📉⛈️ Stochastic Differential Equation (SDE) Models: Modeling Noise in Epidemic Trajectories

Stochastic Differential Equation (SDE) models constitute a powerful class of stochastic compartmental models that extend deterministic epidemic frameworks by explicitly incorporating continuous random perturbations. Rather than assuming fixed transmission and recovery rates, SDEs recognize that epidemiological processes are influenced by ongoing environmental variability, demographic fluctuations, and unobserved behavioral changes. By embedding noise terms directly into classical compartmental equations, SDE models provide a mathematically rigorous way to analyze how uncertainty alters epidemic trajectories, equilibrium stability, and long-term disease persistence.

These models serve as an intermediate level of description between deterministic ordinary differential equation (ODE) models and fully discrete stochastic simulations, offering analytical tractability while retaining essential stochastic realism.

────────────────────────────────────────────

🔄 1. Compartmental Structure and Flow

SDE models are typically formulated by augmenting standard compartmental structures such as the Susceptible–Infectious–Recovered (SIR) or Susceptible–Infectious–Recovered–Susceptible (SIRS) models. The compartment variables S(t), I(t), and R(t) are continuous-valued functions of time, representing population sizes or proportions.

Compartmental Flow with Stochastic Perturbations:

1. Susceptible (S): Individuals become infected through contact with infectious individuals. The infection process follows the classical mass-action structure but is continuously perturbed by random fluctuations.
2. Infectious (I): Infected individuals recover at a mean rate, with stochastic variability accounting for unpredictable changes in disease progression or intervention effectiveness.
3. Recovered (R): Individuals are removed from the infectious process, with potential stochastic variation in recovery dynamics.

The fundamental flow structure mirrors that of deterministic models, but each transition is influenced by continuous noise that modifies the trajectory away from the mean-field path.

────────────────────────────────────────────

🧪 2. Mathematical Formulation

An SDE model is constructed by taking a deterministic compartmental system and adding stochastic terms that represent continuous random perturbations. Each compartment evolves according to a drift term, describing the average deterministic behavior, and a diffusion term, describing stochastic variability.

For a stochastic SIR model with total population size N, the system can be written as:

Change in Susceptible population:
dS(t) = − [β · S(t) · I(t) / N] dt + σ_S(S, I) dW₁(t)

Change in Infectious population:
dI(t) = [β · S(t) · I(t) / N − γ · I(t)] dt + σ_I(S, I) dW₂(t)

Change in Recovered population:
dR(t) = [γ · I(t)] dt + σ_R(I) dW₃(t)

Here, the deterministic terms describe average transmission and recovery, while the stochastic terms introduce continuous random noise. The processes W₁(t), W₂(t), and W₃(t) represent independent Wiener processes, modeling random environmental or demographic fluctuations.

These equations allow researchers to study how noise alters epidemic stability, persistence, and extinction behavior relative to deterministic predictions.

────────────────────────────────────────────

📋 3. Parameter Definitions

Parameter	Definition	Role in the Model
β	Transmission rate	Governs the average rate at which susceptible individuals become infected
γ	Recovery rate	Governs the average rate at which infectious individuals recover
σ_J(X)	Noise intensity function for compartment J	Determines the magnitude of stochastic fluctuations affecting compartment J
W(t)	Wiener process	Represents continuous-time random perturbations (white noise)

────────────────────────────────────────────

📊 4. Parameter Ranges (General Viral Disease Context)

The deterministic parameters in SDE models align with classical compartmental models, while stochastic parameters are calibrated to represent realistic levels of environmental variability.

Parameter	Typical Range	Unit	Context
Recovery rate (γ)	0.07 – 0.14	day⁻¹	Corresponds to an infectious period of approximately 7 to 14 days
Transmission rate (β)	0.3 – 0.6	day⁻¹	Produces moderate to high transmissibility consistent with common viral infections
Noise intensity (σ)	10⁻⁴ – 10⁻²	dimensionless or rate-scaled	Reflects low to moderate continuous stochastic forcing

────────────────────────────────────────────

🛑 5. Applicability and Limitations

Applicability – When to Use SDE Models:

1. Stability and Persistence Analysis:
   SDE models are particularly valuable for analyzing the stability of disease-free and endemic equilibria under random perturbations, including conditions for stochastic persistence or extinction.
2. Continuous Environmental Variability:
   They are well suited for situations where randomness acts continuously over time, such as fluctuating contact rates, seasonal environmental effects, or gradual behavioral changes.
3. Large-Population Stochastic Approximation:
   SDEs provide an efficient approximation to discrete stochastic models in large populations, capturing noise effects without the computational cost of simulating individual events.

Key Assumptions and Weaknesses:

1. Diffusion Approximation:
   SDEs are often derived as approximations of discrete stochastic processes and may lose accuracy when population sizes are small or compartment counts approach zero.
2. Homogeneous Mixing:
   As extensions of classical ODE models, SDEs typically assume mass-action mixing and do not account for network structure or spatial heterogeneity.
3. Gaussian Noise Assumption:
   The use of Wiener processes implies normally distributed fluctuations, which may not fully capture abrupt or highly skewed stochastic events observed in real epidemics.

────────────────────────────────────────────

📚 References

Allen, L. J. S. (2008). An introduction to stochastic epidemic models.

Anderson, R. M., & May, R. M. (1991). Infectious Diseases of Humans: Dynamics and Control.

Jiang, D., Yu, J., Ji, C., & Shi, N. (2011). Asymptotic behavior of global positive solution to a stochastic SIR model.

Lu, Q. (2009). Stability of SIRS system with random perturbations.

Tornatore, E., Maria Buccellato, S., & Vetro, P. (2005). Stability of a stochastic SIR system.