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) model, typically assume that populations are homogeneous—meaning every individual is equally likely to contract a disease. However, real-world populations are inherently diverse due to genetic, behavioral, and environmental factors. This paper demonstrates that incorporating "heterogeneity in susceptibility" fundamentally transforms the mathematical description of an outbreak from a linear process to a power-law one. This shift explains why traditional first-order models often systematically over-predict the final severity of an epidemic.
The Problem: The Assumption of Homogeneous Mixing
The classic Kermack–McKendrick SIR model relies on the simplification of homogeneous mixing. In reality, susceptibility ($\epsilon$) varies widely. As a virus spreads, it acts as a filter, infecting the most susceptible individuals first. This "sculpting" of the population means the remaining pool of susceptible people becomes increasingly resistant over time. Standard first-order models fail to capture this dynamic, leading to significant inaccuracies in long-term predictions.
Methods: Tracking Susceptibility Evolution
The researchers moved beyond standard equations by defining an individual susceptibility variable, $\epsilon$. They used analytical derivations and numerical simulations to track how the distribution of $\epsilon$ changes as people transition from the Susceptible to the Infectious compartment.
Key methodological developments include:
- The Draw Probability: Deriving the conditional probability that a newly infected person has a specific susceptibility, which is consistently higher than the population average.
- Distribution Sculpting: Using Beta distributions to model various initial population shapes (bimodal, uniform, unimodal) and tracking their evolution.
- Analytical Power-Laws: Proving that if a population starts with a Gamma or Exponential distribution of susceptibility, the infection rate follows a power-law ($r_I = \beta I S^p$) rather than the standard linear rate ($r_I = \beta I S$).
- Order Parameter ($\mathcal{K}$): Introducing an "instantaneous order parameter" to quantify deviations from the standard first-order SIR model.
Key Insights: Sculpting and Eigen-distributions
One of the most significant findings is the "eigen-distribution" phenomenon. Exponential and Gamma distributions are mathematically resilient; they maintain their shape even as the mean susceptibility of the population drops. Most other distributions eventually "sculpt" themselves into a Gamma-like shape as the outbreak progresses.
The infection process acts as a natural selection mechanism for the susceptible pool. Because highly susceptible individuals are removed early, the epidemic slows down much faster than a standard model would predict. For instance, if 20% of the population accounts for 80% of the susceptibility (the "80/20 rule"), the resulting model is "5th-order," a massive departure from the 1st-order assumption.
Key Results: Higher-Order Models and Over-prediction
The researchers validated their framework using data from the 2009 H1N1 influenza outbreak. They found that 2nd-order and 5th-order models were significantly more consistent with actual serological results than the standard 1st-order model.
Calibrating a model during the initial exponential growth phase can be a "trap." During this stage, 1st-order and higher-order models appear identical. However, as the sculpting effect kicks in, the paths diverge. The standard SIR model predicted a final infection size of roughly 60% for H1N1, while the 5th-order model predicted a final size closer to 15-20%, aligning closely with observed data. Furthermore, heterogeneity significantly lowers the herd immunity threshold, defined by the new formula $p_C = 1 - 1/R_0^{1/p}$.
Limitations and Future Work
The current model assumes static susceptibility and does not account for behavioral changes (e.g., social distancing) or waning immunity. Additionally, susceptibility and infectiousness are treated as independent variables. If these traits are positively correlated (e.g., a "super-spreader" is also highly susceptible), the sculpting effect would be even more pronounced. Future research could integrate these dynamics into spatial or network-based models.
Summary updated