Predicting if an infection will persist using early data

Early observations of a disease model can inform the epidemic threshold even when individual rates are uncertain.

This work studies Bayesian inference of persistence from early observations of a heterogeneous susceptible--infected--susceptible (SIS) reaction--diffusion model. The model uses standard incidence.

The leading prevalence dynamics depend on unknown coefficients through the difference between transmission and recovery rates and the infected diffusivity. This pair determines the principal eigenvalue. The sign of the principal eigenvalue governs persistence.

The study proves global weak well-posedness and quantitative parameter stability for the SIS system. It uses observation-map estimates to establish Bayesian well-posedness.

For an initial infection of size ε\varepsilon, a uniform second-order expansion describes how separate rates enter prevalence. Incidence depends explicitly on transmission at leading order.

When the noise standard deviation is proportional to ε\varepsilon, the nonlinear posteriors converge in Hellinger distance to low-prevalence limits. This convergence is O(ε)O(\varepsilon) for both prevalence alone and joint prevalence and incidence data.

The principal eigenvalue is bounded using the growth rate of the infected L2L^2 norm between two times. This rate provides a lower bound for the full nonlinear system.

Key facts

  • Early prevalence can inform the threshold even if individual rates remain uncertain.
  • The sign of the principal eigenvalue governs persistence.
  • Nonlinear posteriors show O(ε)O(\varepsilon) convergence in Hellinger distance to low-prevalence limits.

All sections

Edition