Comparing Contact Tracing Through Bluetooth and GPS Surveillance Data: Simulation-Driven Approach.

Area of Science:

• Epidemiology
• Computational Biology
• Network Science

Background:

• Epidemiological simulations are vital for pandemic response, but rely on accurate contact data.
• Estimating individual contact parameters is challenging with traditional methods.
• Digital contact tracing (Bluetooth, GPS) offers higher-resolution data but has limitations like false positives/negatives.

Purpose of the Study:

• To compare the impact of GPS-colocated versus Bluetooth-beacon-derived proximity contact data on disease transmission models.
• To assess how different sensing modalities affect simulation outcomes for various diseases.

Main Methods:

• Collected proximity contact data using both Bluetooth and GPS on smartphones over one month from university students and staff.
• Compared contact network structures derived from GPS and Bluetooth data.
• Assessed the impact of sensing modalities on agent-based susceptible-exposed-infectious-removed (SEIR) transmission models for four diseases, analyzing incidence, attack rate, and infection risk.

Main Results:

• Contact networks derived from GPS and Bluetooth showed significant structural differences.
• Sensing modality significantly impacted estimated attack rates, with differences noted across distance thresholds, modalities, and simulated diseases (P<.001).
• Violin plots and Kullback-Leibler divergence revealed discernible differences in attack rates and individual infection risks between sensing modalities.

Conclusions:

• Contact networks generated by GPS and Bluetooth differ substantially, leading to significantly different disease attack rates.
• While both digital methods improve contact tracing resolution, the choice of sensing modality critically influences simulation outcomes.
• Future studies using single-measure contact data must acknowledge and address the differential impact of measurement modalities.