Abstract
Superspreading is an important feature of SARS-CoV-2, though few studies have investigated quantitatively how transmission characteristics can vary by setting. Using detailed clustering data comprising 8635 SARS-CoV-2 cases confirmed in Hong Kong between 2020–2021 and a negative binomial cluster size model, we estimate the mean number of new infections expected in a cluster CZ and the degree of overdispersion (k) by setting. Estimates of CZ ranged between 0.3–6.1 across eight distinct transmission settings. Close-social indoor (e.g. bars and clubs) and elderly care home settings had the highest CZ around 6, meaning for every introduction an average of six new infections is expected. Overdispersion also differed by setting, ranging from extremely heterogeneous (k = 0.05) to less heterogeneous (k = 1), and was highest in retail, close-social indoor, and care homes settings (k < 0.1), where lower values of k indicate higher superspreading potential. We found that the mean generation interval (GI) also varied by setting (range: 4.4–7.2 days), and settings with shorter mean GIs were associated with smaller cluster sizes. Our results explicitly quantify and demonstrate that superspreading potential and transmission parameters such as the GI can vary across settings, which highlights the need of setting-specific interventions for effective outbreak control.