Graph Learning for Bidirectional Disease Contact Tracing on Real Human Mobility Data
Sofia Hurtado, Radu Marculescu · 2025
Read Graph Learning for Bidirectional Disease Contact Tracing on Real Human Mobility Data by Sofia Hurtado, Radu Marculescu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-18531.txt.
Introduction
Hurtado and Marculescu applied graph learning to contact tracing on real human mobility data in twenty twenty-five. For fast-spreading diseases with many silent cases, tracing must be swift and automated. Exposure apps warn individuals, but reconstructing the actual routes of transmission needs the contact network itself.
The authors define a new network measure, Infectious Path Centrality, which scores contacts by their role on likely transmission paths. A graph learning classifier using that measure identified important transmission events with an F1 score of ninety-four percent. They also tested bidirectional tracing, quarantining contacts both backward and forward in time. When only thirty percent of symptomatic people were tested, bidirectional tracing cut the effective reproduction rate by seventy-one percent, far outperforming forward-only tracing.
Catalog record
| CREATOR | Sofia Hurtado, Radu Marculescu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-18531
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Graph Learning for Bidirectional Disease Contact Tracing on Real Human Mobility Data Sofia Hurtado, Radu Marculescu Abstract For rapidly spreading diseases where many cases show no symptoms, swift and effective contact tracing is essential. While exposure notification applications provide alerts on potential exposures, a fully automated system is needed to track the infectious transmission routes. To this end, our research leverages large-scale contact networks from real human mobility data to identify the path of transmission. More precisely, we introduce a new Infectious Path Centrality network metric that informs a graph learning edge classifier to identify important transmission events, achieving an F1-score of 94%. Additionally, we explore bidirectional contact tracing, which quarantines individuals both retroactively and proactively, and compare its effectiveness against traditional forward tracing, which only isolates individuals after testing positive. Our results indicate that when only 30% of symptomatic individuals are tested, bidirectional tracing can reduce infectious effective reproduction rate by 71%, thus significantly controlling the outbreak. Fig. 1: Manual contact tracing involves collecting past interactions to construct a directed acyclic graph (DAG), where parent nodes are potential sources of infection for their child nodes (forming a contact tracing network). When identifying superspreaders, potential infections, or vaccination candidates, most studies use network analysis techniques such as betweenness centrality on networks with static interactions but dynamic node labels (i.e., health status). However, as illustrated in the contact networks from Day 1 and Day 2 (left), nodes with the highest betweenness centrality [ 1 ] do not necessarily hold significant roles in the contact tracing network (right). Instead, nodes with the highest value of our proposed metric, Infectious Path Centrality—which measures the number of paths connecting two positive leaf nodes—are often the most recent common ancestors, making them (and their offspring) crucial for targeted quarantines. We evaluate our metric by comparing its effectiveness in a bidirectional graph learning mitigation framework, which uses this new transmission network metric to identify and quarantine unseen branches of the disease, against traditional forward contact tracing that quarantines those who test positive. I Introduction Contact tracing has long been a cornerstone public health strategy for managing outbreaks of highly traceable diseases such as Ebola and Rabies [ 2 ] [ 3 ] . In such instances, the transmission routes are typically clear, as infectees can recall specific events like animal bites or direct contact with infectious individuals. However, managing pathogens with aerosol transmission and asymptomatic cases like SARS-CoV-2 [ 4 ] , RSV [ 5 ] , or Influenza [ 6 ] presents greater disease containment challenges, as seen during the COVID-19 pandemic. To contain a disease outbreak, manual contact tracers need to contact each person who tests positive to identify their recent contacts from the past 24-48 hours [ 7 ] . However, the effectiveness of manual tracing depends on the accuracy of people’s memories, their awareness of their surroundings, and the tracer’s ability to reach out to these contacts [ 8 ] . In the beginning of an outbreak, the goal of manual contact tracers is to determine the initial source, often referred to as ”patient zero.” However, as an outbreak escalates into an epidemic, manual tracing can become overwhelmed by rapidly spreading pathogens, especially when community spread occurs [ 9 ] [ 10 ] . This is because numerous undetected transmission events within the community can make it difficult for individuals to trace their infections to the original source. The goal of contact tracers then pivots to identifying untested paths of transmission to notify people who may be infectious. In response to the rapidly evolving SARS-CoV-2 virus, companies raced to digitize disease contact tracing. By monitoring interactions among individuals, modern systems can function as continuous disease surveillance tools. Exposure notification applications, for instance, alert users to recent contacts with infected individuals, leveraging advancements in location-tracking technology [ 11 ] . Despite these advances, a critical question remains: Can we effectively backtrack the chain of transmission, particularly for rapidly spreading diseases where many carriers are asymptomatic? With access to the Foursquare Mobility dataset [ 12 ] that contains visits at various Points of Interest (POIs), we present an always-ready disease surveillance system that tracks the past interactions and performs retroactive disease path detection. In addition, we propose a new propagation network metric, Infectious Path Centrality , that characterizes the centrality of a node along a path of a transmission on a contact tracing network; this is in contrast to the static betweenness centrality on the instantaneous contact networks [ 13 ] (Fig. 1 ). In this research, we have developed a graph learning framework that utilizes our newly proposed metric (Infectious Path Centrality) as node features. This framework automates the bidirectional contact tracing process, allowing it to identify and quarantine more individuals along the suspected transmission paths. To this end, our contributions are as follows: 1. We introduce a novel Infectious Path Centrality metric, which measures how central a node is among all infectious individuals in a contact tracing network. 2. We provide an automated bidirectional contact tracing graph learning framework using real mobility data that identifies transmission events along various paths. 3. We show that bidirectional contact tracing is more effective than forward contact tracing at reducing the disease’s effective reproduction rate. Taken together, our contributions can help build the technology needed to mitigate the next unknown disease outbreak. The remainder of this paper is organized as follows: Section II discusses prior work, Section III describes our approach, Section IV presents our experimental results. Finally, Section V summarizes our contributions. Fig. 2: Our experiment starts by processing Foursquare mobility data containing device dwell times at POIs into person-to-person contact networks (step 1). Two devices (i.e., people) are connected when they visit the same location within the same hour. We then apply an agent-based epidemiological SEIR ( Susceptible - Exposed - Infectious - Recovered ) model [ 14 ] on the dynamic contact networks (step 2). To keep track of transmission events, we form a contact tracing network where the parent is a potential source of infection to the child (step 3). For every transmission event (infection in step (2)), we add the infectious interaction to the contact tracing network, as well as all other interactions the infectee has on the day of infection. This mimics the manual contact tracing where someone recalls all of their interactions on the day they got infected. Step 4 consists of calculating our proposed Infectious Path Centrality metric that is then used as features in edge classification (step 5). The graph learning module classifies the edges in the contact tracing network as being ’infectious’ (i.e., a true transmission event), or ’non-infectious). Finally, we test the efficacy of our approach by comparing a population seeded with the same infectious individuals that undergoes no mitigation, forward contact tracing, and bidirectional contact tracing using our mitigation framework (step 6). II Prior Work and Novel Contribution In this section we present the relevant prior work in disease contact tracing, targeting strategies, and graph learning. II-A Disease Contact Tracing Traditionally, contact tracing has been a manual process, relying heavily on interviews conducted by trained health workers to identify individuals who might have been exposed to an infectious person. This method was instrumental during