Sensor-based surveillance for digitising real-time COVID-19 tracking in the USA (DETECT): a multivariable,...

Area of Science:

• Digital epidemiology
• Wearable technology in public health
• Infectious disease surveillance

Background:

• Traditional viral surveillance methods are limited by in-person data collection and outdated technology.
• Physiological and behavioral changes detected by sensors may precede COVID-19 symptom onset and diagnosis.

Purpose of the Study:

• To assess if continuous sensor data from smartwatches and fitness trackers can provide an early warning signal for COVID-19 activity.
• To evaluate the utility of passively collected physiological data for real-time disease tracking and forecasting.

Main Methods:

• A multivariable, population-based modeling study using data from adult participants (≥18 years) in the USA with connected smartwatches/fitness trackers.
• Collected daily resting heart rate and step count, identifying anomalous sensor days.
• Developed a negative binomial model predicting 7-day moving averages of COVID-19 case counts using historical data and anomalous sensor data.

Main Results:

• The model incorporating sensor data (H1) significantly outperformed the model without sensor data (H0) in predicting COVID-19 case counts.
• Pearson correlation coefficients for 12-day future predictions improved by 32.9% in California and 12.2% in the USA.
• Validation models showed significant correlations for real-time, 6-day, and 12-day future predictions.

Conclusions:

• Passively collected sensor data from consenting participants can provide real-time disease tracking and forecasting.
• Integration of wearable sensor data into viral surveillance programs is a promising strategy for enhanced public health monitoring.