This study will reach out to patients who have undergone diagnostic testing for the following respiratory illnesses from January 1st, 2018 to July 9th, 2023: COVID-19, Influenza, Rhinovirus, and Respiratory Syncytial Virus. This study aims to develop a forecasting model to predict infection onset prior to symptom onset using wearable device data and known symptom onset and test dates.
DUHS patients who have diagnostic testing for Influenza, COVID-19, Respiratory syncytial virus, and Rhinovirus testing within the past 5 years will be initially screened for an email address. Participants will learn about this study via email with a link to complete the survey. A Study ID will be generated for all individuals with an email. Participants will be asked to complete an e-consent via a REDCap survey. If participants have questions, they are provided with study contact information via e-mail. Participants will complete the survey which will have questions on prior symptoms and device ownership (anticipated time to complete: 5 minutes). If the participant owns one of the following wearable devices (Fitbit, Garmin, or Apple Watch), they will be sent to a redirect URL to login into their device account (for Fitbit or Garmin) or be provided with instructions to export their Healthkit data and dump their data into a unique Strongbox link (for Apple Watch). If participants choose to contribute their wearable device data to the study and the data obtained pass through data quality thresholds, they will receive compensation. There is no compensation for survey completion. The investigators will ask participants if they wish to be re-contacted for future studies related to this project. The investigators will collect endpoint data values from the wearable. These data will be used to estimate daily activity amounts and intensity (i.e., exercise and walking), standing, sleep amounts, sleep quality, heart rate variability, SpO2, respiratory rate, and heart rate. All of the wearable device data will be identified using a Study ID. The investigators will use statistical and machine learning models to develop personalized "baseline" models of health and detect anomalies that can help in identifying COVID-19 infection. The investigators will validate and test the sensitivity and specificity of our mode for detecting respiratory infection vs. no infection against symptom surveys and diagnostic testing as ground truth. The model testing and validation will be done separately for each brand of device and will be further modified according to the type of respiratory infection.
Study Type
OBSERVATIONAL
Enrollment
10,034
Duke University
Durham, North Carolina, United States
Develop a forecasting model to predict infection onset prior to symptom onset using the amount of time between known symptom onset and test dates
Known symptom onset and test dates will serve to validate the model
Time frame: 18 Months
Determine if there are signal differences that can differentiate the type of respiratory infection (e.g., COVID-19 vs. Influenza)
Time frame: 18 Months
Percentage of missingness in the wearable device data
Used to determine the performance of the forecasting model.
Time frame: 18 Months
Determine the performance of the forecasting model on a new viral strain through transfer learning
Time frame: 18 Months
Determine if there are physiological differences between initial infection and reinfection
Time frame: 18 Months
Determine if there are physiological differences between varying respiratory infections over time
Time frame: 18 Months
Determine the performance of the forecasting model based on the severity of symptoms
Time frame: 18 Months
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.