Getting started with {superspreading} • superspreading

The {superspreading} R package provides a set of functions for understanding individual-level transmission dynamics, and thus whether there is evidence of superspreading or superspreading events (SSE). Individual-level transmission is important for understanding the growth or decline in cases of an infectious disease accounting for transmission heterogeneity, this heterogeneity is not accounted for by the population-level reproduction number (\(R\)).

Definition

Superspreading describes individual heterogeneity in disease transmission, such that some individuals transmit to many infectees while other infectors infect few or zero individuals (Lloyd-Smith et al. 2005).

Code

library(superspreading)
library(epiparameter)

The {epiparameter} R package is loaded to provide a library of epidemiological parameters. These parameters can be used by the {superspreading} functions. See the Empirical superspreading section of this vignette for its usage.

As an example, offspring distributions are stored in the {epiparameter} library which contain estimated parameters, such as the reproduction number (\(R\)), and in the case of a negative binomial model, the dispersion parameter (\(k\)).

The offspring distribution is the distribution of the number of infectees (secondary case or offspring) that each infector (primary case) produces.

Probability of epidemic

The probability that a novel disease will cause a epidemic (i.e. sustained transmission in the population) is determined by the nature of that diseases’ transmission heterogeneity. This variability may be an intrinsic property of the disease, or a product of human behaviour and social mixing patterns.

For a given value of \(R\), if the variability is high, the probability that the outbreak will cause epidemic is lower as the superspreading events are rare. Whereas for lower variability the probability is higher as more individuals are closer to the mean (\(R\)).

Here we use \(R\) to represent the reproduction number (number of secondary cases caused by a typical case). Depending on the situation, this may be equivalent to the basic reproduction number (\(R_0\), representing transmission in a fully susceptible population) or the effective reproduction number at a given point in time (\(R_t\), representing the extent of transmission at time \(t\)). Either can be input into the functions provided by {superspreading}

The probability_epidemic() function in {superspreading} can calculate this probability. \(k\) is the dispersion parameter of a negative binomial distribution and controls the variability of individual-level transmission.

Code

probability_epidemic(R = 1.5, k = 1, num_init_infect = 1)
#> [1] 0.3333333
probability_epidemic(R = 1.5, k = 0.5, num_init_infect = 1)
#> [1] 0.2324081
probability_epidemic(R = 1.5, k = 0.1, num_init_infect = 1)
#> [1] 0.06765766

In the above code, \(k\) values above one represent low heterogeneity (in the case \(k \rightarrow \infty\) it is a poisson distribution), and as \(k\) decreases, heterogeneity increases. When \(k\) equals 1, the distribution is geometric. Values of \(k\) less than one indicate overdispersion of disease transmission, a signature of superspreading.

When the value of \(R\) increases, this causes the probability of an epidemic to increase, if \(k\) remains the same.

Code

probability_epidemic(R = 0.5, k = 1, num_init_infect = 1)
#> [1] 0
probability_epidemic(R = 1.0, k = 1, num_init_infect = 1)
#> [1] 0
probability_epidemic(R = 1.5, k = 1, num_init_infect = 1)
#> [1] 0.3333333
probability_epidemic(R = 5, k = 1, num_init_infect = 1)
#> [1] 0.8

Any value of \(R\) less than or equal to one will have zero probability of causing a sustained epidemic.

Finally, the probability that a new infection will cause a large epidemic is influenced by the number of initial infections seeding the outbreak. We define a to be this number of initial infections.

Code

probability_epidemic(R = 1.5, k = 1, num_init_infect = 1)
#> [1] 0.3333333
probability_epidemic(R = 1.5, k = 1, num_init_infect = 10)
#> [1] 0.9826585
probability_epidemic(R = 1.5, k = 1, num_init_infect = 100)
#> [1] 1

Empirical superspreading

Given probability_epidemic() it is possible to determine the probability of an epidemic for diseases for which parameters of an offspring distribution have been estimated. An offspring distribution is simply the distribution of the number of secondary infections caused by a primary infection. It is the distribution of \(R\), with the mean of the distribution given as \(R\).

Here we can use {epiparameter} to load in offspring distributions for multiple diseases and evaluate how likely they are to cause epidemics.

Code

sars <- epidist_db(
  disease = "SARS",
  epi_dist = "offspring distribution",
  single_epidist = TRUE
)
#> Using Lloyd-Smith J, Schreiber S, Kopp P, Getz W (2005). "Superspreading and
#> the effect of individual variation on disease emergence." _Nature_.
#> doi:10.1038/nature04153 <https://doi.org/10.1038/nature04153>.. 
#> To retrieve the short citation use the 'get_citation' function
evd <- epidist_db(
  disease = "Ebola Virus Disease",
  epi_dist = "offspring distribution",
  single_epidist = TRUE
)
#> Using Lloyd-Smith J, Schreiber S, Kopp P, Getz W (2005). "Superspreading and
#> the effect of individual variation on disease emergence." _Nature_.
#> doi:10.1038/nature04153 <https://doi.org/10.1038/nature04153>.. 
#> To retrieve the short citation use the 'get_citation' function

The parameters of each distribution can be extracted:

Code

sars_params <- get_parameters(sars)
sars_params
#>       mean dispersion 
#>       1.63       0.16
evd_params <- get_parameters(evd)
evd_params
#>       mean dispersion 
#>        1.5        5.1

Code

family(sars)
#> [1] "nbinom"
probability_epidemic(
  R = sars_params[["mean"]],
  k = sars_params[["dispersion"]],
  num_init_infect = 1
)
#> [1] 0.1198705
family(evd)
#> [1] "nbinom"
probability_epidemic(
  R = evd_params[["mean"]],
  k = evd_params[["dispersion"]],
  num_init_infect = 1
)
#> [1] 0.5092324

In the above example we assume the initial pool of infectors is one (num_init_infect = 1) but this can easily be adjusted in the case there is evidence for a larger initial seeding of infections, whether from animal-to-human spillover or imported cases from outside the area of interest.

We can see that the probability of an epidemic given the estimates of Lloyd-Smith et al. (2005) is greater for Ebola than SARS. This is due to the offspring distribution of Ebola having a larger dispersion (dispersion \(k\) = 5.1), compared to SARS, which has a relatively small dispersion (\(k\) = 0.16).

References

Lloyd-Smith, J. O., S. J. Schreiber, P. E. Kopp, and W. M. Getz. 2005. “Superspreading and the Effect of Individual Variation on Disease Emergence.” Nature 438 (7066): 355–59. https://doi.org/10.1038/nature04153.