Should the rate term in the basic epidemiology models be second-order?

Andrew A. Peterson, C. Franklin Goldsmith, Christopher F. Rose, Andrew J. Medford, Tejs Vegge

Abstract

The fundamental models of epidemiology describe the progression of an infectious disease through a population using compartmentalized differential equations, but do not incorporate population-level heterogeneity in infection susceptibility. We show that variation strongly influences the rate of infection, while the infection process simultaneously sculpts the susceptibility distribution. These joint dynamics influence the force of infection and are, in turn, influenced by the shape of the initial variability. Intriguingly, we find that certain susceptibility distributions (the exponential and the gamma) are unchanged through the course of the outbreak, and lead naturally to power-law behavior in the force of infection; other distributions often tend towards these eigen-distributions through the process of contagion. The power-law behavior fundamentally alters predictions of the long-term infection rate, and suggests that first-order epidemic models that are parameterized in the exponential-like phase may systematically and significantly over-estimate the final severity of the outbreak.

Introduction

Standard epidemiological models, such as the Susceptible-Infectious-Recovered (SIR) framework, typically rely on the "Law of Mass Action." This principle assumes that the rate of new infections is directly proportional to the product of the number of susceptible individuals ($S$) and infectious individuals ($I$), leading to first-order kinetics ($r = \beta x_I x_S$). However, this assumption treats populations as homogeneous "gas particles," ignoring significant variation in biological immune responses, age, and social behaviors.

While "super-spreaders" are well-recognized, this research focuses on "super-catchers"—individuals with high susceptibility. In a heterogeneous population, these highly vulnerable individuals are infected and removed from the susceptible pool early in an outbreak. This process effectively "cools" the average susceptibility of the remaining population, a phenomenon standard models fail to capture. Consequently, models parameterized on early exponential growth data often systematically over-predict the ultimate severity of an epidemic.

Theoretical Framework: The Second-Order Emergence

By applying principles from statistical physics, we can define an individual susceptibility ($\varepsilon$). If we assume $\varepsilon$ follows an exponential distribution—analogous to the Boltzmann distribution—the mean susceptibility of the population ($\bar{\varepsilon}$) evolves as individuals are withdrawn.

This process is akin to "evaporative cooling": just as the most energetic molecules leave a liquid, lowering its temperature, the most susceptible individuals leave the pool first. Mathematically, if susceptibility is exponentially distributed, the mean susceptibility $\bar{\varepsilon}$ becomes directly proportional to the fraction of the population remaining ($x_S$). Substituting this into the rate equation transforms the kinetics from first-order ($x_S$) to second-order ($x_S^2$).

Comparison of the dynamics between first-order ($r = \beta x_I x_S$) and second-order ($r = \beta x_I x_S^2$) SIR models. The second-order model predicts a significantly lower final infection rate and a dampened peak.

Intriguingly, the exponential distribution acts as an "eigen-distribution" or attractor. Even when starting with non-exponential distributions (e.g., uniform), the contagion process naturally sculpts the remaining population toward an exponential shape, reinforcing the second-order behavior.

Methods and Validation

The study utilized a multi-pronged approach involving mathematical derivation, stochastic Monte Carlo simulations (100,000 individuals), and deterministic binned ODE simulations. These methods were used to observe how second-order limits emerge naturally as the number of susceptibility categories increases. Real-world data from the 2003 SARS outbreak and the 2009 H1N1 pandemic were used to ground these theoretical findings.

Key Results

The shift to second-order kinetics fundamentally alters long-term predictions. In the early phase of an outbreak ($x_S \approx 1$), both first- and second-order models are nearly indistinguishable, explaining why traditional models appear accurate initially. However, as the outbreak progresses, the second-order model predicts a much "flatter" curve.

Comparison of final infection size predictions for the 2009 H1N1 pandemic. The second-order model aligns much more closely with actual blood serum measurements than the traditional first-order SIR model.

For the 2009 H1N1 pandemic, classic first-order models predicted a final infection size exceeding 50%. Serological studies later revealed the actual size was approximately 20%. The proposed second-order model predicted ~30%, drastically reducing the over-prediction error. Furthermore, the threshold for herd immunity is redefined: in a second-order model, it is reached at $1/\sqrt{R_0}$ rather than $1/R_0$, suggesting natural immunity may stabilize a population much earlier than previously thought.

Conclusion

Incorporating population-level heterogeneity through second-order kinetics provides a more realistic framework for modeling infectious diseases. While this study uses a basic SIR framework, the insights suggest that more complex models should also account for the "super-catcher" effect to avoid systematic over-estimation of epidemic severity.

Summary updated

Abstract

The fundamental models of epidemiology describe the progression of an infectious disease through a population using compartmentalized differential equations, but do not incorporate population-level heterogeneity in infection susceptibility. We show that variation strongly influences the rate of infection, while the infection process simultaneously sculpts the susceptibility distribution. These joint dynamics influence the force of infection and are, in turn, influenced by the shape of the initial variability. Intriguingly, we find that certain susceptibility distributions (the exponential and the gamma) are unchanged through the course of the outbreak, and lead naturally to power-law behavior in the force of infection; other distributions often tend towards these eigen-distributions through the process of contagion. The power-law behavior fundamentally alters predictions of the long-term infection rate, and suggests that first-order epidemic models that are parameterized in the exponential-like phase may systematically and significantly over-estimate the final severity of the outbreak.
BibTeX
@article{peterson2021should,
  author = {Andrew A. Peterson and C. Franklin Goldsmith and Christopher F. Rose and Andrew J. Medford and Tejs Vegge},
  title = {{Should the rate term in the basic epidemiology models be second-order?}},
  journal = {Journal of Theoretical Biology},
  year = {2021},
  doi = {10.1016/j.jtbi.2021.110839},
}
Figure