How to Estimate Epidemic Risk from Incomplete Contact Diaries Data?

An algorithm to build synthetic temporal contact networks based on close-proximity interactions data

Small populations (e.g., hospitals, schools or workplaces) are characterised by high contact heterogeneity and stochasticity affecting pathogen transmission dynamics. Empirical individual contact data provide unprecedented information to characterize such heterogeneity and are increasingly available, but are usually collected over a limited period, and can suffer from observation bias. We propose an algorithm to stochastically reconstruct realistic temporal networks from individual contact data in healthcare settings (HCS) and test this approach using real data previously collected in a long-term care facility (LTCF). Our algorithm generates full networks from recorded close-proximity interactions, using hourly inter-individual contact rates and information on individuals' wards, the categories of staff involved in contacts, and the frequency of recurring contacts. It also provides data augmentation by reconstructing contacts for days when some individuals are present in the HCS without having contacts recorded in the empirical data. Recording bias is formalized through an observation model, to allow direct comparison between the augmented and observed networks. We validate our algorithm using data collected during the i-Bird study, and compare the empirical and reconstructed networks. The algorithm was substantially more accurate to reproduce network characteristics than random graphs. The reconstructed networks reproduced well the assortativity by ward (first-third quartiles observed: 0.54-0.64; synthetic: 0.52-0.64) and the hourly staff and patient contact patterns. Importantly, the observed temporal correlation was also well reproduced (0.39-0.50 vs 0.37-0.44), indicating that our algorithm could recreate a realistic temporal structure. The algorithm consistently recreated unobserved contacts to generate full reconstructed networks for the LTCF. To conclude, we propose an approach to generate realistic temporal contact networks and reconstruct unobserved contacts from summary statistics computed using individual-level interaction networks. This could be applied and extended to generate contact networks to other HCS using limited empirical data, to subsequently inform individual-based epidemic models.

Reporting delays: A widely neglected impact factor in COVID-19 forecasts

Epidemic forecasts are only as good as the accuracy of epidemic measurements. Is epidemic data, particularly COVID-19 epidemic data, clean, and devoid of noise? The complexity and variability inherent in data collection and reporting suggest otherwise. While we cannot evaluate the integrity of the COVID-19 epidemic data in a holistic fashion, we can assess the data for the presence of reporting delays. In our work, through the analysis of the first COVID-19 wave, we find substantial reporting delays in the published epidemic data. Motivated by the desire to enhance epidemic forecasts, we develop a statistical framework to detect, uncover, and remove reporting delays in the infectious, recovered, and deceased epidemic time series. Using our framework, we expose and analyze reporting delays in eight regions significantly affected by the first COVID-19 wave. Further, we demonstrate that removing reporting delays from epidemic data by using our statistical framework may decrease the error in epidemic forecasts. While our statistical framework can be used in combination with any epidemic forecast method that intakes infectious, recovered, and deceased data, to make a basic assessment, we employed the classical SIRD epidemic model. Our results indicate that the removal of reporting delays from the epidemic data may decrease the forecast error by up to 50%. We anticipate that our framework will be indispensable in the analysis of novel COVID-19 strains and other existing or novel infectious diseases.

Estimating household contact matrices structure from easily collectable metadata

Contact matrices are a commonly adopted data representation, used to develop compartmental models for epidemic spreading, accounting for the contact heterogeneities across age groups. Their estimation, however, is generally time and effort consuming and model-driven strategies to quantify the contacts are often needed. In this article we focus on household contact matrices, describing the contacts among the members of a family and develop a parametric model to describe them. This model combines demographic and easily quantifiable survey-based data and is tested on high resolution proximity data collected in two sites in South Africa. Given its simplicity and interpretability, we expect our method to be easily applied to other contexts as well and we identify relevant questions that need to be addressed during the data collection procedure.

Spatial immunization to abate disease spreading in transportation hubs

Proximity social interactions are crucial for infectious diseases transmission. Crowded agglomerations pose serious risk of triggering superspreading events. Locations like transportation hubs (airports and stations) are designed to optimize logistic efficiency, not to reduce crowding, and are characterized by a constant in and out flow of people. Here, we analyze the paradigmatic example of London Heathrow, one of the busiest European airports. Thanks to a dataset of anonymized individuals' trajectories, we can model the spreading of different diseases to localize the contagion hotspots and to propose a spatial immunization policy targeting them to reduce disease spreading risk. We also detect the most vulnerable destinations to contagions produced at the airport and quantify the benefits of the spatial immunization technique to prevent regional and global disease diffusion. This method is immediately generalizable to train, metro and bus stations and to other facilities such as commercial or convention centers.

C, Galloway D, Schweizer ML, Guclu H, Cousins J, Lingle C, Yearwood GMH, Li K, Calderone P, Noble E, Gao H, Rainey J, Uzicanin A, Read JM. Age-specific social mixing of school-aged children in a US setting using proximity detecting sensors and contact surveys

Comparisons of the utility and accuracy of methods for measuring social interactions relevant to disease transmission are rare. To increase the evidence base supporting specific methods to measure social interaction, we compared data from self-reported contact surveys and wearable proximity sensors from a cohort of schoolchildren in the Pittsburgh metropolitan area. Although the number and type of contacts recorded by each participant differed between the two methods, we found good correspondence between the two methods in aggregate measures of age-specific interactions. Fewer, but longer, contacts were reported in surveys, relative to the generally short proximal interactions captured by wearable sensors. When adjusted for expectations of proportionate mixing, though, the two methods produced highly similar, assortative age-mixing matrices. These aggregate mixing matrices, when used in simulation, resulted in similar estimates of risk of infection by age. While proximity sensors and survey methods may not be interchangeable for capturing individual contacts, they can generate highly correlated data on age-specific mixing patterns relevant to the dynamics of respiratory virus transmission.

Ordinal Decision-Tree-Based Ensemble Approaches: The Case of Controlling the Daily Local Growth Rate of the COVID-19 Epidemic

In this research, we develop ordinal decision-tree-based ensemble approaches in which an objective-based information gain measure is used to select the classifying attributes. We demonstrate the applicability of the approaches using AdaBoost and random forest algorithms for the task of classifying the regional daily growth factor of the spread of an epidemic based on a variety of explanatory factors. In such an application, some of the potential classification errors could have critical consequences. The classification tool will enable the spread of the epidemic to be tracked and controlled by yielding insights regarding the relationship between local containment measures and the daily growth factor. In order to benefit maximally from a variety of ordinal and non-ordinal algorithms, we also propose an ensemble majority voting approach to combine different algorithms into one model, thereby leveraging the strengths of each algorithm. We perform experiments in which the task is to classify the daily COVID-19 growth rate factor based on environmental factors and containment measures for 19 regions of Italy. We demonstrate that the ordinal algorithms outperform their non-ordinal counterparts with improvements in the range of 6–25% for a variety of common performance indices. The majority voting approach that combines ordinal and non-ordinal models yields a further improvement of between 3% and 10%.

Age-specific social mixing of school-aged children in a US setting using proximity detecting sensors and contact surveys

Comparisons of the utility and accuracy of methods for measuring social interactions relevant to disease transmission are rare. To increase the evidence base supporting specific methods to measure social interaction, we compared data from self-reported contact surveys and wearable proximity sensors from a cohort of schoolchildren in the Pittsburgh metropolitan area. Although the number and type of contacts recorded by each participant differed between the two methods, we found good correspondence between the two methods in aggregate measures of age-specific interactions. Fewer, but longer, contacts were reported in surveys, relative to the generally short proximal interactions captured by wearable sensors. When adjusted for expectations of proportionate mixing, though, the two methods produced highly similar, assortative age-mixing matrices. These aggregate mixing matrices, when used in simulation, resulted in similar estimates of risk of infection by age. While proximity sensors and survey methods may not be interchangeable for capturing individual contacts, they can generate highly correlated data on age-specific mixing patterns relevant to the dynamics of respiratory virus transmission.

Evaluating How Smartphone Contact Tracing Technology Can Reduce the Spread of Infectious Diseases: The Case of COVID-19

Detecting and controlling the diffusion of infectious diseases such as COVID-19 is crucial to managing epidemics. One common measure taken to contain or reduce diffusion is to detect infected individuals and trace their prior contacts so as to then selectively isolate any individuals likely to have been infected. These prior contacts can be traced using mobile devices such as smartphones or smartwatches, which can continuously collect the location and contacts of their owners by using their embedded localisation and communications technologies, such as GPS, Cellular networks, Wi-Fi, and Bluetooth. This paper evaluates the effectiveness of these technologies and determines the impact of contact tracing precision on the spread and control of infectious diseases. To this end, we have created an epidemic model that we used to evaluate the efficiency and cost (number of people quarantined) of the measures to be taken, depending on the smartphone contact tracing technologies used. Our results show that in order to be effective for the COVID-19 disease, the contact tracing technology must be precise, contacts must be traced quickly, and a significant percentage of the population must use the smartphone contact tracing application. These strict requirements make smartphone-based contact tracing rather ineffective at containing the spread of the infection during the first outbreak of the virus. However, considering a second wave, where a portion of the population will have gained immunity, or in combination with some other more lenient measures, smartphone-based contact tracing could be extremely useful.

Estimating the outcome of spreading processes on networks with incomplete information: A dimensionality reduction approach

Recent advances in data collection have facilitated the access to time-resolved human proximity data that can conveniently be represented as temporal networks of contacts between individuals. While the structural and dynamical information revealed by this type of data is fundamental to investigate how information or diseases propagate in a population, data often suffer from incompleteness, which possibly leads to biased estimations in data-driven models. A major challenge is thus to estimate the outcome of spreading processes occurring on temporal networks built from partial information. To cope with this problem, we devise an approach based on non-negative tensor factorization, a dimensionality reduction technique from multilinear algebra. The key idea is to learn a low-dimensional representation of the temporal network built from partial information and to use it to construct a surrogate network similar to the complete original network. To test our method, we consider several human-proximity networks, on which we perform resampling experiments to simulate a loss of data. Using our approach on the resulting partial networks, we build a surrogate version of the complete network for each. We then compare the outcome of a spreading process on the complete networks (nonaltered by a loss of data) and on the surrogate networks. We observe that the epidemic sizes obtained using the surrogate networks are in good agreement with those measured on the complete networks. Finally, we propose an extension of our framework that can leverage additional data, when available, to improve the surrogate network when the data loss is particularly large.

Estimating the epidemic risk using non-uniformly sampled contact data

Many datasets describing contacts in a population suffer from incompleteness due to population sampling and underreporting of contacts. Data-driven simulations of spreading processes using such incomplete data lead to an underestimation of the epidemic risk, and it is therefore important to devise methods to correct this bias. We focus here on a non-uniform sampling of the contacts between individuals, aimed at mimicking the results of diaries or surveys, and consider as case studies two datasets collected in different contexts. We show that using surrogate data built using a method developed in the case of uniform population sampling yields an improvement with respect to the use of the sampled data but is strongly limited by the underestimation of the link density in the sampled network. We put forward a second method to build surrogate data that assumes knowledge of the density of links within one of the groups forming the population. We show that it gives very good results when the population is strongly structured, and discuss its limitations in the case of a population with a weaker group structure. These limitations highlight the interest of measurements using wearable sensors able to yield accurate information on the structure and durations of contacts.

Disease Spreading in Time-Evolving Networked Communities

Human communities are organized in complex webs of contacts that may be represented by a graph or network. In this graph, vertices identify individuals and edges establish the existence of some type of relations between them. In real communities, the possible edges may be active or not for variable periods of time. These so-called temporal networks typically result from an endogenous social dynamics, usually coupled to the process under study taking place in the community. For instance, disease spreading may be affected by local information that makes individuals aware of the health status of their social contacts, allowing them to reconsider maintaining or not their social contacts. Here we investigate the impact of such a dynamical network structure on disease dynamics, where infection occurs along the edges of the network. To this end, we define an endogenous network dynamics coupled with disease spreading. We show that the effective infectiousness of a disease taking place along the edges of this temporal network depends on the population size, the number of infected individuals in the population and the capacity of healthy individuals to sever contacts with the infected, ultimately dictated by availability of information regarding each individual’s health status. Importantly, we also show how dynamical networks strongly decrease the average time required to eradicate a disease.