Utilizing Wearable Device Data for Syndromic Surveillance: A Fever Detection Approach

Area of Science:

• Digital epidemiology
• Wearable technology
• Public health surveillance

Background:

• Wearable devices offer continuous physiological monitoring, showing potential for detecting infectious diseases, especially those with fever.
• Previous research indicates wearables can identify fever onset, but their utility for broader syndromic surveillance requires evaluation.

Purpose of the Study:

• To assess the effectiveness of wearable device data for general syndromic surveillance.
• To develop and evaluate a model for detecting self-reported fevers using physiological data from Oura Rings.

Main Methods:

• Utilized data from 63,153 participants wearing Oura Rings, combined with daily questionnaire responses.
• Constructed a dataset classifying days as fever (positive class) or non-fever (negative class) based on self-report and temperature.
• Trained a tree-based classifier using skin temperature, heart rate, and sleep data from nights preceding and following fever days.

Main Results:

• The model achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.85 and an Average Precision (AP) of 0.25.
• At 50% sensitivity, the calibrated model demonstrated a low false positive rate of 0.8%.
• Data from 16,794 participants contributed valid ground truth days, including 724 fever days and 342,430 non-fever days.

Conclusions:

• Leveraging wearable device data for live fever surveillance is feasible at a public health level.
• Implementing such models can enhance real-time detection of disease prevalence and spread during outbreaks.
• Wearable technology holds significant promise for advancing public health surveillance capabilities.