🧠 The SEIR Epidemiological Model: A Popular‑Science Introduction
🩺 Abstract
Infectious diseases often exhibit a latent or incubation period during which people are infected but not yet able to transmit the pathogen. To capture such dynamics, epidemiologists extend the classic SIR (Susceptible–Infected–Recovered) framework by adding an Exposed class, producing the SEIR model. This article introduces the SEIR model in an accessible way, explains its underlying equations and parameters, discusses when it should be used, and surveys common extensions that incorporate demographic processes, waning immunity and vaccination. We conclude with a brief overview of its importance for public‑health planning and research.
🌍 1. Introduction
Compartmental models form the backbone of modern infectious‑disease epidemiology. The simplest of these is the SIR model, which assumes that individuals who recover from infection gain lasting immunity. For many diseases, however, there is a latent period – a delay between infection and the onset of infectiousness – and sometimes immunity wanes. Diseases such as measles, influenza, COVID‑19 and Ebola spend a non‑negligible time in hosts before they become contagious. The SEIR (Susceptible–Exposed–Infectious–Removed) model addresses this by adding an Exposed (E) compartment to track those incubating the disease but not yet capable of transmitting it.
The SEIR model has been widely used to analyse past epidemics and inform control strategies. During the 1918 influenza pandemic, mathematicians like Kermack and McKendrick began to formalise such models; since then, the SEIR framework has guided responses to outbreaks ranging from measles to COVID‑19. Its popularity lies in balancing realism – by incorporating an incubation period – with mathematical tractability.
⚙️ 2. Model Description
The SEIR model divides a fixed population N into four compartments:
- S(t) – susceptible individuals who can contract the disease;
- E(t) – exposed individuals who are infected but not yet infectious;
- I(t) – infectious individuals who can transmit the disease;
- R(t) – removed individuals who are no longer infectious (either immune or deceased).
Because the population is closed (no births or deaths), the totals satisfy S(t) + E(t) + I(t) + R(t) = N for all times t. Transitions between compartments are governed by the following ordinary differential equations:
dS/dt = – β ⋅ S ⋅ I / N
dE/dt = β ⋅ S ⋅ I / N – σ ⋅ E
dI/dt = σ ⋅ E – γ ⋅ I
dR/dt = γ ⋅ I
Here β (beta) is the transmission rate, σ (sigma) is the progression rate from exposed to infectious (the inverse of the average latent period), and γ (gamma) is the recovery rate (the inverse of the average infectious period). New infections occur at rate βSI/N; exposed individuals progress to the infectious state at rate σE; and infectives recover or are removed at rate γI.
📊 Parameter Definitions and Typical Values
| Parameter | Meaning | Typical range and notes |
| β | Transmission rate (contacts × transmission probability) | 0.1 – 2.0 per day, depending on the disease and social behaviour. Higher β means more contacts or higher infectivity. |
| σ | Progression rate from E to I | Equal to 1 divided by the average incubation period. For example, a 5‑day incubation implies σ ≈ 0.2 day⁻¹. |
| γ | Recovery/removal rate | Equal to 1 divided by the average infectious period. For COVID‑19, an infectious period of ~10 days gives γ ≈ 0.1 day⁻¹. |
| N | Total population size | Assumed constant in the basic SEIR model. |
| S₀, E₀, I₀, R₀ | Initial numbers in each compartment | Often normalised so that s₀ + e₀ + i₀ + r₀ = 1. For example, a small outbreak might start with 1 % exposed, 0 % infectious and the rest susceptible. |
The latent period is 1/σ and the infectious period is 1/γ. Typical values differ across diseases: measles has a latent period of ~8 days (σ ≈ 0.125 day⁻¹) and an infectious period of ~5 days (γ ≈ 0.2 day⁻¹), while seasonal influenza has a latent period of 1–2 days (σ ≈ 0.5–1.0 day⁻¹) and an infectious period of 3–5 days (γ ≈ 0.2–0.33 day⁻¹). For COVID‑19, the latent period averages 5–6 days and the infectious period is roughly 7–10 days, leading to σ ≈ 0.17–0.2 day⁻¹ and γ ≈ 0.1–0.14 day⁻¹.
📈 Reproduction Number and Disease Dynamics
The basic reproduction number (R₀) measures the expected number of secondary infections caused by one infectious case introduced into a wholly susceptible population. For the SEIR model without births or deaths, it is
R₀ = β / γ
just as in the SIR model, because the latent stage does not affect the number of people an infectious case eventually infects. If R₀ > 1, the infection can invade; if R₀ < 1, outbreaks die out. The latent period delays the epidemic peak but does not change this threshold. The time to peak and the shape of the outbreak curve depend on σ and γ: longer latent periods (smaller σ) delay the rise, while faster recovery (larger γ) suppresses the peak.
🧪 3. Assumptions and Applicability
Like all compartmental models, the SEIR framework makes simplifying assumptions:
- Closed population: No births, deaths (other than disease‑induced), or migration; hence N is constant.
- Homogeneous mixing: Every individual has an equal chance of contacting every other individual.
- Exposed individuals are not infectious: People in the E compartment cannot transmit the disease until they enter I.
- Constant rates: The transmission rate β, progression rate σ and recovery rate γ are constants over time.
- Deterministic dynamics: The equations represent average behaviour; stochastic fluctuations are ignored.
These assumptions mean the SEIR model applies best to large, well‑mixed populations experiencing diseases with incubation periods and short‑lived immunity. It may not capture heterogeneous contact patterns, age structure, or spatial effects. For small populations or rare diseases, stochastic models that incorporate randomness may be more appropriate.
🔧 4. Model Variants and Improvements
🔄 a) SEIRS Model (Waning Immunity)
Many diseases confer only temporary immunity. The SEIRS model extends SEIR by allowing recovered individuals to lose immunity at rate ρ (rho) and return to the susceptible class:
dS/dt = – β S I / N + ρ R
dE/dt = β S I / N – σ E
dI/dt = σ E – γ I
dR/dt = γ I – ρ R
This model is useful for infections like pertussis or seasonal coronaviruses where immunity wanes over months or years.
👶 b) SEIR with Demography
To model growing or shrinking populations, one can incorporate births and natural deaths. Suppose individuals are born susceptible at rate μN and die at per‑capita rate μ. The equations become:
dS/dt = μN – β S I / N – μ S
dE/dt = β S I / N – σ E – μ E
dI/dt = σ E – γ I – μ I
dR/dt = γ I – μ R
The basic reproduction number generalises to
R₀ = (σ / (σ + μ)) × (β / (γ + μ)).
Vital dynamics allow for endemic equilibria in which the infection persists indefinitely at a constant prevalence.
💉 c) SEIR with Vaccination (SEIRV)
Vaccination campaigns move people from S to an immunised class V, reducing the susceptible pool. The simplest SEIRV model introduces a vaccination rate ν:
dS/dt = – β S I / N – ν S
dE/dt = β S I / N – σ E
dI/dt = σ E – γ I
dR/dt = γ I
dV/dt = ν S
Vaccination lowers R₀ proportionally and can avert epidemics when coverage exceeds the herd immunity threshold (1 – 1/R₀).
🛡️ d) SEIQR Model (Quarantine/Isolation)
During outbreaks, public health authorities often identify and isolate exposed or infected individuals. The SEIQR model adds a Quarantined (Q) compartment. Exposed individuals are moved to Q at rate δ; quarantined individuals may still progress to infectiousness but at reduced transmission. Such models help evaluate the impact of testing, contact tracing and isolation on outbreak control.
🕸️ e) Age‑Structured and Network SEIR Models
Real populations are heterogeneous. Age‑structured SEIR models divide each compartment into age classes with different contact patterns and susceptibility; this is crucial for diseases like measles or COVID‑19, where children and adults have different contact rates. Network SEIR models place individuals on a network of interactions, capturing community structure and clustering. These extensions often require numerical simulation but provide more realistic predictions.
⚗️ f) Stochastic SEIR Models
When case numbers are low, randomness matters. Stochastic SEIR models treat transitions as probabilistic events (e.g., using continuous‑time Markov chains). They capture the chance of extinction even when R₀ > 1 and provide confidence intervals for epidemic trajectories.
🧾 5. Conclusion
The SEIR model is a powerful tool for understanding epidemics with incubation periods. By explicitly modelling the Exposed class, it captures diseases like measles and COVID‑19 more realistically than the SIR model. Its simple structure allows researchers and public‑health officials to estimate reproduction numbers, predict outbreak trajectories and assess the impact of interventions such as vaccination and isolation. Extensions incorporating waning immunity, demography, vaccination, quarantine, age structure and stochastic effects broaden its applicability. As with all models, however, the SEIR framework rests on assumptions that may not hold in all contexts; careful interpretation and, where necessary, enhancement with more detailed models are essential.
📚 6. References
- Hethcote, H. W. (2000). The mathematics of infectious diseases. SIAM Review, 42(4), 599–653. https://doi.org/10.1137/S0036144500371907
- Keeling, M. J., & Rohani, P. (2008). Modeling infectious diseases in humans and animals. Princeton University Press. http://assets.press.princeton.edu/chapters/s9048.pdf
- Anderson, R. M., & May, R. M. (1991). Infectious diseases of humans: Dynamics and control. Oxford University Press. https://doi.org/10.1093/oso/9780198540403.001.0001
- Jay, G. (2020). The SEIR model of infectious diseases. Portland State University. https://web.pdx.edu/~gjay/teaching/mth271_2020/html/09_SEIR_model.html
- World Bank. (2021). An introduction to deterministic infectious disease models. https://documents1.worldbank.org/curated/en/888341625223820901/pdf/An-Introduction-to-Deterministic-Infectious-Disease-Models.pdf
- Park, M., et al. (2019). Directly transmitted viral diseases: modeling the dynamics of transmission. PLOS Pathogens, 15(12), e1008137. https://doi.org/10.1371/journal.ppat.1008137
- Chowell, G., & Vargas, P. (2020). Modeling the COVID-19 pandemic: A review. Infectious Disease Modeling, 5, 527–537. https://doi.org/10.1016/j.idm.2020.08.002
- Arino, J., & van den Driessche, P. (2003). A multi-city epidemic model. Mathematical Population Studies, 10(3), 175–193. https://doi.org/10.1080/08898480304821
- Kucharski, A. J., Russell, T. W., et al. (2020). Early dynamics of transmission and control of COVID-19: a mathematical modelling study. The Lancet Infectious Diseases, 20(5), 553–558. https://doi.org/10.1016/S1473-3099(20)30144-4
- Li, M. Y., & Muldowney, J. S. (1995). Global stability for SEIR models in epidemiology. Mathematical Biosciences, 125(2), 155–164. https://doi.org/10.1016/0025-5564(95)92756-5