Generation interval contraction and epidemic data analysis

The generation interval is the time between the infection time of an infected person and the infection time of his or her infector. Probability density functions for generation intervals have been an important input for epidemic models and epidemic data analysis. In this paper, we specify a general stochastic SIR epidemic model and prove that the mean generation interval decreases when susceptible persons are at risk of infectious contact from multiple sources. The intuition behind this is that when a susceptible person has multiple potential infectors, there is a ``race’’ to infect him or her in which only the first infectious contact leads to infection. In an epidemic, the mean generation interval contracts as the prevalence of infection increases. We call this global competition among potential infectors. When there is rapid transmission within clusters of contacts, generation interval contraction can be caused by a high local prevalence of infection even when the global prevalence is low. We call this local competition among potential infectors. Using simulations, we illustrate both types of competition. Finally, we show that hazards of infectious contact can be used instead of generation intervals to estimate the time course of the effective reproductive number in an epidemic. This approach leads naturally to partial likelihoods for epidemic data that are very similar to those that arise in survival analysis, opening a promising avenue of methodological research in infectious disease epidemiology.

💡 Research Summary

This paper investigates how the generation interval—the time between an infector’s infection and that of his or her infectee—behaves in realistic epidemic settings where susceptibles may be exposed to multiple potential infectors simultaneously. The authors first formalize a general stochastic SIR (Susceptible‑Infectious‑Recovered) model in which each infectious individual generates infectious contacts according to a time‑varying hazard function. For any susceptible, the times of first contact from each infectious source are independent exponential variables, and the actual infection occurs at the minimum of these times. By elementary properties of minima of exponential variables, the expected waiting time until infection, and consequently the expected generation interval, is shown to be inversely proportional to the sum of the individual hazards. Hence, as the number of concurrent infectious sources rises, the mean generation interval contracts.

Two distinct “competition” mechanisms are distinguished. Global competition refers to the situation where the overall prevalence of infection in the population is high, so that a typical susceptible is likely to be simultaneously exposed to many infectors. In this case the contraction of the generation interval is a population‑wide phenomenon that intensifies as the epidemic grows. Local competition, on the other hand, arises when transmission is rapid within tightly knit contact clusters (e.g., households, schools, workplaces). Even if the global prevalence remains low, a susceptible inside a highly infected cluster experiences many local sources, leading to a pronounced contraction of the generation interval within that sub‑population.

The authors illustrate both mechanisms using simulation studies. In a fully mixed model, the mean generation interval steadily declines as the proportion of infectious individuals increases, confirming the global competition effect. In network‑based simulations (small‑world and scale‑free graphs), they vary clustering coefficients to manipulate local prevalence. High clustering produces a sharp early drop in the generation interval, demonstrating that local competition can dominate early epidemic dynamics. The simulated means match the analytical expectations derived from the hazard‑based formulation.

Beyond describing the contraction phenomenon, the paper proposes a novel inference framework that replaces the traditional reliance on observed generation‑interval distributions with direct modeling of the infectious‑contact hazard. By treating each infection event as a “failure” in a survival‑analysis sense, a partial likelihood can be constructed: for event i occurring at time t_i, the contribution is λ_i(t_i) exp(−∫_0^{t_i} λ_i(s) ds), where λ_i(t) is the instantaneous hazard for the specific susceptible‑infector pair. This likelihood is mathematically identical to that of a Cox proportional‑hazards model, allowing the extensive toolbox of survival analysis to be applied to epidemic data. Using this approach, the time‑varying effective reproductive number R_t can be estimated directly from the hazard parameters, without needing to de‑convolve a potentially biased generation‑interval distribution. Simulation results show that hazard‑based estimates of R_t are substantially less biased and have lower mean‑squared error than estimates that ignore generation‑interval contraction.

The discussion emphasizes the practical implications of generation‑interval contraction for real‑time epidemic monitoring and for evaluating interventions that alter contact patterns (e.g., school closures, targeted quarantines). Because the hazard‑based partial likelihood requires only infection times and the identity of infector‑infectee pairs, it can be applied to many existing surveillance datasets, potentially reducing data‑collection burdens. The authors outline several avenues for future research: extending the framework to time‑varying contact networks, integrating Bayesian hierarchical models to capture uncertainty, and incorporating additional covariates such as vaccination status or pathogen strain.

In conclusion, the paper provides a rigorous mathematical proof that multiple concurrent sources of infection inevitably shorten the mean generation interval, distinguishes between global and local competition mechanisms, validates these concepts through extensive simulations, and introduces a hazard‑based partial‑likelihood methodology that bridges epidemic modeling with survival analysis. This work offers both theoretical insight and a practical statistical toolset for more accurate estimation of epidemic dynamics, especially the effective reproductive number, in complex real‑world settings.