the U.S.)

• Process development and compliance (e.g. usability engineering procedure creation, complaint handling refinement to include human factors considerations, and gap analysis and compliance documentation for IEC 62366/IEC 62366-1, IEC 60601-1-6, & FDA Human Factors Guidances)

The global market for COVID-19 detection kits is growing at a compounded annual growth rate of 17.3% and is projected to reach nearly $8.5B by 2026. The manufacturing capacities of large companies collectively are not sufficient to meet the demand of COVID-19 tests; according to experts, the U.S. will need to deliver as many as 20M tests per day to fully remobilize the economy by late July. As a software-only, at-home diagnostic medical device for COVID-19, COVID19orNo.com will excel in this rapidly growing market.

COVID19orNo.com will primarily be marketed towards healthcare facilities. There are currently over 6,100 hospitals in the U.S. and approximately 93% of these allow patient access to electronic health records online. These healthcare facilities that currently utilize online platforms for patient services will comprise our initial target market.

COVID19orNo.com will be sold to healthcare facilities as a diagnostic medical device software license informed by patient-reported data to support clinical decision making for COVID-19. Healthcare facilities will distribute the test to their patients, as a quantitative tool to determine if the patient should come in for treatment and/or for in-vitro testing. Patients can access the user-friendly platform from any device with access to the internet.

As new information about COVID-19 is discovered, software will be updated by the provider and received by the subscribers using the SaaS cloud deployment model. Since the product is a diagnostic medical device, any software updates will comply with federal regulations and guidances on software as a medical device.

We are currently fundraising and applying for grants and awards to fund a questionnaire-based clinical study as well as development of the ML for COVID19orNo.com. After obtaining U.S. FDA Emergency Use Authorization for the product, we will proceed to sell it to healthcare facilities as a software-only medical device using a Software as a Service (SaaS) and Per User Per Month (PUPM) hybrid revenue model. Healthcare facilities will pay a rental fee for the software license to distribute to their patients. Patients of healthcare facilities without a COVID19orNo.com software license can access the software on their own and pay a fee to take the test.  

The software can also be marketed as a clinical study recruitment tool for future efforts. COVID19orNo.com users may choose to leave their email addresses for later contact regarding participation in clinical studies, or users may opt into allowing their results to be shared to a COVID-19 information database. Leveraging the software’s ability to be used as a fully remote recruitment tool in addition to an at-home diagnostic device will allow clinical studies an opportunity to accelerate recruitment efforts and generate additional revenue for the company.

COVID19orNo will initially focus on sales in the U.S., then scale globally to healthcare facilities and individual users worldwide.

Our team would greatly benefit from the support and expertise of the Solve and MIT networks, particularly in the areas of AI and business model development. The financial support and media exposure from being accepted into the Solve community would help launch COVID19orNo into the next level of commercialization, exposing people around the world to our industry-changing, reliable and remote diagnostic medical device.

Our organization would be honored to work with MIT faculty to collaborate on our clinical research study and technological development. 

MIT Computer Science & Artificial Intelligence Lab Clinical Decision-Making Group’s aim to bolster the urgent and critical nature of medical practice and timely clinical decision-making also agrees with COVID19orNo’s mission to address the increasing caseload of COVID-19 with a reliable, instantaneous, online COVID-19 diagnostic tool. Peter Szolovits and his team’s expertise with machine learning for clinical predictive analytics would also greatly supplement our clinical research study and technological development. Particularly, this research group’s familiarity with the proposed ML models: regression, decision trees, and support vector machines would support UserWise’s need for AI and ML expertise. 

The Clinical Machine Learning Group’s initiative to use machine learning and artificial intelligence to transform health care directly aligns with COVID19orNo’s goals. David Sontag and his team’s experience analyzing how various models use EHR data to create symptom correlations to disease would greatly inform our process. The group’s knowledge of how ML models performed for different demographic groups for various diseases would provide valuable insight for our team and help us to avoid potential setbacks and optimize performance when choosing the best models for our purposes. 

In addition to AI and ML support, organizations or other MIT faculty that have connections with healthcare facilities will be beneficial in finding research participants.

Now more than ever is it important to develop life-saving technologies that help diagnose and treat patients during this COVID-19 pandemic. Our remote, at-home diagnostic tool directly applies to the Health Security and Pandemics Challenge by utilizing Artificial Intelligence to diagnose and thus mitigate the spread of COVID-19. Our innovative technology will predict a questionnaire respondent’s likelihood of COVID-19 and provide recommendations and resources to patients necessary to succeed in these trying times. Our tool will be useful outside of being a diagnostic tool as well, as it will track disease progression and reduce emerging outbreaks.

Currently, COVID19orNo.com allows patients to take a symptoms quiz, see how their symptoms compare to reported COVID-19 symptoms, and receive recommendations based on the CDC and WHO. Our website offers a symptom tracker that is presented in an interactive, easy to understand format. Our goal is to make the website into a reliable, remote COVID-19 diagnostic test. Collection of clinical data and development of machine learning will transform the current platform website into a diagnostic medical device. The AI for Humanity Prize and Solver program will help us greatly in our product’s development and marketing efforts and provide the resources we need to facilitate this transformation. Once the diagnostic software is complete, we will attain Emergency Use Authorization by the U.S. Food and Drug Administration so that we can offer our service as a diagnostic COVID-19 test.

COVID19orNo.com will help reduce the shortage of PPE, address the shortage of COVID-19 test kits, reduce hospital congestion, help people stay healthy and identify relevant symptoms earlier/faster. Having a remote test will also allow people to maintain social distancing and reduce exposure while still getting the diagnostic care that they need. One unique feature of COVID19orNo.com is that it tracks symptoms and location of users. Using AI, we will track disease progression in cities and identify areas where symptoms are emerging. With this knowledge, hospitals, cities, and the government can work together to address issues proactively, rather than retroactively. This award will be able to help us elevate humanity by improving the health and safety of everyone, educating the public and helping fight against the spread of COVID-19.