The basic reproduction number, R0, tells a reader how many people one infected person will infect on average. But an average can be a poor guide to what actually happens in the field. In outbreak after outbreak, most infected people transmit to few or no one, while a small minority ignite disproportionate chains of infection. Epidemiologists capture this unevenness with a second, less-publicized figure: the dispersion parameter, k. Understanding k is what separates a model that describes a typical case from one that describes how an epidemic actually spreads.
What the Dispersion Parameter K Actually Measures
R0 is a mean. k describes the variance around that mean — specifically, how unevenly secondary infections are distributed across the infected population. A transmission model with a high k assumes something close to homogeneous mixing: most infected people generate roughly the same number of secondary cases, clustered near the average. A low k means the opposite: transmission is highly variable, most cases produce zero or one further infection, and a small subset produce many. This second pattern is called overdispersion, and it is the mathematical signature of superspreading.
The two numbers describe genuinely different things. A disease can have a modest R0 and still be dominated by superspreading if k is low enough, and two pathogens can share an identical R0 while behaving completely differently in the field — one spreading in a steady chain, the other erupting in clusters tied to specific gatherings or settings. A field team that only asks "what is R0" is missing the variable that often determines whether an intervention should target the general population or a narrower set of high-risk transmission events.
The 20/80 Pattern Behind Many Outbreaks
Overdispersion is frequently summarized as a rough 20/80 pattern: a minority of infected individuals — often cited as around 20 percent — account for roughly 80 percent of onward transmission, while the rest of the infected population contributes comparatively little. It is a heuristic, not a fixed law, and the exact split varies by pathogen and setting, but it captures the operational point: chains of transmission are lumpy, not smooth. A single superspreading event, whether it occurs in a household, a healthcare facility, or a crowded gathering, can do more to shape an epidemic curve than dozens of ordinary, low-transmission cases combined.
This has a direct implication for surveillance and response design covered elsewhere in this desk’s field epidemiology coverage: interventions that specifically identify and interrupt high-transmission settings and events can be disproportionately effective compared with uniform, population-wide measures, particularly when resources for contact tracing or restrictions are limited.
What the Documented K Values Show
Several well-studied outbreaks illustrate how far dispersion estimates can vary by pathogen, and how consistently overdispersion has shown up in the coronavirus family in particular.
| Pathogen / Outbreak | Documented k estimate | What it indicates |
|---|---|---|
| SARS (2003) | Early estimates around 0.16 | Strong overdispersion; transmission concentrated in a minority of cases |
| MERS | Most study estimates below 1 (one outlier estimate of 2.94, wide confidence interval) | Overdispersion the norm, though estimates varied by dataset |
| COVID-19 (SARS-CoV-2) | Multiple studies on the order of 0.1; a systematic review reported a median of 0.31 across studies, with 93 percent of reviewed estimates below 1 | Consistent overdispersion across independent analyses and settings |
| Measles | Very high R0 (commonly cited as roughly 12–18); k less consistently documented | High average transmissibility; heterogeneous mixing around susceptible clusters still recognized as a factor |
Two individual analyses illustrate how these estimates were derived in practice. One study, using aggregated case data drawn from a World Health Organization situation report early in the COVID-19 pandemic, calculated a k of roughly 0.1. A separate analysis built on contact-tracing data from Shenzhen, China, found that only about 9 percent of infected individuals accounted for 80 percent of the onward infections identified through tracing — a concrete field measurement of the same 20/80 pattern described above.
Reading Overdispersion Without Overreading It
A low k value is not a claim about who is dangerous; it is a statement about the distribution of transmission events, which are shaped as much by setting, ventilation, timing relative to symptom onset, and contact patterns as by anything intrinsic to the infected individual. The same person might be a superspreader at one gathering and transmit to no one else across the rest of their infectious period. This distinction matters for how the concept gets used in practice: overdispersion is a case for reviewing high-risk settings and events, not a case for singling out individuals after the fact.
It also explains a pattern that puzzles non-specialists reading early outbreak reports: why case counts can appear to stall for stretches and then jump sharply. Under high overdispersion, whether a chain of transmission produces a cluster or fizzles out depends heavily on whether an early case happens to become a superspreading event — a matter of chance layered on top of the underlying average. Two outbreaks that start with an identical R0 can look very different a month later purely because of where the dispersion landed.
Conclusion
R0 answers the question "how transmissible is this pathogen on average." The dispersion parameter k answers a different and, for response planning, often more actionable question: how unevenly is that average transmission distributed across cases. Outbreaks with low k — as documented for SARS, largely for MERS, and repeatedly for COVID-19 — are shaped disproportionately by superspreading events, which means surveillance and response strategies built around identifying and interrupting high-transmission settings can outperform approaches based on the population average alone. Reading an epidemiological data set with both numbers in view, rather than R0 in isolation, is what separates a model of the typical case from a model of how the outbreak is actually unfolding.