Research On The Distribution Of The Incubation Period Of COVID-19

The incubation period is the timeinterval between exposure to a pathogen and symptomatic disease, its lengthdepends on the disease and the vector, and its probability density function isusually asymmetric and skewed to the right. The importance of diseaseincubation period distribution research is reflected in the following aspects:First, incubation period distribution is often used together with existinginfectious disease models or backtracking algorithms, which is very importantfor predicting the development process of epidemics. If the incubation periodis long, infection It is highly likely that the population will unknowinglytransmit the virus to others; second, identifying covariates that may lead toprolonged incubation periods can lead to effective improvements in preventionand control measures; third, information on incubation periods is also a keyparameter in vaccine clinical trials.

The earliest research on theincubation period of diseases can be traced back to 1950. Sartwell et al.(1950) were the first to discuss the distribution function of the incubationperiod of infectious diseases, carried out pioneering work on the distributionof the incubation period, and proposed that lognormal distribution functioncould be used to describe the distribution of the incubation period. Armenianet al. (1974) believed that the incubation period of infectious diseases shouldfollow lognormal distribution and pointed out that the comprehensiveepidemiological investigation of any disease should include the study of theincubation period. Since then, lognormal distribution has been widely used tosimulate the distribution of incubation periods of infectious diseases.Meanwhile, Armenian et al. (1983) further pointed out that the incubationperiod of infectious diseases should follow lognormal distribution, andproposed that the comprehensive epidemiological investigation of any diseaseshould include the study of the incubation period. In addition, the researchersalso studied the distributions of incubation periods for various other diseasesand suggested that other distributions, such as the gamma, Weibull, andlog-Logistic distributions, could also be used to fit the observed incubation periodsfor infectious diseases. Since the development of COVID-19, some scholars haveconducted studies on the incubation period of COVID-19. In the early stage ofthe epidemic, Li et al. (2020) studied the distribution of the incubationperiod of COVID-19 based on the observation information of 10 confirmedCOVID-19 cases in Hubei Province and suggested that the exposed population bequarantined for medical observation for 14 days, which later became thestandard days for relevant departments to implement quarantine policies andplayed an important role in actual prevention and control. Based on the data of1,099 patients from 552 hospitals in 31 provinces and municipalities beforeJanuary 29, 2020, Zhong's team described the clinical characteristics of thedisease in detail, noting that the median incubation period of COVID-19 was 3.0days (the incubation period ranged from 0 to 24 days). Wang et al. (2020) usedlognormal distribution to fit the incubation data of confirmed COVID-19 casesin Henan Province, and the results showed that the incubation loci of 0.25 and0.75 were 4.1 and 9.4 days, respectively. Recently, Backer et al. (2020) usedWeibull distribution, gamma distribution and lognormal distribution tore-estimate the incubation period of COVID-19 based on the travel history andsymptom onset date of 88 patients with early confirmed cases in Wuhan. Theresults showed that Weibull distribution was the most consistent, and theaverage incubation period of COVID-19 was estimated to be 6.4 days. The 95%confidence interval was 5.6 to 7.7 days, the incubation period varied from 2.1to 11.1 days, and the 97.5% sublocus of the incubation period distribution was 11.1 days.

Different from the above-mentionedestimation methods of incubation period, Qin et al. (2020) proposed a novelmethod to estimate the distribution of incubation period from anotherperspective, and conducted a follow-up survey by continuously trackingasymptomatic people when they left Wuhan until they developed symptoms , thatis, the incubation period is regarded as an update process, and the timedifference between the crowd leaving Wuhan and the emergence of symptoms isregarded as a forward recursive time. By reducing recall bias, this method doesnot require traceability of exposure history. It only needs to pass the timewhen the observable cases left Wuhan to the time of onset, and utilize theabundant and easily available forward time data to improve the accuracy ofestimation, and obtain unobservable cases. Distribution of incubation periods.The results of the study showed that the 90% split point of the incubationperiod distribution was 16.7 days, which was considerably higher than theresults in the existing literature.

However, most of the existingliterature is a retrospective study on the incubation period of emergent publichealth diseases, and there is little mention of how to accurately estimate theincubation period during the development of the disease. In addition, in theexisting literature on sudden infectious diseases, researchers mostly use themethod under the classical assumption that "the sample is unbiased and canreflect the overall situation of the real incubation period" to estimatethe distribution of the incubation period data, ignoring The sample of data observationscollected in the early stages of a disease outbreak is biased and does notreflect the true overall picture.