Simultaneous estimation of the effective reproducing number and the detection rate of COVID-19

TL;DR

This paper introduces a Bayesian model to simultaneously estimate the effective reproduction number and detection rate of COVID-19, providing insights into epidemic dynamics and policy effectiveness.

Contribution

The paper presents a novel Bayesian approach for joint estimation of COVID-19's reproduction number and detection rate, accounting for asymptomatic cases and short-term shocks.

Findings

The model accurately estimates R and detects detection rate shocks in synthetic data.

Application to real data suggests moderate measures can reduce R below 1.

Analysis indicates a downward trend in R in Japan under current policies.

Abstract

A major difficulty to estimate $R$ (the effective reproducing number) of COVID-19 is that most cases of COVID-19 infection are mild or asymptomatic, therefore true number of infection is difficult to determine. This paper estimates the daily change of $R$ and the detection rate simultaneously using a Bayesian model. The analysis using synthesized data shows that our model correctly estimates $R$ and detects a short-term shock of the detection rate. Then, we apply our model to data from several countries to evaluate the effectiveness of public healthcare measures. Our analysis focuses Japan, which employs a moderate measure to keep "social distance". The result indicates a downward trend and now $R$ becomes below $1$. Although our analysis is preliminary, this may suggest that a moderate policy still can prevent epidemic of COVID-19.