Age-structured compartmental models are essential tools in mathematical epidemiology for moving beyond the simplification of homogeneous mixing to capture realistic variations in disease transmission, contact patterns, susceptibility, and clinical outcomes across different demographic groups. By partitioning the population into discrete or continuous age classes, these models provide the high-resolution necessary for accurate policy evaluation, particularly for vaccine-preventable and highly heterogeneous diseases like influenza and COVID-19.
📐 Compartmental Structure and Flow Explanation
Age-structured models integrate demographic structure (age classes) with infection status (e.g., Susceptible, Infectious, Removed). The flow between epidemiological compartments (such as S → E → I → R) remains standard, but the transition rates (β, γ) become age-specific and depend on mixing patterns between age groups.
For an Age-Structured SEIR model with n age groups, populations in each group are represented as:
Sᵢ, Eᵢ, Iᵢ, Rᵢ (for age group i)
Infection Flow:
Susceptible individuals in group i acquire infection based on the age-specific force of infection λᵢ, which aggregates contributions from infectious individuals in all groups j.
Aging Flow:
Individuals progress into the next age class at rates cᵢ (e.g., Sᵢ → Sᵢ₊₁).
Demographic Flow:
Births enter the youngest age group, while natural deaths occur in all groups, enabling long-term endemic modeling.
🧮 Mathematical Formulation (ODE System)
Let sᵢ, eᵢ, iᵢ, rᵢ represent the fractions in each compartment of age group i.
Assume the age-specific force of infection is:
λᵢ = bᵢ × Σⱼ ( b̃ⱼ × iⱼ )
The coupled ODE system becomes:
dsᵢ/dt = cᵢ₋₁ sᵢ₋₁ − ( λᵢ + cᵢ + dᵢ + q ) sᵢ
deᵢ/dt = λᵢ sᵢ + cᵢ₋₁ eᵢ₋₁ − ( εᵢ + cᵢ + dᵢ + q ) eᵢ
diᵢ/dt = εᵢ eᵢ + cᵢ₋₁ iᵢ₋₁ − ( γᵢ + cᵢ + dᵢ + q ) iᵢ
drᵢ/dt = γᵢ iᵢ + cᵢ₋₁ rᵢ₋₁ − ( cᵢ + dᵢ + q ) rᵢ
where:
• c₀ = 0
• newborns enter s₁
• q is the population growth rate
• cᵢ denotes aging from group i to i+1
⚗ Parameter Definitions and Typical Ranges
Age-structured models require demographic, epidemiological, and behavioral parameters for each group. Typical values for viral respiratory disease modeling include:
| Parameter | Interpretation | Typical Range |
|---|---|---|
| cᵢ | Aging rate | 1 / (age-group duration) |
| dᵢ | Age-specific natural death rate | 0.005 – 0.04 per year |
| εᵢ | Rate of leaving exposed state | 0.2 – 0.5 per day |
| γᵢ | Recovery/removal rate | 0.1 – 0.5 per day |
| bᵢ, b̃ⱼ | Contact-rate components for λᵢ | Varies by age-mixing intensity |
| R₀ | Basic reproduction number | 1.5 – 4.0 |
R₀ is typically computed as the dominant eigenvalue of the Next Generation Matrix (NGM), where demographic transitions (cᵢ + dᵢ + q) shape the spectrum of the linearized system.
🎯 Applicability and Limitations
| Use Case | Assumptions and Limitations |
|---|---|
| Evaluating targeted interventions (vaccination, school closure) | Requires detailed empirical age-specific contact matrices (bᵢ, b̃ⱼ), which may be difficult to obtain or uncertain. |
| Studying susceptibility and severity differences by age | Relies on mixing assumptions (e.g., separable or proportionate mixing) that may not fully represent real social networks. |
| Forecasting heterogeneous outcomes (hospitalization, mortality) | High dimensionality: many coupled ODEs increase computational cost, especially in stochastic simulations. |
📚 References
- Anderson, R. M., & May, R. M. Infectious Diseases of Humans: Dynamics and Control. Oxford University Press (1991).
- Davies, N. G., Klepac, P., Liu, Y., et al. Age-dependent effects in the transmission and control of COVID-19 epidemics. Nature Medicine (2020).
- Hethcote, H. W. The mathematics of infectious diseases. SIAM Review (2000).
- Prem, K., Cook, A. R., & Jit, M. Projecting social contact matrices in 152 countries. PLoS Computational Biology (2017).
- Schenzle, D. An age-structured model of measles transmission. IMA Journal of Mathematics Applied in Medicine and Biology (1984).