Network Epidemics 1 (Chair: BRENNAN KLEIN)

As contact tracing is costly and usually provides subjective and fragmented data, study of infectious contact networks is difficult, especially for systems with complex transmission patterns. The genetic distance between pathogen sequence data, is correlated to the proximity of contacts on a transmission tree and is a direct insight into the contact patterns driving the spread of infection in a community or ecosystem. As such genetic distance can be used to determine the size of aggregation areas in spatial network by finding the best scales for aggregations of variables. The transmission dynamics of Highly Pathogenic Avian Influenza (HPAI) is complex as it spans multiple bird species and production systems. To quantify the scales of potentially infectious interactions we use the genetic distances between HPAI H5N1 sequences in the 2020-2022 from Great Britain with the set of other variables (i.e. number of cases, wild bird abundance and poultry farms count) calculated over various geographical scales1. We demonstrate that spatial scales of aggregation depend both on the variable and the time of analysis. Variables that are directly describing HPAI outbreaks (number of wild and domestic bird cases) have different scales of aggregation for the variable type and period of analysis, while indirectly correlated, constant in time variables (wild bird abundance and farm density) generally have the same scale for both analysed periods. Our study shows that the interactions driving complex transmission systems are variable, and separate spatial networks may need to be considered to represent a range of epidemiological interactions. The network with time dynamic spatial aggregation may be necessary for variables directly describing temporal changes of outbreaks.1 Gamża, A., Lycett, S., Sanchez, A. & Kao, R. Using sequence data to study spatial scales of interactions driving spread of Highly Pathogenic Avian Influenza in Great Britain.arXiv preprint arXiv:2411.10424(2024)

A new strain of the mpox virus (clade Ib) emerged in the eastern Democratic Republic of Congo (DRC) in September 2023, spreading rapidly to neighboring countries. While the transmission is primarily driven by sexual contacts, the prolonged reporting of mpox cases among young children suggests a significant role of household contacts. This outbreak in the DRC has driven the development of a network epidemic model that incorporates both age and two transmission modes - household and sexual - to project the required control effort and compare targeted intervention strategies under various scenarios. In our model, sexual contacts are repeated contacts within partnerships, and household contacts are treated as one-off contacts. Additionally, contact patterns are age dependent. The spread of an susceptible-exposedinfectious-recovered (SEIR) infection is superimposed on the network. This network model, combining the two types of contacts, is described by a set of ordinary differential equations (ODE), incorporating a configuration network structure and mass action contact structure, allowing for mathematical analysis. Model parameters were selected to reflect the clade Ib mpox outbreak setting in the DRC. Our approach enables us to compare model outcomes to collected data, such as age-specific incidence in the initial phase of an epidemic. Furthermore, we can explore optimal vaccination strategies at different phases of the epidemic, determine when the dominant mode of transmission might switch from sexual to household transmission, potentially requiring a change in intervention priorities. Our findings demonstrate that this analytically tractable mathematical model of epidemic spread on a network combining household and sexual contacts can provide valuable insights into real-life outbreaks. The clade Ib mpox outbreak in the DRC serves as a practical application, highlighting its potential for public health insights into optimal control strategies.

Background. The 2022 mpox outbreak saw a rapid case surge among men-who-have-sex-with-men (MSM) in previously unaffected regions, driven by heterogeneity in sexual networks. A sudden decline followed, but its drivers remain unclear as it is difficult to distinguish the roles of vaccination, herd immunity, and behavioral changes.Methods. We developed a network model of mpox transmission among MSM based on sexual behavior data (Figure 1a) and fitted it to the Paris region epidemic. We studied whether the decline was driven by post-exposure prophylaxis (PEP) vaccination, immunity among highly active MSM, or behavioral changes. Behavioral shifts were modeled as either uniform or based on individual risk factors, like sexual activity or exposure to diagnosed cases. We used the cross-sectional 2023 ERAS survey to validate findings.Findings. Behavioral changes adopted by 49% (95%CI 47-51%) of MSM regardless of individual risk factors best explained the observed epidemic decline (Figures 1b, 1c). These changes prevented an estimated 73% (28-99%) of mpox cases in summer 2022. Findings aligned with the ERAS survey data, showing that 46% (45-48%) of MSM reduced sexual partners (Figure 1c). On the contrary, PEP vaccination and immunity among highly active MSM were insufficient to curb the outbreak.Interpretation. Widespread behavioral change was the primary driver of the mpox epidemic decline in the Paris region, before preventive vaccination or immunity could affect epidemic spread. These findings highlight the importance of effective risk communication and community engagement in outbreak management. Tailored public health responses that encourage adaptive behaviors, especially as vaccination efforts ramp up, are essential for supporting affected communities.

Respiratory Syncytial Virus (RSV) infection is a major cause of acute respiratory hospitalizations in young children and older adults. In early 2020 most countries implemented non-pharmaceutical interventions (NPIs) to slow the spread of SARS-CoV-2. COVID-19 NPIs disrupted the transmission of RSV on a global scale, and many locations did not experience widespread recirculation until late 2020 or 2021. Here, we use a dynamic mathematical model to investigate the impacts of human mobility on the reemergence of RSV in Seattle, Washington.We characterized within-city mixing, in-flows, and foot traffic to points of interest (POIs) using SafeGraph mobility data. We calibrated an age-structured epidemic model to data on weekly RSV hospitalizations in Seattle-King County, WA (2017–2023), allowing for fluctuations in transmission due to changes in mobility during the pandemic. We focused on mobility metrics that may best approximate transmission-relevant contacts, spatial spread, or external case importations and used maximum likelihood to determine the best-fitting models.COVID-19 NPIs perturbed RSV seasonality from 2020 to 2022. Seattle experienced a small out-of-season outbreak in Summer 2021 and an atypically large and early wave in Fall 2022. Transmission models incorporating mobility network connectivity (measured as the mean shortest path length between Seattle neighborhoods) or the inflow of visitors residing outside of Seattle most accurately captured the timing and magnitude of the first two post-pandemic waves, outperforming models that included foot traffic to schools or child daycares (Figure 1).These results suggest that case importations from other regions and local spread between neighborhoods had the greatest influence on the timing of RSV reemergence in Seattle. Our study contributes crucial insights into the behavioral factors underlying RSV epidemic spread and could potentially inform preventative measures, such as the timing of immunizations.

Epidemic models are indispensable tools for operational tasks, for instance by revealing the infection dynamics between age groups during a pandemic situation. However, evaluating the performance of these models is often hindered by limited data availability; while advanced surveillance systems may report age-stratified case numbers, the number of infections between different age groups is generally unknown. We address this limitation by incorporating an additional dataset, clinical genetic sequences with age metadata, to infer age-infection matrices (AIMs) through a machine learning approach.As training data, we generate synthetic AIMs and genetic sequence data using the Susceptible-Infected-Recovered epidemic models on complete networks weighted by the age-contact structure collected in the POLYMOD survey in 9 European countries in 2006. We validate our estimator on synthetic data generated from the CoMix survey, collected during the COVID-19 pandemic. We apply our trained model to Belgian genetic sequences downloaded form the GISAID database, and obtain AIMs for each month of 2021. The obtained results are in agreement with, and extend the information in reported age-stratified case numbers. As an example, they reveal the diagonal and off-diagonal structures of the AIMs in the months of January and October 2021, respectively. These patterns match the structure of the contact matrices recorded by the CoMix survey at the same time, even though the CoMix matrices were not part of the training data. Beyond acting as a rough proxy for contact matrices when they are unavailable, the estimated AIMs can reveal additional factors influencing infection patterns, such as age-dependent susceptibility and immunity within the population. We believe that our proposed method is an ambitious and important step towards a new generation of operational epidemic models, which integrate genetic sequence data in addition to traditional surveillance reports.