Improved estimation of time-varying reproduction numbers at low case incidence and between epidemic waves

Abstract

AbstractWe construct a recursive Bayesian smoother, termed EpiFilter, for estimating the effective reproduction number, R, from the incidence of an infectious disease in real time and retrospectively. Our approach borrows from Kalman filtering theory, is quick and easy to compute, generalisable, deterministic and unlike many current methods, requires no change-point or window size assumptions. We model R as a flexible, hidden Markov state process and exactly solve forward-backward algorithms, to derive R estimates that incorporate all available incidence information. This unifies and extends two popular methods, EpiEstim, which considers past incidence, and the Wallinga-Teunis method, which looks forward in time. We find that this combination of maximising information and minimising assumptions significantly reduces the bias and variance of R estimates. Moreover, these properties make EpiFilter more statistically robust in periods of low incidence, where existing methods can become destabilised. As a result, EpiFilter offers improved inference of time-varying transmission patterns that are especially advantageous for assessing the risk of upcoming waves of infection in real time and at various spatial scales.Author SummaryInferring changes in the transmissibility of an infectious disease is crucial for understanding and controlling epidemic spread. The effective reproduction number, R, is widely used to assess transmissibility. R measures the average number of secondary cases caused by a primary case and has provided insight into many diseases including COVID-19. An upsurge in R can forewarn of upcoming infections, while suppression of R can indicate if public health interventions are working. Reliable estimates of temporal changes in R can contribute important evidence to policymaking. Popular R-inference methods, while powerful, can struggle when cases are few because data are noisy. This can limit detection of crucial variations in transmissibility that may occur, for example, when infections are waning or when analysing transmissibility over fine geographic scales. In this paper we improve the general reliability of R-estimates and specifically increase robustness when cases are few. By adapting principles from control engineering, we formulate EpiFilter, a novel method for inferring R in real time and retrospectively. EpiFilter can potentially double the information extracted from epidemic time-series (when compared to popular approaches), significantly filtering the noise within data to minimise both bias and uncertainty of R-estimates and enhance the detection of salient changepoints in transmissibility.