Detecting Pathogen Exposure During the Non-symptomatic Incubation Period Using Physiological Data: Proof of Concept...

Area of Science:

• Biomedical Engineering
• Infectious Disease Surveillance
• Computational Biology

Background:

• Early detection of infections is crucial for patient care and public health measures.
• Current methods often rely on overt clinical symptoms, delaying intervention.
• Developing methods for pre-symptomatic detection is a key challenge.

Purpose of the Study:

• To determine the limits of early warning detection using physiological measurements.
• To investigate the specificity of physiological responses to different pathogens.
• To assess the feasibility of using wearable devices for early infection detection.

Main Methods:

• A supervised random forest algorithm was developed to detect pathogen exposure.
• High-resolution physiological telemetry data (blood pressure, intrathoracic pressure, ECG, core temperature) from non-human primate models were used.
• Algorithm performance was evaluated across multiple viral and bacterial pathogen exposures and with ECG-only data.

Main Results:

• The algorithm provided a mean early warning of 51 ± 12 hours with an AUC of 0.93 ± 0.01.
• Performance was comparable across different pathogens (mean early warning 51 ± 14 h, AUC 0.95 ± 0.01).
• Using only ECG data resulted in minimal performance degradation (mean early warning 46 ± 14 h, AUC 0.91 ± 0.001).

Conclusions:

• Physiological measurements can provide over two days of pre-symptomatic infection warning.
• The detected physiological response is indicative of the host's immune response, not pathogen-specific.
• ECG-based detection is feasible, suggesting potential for non-invasive wearable devices.