A Smartphone Magnetometer-Based Diagnostic Test for Automatic Contact Tracing in Infectious Disease Epidemics

Abstract

Smartphone magnetometer readings exhibit high linear correlation when two phones coexist within a short distance. Thus, the detected coexistence can serve as a proxy for close human contact events, and one can conceive using it as a possible automatic tool to modernize the contact tracing in infectious disease epidemics. This paper investigates how good a diagnostic test it would be, by evaluating the discriminative and predictive power of the smartphone magnetometer-based contact detection in multiple measures. Based on the sensitivity, specificity, likelihood ratios, and diagnostic odds ratios, we find that the decision made by the smartphone magnetometer-based test can be accurate in telling contacts from no contacts. Furthermore, through the evaluation process, we determine the appropriate range of compared trace segment sizes and the correlation cutoff values that we should use in such diagnostic tests. INDEX TERMS Mobile sensing, human contact tracing, smartphone magnetometer, infectious disease epidemic, diagnostic test.

Epidemic contact tracing with smartphone sensors

“…The two vital observations of this work were that the electric currents powering the rail lines would alter the on-board magnetic field in such a way that people in different carriages experience various non-deterministic measures; and the fact that passengers must share the same journey between at least two consecutive stations. Similar works utilising the magnetic field to detect colocation of the users were reported in other environments, especially indoors with a high degree of magnetic anomalies due to the building infrastructure (Jeong, Kuk, and Kim 2019;Nguyen et al 2019a).…”

Technical Perspectives of Contact-Tracing Applications on Wearables for COVID-19 Control

“…For example, when a tagged user A emits a wireless signal, the receiving users in proximity could first estimate the distance based on a number of available characteristics as presented in [14]. In case the distance is lower than a predefined threshold, e.g., the value is lower than 2 m, as the safety threshold adopted by many research papers [2], [15], [16], then the receiving user will store the anonymized ID from user tagged user and the corresponding timestamps. Therefore, a ledger of neighboring users could be created per node.…”

“…Undoubtedly, P p depends on several other parameters, which are described in Fig. 2: i) the joint probability of two users found in vicinity of each other to use the same contact tracing application, e.g., with independent users and individual probability P u per user, the joint probability becomes P u *P u , ii) the false alarm P f a and misdetection P md probability of estimating that two users are within infectious distance from each other, e.g., at less than 2 m for more than 15' [2], [15], [16]), iii) the probability P c that the connectivity to the cloud server works properly, e.g., device of user A has access to long-range wireless connectivity to the server storing information about the temporary IDs of COVID-19 positive persons during their period of being infectious, and iv) the illness probability P i (i.e., the actual probability that user B gets the disease if (s)he was within infectious distance from a COVID-19 positive user A for a duration exceeding a threshold).…”

“…Large-scale fading shadowing models typically model such random fluctuations as described in [17], [18], where random signal effects in decibels (dB) or logarithmic scales are usually characterized by a random Gaussian distribution of zero mean and certain standard deviation. Such random fluctuations are especially crucial for RSS-based distance estimates, as a user at a far distance, i.e., outside the typically 'infectious range', can be easily mistaken to be at a nearby distance, i.e., inside the typically 'infectious range' of 2m [15].…”

“…The authors in [15] implemented a completely different approach by using the magnetic field intensity from magnetometers installed on smartphones instead of BLE RSS measurements. A centralized server processes the magnetometerbased metric in order to detect similar patterns of magnetic field intensities between users.…”