key: cord-1043514-6ydefqmc
authors: Pak, D.; Langohr, K.; Ning, J.; Cortes, J.; Gomez, G.; Shen, Y.
title: Age Associated Coronavirus Disease 2019 Incubation Period: Impact on Quarantine Policy
date: 2020-06-28
journal: nan
DOI: 10.1101/2020.06.27.20141002
sha: 2ac8a8ae6a5040b664b39a4262ac1771af3d4b1b
doc_id: 1043514
cord_uid: 6ydefqmc

A unified 14-day quarantine may not effectively contain the spread of COVID-19. With the current recommendation of a uniform 14-day quarantine, up to 9.2% of patients 42 or younger and 18.6% of patients older than 42 years could develop symptoms after the end of the period.

The coronavirus disease 2019 (COVID-43 19) was first reported in Wuhan, China, in December 2019; an outbreak rapidly spread worldwide. The novel virus infection can be asymptomatic or unapparent during a certain period and asymptomatic persons could spread the virus unknowingly. Among patients who develop symptoms, the incubation period is defined as the elapsed time between infection and appearance of the first symptom. Knowledge of the incubation period is essential for disease prevention, facilitating an optimal quarantine guideline to confine the spread.

The incubation period of COVID-19 was investigated in several reports. [1, 2, 3, 4] Despite its importance, it remains unclear how the incubation distribution could vary by gender and age and whether the current 14-day quarantine period ignoring the patient demographic factors would be sufficient for the containment of COVID-19. A more accurate estimation of the incubation period using these factors can optimize COVID-19 surveillance guidelines. We estimate distribution of the incubation periods of COVID-19, and its association with gender and age, using the best available information to model uncertain dates of infection and symptom onset.

Our analysis was based on publicly reported, clinically confirmed cases with symptoms from two sources, available as of March 30, 2020 [5, 6] . To obtain the period of potential exposure, we included the following patients in the analytic dataset: 1) non-residents of Wuhan who visited Wuhan since December 1, 2019 and for whom the exposure period was the time between the earliest possible arrival to and the latest possible departure from Wuhan; 2) people with travel history of visiting the coronavirus-affected nations known at the time or taking a Diamond Princess cruise; 3) non-travel related cases with an exposure history based on contact with infected persons. Among these, 312 cases had recorded exposure periods, or at least an exposure ending date, and dates of symptom onset or hospitalization, as well as gender and age; this number reflects removal of 34 potentially duplicated cases (Supplement S1).

There were 111 cases with complete information on the exposure period (Supplement S2). For cases without an exact starting date of exposure interval, the initial date was set to have a maximum of thirty days of exposure 2 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . or December 1, 2019, whichever one was later. The ending date of exposure was set to precede the known dates of symptom onset and hospitalization. If the exact date of symptom onset was unknown, it was assumed to have occurred before hospitalization (i.e., left-censored).

In a total of 312 patients, the median age was 42 (interquartile range 33-55) years and 126 (40.4%) were women. Age was dichotomized at its median prior to the analysis. The effects of gender and age on the incubation period were estimated under the generalized odds-rate class of regression models, which consists of general parametric models including the log-logistic proportional odds models and the Weibull proportional hazards models, as special cases. [7] We estimated the distribution of incubation period by modelling the interval-censored exposure duration and the possible left-censored symptom onset time (Supplement S3). Figure 1 depicts four types of observed patient cases used for modelling.

Patients older than 42 years of age have, on average, longer incubation periods, compared to 42-year-old or younger patients (p=0.036), whereas gender has no effect on the incubation period (p=0.417). The median incubation period is 6.3 (95% CI 5.4-7.2) days overall, being 5.5 (95% CI 4.5-6.5) days and 7.2 (95% CI 5.8-8.6) days for younger and older patients, respectively. The mean incubation period is 6.7 (95% CI 5.5-7.9) days for the younger patients, and 8.7 (95% CI 7.2-10.2) days for the older patients (Supplement S4). Table 1 shows the estimated differences of days between the quantiles of incubation distribution from two age groups. The incubation period difference between the two age groups was not obvious at the 25th quantile or lower; however, it becomes conspicuous after the medians, leading to 6-day differences between the two age groups at the 97.5th quantile. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. 

The finding that persons older than 42 years have longer incubation periods than those who are younger has important implications for enacting age-specific quarantine policies. This result agrees with previous studies on Severe Acute Respiratory Syndrome that showed a relationship between age 4 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . and the incubation period. [8] Using the uniform 14-day quarantine policy recommended by the World Health Organization and the US Centers for Disease Control and Prevention, [9, 10] our estimators imply that 9.2% (95% CI 3.5-14.8%) of COVID-19 patients younger than 42, and 18.6% (95% CI 11.3-25.9%) of older patients may pose a risk of infection to others before onset of their symptoms in the worst-case scenario. Using a 21-day quarantine, these percentages reduced to 1.7% (95% CI 0-3.7%) and 5.6% (95% CI 1.5-9.7%), respectively. To ensure that at least 90% of cases' symptoms being manifested during quarantine periods, the required durations are estimated to be 14 (95% CI 11-16) days for patients 42 years of age or younger and 18 (95% CI 15-21) days for patients older than 42 years. As such, a unified quarantine policy could be inefficient during a viral outbreak. These estimates were derived from the conservative assumption that the quarantine periods started immediately after infection.

We used a larger data cohort consisting of more confirmed COVID-19 cases than other published papers in this field; the cases may not have the exact symptom onset time (left censored) and include an interval-censored incubation period. Our choice of a parametric model for the incubation period has more flexibility than other commonly used distributions. This flexible model was reduced to the Weibull distribution, a special case as reported in Backer et al [1] and Linton et al [4] (Supplement S5). We also used a descriptive method to check the model fitting by comparing a nonparametric estimation of the incubation distribution for interval-censored data with the model-based estimation under the Weibull distribution by age groups. As shown in Supplement S6, the Weibull distribution fits the data adequately. To examine how sensitive the estimates were to the assumption of 30 days of maximum exposure for cases with missing starting date of exposure interval, we performed the analyses by changing this maximum exposure duration to 20 days or by shifting the lower bound back and forth within 15 days. The changed assumptions had little effect on analysis results (Supplement S7). This analysis has some notable limitations. Our inferences relied on publicly reported confirmed cases that might over represent more severely symptomatic patients. Moreover, the definition of COVID-19 symptoms and hospitalization criteria could differ by country, especially during the initial outbreak. We combined the data sets from two different sources, and the potential variation in source criteria for tracing infected cases may lead to different exposure records. However, we obtained similar findings when fitting the model to each dataset separately, although one indicated no signif-5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 28, 2020. . icant differences but the same trend in the incubation period between the age groups (Supplement S8). An optimal differentiating cut-off age for the incubation period was not identified here due to the limited available data, which is worth pursuing to reduce risks to public health most effectively.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10.1101/2020.06.27.20141002 doi: medRxiv preprint

Incubation period of 2019 novel coronavirus (2019-ncov) infections among travellers from wuhan, china

Clinical characteristics of coronavirus disease 2019 in china

The incubation period of coronavirus disease 2019 (covid-19) from publicly reported confirmed cases: estimation and application

Incubation period and other epidemiological characteristics of 2019 novel coronavirus infections with right truncation: a statistical analysis of publicly available case data

COVID-19 Epidemic situation in real time website

Epidemiological data from the covid-19 outbreak, real-time case information

Estimation and testing in a twosample generalized odds-rate model

Alternative methods of estimating an incubation distribution: Examples from severe acute respiratory syndrome

Considerations for quarantine of individuals in the context of containment for coronavirus disease (COVID-19) website

Press Briefing by Members of the President's Coronavirus Task Force website