Background
Neither specific treatment nor vaccination is available to norovirus illness. Despite a highly contagious virus that causes gastroenteritis worldwide, the monitoring of NoV is elusive. In the United States, NoV was responsible for annual incidence of 21 million [
1], 71,000 hospitalizations [
2] and 800 deaths [
1,
3]. In developing countries, NoV outbreak reports are limited and NoV was found to be responsible for diarrheal disease in The Gambia [
4]. The paradoxical higher prevalence of NoV in controls than in cases [
4] was suggested as a result of reinfection by mathematical modeling [
5]. Although the capacity of norovirus nucleic acid detection is becoming available in China, the understandings towards this pathogen remain accumulating, more formal reports of NoV outbreaks are needed [
6‐
8]. Waterborne NoV outbreaks were reported worldwide and some of the reported outbreaks since 2000 are summarized in Additional file
1: Table S1. Most of the outbreaks were related to potable water contamination. Dubbed winter vomiting bug, waterborne NoV outbreaks indeed appeared to occur all year round.
Even for developed countries, mathematical modeling studies of NoV are limited. Neither estimates of the basic reproduction number
R0 nor other relevant parameters were estimated for developing countries [
5]. An estimate of
R0 = 3.74 was obtained by fitting a compartmental transmission model to a NoV outbreak data involving 39 individuals in Belgium [
9].
An uncontained outbreak could have severe economic consequences [
10]. Ward closure to control nosocomial NoV outbreak was also not advocated, concerning bed-days and revenue loss [
11‐
13]. Campus NoV outbreak was also reported in California, Michigan and Wisconsin in 2008, widespread infection and rapid transmission prompted closure of the Michigan campus [
14]. School closure is a common method to contain NoV transmission in China. However, the effectiveness is unknown. Furthermore, despite China's improvement in water, sanitation and hygiene, more than 320 million people still lacked access to sanitized water [
15]. The vulnerability is to be explored.
Although the factors of incubation period, latent period, infectious period, the proportion of latent infection for NoV were investigated [
16‐
25], the corresponding factors of NoV transmission in the community, schools and via water and the contribution of asymptomatic individuals to the transmission have not been elucidated. In China, there is a lack of quantitative assessment of the effectiveness of interventions such as isolation, water disinfection and school closure. The aforementioned factors and effectiveness of countermeasures are difficult to characterize by conventional epidemiological methods. Mathematical modeling and simulation can complement the analysis [
26,
27]. Here we report two outbreaks in a China metropolis, Changsha City, and compared the outbreak reports with those of the worldwide. Basing on the natural history of diarrhea due to NoV infection, we built an ordinary differential equation (ODE) model to characterize the NoV interpersonal and waterborne transmission dynamics and the effectiveness of intervention in schools and communities.
Methods
The study and investigation were in response to an acute public health emergency event, and no personal identifiable information was include. Ten to fifty percent of the total population at risk were used for space-time scanning by SaTScan™ v9.4.2 (July 2015, Boston and Information Management Services Inc., Calverton Maryland). The significance of mode of transmission was estimated by permutation tests on the basis of Monte Carlo simulations [
28]. Random walk was used to sample the probability distribution of interpersonal transmission [
29]. Two states “interpersonal” and “non-interpersonal” were modeled. In the first incident, random walk modeling was used to assess the proportion between interpersonal and waterborne transmissions. In the second incident, the visit frequency to the potential source of infection was estimated.
A Susceptible-Exposed-Infectious/asymptomatic-Removed-Water (SEIARW) model, similar to the one modeling shigellosis outbreak [
26], was used to characterize NoV transmission epidemics. Berkeley Madonna 8.3.18 (developed by Robert Macey and George Oster of the University of California at Berkeley. Copyright ©1993-2001 Robert I. Macey & George F. Oster) was employed for model simulation. The Runge-Kutta method of order 4 with the tolerance set at 0.001 was used to perform curve fitting on prevalence [
30]. Sensitivity analysis was tested by varying parameters into 1000 values ranging from the minimum to maximum of the reported values in literature (Additional file
2: Table S2).
Discussions
To our knowledge, this is the first study to characterize the transmission dynamics and control of a waterborne NoV outbreak. Our data also enhances our understanding towards the NoV outbreak situation in China.
Inefficient transmission of asymptomatic individuals reconciles the finding of the paradoxical higher prevalence of NoV in controls than in cases of the recent Global Enteric Multicenter Study [
4]. Our data also enrich the dataset of waterborne NoV outbreaks in China. School closure was not predicted to be an effective control. Identification of target affected population is as important as locating infection source and mode of transmission in order to take prompt countermeasures. In resource-limited setting, a balance between hazards and benefits calls for action to enforce sanitation and water disinfection; and promote personal hygiene to proactively prevent and contain NoV and other waterborne disease outbreaks.
Contaminated water was 14- to 500-fold more infectious than transmission associated with person-to-person contact. Usually water is far less effective in the transmission of NoV. Despite a much lower transmissibility by water (
b
W
=0.01) than by the infected in a local community (
b = 0.82) or a school (
b = 1.63), the spread of NoV by water was still the most rapid (
R0 = 4.91), versus
R0 = 1.94 in community and
R0 = 3.44 in school. A previous mathematical modeling study of NoV suggested that asymptomatic prevalence could be varying [
5], however, our results predicted that contribution of asymptomatic individuals to the spread of NoV was minimal, with low interpersonal transmissibility coefficients
k = 3.73 × 10
-9 and
k = 4.15 × 10
-11 under the school and the community settings respectively. Their shedding capability was also only about 18 % of the symptomatic individuals. Latent infection devoid of gastroenteritis symptoms such as vomiting and/or watery diarrhea may have significantly lowered the transmissibility. Even if they do shed NoV, the titer is lower [
23‐
25]. Therefore, our results suggest a possible alternative explanation for the paradoxical higher prevalence of NoV in controls than in cases of the recent Global Enteric Multicenter Study [
4].
Water-to-person transmission was predicted to be the fastest, followed by interpersonal transmission in school, and then interpersonal transmission in the community. Wide coverage of water supply, daily contact and low infectious dose [
31] render water an effective vector for NoV infection. In turn, population density and contact frequency are both higher in schools than in community, which accounts for higher interpersonal transmissibility in the former.
In the NoV outbreak due to drinking water source contamination, isolation alone did not contain the outbreak; TAR was predicted to be about the same, but with a longer DO. Similarly, water disinfection alone could only reduce TAR and DO to about 23 % and 58 % respectively. Only a combined measure of both could bring down the TAR and DO from 70.00 % (no intervention) to 24.93 % and 154 (no intervention) to 23 days.
The actual NoV outbreak TARs reported in China was significantly lower than worldwide. Two reasons may be responsible for this finding. Firstly, the number of the affected population in China was reported significantly larger than that worldwide (3915 versus 1845,
P < 0.05) while the number of infected population was not (294 versus 158,
P = 0.96). If reported target populations were non-specific, the reported TARs were indeed deflated. Although the Chinese population is large and reports in India and Russia were unavailable, the number of infected was not correspondingly high. Secondly, many worldwide reports of large TAR were specific incidences in a focused area such as excursion or leisure activities in resorts, hotels or small towns. High cost of viral monitoring outbreaks and control [
32], insufficient public perception and awareness of NoV, health-seeking motivation and practices of gastroenteritis, immature information-sharing, data collection and dissemination may hinder reports from local schools, departments or other entities. Experience accumulation, regular health structures and expert epidemiologic support at local level are the first steps towards a well-designed sentinel system for health promotion and disease transmission blockage. In the second incident of NoV outbreak, no positive results were obtained from water samples and those collected from food processing staff. Not a single one of the staff indeed manifested symptoms of NoV infection. The relatively high visit frequency obtained by random walk analysis suggests that toilet might act as the hotbed for transmission. In certain areas of China, hand washing after toilet visits before eating or preparing food is not deemed necessary. Moreover, vomitus and stool in the toilet may not be discarded properly due to the prevalence of dry toilets [
33]. In future, flushing and keeping the surrounding area of toilets clean should be advocated. Personal hygiene should also be emphasized.
A limitation of this study was that we analyzed data of only two outbreaks, and that five parameters had to be taken from the literature. Nevertheless, our mathematical models reflect well the actual epidemic. Moreover, sensitivity analysis confirms that our model was relatively insensitive to the change in parameter. Finally, we also characterized the transmissibility and viral shedding coefficient during the asymptomatic phase of the disease. Further field test and pathological investigation are warranted for infection control of NoV.