Effects of prolonged incubation period and centralized quarantine on the COVID-19 outbreak in Shijiazhuang, China: a modeling study

Coronavirus disease 2019 (COVID-19) emerged in Wuhan, Hubei Province, China during early December 2019 [1]. The causative virus, SARS-CoV-2, could be spread by asymptomatic, presymptomatic, and symptomatic patients via droplets during close face-to face contact [2]. As of 27 April 2021, there were more than 147 million confirmed cases and more than 3 million deaths worldwide [3]. China had 103,503 confirmed cases [3], about half of which were reported in Wuhan during the first wave of the pandemic.

Because of serial strict non-pharmaceutical interventions (NPIs) [4], the first wave of the COVID-19 pandemic in China was controlled by the end of March 2020. However, increasing travels among China and foreign countries and rising infected cases in foreign countries increased the spread of infection, and there were many COVID-19 outbreaks elsewhere in China. The increase in domestic population mobility and relaxation of NPIs had also contributed to local outbreaks. Since the end of March 2020, there were reports of sporadic and localized outbreaks in some important international transportation hub cities and in cities with large populations, such as Beijing [5], Urumqi [6], Dalian [7], Qingdao, Kashgar, and Shanghai. Based on the initial success in controlling the first wave of the pandemic in Wuhan, the introduction of non-pharmaceutical interventions (detailed and rapid close contact tracing; centralized quarantine/isolation; home quarantine; managed closing of communities; traffic restrictions) helped to prevent virus transmission during these subsequent local outbreaks. Moreover, mass nucleic acid testing was implemented to help control these outbreaks. In addition, the virus has mutated because of the global nature of the pandemic; before the herd immunity is established, NPIs continue to play significant roles in controlling this global pandemic.

Shijiazhuang, the capital of Hebei Province, is located in North China and has a population of 10 million and consists of 22 administrative districts. On 2 January 2021, a local outbreak of COVID-19 occurred in Shijiazhuang [8]. This outbreak differed in some ways from previous localized outbreaks. Analysis of the COVID-19 outbreak in Shijiazhuang may therefore be useful as a reference for controlling local outbreaks of COVID-19 in other cities and provide important new information about this disease.

We used publicly available data to analyze the characteristics of the local outbreak of COVID-19 in Shijiazhuang. In particular, we estimated the incubation period and serial interval, used the SEIR^+q model to simulate the spread and transmission of COVID-19, and assessed the impact of different non-pharmaceutical interventions.

From 2 January to 14 February 2021, there were 869 patients with symptomatic COVID-19 in the 14 districts of Shijiazhuang (there were no cases in 8 districts), 80.7% of these patients were from Gaocheng District, and 90.9% of the patients from Gaocheng District lived in Zengcun (Fig. 2A). Among all 869 patients, 78.1% were immediately confirmed cases and 21.9% were initially asymptomatic but subsequently classified as symptomatic (later becoming symptomatic cases), 40.2% were male, the median age was 47.0 years, most patients (81.2%) were 20 to 79 years-old, and 47.8% were centrally quarantined before diagnosis (Table 1). There was no significant difference in sex between the immediately confirmed cases and later becoming symptomatic cases. However, the later becoming symptomatic cases were younger (P < 0.001) and a greater percentage of immediately confirmed cases were centrally isolated (P < 0.001).

Most (89.6%) COVID-19 patients were confirmed before 20 January, namely, before the third-round of the city-wide nucleic acid testing (Table 1, Fig. 2B). However, more of the later becoming symptomatic cases were clinical diagnosed after 20 January (P = 0.007). As a result of the first and second rounds of city-wide nucleic acid testing, most patients (93.9%) were identified and isolated before 20 January (Fig. 2C), leading to timely control of the source of infection. For all patients, the mean lag time from a positive nucleic acid test to clinical confirmation (namely make a definite diagnosis, which was used for the distinguish of asymptomatic cases and later becoming symptomatic cases) was 2.7 days. Notably, the mean lag time from a positive nucleic acid test to clinical confirmation was longer for later becoming symptomatic cases than immediately confirmed cases (7.7 days vs. 1.3 days, P < 0.001).

According to the LooIc criterion, the Weibull distribution provided the best fits to data for all cases, immediately confirmed cases, and later becoming symptomatic cases (Additional file 1: Table S2). The average incubation period was 11.6 days (95% CI 10.6, 12.7 days) for all cases, 10.6 days (95% CI 9.5, 11.7 days) for immediately confirmed cases, and 15.8 days (95% CI 13.7, 17.8 days) for later becoming symptomatic cases (Table 2). We also analyzed the time between symptom onset in 74 consecutive generations of patients and 64 corresponding index patients, with exclusion of those who reported the same date of symptom onset as the index patients. The serial intervals of all cases, immediately confirmed cases, and later becoming symptomatic cases followed Gamma distributions, and the estimated means were 6.6 days (all cases), 5.1 days (immediately confirmed cases), and 10.4 days (later becoming symptomatic cases; Table 2). A small number of consecutive generations of patients may have had another exposure to an unidentified infection source, which couldn’t be excluded in our analysis. Thus, we performed a sensitivity analysis using different delay times between symptom onset of index patients and consecutive generations of patients. The estimated results were 6.6 to 9.3 days for all cases, 5.1 to 7.6 days for immediately confirmed cases, and 10.4 to 12.3 days for later becoming symptomatic cases (Additional file 1: Table S3). We also compared the distributions of the incubation period and the serial interval for each of these groups (Fig. 3). In each case, the serial interval was shorter than incubation period. Incubation period and serial interval of female were less than those of male (Fig. 4A). Except in the immediately confirmed cases, incubation period and serial interval of cases aged 19+ were less than those of cases aged 0–18 (Fig. 4B).

The SEIR^+q model provided a good fit to the data (Fig. 5A). We used the fitted model to simulate the effects of changes of four key parameters (Q, α, ω, ρ_E) that correspond to different interventions. The range of the incubation period (1/α) was 5 to 28 days, the range of the isolation period (1/ω) was 14 to 28 days, and the ranges of Q (effectiveness of comprehensive quarantine measures) and ρ_E (effectiveness of nucleic acid detection) were 0.8 times to 1.2 times their initial values. Due to the reduced infectiousness of presymptomatic and asymptomatic individuals [17, 18], use of the same prevention and control measures but a longer incubation period, the cumulative number of cases reduced but the duration of the epidemic increased (Fig. 4B, Additional file 1: Table S4). If the isolation period and the effectiveness of the comprehensive quarantine measures decreased, the cumulative number of cases increased and the epidemic period was prolonged (Fig. 5C, D; Additional file 1: Table S4). Increasing the effectiveness of the nucleic acid test reduced the cumulative number of cases, but had little effect on the duration of the epidemic (Fig. 5E, Additional file 1: Table S4).

If the incubation period was shorter, more stringent prevention and control measures were needed to curb the epidemic and minimize the cumulative number of cases (Fig. 5F–H). A comparison of the effect of the three measures under the same change of incubation period indicated that strengthening comprehensive quarantine measures provided the most effective intervention, followed by improving nucleic acid testing; extending the isolation period had limited effect. Thus, if the incubation period was less than 7.5 days, more than 28 days of quarantine/isolation were needed to control the outbreak (Fig. 5F). If the isolation time decreased, more stringent quarantine measures and nucleic acid testing were needed to control the outbreak (Fig. 5I, J). To minimize the cumulative number of cases, more improvements were needed for nucleic acid testing than quarantine. If the effectiveness of the comprehensive quarantine measures declined by 10%, a 20% or more increase in the effectiveness of nucleic acid testing was needed to minimize the cumulative number of cases (Fig. 5K).

We identified 869 patients with confirmed symptomatic COVID-19 and found a higher proportion of females than males, consistent with previous studies [4, 21, 22]. In comparison with previous studies, our COVID-19 patients were younger, with a higher proportion of patients aged 0 to 19 years old. Some studies suggested that the risk of infection from family members was greater than that from other contacts [23‐25]. Therefore, after closure of schools on 6 January, children and adolescents may have had a greater risk of infection, and this might have contributed to the relatively high proportion of young patients. We also found that the later becoming symptomatic cases were younger than the immediately confirmed cases, meaning that symptoms might be milder or appear later in children and young people than in adults [26, 27]. However, there was no significant difference in the percentages of later becoming symptomatic cases and immediately confirmed cases among males and females.

We estimated that the mean incubation period was 11.6 days, longer than reported in previous studies [1, 2, 9, 10, 28]. Obviously, the mean incubation period of later becoming symptomatic cases (15.8 days) was much longer than that of immediately confirmed cases (10.6 days). The mean serial interval was 6.6 days, shorter than the serial interval (7.5 days) during the first COVID-19 epidemic wave in Wuhan [1], but longer than other studies [9, 29, 30]. And the serial interval of our later becoming symptomatic cases was much longer (10.4 days). There were also several studies reported that the incubation periods of cases associated with the Wuhan epidemic were longer than 14 days [31, 32]. Besides, taking Shanghai as another example, based on Zhang. et al.’s work [9] with averaged incubation period 5.2 days, we estimate that the average number of patients’ incubation period larger than 14-day centralized quarantine should be around only one among 2124 imported cases as of 4 September 2021 [33]. However, more than five onset COVID-19 cases after 14-day centralized quarantine have been reported in the last 1.5 years in Shanghai. In addition, other cities, such as Guangzhou, Beijing, and Chengdu, also reported several import cases that onset after 14-day centralized quarantine [34]. Besides, there were increasingly number of imported cases in the winter period who were diagnosed with COVID-19 after lifting of the 14-day centralized quarantine in China [34]. Those facts suggested the possibility of a prolonged incubation period.

There have been evidence that longer incubation period of SARS-CoV-2 corresponded to less severe symptoms [35‐37], and the young age of host might also have influences on the longer incubation period [38]. A majority of patients in this study were young, which might present with mild symptoms, a fact that may lead to the longer incubation period. Besides, increased virus mutations may have contributed to the longer incubation period in our patients. The mutation of virus, such as higher number of protein-coding genes or GC content (namely, contents of Guanine (G) and Cytosine (C) in viral RNA), corresponds to a longer incubation time [35]. Higher GC content leads to stable secondary structures in the virus RNA, which means that the ribosome needs to disrupt higher kinetic barriers during translation and needs longer translation time [39]. The longer cumulative translation time extends the replication cycle of virus and finally results the longer incubation period [35].

A longer incubation period indicates the need for a longer centralized quarantine period to identify infected individuals and control the outbreak. The large number of covert cases and the high transmissibility of SARS-CoV-2 are two key features of COVID-19 outbreaks [40], and a longer incubation period exacerbates these features. A longer incubation period makes it more difficult to identify infected individuals, and a longer infectious period may lead to infection of more people before diagnosis or quarantine. The incubation period was much longer than serial interval, indicating that mass transmission might occur before symptom onset. In rural areas, routine surveillance and prevention procedures are more difficult [41, 42], and this may lead to failure in the timely detection of new cases. The first COVID-19 case in the Shijiazhuang outbreak was reported on 2 January 2021; however, 196 (22.6%) infected individuals were identified by close contact tracking by 6 January. This indicates that the virus had already spread in the population before the first case was identified, presumably because of the long incubation period.

Due to the timely response in Shijiazhuang, including the rapid implementation of quarantine and lockdown, most cases of COVID-19 were in Gaocheng District and there were no cases related to this outbreak in other provinces of China. Moreover, because of the massive use of city-wide nucleic acid testing, many infected people were found before symptom onset and were isolated before January 15. These interventions had a significant impact on controlling the transmission and spread of the virus.

The results of our SEIR^+q model showed that quarantine measures, which included close contact tracking, centralized isolation, home quarantine, and distant centralized quarantine, had a more significant effect on the outbreak than massive nucleic acid testing. Our model also indicated that if the comprehensive quarantine intensity was reduced by 10%, then at least 20% or more nucleic acid tests were needed to minimize the cumulative number of cases. If so, this would lead to additional expenditures of $18 million, because nucleic acid testing costs $3 per sample [43]. Although nucleic acid testing played a significant role in identification of cases, especially asymptomatic and presymptomatic cases in the community, the possibility of false-negatives [44] and the delay in receiving the results [45] are two main limitations. Due to false-negative results, some asymptomatic or presymptomatic individuals who are infected may not be isolated and might infect others [44]. To avoid the false-negative results, it is usually necessary to test the same person several times, which needs intensive manpower and resources and depends on the economic strength of the country/city. Thus, repeat testing is not applicable in all cities. Therefore, efforts are needed to improve the sensitivity of the nucleic acid testing, and standard sampling and preservation of samples are also needed to reduce the false-negative results. In addition, the effects of nucleic acid testing and massive quarantine rely on cooperation and compliance of the general population, so education of the public on these important interventions is also important.

This study had limitations. First, there was no detailed information (exposure interval, date of symptom onset, etc.) of individuals with asymptomatic infections, nor was there information on clinical severity of symptomatic cases, which might have biased our estimates of the incubation period and serial interval. Besides, incubation period might be overestimated, due to the cases may have multiple contacts with index cases and the exact dates of exposure were unknown. Although it is worthwhile to further study, however, a number of published papers have clarified that the incubation period of asymptomatic patients is longer that of symptomatic patients [31, 46]. Second, in the clusters used to estimate serial interval, infectors of some cases (especially those who develop symptoms 1 to 3 days after the date of symptom onset of the index cases) might be unidentified. Sensitivity analysis with censoring cluster data [9] have been conducted to overcome the limitation. Further, our study has similar limitations as other studies using SEIR model. For example, we do not consider the heterogeneity of the population and the randomness of infection in the SEIR^+q model. Age stratified SEIR model or agent-based model would be preferable to analyse the characteristics of the outbreak and simulate more real-world interventions, which also require individual information in more details.

In conclusion, our study of the COVID-19 outbreak in Shijiazhuang estimated that the mean incubation period was 11.6 days. This suggests it is necessary to prolong the isolation/quarantine period currently so that all potentially infected individuals can be followed. Importantly, a long incubation period may lead to significant transmission of the virus before identification of initial cases. The serial interval (6.6 days) was substantial shorter than the incubation period, suggesting that a significant amount of secondary transmission may occur prior to illness onset. Although resource intensive, the use of comprehensive quarantine measures and mass nucleic acid testing were effective to reduce the disease transmission quickly.