You are on page 1of 9

Clinical Infectious Diseases

Review Article

The Incubation Period Distribution of Coronavirus


Disease 2019: A Systematic Review and Meta-analysis
Hualei Xin,1 Jessica Y. Wong,1 Caitriona Murphy,1 Amy Yeung,1 Sheikh Taslim Ali,1,2 Peng Wu,1,2, and Benjamin J. Cowling1,2
1
World Health Organization Collaborating Centre for Infectious Disease Epidemiology and Control, School of Public Health, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong
Special Administrative Region, China; and 2Laboratory of Data Discovery for Health Limited, Hong Kong Science Park, New Territories, Hong Kong Special Administrative Region, China

Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021


Incubation period is an important parameter to inform quarantine period and to study transmission dynamics of infectious diseases.
We conducted a systematic review and meta-analysis on published estimates of the incubation period distribution of coronavirus
disease 2019, and showed that the pooled median of the point estimates of the mean, median and 95th percentile for incubation pe-
riod are 6.3 days (range, 1.8–11.9 days), 5.4 days (range, 2.0–17.9 days), and 13.1 days (range, 3.2–17.8 days), respectively. Estimates
of the mean and 95th percentile of the incubation period distribution were considerably shorter before the epidemic peak in China
compared to after the peak, and variation was also noticed for different choices of methodological approach in estimation. Our find-
ings implied that corrections may be needed before directly applying estimates of incubation period into control of or further studies
on emerging infectious diseases.
Keywords.  COVID-19; SARS-CoV-2; incubation period; systematic review; meta-analysis.

The incubation period of an infectious disease describes the selection bias during the exponential phase of an epidemic [11],
time delay between infection and onset of illness. Incubation determination of illness onset by the predetermined syndromic
periods can vary from one individual to another for various definitions [12], transmission occurring in different settings [11],
reasons including the difference in pathogen transmission and statistical methods and underlining assumptions [10] would
route, the dose of exposure to a pathogen, and the functionality have influenced estimation of the incubation period.
of the host immune system. Incubation periods are often char- Here, we reviewed published estimates of the incubation pe-
acterized by parametric probability distributions such as the riod distribution with a particular interest in the methodology
log-normal, Weibull, and gamma distributions. Coronavirus applied and potential biases that might have affected the esti-
disease 2019 (COVID-19) was first detected in China in mated mean and the right-hand tail measure of the incubation
December 2019 and rapidly spread worldwide, causing a global period distribution.
pandemic with >1.9 million deaths by 13 January 2021. The in-
cubation period distribution of COVID-19 has been a critical METHODS
epidemiological parameter in characterizing the transmission
This systematic review and meta-analysis followed the check-
dynamics, particularly the role of presymptomatic transmission
list of the Preferred Reporting Items for Systematic Reviews and
[1–4] and determination of the quarantine period for persons
Meta-Analyses (PRISMA) guidelines [13].
who might have been exposed to infection [5, 6].
A number of studies have reported estimates of the incubation Search Strategy and Selection Criteria
period distribution, with quite considerable variability between Articles reporting estimates of the incubation period distri-
studies [7, 8]. However, a number of issues can complicate esti- bution of COVID-19 were extracted from the database of
mation of the incubation period distribution and lead to partic- Medline (PubMed; US National Library of Medicine, Bethesda,
ular biases. First, data on the time of infection or exposure are Maryland) with the publication date between 1 February 2020
typically interval-censored rather than exactly known [9, 10], and and 25 September 2020. The following search strategy was used
the probability of infection over time within an exposure interval to search in “All Fields” of PubMed:
may not follow a uniform distribution [4]. Furthermore, the
1. “incubation period”

2. “coronavirus” OR “nCoV” OR “COVID” OR “COVID-19”
Received 15 January 2021; editorial decision 22 May 2021; published online 12 June 2021. OR “2019-nCoV” OR “2019 novel coronavirus” OR “nCoV-
Correspondence: P.  Wu, School of Public Health, Li Ka Shing Faculty of Medicine, The
University of Hong Kong, 7 Sassoon Road, Pokfulam, Hong Kong (pengwu@hku.hk). 2019” OR “SARS-CoV-2” OR “Wuhan pneumonia”
Clinical Infectious Diseases®  2021;XX(XX):0–0 3.
(“2020/02/01”[Date - Publication]: “2020/09/25”[Date
© The Author(s) 2021. Published by Oxford University Press for the Infectious Diseases Society
of America. All rights reserved. For permissions, e-mail: journals.permissions@oup.com.
- Publication])
DOI: 10.1093/cid/ciab501 4. #1 AND #2 AND #3

Incubation Period Distribution of COVID-19  •  cid 2021:XX (XX XXXX) • 1


Full texts of the articles identified via the search strategy were December 2019 when the first case in China was reported. We
reviewed for eligibility. Articles in English and Chinese were categorized estimates of the incubation period in China for
selected for assessment if the mean or median estimate of the time periods before and after the epidemic peak on 27 January
incubation period distribution was reported or if detailed indi- 2020 [17]. We also categorized the estimates based on the fitting
vidual data were included in the article allowing us to calculate approaches of the probability distribution (log-normal, gamma,
the median of the incubation period distribution. We excluded Weibull, normal) used in the studies.
studies that reported estimates derived from systematic reviews, Mean and median incubation periods were described and
studies that only reported the incubation period for a subgroup compared between the studies. Reported estimates of the mean
of the population (eg, children, older adults, persons with severe and the 95th percentile of the incubation period distribution of
illness or having certain types of underlying conditions), and COVID-19 for different subgroups (different study period and
studies that were published as letters or conference proceedings. statistical approach) were combined using a random-effects

Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021


All articles identified were reviewed by 2 authors (H. X.  and model, with the statistical heterogeneity assessed by the I2 sta-
C. M.) independently, and the Strengthening the Reporting of tistic. Meta-regression analyses using a mixed-effects model
Observational Studies in Epidemiology (STROBE) checklist were conducted to quantify the association between study pe-
was used to evaluate quality of the studies [14] (Supplementary riod and statistical approach and the estimates of the mean and
Table 1). the 95th percentile of incubation period [18]. In addition, we
retrieved open-source data on individual cases to reestimate
Data Extraction the incubation period for some of these studies to examine and
All data were extracted independently and entered in a stand- validate the results of our meta-analysis in particular. Finally,
ardized form by 2 authors. Information was extracted on the es- we derived a prospective correction on the estimates of incuba-
timates of the mean, the median, and the 95th percentile, and the tion period to address possible sampling bias during the early
corresponding 95% confidence interval (CI) of the incubation exponential growth phase of the epidemic (see Supplementary
period distribution of COVID-19. For studies reporting both Materials for details). Analyses were conducted in R version
the mean and the median, we selected the one with an estimated 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria)
95% CI while the mean was selected if the 95% CI was available with the metafor package.
for both or neither of the 2 measures. We used a normal ap-
proximation to derive the 95% CI for a reported mean incuba-
RESULTS
tion period without an estimated 95% CI when the interquartile
range and sample size were available [15]. For studies providing By applying the selection criteria, we included 72 of 311 iden-
individual data on the duration between infection and onset tified studies into our analysis, with 75 central tendency esti-
instead of an estimate of the incubation period, a normal per- mates, including 34 means and 41 medians (Figure 1). Most
centile method was used to estimate the median and the 95th (58/72 [80.6%]) of the publications were in English, and 62
percentile. For studies without 95th percentile but reporting the (86.1%) studies reported estimates of the incubation period
fitted distribution parameters, we calculated the 95th percentile using the data collected from China. The reported sample sizes
based on these parameters using fitdistrplus package in R. Other ranged between 4 and 2907 with only 13 (18.1%) studies having
information such as study language, study country, sample size, a sample size >200 cases (Supplementary Table 2).
start and end dates of the study period, parametric distribution, The mid-dates of study period for the 34 estimates of the
or statistical model used in estimation of the incubation period mean incubation period were between 11 January 2020 and
(log-normal, gamma, Weibull, normal) was also extracted for 13 April 2020. Of the 29 means reported from China, the mid-
each selected study. Some studies provided multiple estimates dates of the study period of 9 studies (31.0%) were before the
of the incubation period based on different parametric distribu- peak of the first epidemic wave while the rest (20/29 [69.0%])
tions; we only selected the best-fitted distribution as indicated. were after the epidemic peak. Various distributions were applied
to parameterize the incubation period, including log-normal (9
Statistical Analysis [26.5%]), gamma (6 [17.6%]), Weibull (7 [20.6%]), and normal
The mid-date of the study period corresponding to each esti- (12 [35.3%]) (Supplementary Table 2). Estimates of the mean
mate was calculated based on the start and end dates of the re- incubation period were found to be higher than estimates of the
ported study period. Several studies did not disclose their exact median (Figure 2), as would be expected from a typically right-
start date. For studies that analyzed cases outside Hubei prov- skewed distribution. The pooled medians of the point estimates
ince in China without reporting a start date of the study, the of the mean and median incubation period were 6.3 days (range,
start date was redefined as 20 January 2020 on which the first 1.8–9.1 days) and 5.4 days (range, 2.0–17.6 days), respectively.
confirmed case outside Hubei was identified [16]. For other Typical delays between the study period and the time of
studies conducted in China, the start date was redefined as 31 publication were a few weeks in early 2020, which increased

2 • cid 2021:XX (XX XXXX) • Xin et al


Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021
Figure 1.  Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram indicating the screening process to identify studies reporting estimates of
the incubation period for coronavirus disease 2019.

to a few months in later 2020 (Figure 3A). The estimates of used to estimate the 95th percentile (Supplementary Table 2).
the mean incubation period appeared to be longer for studies The pooled median of the point estimates of the 95th percen-
using data collected in a later period (Figure 2A). The pooled tile of the incubation period was 12.5  days (range, 3.2–18.3).
estimate of mean incubation periods before the epidemic peak Similar to the mean incubation period, lower estimates of the
in China was 5.2 days (9 studies; 95% CI, 4.8–5.7; I2 = 56.5%), 95th percentile were reported in the early phase of the epi-
significantly lower than the pooled estimate for studies con- demic compared to the later phase of the epidemic in China
ducted after the peak (18 studies; 7.2 days [95% CI, 6.6–7.8]; (Figure 3D), with the pooled estimate of 95th percentile to be
I2 = 89.5%) (Figure 3B, Figure 3C). We also identified an asso- 11.0 days (5 studies; 95% CI, 9.9–12.0; I2 = 0.0%) before peak
ciation between the choice of parametric distribution and the and 14.6 days (7 studies; 95% CI, 13.7–15.5; I2 = 66.9%) after
estimated mean incubation period (Figure 4B). The shortest peak. Figure 4D shows variations in the 95th percentile with
estimate of the pooled mean incubation period was from different distributional assumptions. Pooled estimates of the
log-normal (9 studies; 5.9 days [95% CI, 5.1–6.7]; I2 = 81.0%), 95th percentile were 12.6  days (7 studies; 95% CI, 11.2–14.0;
followed by gamma distribution (6 studies; 6.5 days [95% CI, I2  =  61.7%) and 14.1  days (5 studies; 95% CI, 12.3–15.8;
4.9–8.1]; I2 = 94.4%), normal distribution (11 studies; 6.7 days I2 = 87.8%) for estimates based on log-normal distribution and
[95% CI, 6.0–7.3]; I2  =  89.8%), and Weibull distribution (6 Weibull distribution, respectively. Our analyses on individual
studies; 7.0 days [95% CI, 5.8–8.2]; I2 = 93.3%). Similar vari- case data available for Singapore and Tianjin produced similar
ations were identified from an analysis of individual case data observations (Supplementary Figure 2).
collected from Singapore and Tianjin (Supplementary Figures The meta-regression analyses were conducted based only
1 and 2). on the data collected from China, which allowed us to ex-
We identified 34 estimates of the 95th percentile of the in- amine the potential variations in the reported incubation
cubation period distribution. The mid-dates of study period periods before and after the epidemic peak. In total, 27 and
for these studies were between 11 January 2020 and 5 March 12 estimates of the mean and 95th percentile were included
2020, with 11 (32.4%) estimates calculated before the local epi- into the analysis after excluding the estimates from other
demic peak. Distributions of log-normal (10 [29.4%]), gamma territories (5 and 13 estimates for mean and 95th percen-
(5 [14.7%]), Weibull (7 [20.6%]), and normal (12 [35.3%]) were tile, respectively), without reporting the standard deviation,

Incubation Period Distribution of COVID-19  •  cid 2021:XX (XX XXXX) • 3


Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021

Figure 2.  Estimates of the incubation period for coronavirus disease 2019 from selected studies by type of measure. A, Forest plot of reported mean (dots) estimates of the
incubation period with 95% confidence intervals (CIs) (bars) ordered by mid-date of the study period. B, Forest plot of reported median (triangles) estimates of the incubation
period with 95% CIs (bars) ordered by mid-date of the study period. C, Boxplot of reported mean and median estimates of the incubation period. Citations in panels A and B
corresponds to the reference numbers in Supplementary Materials.

4 • cid 2021:XX (XX XXXX) • Xin et al


Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021
Figure 3.  Reported estimates of the mean and the 95th percentile of the incubation period for coronavirus disease 2019 (COVID-19) by study period. A, The study pe-
riod (bars) and publication date (dots) for studies included in the analysis, ordered by mid-date of the study period. Citations corresponds to the reference numbers in
Supplementary Materials. B, Estimates of the mean (circles) and the 95th percentile (triangles) by study mid-date from studies using the data collected from China. Gray bars
indicate daily numbers of confirmed COVID-19 cases in China reported to the World Health Organization. Red dotted line indicates the date of 27 January 2020 to classify
the time periods before and after the epidemic peak in China. C, Boxplot and reported individual estimates (dots and triangles) of the mean before and after the peak of an
epidemic based on the dichotomized mid-date of the study period. D, Boxplot and reported individual estimates (dots and triangles) of the 95th percentile before and after
the peak of an epidemic based on the dichotomized mid-date of the study period.

standard error, or 95% CI (1 and 8 estimates), and an ob- We examined the estimates from 4 studies with the mid-point
vious outlier (1.8 days and 3.2 days for the mean and 95th of the study period before the epidemic peak characterizing
percentile). Table 1 shows that the estimates of the mean the incubation period with gamma distributions to explore
and the 95th percentile after the epidemic peak were sig- possible sampling bias in early stage of pandemic. The pooled
nificantly higher than estimates before the peak. Compared mean incubation period from these 4 studies was 5.4  days
to studies using log-normal distribution, no significant dif- (95% CI, 4.4–6.5). If allowing for an exponential growth rate
ferences in the estimates were identified for studies using to be 0.10 [5], the corrected mean incubation period would be
either gamma, Weibull, or normal distribution. 6.4 days (95% CI, 5.2–7.6) (Figure 5). A higher growth rate was

Incubation Period Distribution of COVID-19  •  cid 2021:XX (XX XXXX) • 5


Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021
Figure 4.  Reported estimates of the mean and the 95th percentile of the incubation period for coronavirus disease 2019 by type of distribution applied in data fitting. A,
Forest plot of reported mean incubation periods with 95% confidence intervals (CIs) (bars) estimated with log-normal, gamma, Weibull, or normal distribution, ordered first
by analysis method and then, within each analysis method, by the mid-date of the study period. B, Boxplot and reported individual estimates of the mean incubation period
by type of analytic approach. C, Forest plot of reported 95th percentiles of the incubation period with 95% CIs (bars) estimated with log-normal, gamma, Weibull, or normal
distribution (color coded), with the estimates ordered by the mid-date of the study period within each category of the analytic approaches. D, Boxplot and reported individual
estimates of the 95th percentiles of the incubation period by type of analytic approach. Citations in panels A and C corresponds to the reference numbers in Supplementary
Materials.

associated with a greater underestimation of the mean incu- between 11 and 14.6 days. Notably, we identified a bias towards
bation period in estimates without a correction (Figure 5 and shorter estimates of the incubation period in the early stage of
Supplementary Materials). the pandemic while case numbers were rising exponentially,
which was also indicated by the temporal patterns of the incu-
bation period estimated throughout the epidemic based on in-
DISCUSSION
dividual case data from Singapore and Tianjin (Supplementary
We reviewed 72 studies on the incubation period distribution Figure 1).
of COVID-19, 62 of which were based on data collected from Variation in estimates of the mean incubation period were
China. The pooled estimates from our analysis indicated that expected to result from bias in case ascertainment that people
the mean incubation period of COVID-19 was likely to fall with shorter incubation periods were more likely to be included
between 5.2 and 7.2  days, and the 95th percentile would be in the growing stage of the pandemic, which could be corrected

6 • cid 2021:XX (XX XXXX) • Xin et al


Table 1.  Results From a Meta-regression Analysis of Variables Potentially Associated With Estimation of the Mean and 95th Percentile of the Coronavirus
Disease 2019 Incubation Period

Mean of Incubation Period 95th Percentile of Incubation Period


(n = 27) (n = 12)

Variable No. of Estimates β (95% CI) No. of Estimates β (95% CI)

Study period
  Before peak 9 0 Referent 5 0 Referent
  After peak 18 1.95 (.92–2.98) 7 3.4 (1.62–5.17)
Approach
 Log-normal 7 0 Referent 7 0 Referent
 Gamma 5 .61 (–.80 to 2.02) … …

Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021


 Weibull 6 .82 (–.54 to 2.18) 5 .43 (–1.18 to 2.05)
 Normal 9 .15 (–1.21 to 1.52) … …

by observing exposure to infection and the development of subsequent analyses and to support health policies [20]. Our
symptoms in a cohort of individuals over the epidemic [11]. In proposed correction for sampling bias might be able to pro-
addition, in the early stage of the pandemic, published studies vide a more accurate estimate of the incubation period from
tended to include patients with well-characterized informa- data collected in the early phase of an epidemic when infections
tion, such as seriousness of infection or exposure details, which are increasing exponentially. Correcting for the growth rate
might lead to a underestimation of the incubation period [19]. can account for the proportion of cases with longer incubation
Other factors could also contribute to the difference, including periods who are relatively less likely to be included in early data
publication bias, overrepresentation of milder cases, and dif- (Supplementary Materials and Figure 5). Shorter estimates of
ferent ways of establishing exposure windows. Few studies ac- the incubation period could also be related to exposure to a
counted for such potential bias in estimation of the incubation high viral load or virus strains with higher virulence [16].
period. The earliest, perhaps biased, estimates of the incubation One of the applications of the incubation period distribution
period distribution have been, however, widely used in many is in determining the time period of quarantine for exposed in-
dividuals. Our study found that the pooled estimates of the 95th
percentile of the incubation period before and after the epi-
demic peak were 11.0 and 14.6 days, respectively. This indicated
that a quarantine period of 14 days could capture at least 95% of
infected individuals who would develop symptoms.
The distinction between the incubation period and the latent
period implies that the latent period of COVID-19 is gener-
ally expected to be shorter than the incubation period because
of the occurrence of presymptomatic transmission [3, 21, 22].
A number of estimates of the latent period have been published,
with the reported means ranging from 2.6 to 3.3 days [23, 24] or
from 1.88 to 7.4 days before symptom onset [3, 22]. The earlier
shedding of virus than symptom onset made it critical to test
exposed individuals during quarantine so that asymptomatic
cases could also be identified [25–28]. Laboratory testing before
exit from 2-week quarantine may identify some infected persons
with the incubation period >14 days. Johansson et al estimated
that the residual risk of transmission after a 10-day quarantine
would be as low as 0.3% for asymptomatic individuals with a
negative laboratory test result at the last day of quarantine [29].
Quilty et  al estimated that the transmission-potential averted
Figure 5.  Prospective correction for the incubation period estimates during the
early phase of the pandemic. For given estimates of incubation period as gamma rates were similar between 14 days of quarantine without poly-
(mean [standard deviation]) in the different studies during the exponential phase of merase chain reaction test and 7 days of quarantine with a neg-
the epidemic, the corrected pooled mean (blue line) and 95% confidence interval ative exit test [30]. On 2 December 2020, the US Centers for
(shaded area) are presented for different possible growth rates (r = 0.00–0.20). The
scenario r = 0 indicates no correction for growth rate and the respective suggested Disease Control and Prevention updated the recommendation
corrections can be attained based on the exponential growth rates, r > 0. on the quarantine period, allowing asymptomatic individuals

Incubation Period Distribution of COVID-19  •  cid 2021:XX (XX XXXX) • 7


to exit quarantine after 10 days, or after 7 days with a negative Notes
virus test at exit [31]. Author Contributions. H.  X., P.  W., and B.  J. C.  conceived the study.
H. X., J. Y. W., C. M., and A. Y. contributed to data collection. H. X. per-
Since incubation periods of most infectious diseases follow
formed the data analyses and wrote the first draft of the manuscript. All
a right-skewed distribution [32], application of the symmetric authors interpreted the data, provided critical review and revision of the
normal distribution would tend to underestimate the right-hand text, and approved the final version of the manuscript.
Acknowledgments. The authors thank Julie Au for technical support.
tail of the incubation period. Variations in the reported estimates
Financial support. This work was supported by the Health and Medical
of incubation period were observed from studies using different Research Fund, Food and Health Bureau, Government of the Hong Kong
distributions whereas no significant difference was identified in Special Administrative Region (grant number COVID190118). B.  J. C.  is
our meta-regression analysis. The estimates of the 95th percentile supported by the AIR@innoHK program of the Innovation and Technology
Commission of the Hong Kong Special Administrative Region government.
of the incubation period might have been more heavily affected Potential conflicts of interest. B.  J. C.  consults for Roche and Sanofi
by the tail probability of the underlining distributions assumed,

Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021


Pasteur. All other authors report no other potential conflicts of interest.
leading to larger estimates of the 95th percentile if a higher prob- All authors have submitted the ICMJE Form for Disclosure of Potential
Conflicts of Interest. Conflicts that the editors consider relevant to the con-
ability was assumed at the tail of the underlying distribution. This
tent of the manuscript have been disclosed.
was further validated by the estimated incubation periods by fit-
ting different distributions based on the same individual data from References
Singapore and Tianjin (Supplementary Figure 1). Several other 1. Nishiura H, Linton NM, Akhmetzhanov AR. Serial interval of novel coronavirus
(COVID-19) infections. Int J Infect Dis 2020; 93:284–6.
studies similarly reported that the 95th percentile of the incuba- 2. Du Z, Xu X, Wu Y, Wang L, Cowling BJ, Meyers LA. Serial interval of COVID-19
tion period was longer with a log-normal distribution than with among publicly reported confirmed cases. Emerg Infect Dis 2020; 26:1341–3.
3. He X, Lau EHY, Wu P, et al. Temporal dynamics in viral shedding and transmissi-
a gamma or Weibull distribution [33–36]. Less-variable estimates
bility of COVID-19. Nat Med 2020; 26:672–5.
might have derived if the exact time of exposure could be obtained 4. Ren X, Li Y, Yang X, et al. Evidence for pre-symptomatic transmission of corona-
[10]. Furthermore, the sample size would also affect the precision virus disease 2019 (COVID-19) in China. Influenza Other Respir Viruses 2021;
15:19–26.
of estimates of the 95th percentile of the incubation period [10]. 5. Li Q, Guan X, Wu P, et al. Early transmission dynamics in Wuhan, China, of novel
Our study had a number of limitations. First, most arti- coronavirus-infected pneumonia. N Engl J Med 2020; 382:1199–207.
6. Linton NM, Kobayashi T, Yang Y, et al. Incubation period and other epidemiolog-
cles included in our study used publicly available data. Some ical characteristics of 2019 novel coronavirus infections with right truncation: a
studies in our review might have used overlapping data, statistical analysis of publicly available case data. J Clin Med 2020; 9:538.
7. Biggerstaff  M, Cowling  BJ, Cucunubá  ZM, et  al; WHO COVID-19 Modelling
leading to double counting in the pooled estimates. Second, Parameters Group. Early insights from statistical and mathematical modeling of
as described above, self-reported information on symptom key epidemiologic parameters of COVID-19. Emerg Infect Dis 2020; 26:e1–e14.
8. Khalili  M, Karamouzian  M, Nasiri  N, Javadi  S, Mirzazadeh  A, Sharifi  H.
onset could be subject to reporting biases as well as affected
Epidemiological characteristics of COVID-19: a systematic review and meta-
by the case definitions and approaches to ascertainment. analysis. Epidemiol Infect 2020; 148:e130.
Third, some factors that could correlate with the incubation 9. Sartwell PE. The distribution of incubation periods of infectious disease. Am J
Hyg 1950; 51:310–8.
period such as age, disease severity, and interval censoring 10. Cowling  BJ, Muller  MP, Wong  IO, et  al. Alternative methods of estimating an
of exposure or illness onset were not included in this study incubation distribution: examples from severe acute respiratory syndrome.
Epidemiology 2007; 18:253–9.
because of data availability. Finally, we were not able to as- 11. Qin J, You C, Lin Q, Hu T, Yu S, Zhou XH. Estimation of incubation period dis-
sess the potential impact of contact tracing strategies on es- tribution of COVID-19 using disease onset forward time: a novel cross-sectional
and forward follow-up study. Sci Adv 2020; 6:eabc1202.
timates of the incubation period, particularly when the time 12. Böhmer MM, Buchholz U, Corman VM, et al. Investigation of a COVID-19 out-
period of contact tracing was changed, given the lack of rel- break in Germany resulting from a single travel-associated primary case: a case
series. Lancet Infect Dis 2020; 20:920–8.
evant information.
13. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred Reporting
In conclusion, our study identified variations in the reported Items for Systematic Reviews and Meta-Analyses: the PRISMA statement. PLoS
estimates of the incubation period for COVID-19 and method- Med 2009; 6:e1000097.
14. Cuschieri S. The STROBE guidelines. Saudi J Anaesth 2019; 13:31–4.
ological factors associated with the estimates, including time of 15. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard devi-
study, type of estimate, and assumed distribution of the param- ation from the sample size, median, range and/or interquartile range. BMC Med
Res Methodol 2014; 14:135.
eter. Shorter incubation periods were more likely to be identi- 16. Leung C. The difference in the incubation period of 2019 novel coronavirus
fied at the early stage of an epidemic or by a certain statistical (SARS-CoV-2) infection between travelers to Hubei and nontravelers: the
need for a longer quarantine period. Infect Control Hosp Epidemiol 2020;
approach, indicating the need for correction on the estimate be- 41:594–6.
fore using it for determination of a valid quarantine period or 17. Ali ST, Wang L, Lau EHY, et al. Serial interval of SARS-CoV-2 was shortened over
time by nonpharmaceutical interventions. Science 2020; 369:1106–9.
for studies on transmission dynamics of COVID-19.
18. Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat
Softw 2010; 36:1–48.
Supplementary Data 19. McAloon C, Collins Á, Hunt K, et al. Incubation period of COVID-19: a rapid
Supplementary materials are available at Clinical Infectious Diseases on- systematic review and meta-analysis of observational research. BMJ Open 2020;
line. Consisting of data provided by the authors to benefit the reader, 10:e039652.
20. Hellewell  J, Abbott  S, Gimma  A, et  al; Centre for the Mathematical Modelling
the posted materials are not copyedited and are the sole responsibility
of Infectious Diseases COVID-19 Working Group. Feasibility of controlling
of the authors, so questions or comments should be addressed to the
COVID-19 outbreaks by isolation of cases and contacts. Lancet Glob Health
corresponding author. 2020; 8:e488–96.

8 • cid 2021:XX (XX XXXX) • Xin et al


21. Ferretti  L, Wymant  C, Kendall  M, et  al. Quantifying SARS-CoV-2 transmission 30. Quilty  BJ, Clifford  S, Flasche  S, Kucharski  AJ, Edmunds  WJ. Quarantine and
suggests epidemic control with digital contact tracing. Science 2020; 368:eabb6936. testing strategies in contact tracing for SARS-CoV-2: a modelling study. Lancet
22. Tindale LC, Stockdale JE, Coombe M, et al. Evidence for transmission of COVID- Public Health 2021; 6:e175–83.
19 prior to symptom onset. Elife 2020; 9:e57149. 31. Centers for Disease Control and Prevention. Options to reduce quarantine for
23. Peirlinck M, Linka K, Sahli Costabal F, Kuhl E. Outbreak dynamics of COVID-19 contacts of persons with SARS-CoV-2 infection using symptom monitoring and
in China and the United States. Biomech Model Mechanobiol 2020; 19:2179–93. diagnostic testing. 2020. Available at: https://www.cdc.gov/coronavirus/2019-
24. Zhao  S. Estimating the time interval between transmission generations when ncov/more/scientific-brief-options-to-reduce-quarantine.html. Accessed 30
negative values occur in the serial interval data: using COVID-19 as an example. December 2020.
Math Biosci Eng 2020; 17:3512–9. 32. Lessler  J, Reich  NG, Brookmeyer  R, Perl  TM, Nelson  KE, Cummings  DA.
25. Ng  Y, Li  Z, Chua  YX, et  al. Evaluation of the effectiveness of surveillance and Incubation periods of acute respiratory viral infections: a systematic review.
containment measures for the first 100 patients with COVID-19 in Singapore— Lancet Infect Dis 2009; 9:291–300.
January 2–February 29, 2020. MMWR Morb Mortal Wkly Rep 2020; 69:307–11. 33. Backer  JA, Klinkenberg  D, Wallinga  J. Incubation period of 2019 novel coro-
26. Li Z, Chen Q, Feng L, et al; China CDC COVID-19 Emergency Response Strategy navirus (2019-nCoV) infections among travellers from Wuhan, China, 20-28
Team. Active case finding with case management: the key to tackling the COVID- January 2020. Euro Surveill 2020; 25:2000062.
19 pandemic. Lancet 2020; 396:63–70. 34. Lai  CKC, Ng  RWY, Wong  MCS, et  al. Epidemiological characteristics of the

Downloaded from https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciab501/6297425 by guest on 20 August 2021


27. Lin J, Huang W, Wen M, et al. Containing the spread of coronavirus disease 2019 first 100 cases of coronavirus disease 2019 (COVID-19) in Hong Kong Special
(COVID-19): meteorological factors and control strategies. Sci Total Environ Administrative Region, China, a city with a stringent containment policy. Int J
2020; 744:140935. Epidemiol 2020; 49:1096–105.
28. Jin H, Liu J, Lu L, Cui M. COVID-19 emergencies in the globe: China’s experi- 35. Dai J, Yang L, Zhao J. Probable longer incubation period for elderly COVID-19
ence in controlling COVID-19 and lessons learned. Int J Qual Health Care 2020; cases: analysis of 180 contact tracing data in Hubei Province, China. Risk Manag
744:mzaa143. Healthc Policy 2020; 13:1111–7.
29. Johansson  MA, Wolford  H, Paul  P, et  al. Reducing travel-related SARS-CoV-2 36. Patrikar  SR, Kotwal  A, Bhatti  VK, et  al. Incubation period and reproduction
transmission with layered mitigation measures: symptom monitoring, quaran- number for novel coronavirus 2019 (COVID-19) infections in India. Asia Pac J
tine, and testing. BMC Med 2021; 19:94. Public Health 2020; 32:458–60.

Incubation Period Distribution of COVID-19  •  cid 2021:XX (XX XXXX) • 9

You might also like