Real-time seroprevalence and exposure levels of emerging pathogens in infection-naive host populations

Abstract

For endemic pathogens, seroprevalence mimics overall exposure and is minimally influenced by the time that recent infections take to seroconvert. Simulating spatially-explicit and stochastic outbreaks, we set out to explore how, for emerging pathogens, the mix of exponential growth in infection events and a constant rate for seroconversion events could lead to real-time significant differences in the total numbers of exposed versus seropositive. We find that real-time seroprevalence of an emerging pathogen can underestimate exposure depending on measurement time, epidemic doubling time, duration and natural variation in the time to seroconversion among hosts. We formalise mathematically how underestimation increases non-linearly as the host's time to seroconversion is ever longer than the pathogen's doubling time, and how more variable time to seroconversion among hosts results in lower underestimation. In practice, assuming that real-time seroprevalence reflects the true exposure to emerging pathogens risks overestimating measures of public health importance (e.g. infection fatality ratio) as well as the epidemic size of future waves. These results contribute to a better understanding and interpretation of real-time serological data collected during the emergence of pathogens in infection-naive host populations.

Figure 1

General seroepidemiological dynamics in both an homogeneous and inhomogeneous host-population. (A) Time series of cumulative incidence (exposed, orange), cumulative seroconverted (purple), incidence (green) and seroconversion (black) daily events. (B, C) Relative exposure underestimation (REU, blue) and absolute exposure underestimation (AEU, red) dependent on time (B) and the percent exposed (C). Outbreak simulated with parameters as default in Table 1. (D) Time series of incidence (green) and seroconversion (black) daily events. (E, F) REU (blue) and AEU (red) dependent on time (E) and the percent exposed (F). (A–C) Vertical dashed line marks the day in which the theoretical herd-immunity threshold is reached. Curves present the output of a single stochastic simulation. (D–F) Lines are the mean and areas the standard deviation of model output across all communities of the meta-population among 100 outbreaks simulated with parameters as default in Table 1 except $nC=144$ ($12^2$ lattice).

Figure 2

Summarizing exposure underestimation during epidemic growth. (left) Examples of simulated outbreaks (A–D) with parameter variations from the default parameter set (Fig. 1A–C, Table 1). Outbreak A with $\mathrm{\Gamma }m=28$. Outbreak B with $\mathrm{\Gamma }s=6$. Outbreak C with nC = 100. Outbreak D with R0 = 3.5. For each, the outbreak’s REU(t) (blue), incidence (green), REU mean and standard deviation for the entire time period (purple) and pREU (cyan) are presented. Time period considered is from the first case to outbreak peak. (right) Mean REU of each outbreak (N = 3100) and pREU shown with a linear regression with R-squared = 0.943.

Figure 3

Generalizing exposure underestimation during epidemic growth. (A–D) Each heatmap of pREU is for an epidemic doubling time (panel title) and combinations of T2S mean and shape (smaller shapes imply larger variance in responses among hosts). The color scale is discretized for visualization. (E) Sensitivity of pREU when varying doubling time, T2S mean and shape. Dashed lines present the limits pREU = 1 and doubling time / T2S mean = 0. T2S mean varied from 1 to 100, T2S shape from 1 to 10, doubling time 1 to 100.