Compartmental models that explicitly incorporate therapeutic interventions—such as SITR (Susceptible–Infectious–Treated–Removed) or latent-phase extensions like SEITR or SLITR—are essential tools for analyzing how treatment, isolation, and diagnostic strategies reshape epidemic trajectories.
By introducing a dedicated Treated (T) compartment with reduced infectiousness, these models enable direct quantification of intervention impact, independent from natural recovery processes.
🔄 Compartmental Structure and Flow
The SEITR model extends the classical SEIR framework by adding both a latent stage (E) and an intervention stage (T).
- S — Susceptible: Individuals who can become infected.
- E — Exposed: Infected but not yet infectious; undergoing incubation.
- I — Infectious: Actively transmitting disease; no isolation.
- T — Treated / Isolated: Diagnosed individuals receiving treatment or isolation, with reduced infectiousness.
- R — Removed: Recovered with immunity or deceased.
The baseline flow is sequential:
S → E → I
From the infectious class I, individuals may either recover naturally (I → R) or undergo intervention (I → T).
From T, individuals either recover or are removed (T → R).
This separation allows explicit measurement of how rapid case identification and isolation reduce transmission intensity.
Mathematical Formulation (SEITR / SLITR ODE System)
Assume a constant population size N, with recruitment Λ and natural mortality μ included in each compartment.
System of Deterministic ODEs
dS/dt = Λ − β S ( I + δ T ) / N − μ S
dE/dt = β S ( I + δ T ) / N − ( σ + μ ) E
dI/dt = σ E − ( γ + γΤ + μ ) I
dT/dt = γΤ I − ( η + μ ) T
dR/dt = γ I + η T − μ R
Parameter Definitions
- β — Transmission rate (contact rate × transmission probability)
- σ — Rate of progression from E to I (inverse of latent period)
- γ — Natural recovery rate from I
- γΤ — Rate at which infectious individuals are detected and moved to T
- η — Removal rate from T (recovery or death)
- δ — Relative infectivity of treated individuals (0 ≤ δ ≤ 1)
- Λ, μ — Birth/recruitment rate and natural death rate
When δ = 0, treatment or isolation is perfect (no transmission from T).
📊 Parameter Ranges for Viral Epidemics
| Parameter | Description | Typical Range (day⁻¹) |
|---|---|---|
| Λ, μ | Birth / natural death rates | ≤ 0.0001 |
| β | Transmission coefficient | 0.10 – 0.50 |
| σ | Latent progression rate (1 / latent period) | 0.10 – 0.33 |
| γ | Natural recovery rate (1 / infectious period) | 0.05 – 0.20 |
| γΤ | Treatment / isolation rate (I → T) | 0.10 – 1.0+ |
| η | Removal rate from T | 0.05 – 0.33 |
| δ | Relative infectivity while treated | 0 – 0.5 |
These ranges reflect typical values for acute viral pathogens (influenza-like illness, SARS-CoV-2, etc.), where latency and rapid isolation critically shape epidemic control.
⚠️ Applicability and Limitations
✔️ Use Cases
- Intervention Efficacy:
Measures how the flow I → T reduces the control reproduction number Rc. - Resource Allocation:
Estimates peak demand for isolation units, antiviral therapy, or hospital beds. - Targeted Disease Control:
Particularly useful when latency matters (SEITR) and infectiousness drops upon detection.
❌ Limitations
- Exponential Timing Assumption:
Residence times in E and I are exponentially distributed, which may not reflect biological reality. - Homogeneous Mixing:
Assumes uniform contact patterns; ignores age, social networks, superspreading. - Parameter Sensitivity:
Early in an outbreak, intervention effectiveness (δ) and detection rates (γΤ) are difficult to estimate. - Deterministic Dynamics:
Cannot capture stochastic extinction events or large fluctuations at low case numbers.
📚 Selected References
Giordano, G. et al. (2020). Modelling the COVID-19 epidemic and interventions in Italy.
Kermack, W. O. & McKendrick, A. G. (1927). A contribution to the mathematical theory of epidemics.
Brauer, F. (2008). Compartmental Models in Epidemiology.
Gumel, A. B. et al. (2004). Modelling strategies for controlling SARS outbreaks.
Diekmann, O., Heesterbeek, J. A. P. & Metz, J. A. J. (1990). On the definition and computation of R₀.