Questions: Network Epidemiology and Disease Transmission

Question 1

Two populations have the same average contact rate (mean degree 5). Population A has nearly uniform contacts — everyone has about 5 connections. Population B is highly heterogeneous — most people have 1–2 contacts, but a few individuals have hundreds. Which is more vulnerable to epidemic spread?

• A. Population A — uniform contact rates create a more predictable and efficient transmission network
• B. Population B — high degree variance inflates the ratio ⟨k²⟩/⟨k⟩, lowering the epidemic threshold well below what the mean contact rate would predict
• C. They are equally vulnerable since the average contact rate is the same
• D. Population A — variance in contacts creates gaps that slow transmission

Answer: (B)

The epidemic threshold in a network model depends on ⟨k²⟩/⟨k⟩ — mean squared degree divided by mean degree — not simply on mean degree. When degree variance is large, mean squared degree is much larger than mean degree squared, inflating this ratio and dramatically lowering the threshold for epidemic spread. Population B's hubs simultaneously receive infection from many sources and transmit to many recipients, making even a weakly transmissible pathogen capable of sustaining an epidemic that a well-mixed model (using only average contact rate) would predict to die out.

Question 2

Public health officials have vaccine for 20% of a highly heterogeneous contact network. Which allocation strategy most efficiently reduces epidemic spread?

• A. Random vaccination of 20% of the population
• B. Vaccinating individuals with the most contacts first (targeted hub vaccination)
• C. Vaccinating geographically clustered groups regardless of contact count
• D. Vaccinating only people who have already been exposed and recovered

Answer: (B)

In heterogeneous networks, hubs are disproportionately responsible for epidemic spread — they receive infection from many sources and transmit to many recipients. Vaccinating hubs removes a disproportionate number of transmission pathways with each dose. Random vaccination of 20% is much less efficient: it mostly removes low-degree nodes that contribute little to epidemic spread. Targeted vaccination of hubs is a core insight of network epidemiology: the structure of the network determines which interventions are most leverage-efficient.

Question 3

In a network model and a well-mixed SIR model with the same average contact rate, the epidemic threshold is identical.

• T. True
• F. False

Answer: False

This is the central correction network epidemiology makes to classical SIR models. In a well-mixed model, the epidemic threshold depends only on mean contact rate. In a heterogeneous network, the threshold depends on ⟨k²⟩/⟨k⟩. When degree variance is large — as in real human contact networks with hubs — this ratio far exceeds the mean degree, meaning the effective threshold is much lower. A pathogen that would die out in a well-mixed population can sustain an epidemic in a heterogeneous network with the same average contact rate.

Question 4

Clustering in a contact network — the tendency for your contacts' contacts to also be your contacts — can slow epidemic spread between communities even while concentrating transmission within them.

• T. True
• F. False

Answer: True

Clustering creates triangles in the contact network: tightly knit groups where everyone knows everyone. Within these groups, transmission spreads rapidly. But the same clustering means that most edges are 'used up' locally — there are fewer long-range ties connecting communities. Long-range ties (bridges between clusters) are the critical conduits for epidemic expansion across communities. High clustering without bridging ties thus accelerates within-cluster spread while slowing between-cluster spread — explaining why community detection and targeted inter-community interventions can be effective.

Question 5

Why does degree variance — not just mean degree — determine epidemic vulnerability in a contact network?

Model answer: High-degree nodes (hubs) are disproportionately important for epidemic spread: they receive infection from many sources and transmit to many recipients simultaneously. When degree variance is high, the ratio ⟨k²⟩/⟨k⟩ is large, lowering the epidemic threshold. This means a pathogen can sustain spread in a heterogeneous network even when mean contact rate is low. Two populations with the same average contact rate but different variance have fundamentally different epidemic dynamics — variance is not noise around the mean, it is the key structural feature driving transmission.

The well-mixed SIR model loses all information about contact structure by collapsing heterogeneity into a single average. Network epidemiology recovers this information by tracking who contacts whom. The ⟨k²⟩/⟨k⟩ ratio appears because the probability that a randomly chosen contact belongs to a high-degree node is proportional to that node's degree — so an infected person is disproportionately likely to have been infected by a hub, and their random contact is disproportionately likely to be a hub as well. This 'friendship paradox' amplifies the epidemiological importance of hubs beyond what their numbers alone suggest.