Developing Predictive Models for COVID-19 with Wearables Data (2021-2022)

Background 

The COVID-19 pandemic has resulted in over nine million infections and over 229,000 deaths in the U.S. alone as of October 30, 2020. To quickly identify and isolate new infections and clusters, public health officials are seeking new tools to target diagnostic testing for individuals who exhibit symptoms.

Previous research has demonstrated that wearable technologies can detect physiologic and behavioral changes when a user becomes infected with influenza, including a heightened resting heart rate, lower heart rate variability, decreased blood oxygen saturation, disturbed sleep, decreased physical activity and changes in wear habits. Together, these “digital biomarkers” form a signature of infection.

In April 2020, Duke launched CovIdentify to test the viability of using wearables to quickly identify individuals who may have contracted the coronavirus. The CovIdentify platform integrates information from widely used wearables with simple daily electronic self-reports on symptoms and social distancing, for up to 12 months. CovIdentify’s overarching objective is to implement existing digital biomarkers and establish new digital biomarkers to develop, validate and translate CovIdentify as a continuous screening tool. Since September 21, 2020, researchers have collected data from over 5,500 individuals. 

Project Description 

This project’s three goals are to develop digital biomarkers associated with COVID-19, provide visualizations of these biomarkers for both participants and researchers, and recruit members of underrepresented groups through community outreach. 

With the development of a large-scale database housing REDCAP survey data, wearable device data and iOS application data, the next step for this project is data visualization and analysis. The team will create dashboards that visualize the digital biomarkers to provide participants with an understanding of their health and give other researchers the tools to develop their own digital biomarkers. 

In addition, since the project will continue to collect users’ wearable and survey data, the team must ensure that the population is representative of the target population. Team members will continue the work with the Duke Clinical and Translational Science Institute (CTSI) Recruitment Innovation Center (RIC) to present to community groups (virtually) and deliver wearable devices to underrepresented groups through the approved in-person, contactless protocol. 

The team will pursue the following aims:

  1. Digital biomarker development: With the redesign of the database that stores wearable and survey   data from participants into an efficient structure, team members will be able develop digital biomarkers. 
  2. Dashboard development: The current iOS application can only pull data that is stored on a participant’s Apple Health Kit. The team will pull each individual’s data from the new database and display information in a simple yet informative manner.
  3. Community outreach: The team will not only seek to diversify the study population by improving the recruitment and outreach efforts but also provide ways to meet the needs of the community groups affected by COVID-19

Learn more about this project team by viewing the team's video.

Anticipated Outputs

Database system for storing, securing and retrieving study data; identification and validation of digital biomarkers; blog posts and publications

Student Opportunities

Ideally, this project team will be comprised of 5 graduate students and 12 undergraduate students. The project will include team members with backgrounds in both technical and communication skills. Students with interest in computer science, data science, public health and communications will be preferred. A key emphasis of this project is communicating the results of our findings to the lay population, and eagerness for learning how to use technology and data to improve people’s access to healthcare is critical. 

Undergraduate students will gain skills in conducting literature reviews, developing data science skills and learning about database structures and management. They will also practice communication skills by presenting during lab meetings and conveying their ideas in a succinct, clear manner. Master’s-level students will have the opportunity to further research developing digital biomarkers from this project and create their own thesis project. All students contributing to the project will participate in writing up at least one publication from this study. 

Current members of the 2020-2021 team are collaborating virtually. The Bass Connections and MIDS capstone members have been divided into four subteams. Each subteam meets with one of the Ph.D. students weekly to go over the team’s progress and the next week’s objectives. There is a large group meeting with all the subteams. During this hour-long meeting, each subteam presents several slides with figures or code chunks they have been working on. The faculty and Ph.D. student leaders also address any issues voiced by the team members and provide troubleshooting guidance. We will employ the same method of subdividing members of the 2021-2022 team. We will continue to have one or two graduate students leading each team and confirm weekly subteam and large team meetings prior to the semester. 

The optional summer component will be 10-12 weeks long and span from mid-to-late May to early-to-mid July 2021. Students will be expected to work 20 hours a week and will partake in virtual daily stand-up meetings.

Peter Cho will serve as the project manager.

Timing

Summer 2021 – Spring 2022 

  • Summer 2021 (optional): Conduct literature review and exploratory data analysis; verify existing biomarkers for COVID-19; implement machine learning algorithms for classifying and predicting illness trajectory; reproduce work for similar studies; develop initial figures for dashboards for visualizing and interacting with data 
  • Fall 2021: Create subteams based on project’s three aims; conduct literature review to gain an understanding of the field; build ML algorithms for digital biomarker development; connect with community groups and hand-off devices
  • Spring 2022: Add dashboard to website and create usernames/passwords for participants to log in on any platform; implement DL time-series-based algorithms for wearable data; establish community advisory board to meet with community leaders and identify needs that we can address

Crediting 

Academic credit available for fall and spring semesters; summer funding available

See related Data+ summer project, CovIdentify (2021), and earlier related team, Equity and Efficiency of Using Wearables Data for COVID-19 Monitoring (2020-2021).

 

Image: Video still from CovIdentify website

Video still from CovIdentify website.

Team Leaders

  • Peter Cho, Pratt–Biomedical Engineering–Ph.D. Student
  • Jessilyn Dunn, Pratt School of Engineering-Biomedical Engineering
  • Ryan Shaw, School of Nursing

/graduate Team Members

  • Oana Enache, Biostatistics - Master, Biomedical Data Sci-Master
  • Israel Golden, Master of Environmental Management
  • Joseph Hsieh, Interdisciplinary Data Science - Masters
  • Alena Kaloditzsa, Interdisciplinary Data Science - Masters
  • Xiao Lu, Interdisciplinary Data Science - Masters
  • Karnika Singh, Biomedical Engineering-PHD
  • Tommy Tseng, Interdisciplinary Data Science - Masters
  • Will Wang, Biomedical Engineering-PHD
  • Jiaman Wu, Interdisciplinary Data Science - Masters

/undergraduate Team Members

  • Yen Dinh, Biology (BS)
  • Ethan Ho, Biomedical Engineering (BSE)
  • Qi Khoo, Economics (BS), Computer Science (BS2)
  • Yvonne Kuo, Computer Science (BS)
  • Amrita Lakhanpal, Computer Science (BS)
  • Libba Lawrence, Electrical & Computer Egr(BSE), Computer Science (BSE2)
  • Aneesh Patil, Economics (BS), Political Science (AB2)
  • Rami Sbahi, Computer Science (BS), Statistical Science (BS2)
  • Amanda Stern, Computer Science (BS)
  • Bilge Tatar, Computer Science (BS), Neuroscience (BS2)
  • Jaclyn Xiao, Biomedical Engineering (BSE), Computer Science (BS2)

/yfaculty/staff Team Members

  • Leonor Corsino, School of Medicine-Medicine
  • Rosa Gonzalez-Guarda, School of Nursing
  • Leatrice Martin, Ctsi-Clinical

/zcommunity Team Members

  • Duke Clinical and Translational Science Institute, Community Engaged Research Initiative