Moving beyond averages: Individual-level variation in disease transmission

It is common practice in disease modeling studies to characterize groups or subgroups using population-average parameters, most importantly the basic reproductive number, R0. This approach overlooks variation at the individual level, which is caused by many factors. In this paper we show evidence of significant individual-level variation in transmission patterns for several diseases, and discuss how this can be incorporated into epidemiological models. We introduce a natural generalization of R0: the ‘individual reproductive number’, ν, which is the expected number of secondary cases caused by a given infected individual. Individual reproductive numbers for a population are drawn from a continuous probability distribution with mean equal to R0 (or to the effective reproductive number, R, if the population is not wholly susceptible). In this framework, superspreading events correspond to extreme values from the right-hand tail of the distribution of ν, and we propose a precise and generalizable definition of superspreading events based on probabilistic considerations. We analyze detailed transmission data for a range of directly-transmitted diseases, and find that conventional models assuming homogeneous transmission cannot account for observed patterns. Analysis of a branching process model incorporating individual-level heterogeneity reveals that observed levels of variation cause invasion dynamics to differ dramatically from predictions based on population averages. We explore the implications of these findings for outbreak control policies, demonstrating that individualspecific control measures are more likely to stop an outbreak than populationwide measures when both have the same effect on R0. We also highlight the effectiveness of measures targeting highly infectious individuals, and discuss how our results relate to recently-proposed surveillance methods for emerging diseases. We conclude by discussing future challenges in empirical and theoretical studies.