There Is No Single R0: The Five Hidden Shapes of Disease Spread

Ask most people what a disease outbreak has in common with the spread of a rumor, and they'll say infection. Ask a mathematician, and they'll say $R_0$ — the basic reproduction number, the single number that decides whether a pathogen fizzles out or becomes a pandemic. It's the same $R_0$ you heard endlessly during COVID-19: the number of people each infected person passes the virus to, on average, in a population with no immunity. Below 1, the outbreak dies. Above 1, it takes off.

But here's what most people never hear: there isn't one $R_0$. There are many. The way $R_0$ is shaped — whether it's a sum, a maximum, a geometric series, or a geometric mean — encodes the biology of transmission itself. Two mathematicians from the University of Texas at San Antonio and UC Berkeley, Zhuolin Qu and Abhi Ashwath, have catalogued these shapes across a sweeping taxonomy of disease models (Qu & Ashwath, 2026). Their review, a rare blend of research synthesis and teaching text, shows that the algebraic form of $R_0$ is not an accident. It's a fingerprint of how an infection moves — and it can tell you things about intervention that the raw number alone cannot.

The insight at the heart of the paper is deceptively simple: when a disease has multiple stages, multiple strains, or a vector like a mosquito in the loop, the formula for $R_0$ rearranges itself into a recognizable shape. A weighted sum appears when stages branch independently. A maximum appears when strains compete. A geometric series appears when a disease can relapse, round after round. Each shape is a message about the transmission pathway, and reading those shapes is a skill in itself.

The Science

Before the taxonomy, the authors walk through the machinery used to compute $R_0$, demonstrating it on the workhorse of epidemiology: the SEIR model, in which people move through Susceptible, Exposed, Infectious, and Recovered compartments (Qu & Ashwath, 2026; Fig.

). The tool is the next-generation matrix (NGM) approach, first devised by Diekmann and collaborators in 1990 and formalized for ordinary differential equation models by van den Driessche and Watmough a dozen years later. The idea: split the "infected subsystem" of the model into two pieces — new infections ($\mathcal{F}$) and transitions that move people through or out of infected states ($\mathcal{V}$). The ratio of these two, assembled as a matrix and summarized by its spectral radius (the largest eigenvalue), is $R_0$.

For the SEIR model this yields a clean product: the probability of surviving the exposed period, times the contact rate and transmission probability per contact, times the average time spent infectious — three factors multiplying into one number. The mechanics matter because they reveal the meaning of each algebraic form.

One crucial caveat is threaded through the paper: $R_0>1$ guarantees an outbreak only locally, near the disease-free state. In some models, an unstable "sub-threshold" equilibrium can persist even when $R_0<1$ — a backward bifurcation in the jargon, where a large enough introduction of infection can take hold despite the threshold saying otherwise (van den Driessche & Watmough, 2002). The nice clean number, in other words, has edges.

With the machinery in hand, the taxonomy unfolds across two grand transmission families: horizontal (infection passing between individuals in the same generation) and vertical (parent to offspring through reproduction).

What They Found

The paper's core contribution is a table of structural forms, each tied to a biological mechanism (Qu & Ashwath, 2026; Table 1). Five shapes do the heavy lifting, and each tells a distinct story.

The weighted sum. Some diseases — HIV is the motivating case — show heterogeneity in infectiousness. Either individuals progress through stages of rising and falling infectiousness over time (the staged progression model), or people are assigned at infection to fixed groups with distinct transmission rates (the differential infectivity model).

Weighted-Sum Form: SP and DI Models Share Structure

Label	Value
Differential infectivity (DI)	3
Staged progression (SP)	3

shows how both produce $R_0$ as a weighted sum:

$$R_0=\sum _{k=1}^n{\mathcal{P}}_k{R}_0^{(k)},$$

where ${\mathcal{P}}_k$ is the probability of reaching stage $k$ and $R_0^{(k)}$ is the secondary infections produced by one individual in that stage. The two models look superficially different — sequential progression versus parallel assignment (Fig.

) — yet collapse to the same algebraic shape, differing only in how the entry probabilities are computed. That's the paper's recurring lesson: the structure of $R_0$ reveals shared transmission logic even across biologically distinct diseases.

The maximum. When multiple strains of a pathogen circulate, transmission preserves identity: a flu variant's secondary cases inherit that variant. The infected subsystem decouples into independent one-dimensional problems, and the global $R_0$ is simply the maximum of the strain-level values (Fig.

):

$$R_0=\underset{i}{\max }{R}_0^{(i)}.$$

The galaxy's fastest-spreading strain dominates. This is the mathematics behind strain replacement — think of more transmissible variants of a pandemic virus outcompeting their predecessors, or drug-resistant strains winning when they carry a transmission advantage.

The geometric series. Some infections relapse. Plasmodium vivax malaria, invasive non-typhoidal Salmonella, and tuberculosis all permit cycles of latency and reactivation within a single host. Each relapse episode contributes to transmission, so $R_0$ must sum an infinite cascade of possible bouts (Fig.

):

$$R_0=\frac{R_0^{\text{cycle}}}{1-P_{\gamma }}=R_0^{\text{cycle}}\sum _{i=0}^{\infty }(P_{\gamma }{)}^i,$$

where $P_{\gamma }$ is the relapse probability. This is where the algebra gets genuinely revealing: the geometric-series form means that interventions targeting relapse — sterilizing latent reservoirs, interrupting hypnozoite reactivation — can be worth as much as blocking primary transmission, because they chop off the entire tail of the series.

The geometric mean. Vector-borne diseases — malaria, dengue, Zika — are bipartite transmission cycles: human to mosquito, mosquito to human. Sexually transmitted infections follow the same two-host structure. The result is a geometric mean (Fig.

):

$$R_0=\sqrt{R_{HV}{R}_{VH}},$$

the product of the human-to-vector and vector-to-human reproduction numbers, square-rooted. The symmetry is telling: a vector-control program that halves human-to-vector transmission and a vaccine that halves vector-to-human transmission don't add — they multiply through the geometric mean, which is why attacking both sides of the cycle is so powerful.

The fitness ratio. This is the paper's contribution to a genuinely modern problem: gene drives (Fig.

).

<｜DSML｜tool_calls> <｜DSML｜invoke name="write_paper_digest"> <｜DSML｜parameter name="body" string="true">Ask most people what a disease outbreak has in common with the spread of a rumor, and they'll say infection. Ask a mathematician, and they'll say $R_0$ — the basic reproduction number, the single number that decides whether a pathogen fizzles out or becomes a pandemic. It's the same $R_0$ you heard endlessly during COVID-19: the number of people each infected person passes the virus to, on average, in a population with no immunity. Below 1, the outbreak dies. Above 1, it takes off.

But here's what most people never hear: there isn't one $R_0$. There are many. The way $R_0$ is shaped — whether it's a sum, a maximum, a geometric series, or a geometric mean — encodes the biology of transmission itself. Two mathematicians, Zhuolin Qu of the University of Texas at San Antonio and Abhi Ashwath of UC Berkeley, have catalogued these shapes across a sweeping taxonomy of disease models (Qu & Ashwath, 2026). Their review, a rare blend of research synthesis and teaching text, shows that the algebraic form of $R_0$ is not an accident. It's a fingerprint of how an infection moves — and it can tell you things about intervention that the raw number alone cannot.

The insight at the heart of the paper is deceptively simple: when a disease has multiple stages, multiple strains, or a vector like a mosquito in the loop, the formula for $R_0$ rearranges itself into a recognizable shape. A weighted sum appears when stages branch independently. A maximum appears when strains compete. A geometric series appears when a disease can relapse, round after round. Each shape is a message about the transmission pathway, and reading those shapes is a skill in itself.

The Science

Before the taxonomy, the authors walk through the machinery used to compute $R_0$, demonstrating it on the workhorse of epidemiology: the SEIR model, in which people move through Susceptible, Exposed, Infectious, and Recovered compartments (Qu & Ashwath, 2026; Fig.

). The tool is the next-generation matrix (NGM) approach, first devised by Diekmann and collaborators in 1990 and formalized for ordinary differential equation models by van den Driessche and Watmough a dozen years later. The idea: split the "infected subsystem" of the model into two pieces — new infections ($\mathcal{F}$) and transitions that move people through or out of infected states ($\mathcal{V}$). The ratio of these two, assembled as a matrix and summarized by its spectral radius (the largest eigenvalue), is $R_0$.

For the SEIR model this yields a clean product: the probability of surviving the exposed period, times the contact rate and transmission probability per contact, times the average time spent infectious — three factors multiplying into one number. The mechanics matter because they reveal the meaning of each algebraic form.

One crucial caveat is threaded through the paper: $R_0>1$ guarantees an outbreak only locally, near the disease-free state. In some models, an unstable "sub-threshold" equilibrium can persist even when $R_0<1$ — a backward bifurcation in the jargon, where a large enough introduction of infection can take hold despite the threshold saying otherwise (van den Driessche & Watmough, 2002). The clean number, in other words, has edges.

With the machinery in hand, the taxonomy unfolds across two grand transmission families: horizontal (infection passing between individuals in the same generation) and vertical (parent to offspring through reproduction).

What They Found

The paper's core contribution is a structural taxonomy of $R_0$ forms, each tied to a biological mechanism (Qu & Ashwath, 2026). Five shapes do the heavy lifting, and each tells a distinct story.

The weighted sum. Some diseases — HIV is the motivating case — show heterogeneity in infectiousness. Either individuals progress through stages of rising and falling infectiousness over time (the staged progression model), or people are assigned at infection to fixed groups with distinct transmission rates (the differential infectivity model).

Weighted-Sum Form: SP and DI Models Share Structure

Label	Value
Differential infectivity (DI)	3
Staged progression (SP)	3

shows how both produce $R_0$ as a weighted sum:

$$R_0=\sum _{k=1}^n{\mathcal{P}}_k{R}_0^{(k)},$$

where ${\mathcal{P}}_k$ is the probability of reaching stage $k$ and $R_0^{(k)}$ is the secondary infections produced by one individual in that stage. The two models look superficially different — sequential progression versus parallel assignment (Fig.

) — yet collapse to the same algebraic shape, differing only in how the entry probabilities are computed. That's the paper's recurring lesson: the structure of $R_0$ reveals shared transmission logic even across biologically distinct diseases.

The maximum. When multiple strains of a pathogen circulate, transmission preserves identity: a flu variant's secondary cases inherit that variant. The infected subsystem decouples into independent one-dimensional problems, and the global $R_0$ is simply the maximum of the strain-level values (Fig.

):

$$R_0=\underset{i}{\max }{R}_0^{(i)}.$$

The fastest-spreading strain dominates. This is the mathematics behind strain replacement — think of more transmissible variants of a pandemic virus outcompeting their predecessors, or drug-resistant strains winning when they carry a transmission advantage.

The geometric series. Some infections relapse. Plasmodium vivax malaria, invasive non-typhoidal Salmonella, and tuberculosis all permit cycles of latency and reactivation within a single host. Each relapse episode contributes to transmission, so $R_0$ must sum an infinite cascade of possible bouts (Fig.

):

$$R_0=\frac{R_0^{\text{cycle}}}{1-P_{\gamma }}=R_0^{\text{cycle}}\sum _{i=0}^{\infty }(P_{\gamma }{)}^i,$$

where $P_{\gamma }$ is the relapse probability. This is where the algebra gets genuinely revealing: the geometric-series form means that interventions targeting relapse — sterilizing latent reservoirs, interrupting hypnozoite reactivation — can be worth as much as blocking primary transmission, because they chop off the entire tail of the series.

The geometric mean. Vector-borne diseases — malaria, dengue, Zika — are bipartite transmission cycles: human to mosquito, mosquito to human. Sexually transmitted infections follow the same two-host structure. The result is a geometric mean (Fig.

):

$$R_0=\sqrt{R_{HV}{R}_{VH}},$$

the square root of the product of the human-to-vector and vector-to-human reproduction numbers. The symmetry is telling: a vector-control program that halves human-to-vector transmission and a vaccine that halves vector-to-human transmission don't simply add — they interact through the product inside the geometric mean. Attacking both sides of the cycle amplifies the effect in a way that attacking only one cannot.

The fitness ratio. This is where the paper extends into genuinely novel territory: gene drives and population replacement. Rather than an infection passing between hosts, a heritable genetic element — here, the bacterium Wolbachia, used to suppress mosquito-borne disease — competes with a resident-type organism for reproductive success. The "reproduction number" becomes a ratio of fitnesses (Fig.

):

$${\mathcal{G}}_{0w}/{\mathcal{G}}_{0u},$$

the introduced cohort's reproductive output versus the resident's.

The Five Structural Forms of R0 in the Taxonomy

Label	Value
Weighted sum	1
Maximum	1
Geometric series (cyclic)	1
Geometric mean	1
Fitness ratio	1

shows how vertical transmission reshapes this ratio. When the ratio exceeds 1, the introduced lineage takes over the population — which is precisely what makes Wolbachia a tool for replacing wild mosquitoes with ones that can't transmit dengue or Zika.

Transmission routes are also transposable. The paper draws a striking structural equivalence between vertical transmission of infection from parent to offspring and the de novo reintroduction of a disease into a population — both are governed by the same mathematics, a "shared structural feature" the authors note. It's a unifying observation: the same framework that describes a mother passing a pathogen to a newborn also frames recurrent infection and re-exposure dynamics.

Why This Changes Things

The taxonomy's power is that it turns $R_0$ from a single opaque number into a diagnostic. When epidemiological data give you a measured $R_0$, the structural form tells you which transmission features must be present to produce it — and therefore which interventions will bite hardest.

Consider the geometric series for relapse diseases like tuberculosis. A researcher seeing a sum over relapse cycles immediately knows that curing active disease alone is insufficient: the $1-P_{\gamma }$ denominator means relapsing individuals re-enter the transmission pool, and the series tail can dominate $R_0$. The mathematics prescribes the intervention: treat latent infection, not just active disease. That's not a modeling nuance; it's the difference between programs that stamp out TB and ones that merely chase it.

The competing-strain maximum carries a similarly blunt message. When $R_0$ is governed by the single most transmissible strain, monitoring an average across strains is worse than useless — it's misleading. Surveillance must track the fastest variant, because that's what decides whether the population tips over the threshold. This is, at heart, the story of every pandemic variant arms race.

The geometric mean for vector-borne disease is arguably the most actionable result of all. Because $R_0=\sqrt{R_{HV}{R}_{VH}}$, the two transmission directions multiply together. A malaria program that combines bed nets (cutting vector-to-human transmission) with treatment that reduces human infectiousness (cutting human-to-vector transmission) reaps a benefit that is multiplicative, not additive. The math says: don't pick one side of the mosquito-human loop. Push both, and the threshold falls faster than either intervention alone could manage.

And the gene-drive fitness ratio moves the framework beyond disease entirely, into population and ecological engineering. When the goal is to replace a mosquito population with one carrying a transmission-blocking agent, the relevant "reproduction number" is no longer about pathogens at all — it's about whether the introduced lineage can outcompete the resident one. The same mathematical skeleton supports pandemic modeling and cutting-edge synthetic biology.

There's also an elegant pedagogical payoff. The authors wrote this as "a self-contained pedagogical reference," in the spirit of prior review texts (Qu & Ashwath, 2026). For newcomers, the field's $R_0$ derivations are scattered across decades of papers with inconsistent notation; this review collects them under one framework. For practitioners, it supplies a checklist: which shape to expect, which mechanism it implies, which intervention it recommends.

What's Next

The most valuable open thread is the direction of inference. The paper works from mechanism to $R_0$: given a model, derive the shape. The natural next step is the inverse problem — given a measured or estimated $R_0$ and its algebraic form, infer the underlying transmission structure. If field data show a geometric series, that's coherent with relapse; if they show a maximum, that's coherent with competing strains. Automating this reverse mapping is a tantalizing, and largely unexplored, research program.

Equally promising is the extension to more complex contact structures. The models surveyed here are largely homogeneous-mixing compartmental models; real outbreaks move through networks, age-stratified populations, and spatial patches. How do the five canonical shapes survive contact heterogeneity? Early work suggests the maximum and geometric-series forms are robust, but the weighted sum can change character when mixing becomes non-random.

The caveats deserve honesty. The authors stress that $R_0$'s guarantee is strictly local — the backward-bifurcation phenomenon means a "safe" $R_0<1$ isn't always safe. And every compartmental model is an abstraction; the taxonomy describes the logic of transmission mechanisms, not the messiness of a particular outbreak. That's not a flaw in the framework; it's a reminder to treat $R_0$ as what it is — a dimensionless lens, not a crystal ball.

What Qu and Ashwath have built is a Rosetta Stone for mathematical epidemiology. Five algebraic shapes, five transmission worlds, one unifying language. Read the shape, and the mechanism — and often the intervention — follows. For students, it's a map of the field. For researchers, it's a tool that reveals structure hiding in plain sight. For anyone who lived through a pandemic, it's a quiet reminder that the numbers in the news were never just numbers. They were encoded biology, waiting to be read.