Epidemiological studies on effective reproductive number and asymptomatic infections of COVID-19
概要
【Background and Objectives】
Coronavirus disease 2019 (COVID-19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that reached pandemic levels in 2020. It remains unclear whether the pandemic will come to an end soon with the potential emergence of new variants of concerns. The key parameters governing infectious dynamics of ongoing epidemics of COVID-19 are 1) the effective reproduction number, R(t), and 2) the relative transmissibility of asymptomatic cases as compared with symptomatic cases. To accurately estimate these critical epidemiological parameters, we must address the issue of “observability”. COVID-19 infection events are generally not directly observable, and full datasets are rarely available. As a result, most studies have leveraged information on illness onset or the serial interval, or both, to estimate R(t). While the estimation of R(t) based on illness onset is conventional, accumulating evidence suggests that pre-symptomatic transmission contributes to the secondary transmission of COVID-19. Estimates of R(t) based on illness onset data seldom consider pre-symptomatic transmission. Similarly, individuals with asymptomatic SARS-CoV-2 infection can propagate the virus without being “noticed” and thus understanding viral transmissibility among asymptomatic individuals is critical for designing strategy controlling COVID-19. However, we seldom have opportunities to investigate on this because asymptomatic transmissions are usually hard to be traced (= observed). To address the issue of “observability” in estimation of each parameter, the objectives of this thesis are 1) to propose an alternative method of estimating R(t) as a function of time of infection using observable data while explicitly incorporating the pre-symptomatic transmission into the model and 2) to estimate the relative reproduction number of asymptomatic cases compared with symptomatic cases.
【Methods】
This thesis consists of two parts: In Section 1, we explored an alternative method of estimating R(t) and proposed a modified renewal equation using observable data while explicitly incorporating the pre-symptomatic transmission into the model. This method was applied to the early epidemic in Osaka prefecture to see the impact of public health and social measures (PHSM) on R(t). The negative log likelihood values of all possible combinations of “event-based” models were compared to account for the loss of single or combination of events. A Joinpoint segmented regression model was also used to assess whether a significant change in R(t) during each wave was associated with any of the start dates of key interventions. In Section 2, the relative reproduction number of asymptomatic cases compared with symptomatic cases was estimated. The data of two early clusters in Japan was used, in which the information of symptomatic status and time of illness onset were meticulously collected. Assuming that the number of secondary cases resulting from either primary symptomatic or asymptomatic cases independently followed negative binomial distributions, the relative reproduction numbers of an asymptomatic case compared with a symptomatic case as well as dispersion parameter was estimated with 95% CI in each cluster. Whether symptomatology was associated with transmission of symptomatic vs. asymptomatic infections was also assessed. We explored the impact of isolation on the transmissibility of asymptomatic cases, using the probability distribution function of the serial interval shortened by isolation.
【Results】
The “simple” alternative method was successful in estimating the R(t) for COVID-19 over the course of the epidemic in Osaka. Based on estimated R(t), the epidemic was found to come under control around 2 April 2020 during the first wave, and 26 July 2020 during the second wave. Consistent patterns were revealed when our estimates were compared with the R(t) estimates based on the back-projected incidence of infection. R(t) did not decline drastically following any single intervention. However, when multiple interventions were combined, the relative reductions in R(t) during the first and second waves were 70% and 51%, respectively. During the second wave, a significant impact on the observed secondary transmission patterns was not produced by the combined effect of interventions that focus on high-risk groups. The reproduction number of symptomatic
cases in Tokyo/Kanagawa cluster and Kyoto cluster was estimated at 1.2 (95% CI: 0.5–2.9) and 1.14 (95% CI: 0.61–2.09), respectively. The relative reproduction number for asymptomatic cases for each cluster was estimated at 0.27 (95% CI: 0.03–0.81) and 0.19 (95% CI: 0.03–0.66), respectively. Any apparent increased tendency for symptomatic primary case to produce symptomatic secondary cases was not found. We also assessed the relative transmissibility in the model using the probability density function of the generation interval adjusted for the isolation period in Kyoto cluster because movement of all identified close contacts was restricted for 14 days. However, the effectiveness of case isolation could not be estimated jointly with other parameters.
【Discussion】
We proposed the first method to estimate R(t) as a function of time of infection using observable data (= incident cases as a function of time of illness onset). The limitations in our method included 1) it could not estimate R(t) for the most recent 12 days and 2) the heterogeneities of host population were not incorporated into the model. The results of PHSM impact on R(t) in second wave might suggest infection control focusing on high-risk groups is not substantial. The relative reproduction number for asymptomatic cases was estimated to be 20 – 30 % of symptomatic cases, which was broadly consistent with the previous report. Contact tracing focusing on symptomatic index cases may be justified when there is limited testing capacity. The sample size was small for each cluster, involving a broad uncertainty bound and a wide 95% CI.
【Conclusion】
Two key epidemiological parameters governing transmission dynamics of COVID-19 were investigated using mathematical modeling approach while addressing the issue of “observability” carefully. The alternative method of estimating R(t) based on observable illness onset data was firstly devised and successful in estimating the R(t) for COVID19. The transmissibility of asymptomatic cases of SARS-CoV-2 infection was shown to be small relative to symptomatic cases. Our findings suggested that concerted efforts would be required to curb the COVID-19 epidemic and contact tracing focusing on symptomatic index cases may be justified when there is limited testing capacity. This study would assist planning the strategy for the efficient and feasible public health actions in future pandemics.