Disease Surveillance with Multi-modal Sensor Network & Data Analytics

We propose a low-cost multi-modal wireless sensor network enabled by real-time data analytics for tracking disease transmissions and outbreaks in human populations and environments. This early warning disease surveillance system will be piloted for COVID-19 surveillance in 4 low-income communities via continuous monitoring of wastewater and air environments.

Dr. Sheree Pagsuyoin; Primary Investigator

We propose a new early warning surveillance system for tracking disease outbreaks and transmissions based on our multi-modal sensor network and data analytics. Infectious disease outbreaks exert tremendous socioeconomic and health burdens on societies – the 2002 SARS epidemic cost US$54 billion in Asia alone, losses from 2014 West Africa Ebola outbreak is estimated at US$53 billion, and the ongoing COVID-19 pandemic has claimed 136 million cases and 2.9 million lost lives globally. While significant progress has been made in the fight against infectious diseases, there is a need for cost-effective strategies to reduce the burdens of disease transmission. Effective disease surveillance systems, particularly in vulnerable and underserved communities, can help detect potential outbreaks before they spiral out of control and cost lives and livelihoods.

Our target markets are local governments that need to monitor and identify potential disease outbreak and transmission within communities. Our sensor network can be deployed in remote areas with limited access to health services, enabling timely emergency response should an outbreak occur. Our sensor’s wireless connectivity allows real-time transmission of results to data centers, enabling rapid tracking of infection spread and hotspots, and informed decisions for mitigation strategies by the local authorities. Our universities, UMass Lowell (UML), Northeastern University (NU), and the University of the Philippines (UP), have a strong presence and well-established connections within our local communities through other  community projects. UML and NU are both doing surveillance of COVID-19 cases on campus using highly efficient protocols for sample collection and public data reporting. UP actively supports the Philippine government in COVID-19 response through several projects. We also have an ongoing year-long wastewater-based surveillance of COVID-19 cases in the Greater Lowell Region in partnership with local authorities. If successful, this project will deliver new sensor systems, which will have significant and meaningful impacts on preemptive outbreak mitigation and on regional economic development.

Our Disease Outbreak Surveillance System is low-cost, low-maintenance, and can be deployed in remote areas with limited access to health services. It continuously collects environmental data to predict the likelihood of outbreaks, and provides real-time data analytics with mobile-based data archiving and visualization for public reporting. These extensive environmental datasets can be used as a surrogate for or as complementary to human disease data, significantly reducing the costs needed to acquire health surveillance data from populations. Data from the surveillance system can be used by community authorities in health decision-making, and inform advisories for public health and safety. The sensor network is flexible and can be adapted for targeted surveillance of specific diseases in populations.

Our Disease Outbreak Surveillance System provides critical and timely information for public health decision making within the community. It functions as a continuous monitoring and early warning system for disease outbreaks. Data from the system can be used directly by health authorities to guide decisions on disease intervention and prevention, prioritization and deployment of health resources, and strategies for increased community resilience against disease outbreaks. In turn, these informed health decisions greatly benefit public health by preventing or containing disease outbreaks, and saving lives.

In Year 1, we will build the hardware and software for our pilot system deployment. In the latter part of Year 1, we will deploy our Surveillance System in a small low-income community (population 5,000) in Lowell, Massachusetts. During this time, we will collect environmental data to calibrate and refine our sensor system. In Year 2, we will expand our pilot sites to 2 other larger locations in Massachusetts (population 40,000 and 200,000), and actively engage with the community health advisory boards to communicate results and better understand the specific needs of the board with respect to disease surveillance. The latter will allow us to tailor the design of the sensor network and data analytics framework to the specific needs of the community. In Year 3, we will deploy our surveillance system in Diliman, Philippines. In the latter Part of Year 3, we will have completed our market feasibility study using the lessons learned from our pilot tests and a market survey to identify strategies for deployment in different settings and scale of monitored community.

We will evaluate our success based on the following metrics

• Sensor performance – high specificity and sensitivity in detecting SARS-CoV-2 in environmental media. Currently, our sensors have high sensitivity and specificity based on industry standards (i.e., able to detect SARS-CoV-2 at levels at or below what is typically present in air and water). During project deployment, we will test our sensors’ performance against new viral strains that may emerge, and compare sensor results with lab-based RT-qPCR.
• System performance – ability of the Surveillance Network to detect and report spikes in COVID-19 cases in pilot sites using field / environmental data. During pilot deployment, we will contrast our System’s performance against reported COVID-19 cases in the pilot sites. This metric mirrors what we currently do with wastewater-based surveillance via RT-qPCR where we compare results with COVID-19 case counts in our sampled community. 
• Community Engagement – We will directly engage with health authorities to improve the design of our surveillance network to better serve their needs for disease outbreak surveillance. We will conduct multiple stakeholder meetings and solicit feedback (group discussions, structured dialogues, opinion surveys) on the performance and usability of out surveillance system, and its impacts on their health decision making.

We identify three areas of potential challenges and have outlined corresponding mitigation strategies:

• Technical (Manufacture of Devices). We have partnered with several companies to help build components of our sensor and data analytics framework. With our industry network, we are confident that we have the resources for large scale manufacture. However, these resources may not be readily available in the Philippines and in other regions. Hence, we will study the feasibility of manufacturing sensors locally in pilot areas and in similar settings. 
• Deployment Logistics and Local Policy. Sensor installations in households and in sewers will require community buy-in and trust, and the support of local authorities to install and monitor the sensors. For these reasons, we chose pilot sites where we have existing strong connections with community leaders. Through regular open dialogues, we will closely collaborate with community leaders improve the performance of our disease surveillance system. We will document lessons learned from community engagement to identify strategies for future engagement with other communities.     
• Financial / Market. While our surveillance system is low-cost by design, we still need to evaluate the financial viability in low-income settings, hence we will study market feasibility in our pilot sites.

• University of Massachusetts Lowell (UMass Lowell)
• Northeastern University 
• University of the Philippines
• Cuebiq

Infectious disease outbreaks exert tremendous socioeconomic and health burdens on societies. Effective disease surveillance systems, particularly in vulnerable and underserved communities, can help detect potential outbreaks before they spiral out of control and cost lives and livelihoods. The support from Trinity Challenge can help build field-deployable units of the CAWM and e-Nose sensors system and implement the pilot test for COVID-19 surveillance in 4 local communities in the United States and the Philippines, which eventually can be scaled up for real-time identification and early warning of potential disease outbreaks in human populations.

Thank you, Trinity has connected UMass Lowell with Trinity Members, Northeastern University and Cuebiq, with whom we partnered in this proposed solution. 

In addition, we plan to partner with Facebook and Google to implement our solution. Our machine learning models and algorithms will require integrating data measurements from our sensor network with other relevant supporting data to model disease spread and identify at-risk communities. These data include but are not limited to: climate data, demographic profiles, population maps, case reports from local health centers, and community behavior (e.g., mobility). Facebook, through its Data for Good initiative, collects and maintains several datasets and tools, including Movement Range Maps, Disease Prevention Maps, and COVID-19 AI forecasting tools,  which are useful in tracking movement and disease spread. Custom local data for our pilot sites can be accessible through partnership with Facebook. Google, through its Cloud Public Datasets program, hosts a suite of COVID-19 related datasets on climate, socio-economic metrics, health, and mobility that can help train/update our machine learning models and algorithms.