Superspreader Events: Why 10% of Cases Drive 80% of Transmission

simulator advanced ~12 min
Loading simulation...
~10% of cases cause 80% of transmission at k=0.1

With dispersion parameter k = 0.1 (similar to SARS-CoV-2 estimates), about 10% of infected individuals generate 80% of all secondary infections. Most infected people transmit to nobody. This extreme overdispersion means outbreaks are driven by rare superspreading events, not average transmission.

Formula

Negative binomial: P(X=x) = C(x+k-1, x) × (k/(k+R₀))^k × (R₀/(k+R₀))^x
Proportion causing fraction f of transmission: 1 - betainc(k/(k+R₀), k, 1+k×(1-f)/R₀)
Extinction probability from one case: P(extinct) = (k/(k+R₀))^k

The 80/20 Rule of Epidemics

Not all infected individuals are equal transmitters. For many respiratory pathogens, a small fraction of cases generates the vast majority of secondary infections. Studies of SARS-CoV-2 suggest that roughly 10-20% of cases cause 80% of all transmission, while 70-80% of infected people transmit to nobody at all. This phenomenon, called overdispersion, fundamentally changes how we should think about epidemic control.

The Dispersion Parameter k

The dispersion parameter k quantifies this heterogeneity. When secondary infections follow a negative binomial distribution with mean R₀ and dispersion k, lower k means more variation. At k = 0.1, the distribution has a long right tail — most cases produce zero secondary infections, but rare superspreading events produce dozens. At k = 1 (Poisson-like), transmission is more uniform. SARS, MERS, and SARS-CoV-2 all show k well below 1.

Anatomy of a Superspreading Event

Superspreading events share common features: enclosed spaces, poor ventilation, prolonged exposure, and activities that enhance aerosol production. Choir practices, meat-packing plants, cruise ships, and indoor religious services have all generated documented superspreading events. The physics is straightforward — an infectious person exhales pathogen-laden aerosols that accumulate in poorly ventilated spaces, exposing everyone present to high viral doses over extended periods.

Implications for Control

This simulation visualizes the negative binomial distribution of secondary cases. Adjust the dispersion parameter to see how the distribution shifts from uniform to highly skewed. The 80/20 fraction readout shows what proportion of cases drives most transmission. Watch the extinction probability change — with low k, most introductions die out, but the ones that do not can be explosive. This insight motivates targeted interventions: restrict superspreading conditions rather than all contact equally.

FAQ

What is a superspreading event?

A superspreading event (SSE) occurs when a single infected individual transmits the pathogen to a disproportionately large number of others — typically 8 or more secondary cases. SSEs often involve crowded indoor settings with poor ventilation, prolonged exposure, and activities that generate aerosols (singing, shouting, heavy breathing). Famous examples include the Skagit Valley choir outbreak (53 of 61 members infected) and the Shincheonji church cluster in South Korea.

What is the dispersion parameter k?

The dispersion parameter k describes the heterogeneity of transmission when secondary cases follow a negative binomial distribution. Lower k means more overdispersion: most cases transmit to few or none, while rare cases infect many. SARS had k ≈ 0.16, SARS-CoV-2 had k ≈ 0.1-0.5, while influenza has k ≈ 1 (more uniform). When k < 1, superspreading plays a dominant role.

Why does overdispersion create a 'stochastic lottery'?

With overdispersed transmission, most infected individuals transmit to zero or one person. This means any single introduction has a high probability of fizzling out (the extinction probability). But occasionally, one introduction triggers a superspreading event that seeds a large cluster. This creates boom-or-bust dynamics where outbreaks seem to appear randomly and grow explosively.

How should control strategies account for superspreading?

Instead of reducing average R₀ uniformly (blanket lockdowns), overdispersion suggests targeting the conditions that enable superspreading: limiting large indoor gatherings, improving ventilation, reducing crowding, and backward tracing (finding the source event, not just forward contacts). Japan's cluster-busting strategy during COVID-19 explicitly targeted superspreading events rather than all transmission.

Sources

Embed

<iframe src="https://homo-deus.com/lab/epidemiological-modeling/superspreader-events/embed" width="100%" height="400" frameborder="0"></iframe>
View source on GitHub