Comparisons of statistical distributions for cluster sizes in a developing pandemic

Area of Science:

• Epidemiology
• Statistical modeling
• Infectious disease dynamics

Background:

• SARS-CoV-2 transmission cluster sizes are often analyzed using the negative binomial distribution.
• The negative binomial distribution is favored over the Poisson distribution due to its longer tail, accommodating variance greater than the mean.
• However, its adequacy in describing superspreading events requires further investigation.

Purpose of the Study:

• To evaluate the effectiveness of extended Poisson process modeling (EPPM) for SARS-CoV-2 transmission cluster sizes.
• To compare EPPM with standard distributions like Poisson and negative binomial.
• To identify potential underestimation of superspreading events by current models.

Main Methods:

• Utilized extended Poisson process modeling (EPPM) with nested models.
• Included Poisson and negative binomial distributions within the EPPM framework.
• Assessed model adequacy using SARS-CoV-2 transmission cluster size data from various settings.

Main Results:

• The Poisson distribution was inadequate for most analyzed SARS-CoV-2 transmission data.
• The negative binomial distribution was also found to be inadequate in several instances.
• EPPM indicated significantly larger tail probabilities than the negative binomial distribution.

Conclusions:

• The negative binomial distribution may underestimate the probability and severity of SARS-CoV-2 superspreading events.
• Settings with large shared accommodation, meals, and work activities present a higher potential for severe superspreading.
• Public health interventions should prioritize transmission prevention in these high-risk settings.