Statistical Deconvolution for Inference of Infection Time Series

Abstract

Accurate measurement of daily infection incidence is crucial to epidemic response. However, delays in symptom onset, testing, and reporting obscure the dynamics of transmission, necessitating methods to remove the effects of stochastic delays from observed data. Existing estimators can be sensitive to model misspecification and censored observations; many analysts have instead used methods that exhibit strong bias. We develop an estimator with a regularization scheme to cope with stochastic delays, which we term the robust incidence deconvolution estimator. We compare the method to existing estimators in a simulation study, measuring accuracy in a variety of experimental conditions. We then use the method to study COVID-19 records in the United States, highlighting its stability in the face of misspecification and right censoring. To implement the robust incidence deconvolution estimator, we release incidental, a ready-to-use R implementation of our estimator that can aid ongoing efforts to monitor the COVID-19 pandemic.

Estimating actual SARS-CoV-2 infections from secondary data

“…Typically, pandemic management relies on diagnostic testing of individuals, reporting the number of positive tests on a given day t as documented daily new infection cases ( N INF ). In fact, there is a time lag between infection and testing that includes both the incubation period and the latency between symptom onset and testing 19 . However, as it has no influence on the derived methodology, we choose to disregard this time lag in the following—thus taking N INF as reported.…”

Comparative epidemiology of outbreaks caused by SARS-CoV-2 Delta and Omicron variants in China

“…To account for the potential of pre-symptomatic transmission for SARS-CoV-2 viruses [14], we reconstructed the epidemic curve by date of infection to estimate R t . The epidemic curve by infection date was reconstructed based on daily case numbers by report date using a deconvolution approach [15], allowing for the delay from infection to report (Supplementary Appendix, p. 2). Then, we estimated the R t based on the Poisson framework developed by Cori et al [16] with daily confirmed case numbers and daily symptomatic case numbers .…”

Constrained numerical deconvolution using orthogonal polynomials

“…In the first example, we apply the method to COVID-19 curves, reconstructing the underlying dynamics of the pandemic and providing valuable insights into the evolution of the virus. This application is particularly relevant in the context of recent studies that have used deconvolution techniques to analyze the spread of infectious diseases [28] , [30] . In the second example, we apply our method to the study of mavoglurant, a drug used to treat neurological disorders.…”