Epidemic centrality - is there an underestimated epidemic impact of network peripheral nodes?

In the study of disease spreading on empirical complex networks in SIR model, initially infected nodes can be ranked according to some measure of their epidemic impact. The highest ranked nodes, also referred to as “superspreaders”, are associated to dominant epidemic risks and therefore deserve special attention. In simulations on studied empirical complex networks, it is shown that the ranking depends on the dynamical regime of the disease spreading. A possible mechanism leading to this dependence is illustrated in an analytically tractable example. In systems where the allocation of resources to counter disease spreading to individual nodes is based on their ranking, the dynamical regime of disease spreading is frequently not known before the outbreak of the disease. Therefore, we introduce a quantity called epidemic centrality as an average over all relevant regimes of disease spreading as a basis of the ranking. A recently introduced concept of phase diagram of epidemic spreading is used as a framework in which several types of averaging are studied. The epidemic centrality is compared to structural properties of nodes such as node degree, k-cores and betweenness. There is a growing trend of epidemic centrality with degree and k-cores values, but the variation of epidemic centrality is much smaller than the variation of degree or k-cores value. It is found that the epidemic centrality of the structurally peripheral nodes is of the same order of magnitude as the epidemic centrality of the structurally central nodes. The implications of these findings for the distributions of resources to counter disease spreading are discussed.

💡 Research Summary

The paper investigates how the epidemic impact of individual nodes in complex networks depends on the dynamical regime of disease spreading, and it proposes a new metric—epidemic centrality—to capture this dependence in a way that is useful for public‑health resource allocation. Using the stochastic Susceptible‑Infected‑Recovered (SIR) model, the authors first demonstrate that the ranking of “superspreaders” (nodes that, when initially infected, generate the largest final outbreak) is not fixed: the same network can produce very different rankings when the transmission probability p and the recovery probability q are varied. This hypothesis is tested on four empirical networks—a condensed‑matter collaboration network (≈27 500 nodes), a US Western‑states power‑grid (≈5 000 nodes), an astrophysics pre‑print co‑authorship network (≈16 700 nodes), and an Internet autonomous‑system (AS) graph (≈23 000 nodes)—as well as on synthetic Erdős‑Rényi graphs. For each node i, the authors compute the average final infected fraction X_i(p,q) for two distinct parameter pairs (p₁,q₁) and (p₂,q₂). Plotting X_i(p₁,q₁) versus X_i(p₂,q₂) yields a cloud of points rather than a monotonic curve, clearly indicating that node rankings can invert when the disease dynamics change. Specific examples (points A and B) illustrate cases where a peripheral node outranks a hub under one parameter set but not under another.

To address the practical problem that policymakers rarely know the exact disease parameters before an outbreak, the authors introduce epidemic centrality (EC). EC is defined as an average of a node’s epidemic impact over all relevant (p,q) values, using the recently proposed phase‑diagram of epidemic spreading as a framework. Three averaging schemes are examined: (i) uniform averaging over a discretized (p,q) grid, (ii) logarithmic weighting that emphasizes extreme regimes, and (iii) weighting based on empirical prevalence of different transmission scenarios. EC thus integrates the full phase diagram into a single scalar per node.

The authors then compare EC with traditional structural centralities—degree, k‑core index, and betweenness. While EC shows a positive trend with degree and k‑core (high‑degree or deep‑core nodes tend to have higher EC), the variation of EC across the network is far smaller than the variation of the structural measures. Crucially, nodes that are structurally peripheral (low degree, low k‑core) often have EC values comparable to those of central hubs. This suggests that peripheral nodes may have an underestimated epidemic impact when only structural metrics are considered.

An analytically tractable example further clarifies the mechanism. The authors construct a tree‑like network consisting of three high‑degree hubs (nodes 1, 2, 3) linked by chains of comparable length. Node 2, despite having a lower degree than nodes 1 and 3, occupies a central position. Using a bipartite infection model from prior work, they derive closed‑form expressions for the expected number of infected nodes when the outbreak starts at node 1 versus node 2. Both expectations depend on a single quantity P = P(X(1)=1), which encapsulates the combined effect of p and q. For small P (low transmissibility) the hub‑starting scenario yields a larger outbreak; for large P (high transmissibility) the centrally‑located lower‑degree node produces a larger outbreak. This illustrates how the same topology can generate opposite superspreader rankings under different dynamical regimes.

Finally, the paper discusses policy implications. Current epidemic preparedness often allocates vaccines, testing resources, or quarantine measures preferentially to high‑degree or high‑k‑core nodes. The findings argue that such strategies may be suboptimal when the disease parameters are uncertain. An EC‑based ranking, by averaging over all plausible regimes, offers a more robust basis for resource distribution, ensuring that peripheral nodes—potentially overlooked by structural metrics—receive appropriate attention. The authors suggest future extensions to dynamic networks, multi‑pathogen scenarios, and validation with real outbreak data.

In summary, the study reveals that superspreader identification is intrinsically dynamical, proposes epidemic centrality as a comprehensive, regime‑agnostic node importance measure, and demonstrates that peripheral nodes can possess epidemic influence comparable to that of network cores, thereby calling for a reassessment of how preventive resources are prioritized in complex networked populations.