Applied Network Science最新文献

Cancer care relies on effective coordination within a multidisciplinary care team. Changes to teams due to departures remain understudied despite rising oncologist turnover in the United States. In this study, we aimed to investigate the impact of oncologist departures on the remaining care team members. We used Medicare claims associated with beneficiaries aged 66-99 to identify physicians involved in care for common cancer types (i.e., breast, lung, and colorectal cancer). We restricted our analysis to medical oncologists, radiation oncologists, and surgeons specializing in oncology (collectively, "oncologists"). We identified oncologists who left a practice location in 2017-2019 using the Medicare Carrier file and linked them to retained oncologists based on shared patients. Multivariable hierarchical linear regression was used to investigate how retained oncologists' patient-sharing patterns changed after a colleague's departure. Our results support that retained rural-practicing oncologists experienced an expansion and restructuring of their patient-sharing ties following oncologist departures while retained urban-practicing oncologists experienced a consolidation. Network restructuring may demonstrate an adaptive response that ensures patient continuity of care, but it may also reflect unique challenges faced by oncologists practicing in rural versus urban settings.

Supplementary information: The online version contains supplementary material available at 10.1007/s41109-025-00762-3.

Temporal networks, whose network topology changes over time, are used to represent, e.g., opportunistic mobile networks, vehicle networks, and social contact networks, where two mobile devices (autos or individuals) are connected only when they are close to (interact with) each other. Such networks facilitate the transfer of information. In this paper, we address the problem of navigation on temporal networks: how to route a traffic demand from a source s to a destination d at time $t_s$ , based on the network observed before $t_s$ ? Whenever the node hosting the information has a contact or interacts with another node, the routing method has to decide whether the information should be forwarded to the contacted node or not. Once the information is forwarded, the contacted node becomes the only node hosting the information. Firstly, we introduce a framework of designing navigation algorithms, in which a distance metric is defined and computed between any node to the target d based on the network observed before $t_s$ . Whenever a hosting node has a contact, it forwards the information to the contacted node if the contacted node is closer to the target than the hosting node according to the distance metric. Secondly, we propose systematically distance metrics of a node pair in the temporal network observed, that capture different network properties of a node pair. Thirdly, these metrics or routing strategies are evaluated in empirical contact networks, from the perspective of the time duration of the routing and the probability that the destination can be reached. Their performance is further explained via the correlation between distance metrics and the stability of each metric in ranking nodes' distance to a target node. This work may serve as inspiration for evaluating and redesigning these strategies in other types of networks beyond physical contact networks.

The recent published literature on linear network autocorrelation models of actor behaviors or other mutable attributes has revealed a curious finding. Irrespective of the size of the network and the status of other network features, likelihood-based estimators (e.g., maximum likelihood and Bayesian) of the autocorrelation parameter ([Formula: see text]) are negatively biased and become increasingly so as the density of the network increases. In this paper we investigate the pattern of bias of estimators of [Formula: see text] when analyzing multiple mutually exclusive sub-networks and directed networks with various levels of reciprocity. In addition to considering the case of a linear network autocorrelation model applied to a binary-valued network, the edges may be weighted and the attribute whose actor-interdependence (or peer-effect) we are interested in may be an event (i.e., a binary outcome), a count, or a rate outcome motivating the use of generalized linear network autocorrelation models. We perform a simulation study that reveals that bias reduces substantially as either the number of sub-networks increases or with increased variation across the network in the edge weights but this pattern is not observed with reciprocity. The findings for generalized linear network autocorrelation models are in general similar to those for linear network autocorrelation models. Finally, we perform a statistical power analysis based on these findings for use in designing future studies whose goal is to estimate or to detect peer-effects.

Fully decentralised federated learning enables collaborative training of individual machine learning models on a distributed network of communicating devices while keeping the training data localised on each node. This approach avoids central coordination, enhances data privacy and eliminates the risk of a single point of failure. Our research highlights that the effectiveness of decentralised federated learning is significantly influenced by the network topology of connected devices and the initial conditions of the learning models. We propose a strategy for uncoordinated initialisation of the artificial neural networks based on the distribution of eigenvector centralities of the underlying communication network, leading to a radically improved training efficiency. Additionally, our study explores the scaling behaviour and the choice of environmental parameters under our proposed initialisation strategy. This work paves the way for more efficient and scalable artificial neural network training in a distributed and uncoordinated environment, offering a deeper understanding of the intertwining roles of network structure and learning dynamics.

We are generally interested in the association between a prescribing physician's position in a physician shared-patient network and their patients' receipt of risky drug combinations. An informal physician network (not restricted to a hospital or a health system) of physicians based in Ohio was constructed based on overlapping care of patients between physicians reflected in face-to-face visits in Fee-for-service Medicare claims for Ohio-residing beneficiaries. Separately, Medicare prescription drug events for beneficiaries receiving opioids, benzodiazepines, or non-benzodiazepine sedative hypnotics (sedative hypnotics) prescribed by these physicians in 2014 were used to map patients' drug status with respect to these three classes. We assigned patient prescription receipt to time-varying drug states and linked each drug state transition to a "responsible" prescribing physician. Outcomes of interest include transitions across drug states, particularly those resulting in combinations of increased risk (e.g., a benzodiazepine or sedative hypnotic with an opioid), and patients' time to discontinuation of overlapping prescriptions of an opioid, benzodiazepine, and a sedative hypnotic while the key predictors of these transitions reflected characteristics of a prescriber's physician network position and physician speciality. We found that among beneficiaries receiving none of the three risky drug groups, patients seeing physicians with higher closeness centrality (shorter average path lengths to other physicians through the network) were less likely to transition to two or three risky drugs; and they were more likely to discontinue overlapping prescriptions of an opioid, benzodiazepine, and sedative hypnotic. Compared to PCPs, psychiatrists appeared more likely to prescribe risky drug combinations, and their patients were less likely to discontinue overlapping three-drug prescriptions. This work demonstrates that characterizing physicians' prescribing behavior in relation to their position in shared-patient networks may reveal strategies for optimizing network-based interventions to improve prescribing quality.

Supplementary information: The online version contains supplementary material available at 10.1007/s41109-025-00721-y.

Hamill and Gilbert (J Artif Soc Soc Simul 12, 1-23, 2009) developed the Social Circles algorithm to generate synthetic networks that have properties of real social networks such as skewed degree distribution, positive clustering coefficient, degree assortativity and short path lengths. To assess the viability of Social Circles as a general network generator, we systematically examine the relationship between algorithm parameters and a broader range of structural properties of the generated networks. We varied social reaches for agents, distribution of social reaches in the population, and node density. We find that edge density and centrality measures can be controlled in a predictable way: longer reaches are associated with denser networks, shorter paths, lower degree assortativity (with some exceptions), and smaller variation in centrality measures. However, these network properties changed together and there is limited capacity to control properties separately. Further, clustering coefficient is insensitive to algorithm inputs. Thus, it cannot be used as a general network generator as it stands. If these properties are important, Social Circles could be used to generate starting networks with reasonable social structure, but further steps would be required to refine the structural properties.

The friendship paradox, initially discussed by Scott Feld in 1991, highlights a counterintuitive social phenomenon where individuals tend to have fewer friends than their friends do on average. The sociological implications of this paradox are profound, as it can create a distorted understanding of social norms and consequently influence beliefs, attitudes, and behaviors, particularly when highly connected individuals present a skewed representation of those norms. In essence, it can lead individuals to misjudge what is typical or desirable within their social circles. This study investigates the temporal dynamics of the friendship paradox using smartphone communication data from over 600 incoming freshmen at the University of Notre Dame participating in the NetHealth project. By tracking the friendship index- the ratio of an individual's friends' average number of friends to their own number of friends- over 119 days during the Fall semester of 2015, we examine how the paradox evolves over time. Our findings reveal that the friendship index stabilizes more rapidly than both the individuals' own degree and the variation among their friends' degrees. Results from the latent growth-curve model (LGCM) confirm that while the friendship index continues to increase, its growth rate declines over time. Moreover, the LGCM identifies individual degrees, ethnic backgrounds, and personality traits as influential factors shaping the manifestation and development of the friendship paradox. By exploring the mechanisms underlying this paradox in a dynamic communication network, this study enhances our understanding of the structural factors influencing the evolution of the friendship paradox in digitally mediated interactions.

Supplementary information: The online version contains supplementary material available at 10.1007/s41109-025-00710-1.

In models of infectious disease dynamics, the incorporation of contact network information allows for the capture of the non-randomness and heterogeneity of realistic contact patterns. Oftentimes, it is assumed that this underlying network is known with perfect certainty. However, in realistic settings, the observed data usually serves as an imperfect proxy of the actual contact patterns in the population. Furthermore, event times in observed epidemics are not perfectly recorded; individual infection and recovery times are often missing. In order to conduct accurate inferences on parameters of contagion spread, it is crucial to incorporate these sources of uncertainty. In this paper, we propose the use of Network-augmented Mixture Density Network-compressed ABC (NA-MDN-ABC) to learn informative summary statistics for the available data. This method will allow for Bayesian inference on the parameters of a contagious process, while accounting for imperfect observations on the epidemic and the contact network. We will demonstrate the use of this method on simulated epidemics and networks, and extend this framework to analyze the spread of Tattoo Skin Disease (TSD) among bottlenose dolphins in Shark Bay, Australia.

We present a novel approach for computing a variant of eigenvector centrality for multilayer networks with inter-layer constraints on node importance. Specifically, we consider a multilayer network defined by multiple edge-weighted, potentially directed, graphs over the same set of nodes with each graph representing one layer of the network and no inter-layer edges. As in the standard eigenvector centrality construction, the importance of each node in a given layer is based on the weighted sum of the importance of adjacent nodes in that same layer. Unlike standard eigenvector centrality, we assume that the adjacency relationship and the importance of adjacent nodes may be based on distinct layers. Importantly, this type of centrality constraint is only partially supported by existing frameworks for multilayer eigenvector centrality that use edges between nodes in different layers to capture inter-layer dependencies. For our model, constrained, layer-specific eigenvector centrality values are defined by a system of independent eigenvalue problems and dependent pseudo-eigenvalue problems, whose solution can be efficiently realized using an interleaved power iteration algorithm. We refer to this model, and the associated algorithm, as the Constrained Multilayer Centrality (CMLC) method. The characteristics of this approach, and of standard techniques based on inter-layer edges, are demonstrated on both a simple multilayer network and on a range of random graph models. An R package implementing the CMLC method along with example vignettes is available at https://hrfrost.host.dartmouth.edu/CMLC/.