Identify Hidden Spreaders of Pandemic over Contact Tracing Networks

Siyu Huang, Jun Sun, Liang Feng, Jiaqin Xie, Dihong Wang, Yuanliang HuarXiv 2021 · DOI · arXiv

Abstract

The COVID-19 infection cases have surged globally, causing devastations to both the society and economy. A key factor contributing to the sustained spreading is the presence of a large number of asymptomatic or hidden spreaders, who mix among the susceptible population without being detected or quarantined. Here we propose an effective non-pharmacological intervention method of detecting the asymptomatic spreaders in contact-tracing networks, and validated it on the empirical COVID-19 spreading network in Singapore. We find that using pure physical spreading equations, the hidden spreaders of COVID-19 can be identified with remarkable accuracy. Specifically, based on the unique characteristics of COVID-19 spreading dynamics, we propose a computational framework capturing the transition probabilities among different infectious states in a network, and extend it to an efficient algorithm to identify asymptotic individuals. Our simulation results indicate that a screening method using our prediction outperforms machine learning algorithms, e.g. graph neural networks, that are designed as baselines in this work, as well as random screening of infection's closest contacts widely used by China in its early outbreak. Furthermore, our method provides high precision even with incomplete information of the contract-tracing networks. Our work can be of critical importance to the non-pharmacological interventions of COVID-19, especially with increasing adoptions of contact tracing measures using various new technologies. Beyond COVID-19, our framework can be useful for other epidemic diseases that also feature asymptomatic spreading.

Introduction

A defining challenge of the COVID-19 pandemic is the "hidden spreader"—individuals who are asymptomatic or presymptomatic. Studies suggest these cases can account for a significant portion of infections, with some estimates reaching up to 80%. Because these individuals show no symptoms, they bypass traditional screening but continue to drive transmission. Current contact tracing often relies on exhaustive, untargeted screening of every close contact, which is resource-intensive and slow. Conversely, simple "neighbor containment" (testing only immediate contacts of known cases) often misses the broader chain of transmission. This research proposes a mathematical framework to rank individuals in a network by their probability of being a hidden spreader, allowing for surgical, high-precision testing interventions.

Methodology: A Physical Spreading Framework

The researchers moved away from "black-box" machine learning and instead embedded the empirical biology of COVID-19 into a Markovian transition framework. This approach models the pandemic as a physical process with fine-grained infection stages.

State Vector Modeling

Instead of a simple binary (Infected/Uninfected), each individual is represented by a state vector $\mathbf{Z}_i(t)$ covering five distinct states:

  • Susceptible (S)
  • Presymptomatic (P)
  • Asymptomatic (A)
  • Symptomatic Infectious (I)
  • Recovered (R)

To capture the specific probability distributions of incubation and recovery times, the $P$, $I$, and $A$ states are further broken down by the number of days an individual has been in that state. The model uses physical equations to calculate the probability of a node moving between states based on the infection status of its neighbors and known biological parameters.

Table of epidemiological parameters and distributions used to model the transition probabilities between infectious states.

As shown in the parameter table above, the model incorporates specific values for the basic reproduction number ($R_0 = 3.50$) and the fraction of asymptomatic infections ($p = 15\%$), alongside fitted distributions for the duration of various infectious states derived from clinical data.

Dynamic Screening Algorithm

The algorithm is iterative: when a person is screened and found positive, the network's risk profile is updated immediately, refining the search for the next most likely spreader. This "time-since-infection" tracking allows the model to capture the window of peak infectivity much more accurately than static models.

Key Results and Insights

The study validated its model using real-world datasets from Singapore (Singapore A and B) and high-resolution empirical social networks (ISA Network).

  • Domain Knowledge vs. Machine Learning: The physical equations derived from epidemiological theory significantly outperformed sophisticated Graph Neural Networks (GNNs) like Node2Vec, GCN, and GAT. This suggests that directly encoding the "rules" of viral transmission is more effective than attempting to learn them from noisy or sparse data.
  • Robustness to Incomplete Data: A critical finding for real-world application is that the method maintains high precision even when up to 80% of the links in the contact network are missing. This is vital because manual contact tracing is rarely 100% complete.
  • Flattening the Curve: In simulations, "neighbor containment" based on this algorithm was far more effective at suppressing daily new cases than standard quarantine strategies, even with limited daily testing capacity (e.g., testing only 2% of the population per day).

Conclusion

By leveraging pure physical spreading equations, this framework provides a high-precision tool for identifying hidden spreaders. While the model requires accurate estimates of biological parameters like $R_0$, its robustness to missing data and superior performance over generic ML models make it a powerful candidate for non-pharmacological intervention. Future work could adapt this framework for other diseases with latent or asymptomatic periods, such as influenza, by updating the transition probability distributions.

Summary updated

Abstract

The COVID-19 infection cases have surged globally, causing devastations to both the society and economy. A key factor contributing to the sustained spreading is the presence of a large number of asymptomatic or hidden spreaders, who mix among the susceptible population without being detected or quarantined. Here we propose an effective non-pharmacological intervention method of detecting the asymptomatic spreaders in contact-tracing networks, and validated it on the empirical COVID-19 spreading network in Singapore. We find that using pure physical spreading equations, the hidden spreaders of COVID-19 can be identified with remarkable accuracy. Specifically, based on the unique characteristics of COVID-19 spreading dynamics, we propose a computational framework capturing the transition probabilities among different infectious states in a network, and extend it to an efficient algorithm to identify asymptotic individuals. Our simulation results indicate that a screening method using our prediction outperforms machine learning algorithms, e.g. graph neural networks, that are designed as baselines in this work, as well as random screening of infection's closest contacts widely used by China in its early outbreak. Furthermore, our method provides high precision even with incomplete information of the contract-tracing networks. Our work can be of critical importance to the non-pharmacological interventions of COVID-19, especially with increasing adoptions of contact tracing measures using various new technologies. Beyond COVID-19, our framework can be useful for other epidemic diseases that also feature asymptomatic spreading
BibTeX
@misc{huang2021identify,
  author = {Siyu Huang and Jun Sun and Liang Feng and Jiaqin Xie and Dihong Wang and Yuanliang Hu},
  title = {{Identify Hidden Spreaders of Pandemic over Contact Tracing Networks}},
  year = {2021},
  eprint = {2103.09390},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2103.09390},
}
Figure