Abstract
Contact-tracing is an essential tool in order to mitigate the impact of pandemic such as the COVID-19. In order to achieve efficient and scalable contact-tracing in real time, digital devices can play an important role. While a lot of attention has been paid to analyzing the privacy and ethical risks of the associated mobile applications, so far much less research has been devoted to optimizing their performance and assessing their impact on the mitigation of the epidemic. We develop Bayesian inference methods to estimate the risk that an individual is infected. This inference is based on the list of his recent contacts and their own risk levels, as well as personal information such as results of tests or presence of syndromes. We propose to use probabilistic risk estimation in order to optimize testing and quarantining strategies for the control of an epidemic. Our results show that in some range of epidemic spreading (typically when the manual tracing of all contacts of infected people becomes practically impossible, but before the fraction of infected people reaches the scale where a lock-down becomes unavoidable), this inference of individuals at risk could be an efficient way to mitigate the epidemic. Our approaches translate into fully distributed algorithms that only require communication between individuals who have recently been in contact. Such communication may be encrypted and anonymized and thus compatible with privacy preserving standards. We conclude that probabilistic risk estimation is capable to enhance performance of digital contact tracing and should be considered in the currently developed mobile applications.
Introduction
The standard "trace-test-isolate" strategy is a cornerstone of public health during early-stage epidemics. However, its effectiveness diminishes rapidly when viral spread outpaces manual tracing efforts. While digital contact tracing (DCT) via Bluetooth-enabled proximity logging offers scalability, early implementations were often "binary" — notifying anyone near a positive case regardless of the actual transmission risk. This paper proposes a more sophisticated approach using Bayesian inference to transform DCT into a precise risk-estimation engine. By calculating the probability of infection based on contact history and health data, authorities can prioritize testing and quarantine more effectively, potentially containing outbreaks even with limited resources.
Methodology
The researchers frame epidemic spread as a stochastic process on a dynamic contact network and introduce two primary algorithms for "Bayesian Epidemic Tracing":
- Belief Propagation (BP): A message-passing algorithm that estimates the marginal probability of infection. It accounts for causal transmission chains and can handle non-Markovian dynamics, such as the specific infectiousness window of COVID-19.
- Mean-Field (MF) Approximation: A computationally lighter version that assumes neighbor infection statuses are independent. While slightly less accurate than BP, it requires less data exchange, enhancing privacy and mobile feasibility.
Both methods are designed to be distributed, requiring only local communication between devices that have been in contact, thus aligning with privacy-preserving standards. The algorithms were evaluated using two simulation environments: a Geometric Contact Model (a 2D random graph) and OpenABM-Covid19 (a realistic agent-based model).
Results and Key Insights
The study demonstrates that a perfect model of the virus is not required for effective inference; a simple Susceptible-Infected-Recovered (SIR) prior performed exceptionally well even in complex simulations. The core advantage of Bayesian inference lies in its ability to identify "silent spreaders" — individuals likely to be infected who serve as common links between confirmed cases.
As shown in the Geometric Contact Model simulations, BP and MF methods significantly outperform random testing (RG) and standard contact tracing (CT). As the number of daily observations (tests) increases, the Bayesian methods are able to suppress the epidemic curve much more effectively, particularly in the household-structured model (HH).
In the more realistic OpenABM simulations, the BP-based strategy demonstrated the ability to control an outbreak with as few as 1,000 to 1,500 tests per day.
Additional findings include:
- Robustness: BP maintained its effectiveness even with a 30% false-negative rate in medical tests.
- Adoption: While 100% adoption is ideal, significant delays in epidemic peaks were observed at 60-70% app adoption.
- Early Intervention: Starting the intervention on day 7 instead of day 10 allowed for control with ten times fewer tests.
Discussion
While promising, the implementation of these algorithms faces practical hurdles. Integrating BP into existing privacy protocols like DP-3T remains an engineering challenge. Furthermore, the computational load on mobile devices and the necessity of high public compliance with quarantine recommendations are critical factors for real-world success. Despite these challenges, probabilistic risk estimation offers a powerful tool for enhancing digital contact tracing and should be considered for future pandemic response strategies.
Summary updated