Comparison of the three waves of avian influenza A(H7N9) virus circulation since live poultry markets were permanently closed in the main urban areas in Zhejiang Province, July 2014‐June 2017

Background The sudden increase in the number of human cases infected with avian influenza A(H7N9) virus after September 2016 raised global concern. Objectives To assess the changes in epidemiological characteristics of H7N9 cases since the massive closure of live poultry markets (LPMs) in the main urban areas in Zhejiang province. Methods We used descriptive statistics to compare epidemiological characteristics of the three distinct waves of H7N9 cases in Zhejiang province. The rural or urban cases were defined according to the location where the patients had exposure within 2 weeks before illness onset. Results Between July 2014 and June 2017, 166 H7N9 cases were reported in Zhejiang province, with 45, 34, and 87 cases reported in the third, fourth, and fifth wave, respectively. Across the three waves, most reported cases were from rural areas. A similar percentage of cases in all three waves reported exposure to LPMs, raising poultry at home or around the house, as well as occupational exposure. Compared to the third (80.00%) and fourth wave (70.59%), a significantly larger proportion of cases from the non‐LPMs closure areas were observed in the fifth wave (89.66%) (P = .034). Conclusion Epidemiological characteristics of human cases infected with avian influenza A(H7N9) virus had generally remained unchanged since the massive closure of LPMs in the main urban area of Zhejiang province. The sudden increase in the number of H7N9 cases in the fifth wave was mainly attributed to the excessive cases reported from areas where LPMs were not permanently closed.


| INTRODUCTION
Since the first outbreak of avian influenza A(H7N9) virus in humans in 2013, five seasonal epidemic waves have been documented in Mainland China, resulting in 1557 confirmed cases and 605 deaths as of July 25,2017. 1 Epidemiological data suggested that avian influenza A(H7N9) infections were associated with exposure to live poultry markets (LPMs). [2][3][4][5][6] To control its impact, temporary or permanent closure of live poultry markets has been implemented by many local governments, leading to a declining trend of laboratory-confirmed cases in the first four waves. [7][8][9][10][11] However, an unprecedented outbreak of human infection with avian influenza A(H7N9) virus, identified as the fifth wave, occurred in China from September 2016, and a number of cases reported in the fifth wave exceeded those reported in the previous waves in China. [8][9][10]12 Zhejiang province, located in southeast China, has the largest pro- With its substantial economic impact to the poultry industry, [15][16][17] massive closure of LPMs across the nation has not been implemented in China. Instead, the closure of LPMs in the central urban areas, which has been mandated in Zhejiang province since 2014, is considered as a more feasible option for the local governments in practice. Therefore, analyzing the changes in epidemiology of avian influenza A(H7N9) virus in Zhejiang province may provide useful information for future control and prevention of H7N9 in China. In this study, we compared the epidemiological characteristics of H7N9 cases among the latest three waves since July 2014. The findings from this study will allow us to address whether there were any significant changes in the epidemiology of avian influenza A(H7N9) virus since LPMs were permanently closed in the main urban areas, particularly with regard to the fifth wave.

| Definition of urban and rural cases
In this study, the rural or urban cases were defined according to the location where the patients had exposure within 2 weeks before illness onset. If a case had exposure in both urban and rural areas, we defined its exposure location based on the exposure's risk to onset, following the order of LPMs, occupational exposure, and finally raising poultry at home or around the house. We have chosen a somewhat different definition of case location because we believed that the exposed location can reflect the source of infection more accurately and that location should be where we need to take control measures. However, we also conducted the exposure analyses by defining the rural/urban cases according to the location of residence to make better comparison with most of the previous studies.

| Ethical approval
The National Health and Family Planning Commission ruled that the collection of data for laboratory-confirmed cases of avian influenza A(H7N9) virus infection was part of a continuing public health investigation of an emerging outbreak. The study was therefore exempt from institutional review board assessment.

| Statistical analysis
The means and standard deviations, or medians and ranges were obtained for continuous variables, and characteristic percentages were calculated for categorical variables. Nonparametric tests, including Kruskal-Wallis tests, were used to compare continuous variables such as times between illness onset and other dates of interest. Cochran-Armitage trend test, chi-square, or Fisher's exact tests were used to analyze categorical variables for the three waves. SAS9.2 (SAS Institute, Cary, NC, USA) was used for analyses. The level of significance was set at 0.05.

| Epidemiology
As of June 30, 2017, a total of 166 human infections with influenza A(H7N9) virus were identified during the third, fourth, and fifth wave in Zhejiang Province. In each respective wave, 45, 34, and 87 cases were confirmed. Although there was no statistically significant difference seen among the three waves in death rate of human infections with H7N9 cases, the death rate in the fifth wave was much lower (33.33%) than that in the fourth (38.24%) and third wave (53.33%) ( Table 1).
No statistically significant differences were observed in terms of age and sex across the three waves. Of the 166 cases, 108 (65.06%) were from rural areas. Rural cases occupy a larger proportion than urban cases in each wave (60.00% rural cases in the third wave, 55.88% rural cases in the fourth wave, and 71.26% rural cases in the fifth wave), but did not significantly differ across the three waves (P = .198). The occupation and underlying medical disease did not differ statistically significantly among the three waves. Compared to the third (80.00%) and fourth wave (70.59%), a statistically significantly larger proportion of cases from the non-LPMs closure areas were observed in the fifth wave (89.66%) (P = .034) ( Table 1). The number of cases from non-LPMs closure urban area and non-LPMs closure rural area in the fifth wave increased dramatically from the third and fourth wave ( Figure 1).

| Temporal distribution
In both fourth wave and fifth wave, the first confirmed case was reported in September, while in the third wave, the first confirmed case occurred in November. Across all the three waves, the number of H7N9 cases consistently peaked in January, although it was evident that the number of infected cases from the fifth wave was significantly higher than that in the third and fourth wave during the peak period ( Figure 2).

| Geographic distribution
The fifth wave has impacted the most cities and districts, with 46 counties in total compared to 32 and 24 counties during the third and fourth wave, respectively. In both third wave and fourth wave, the affected areas were concentrated in the north part of Zhejiang Province. However, cases identified in the fifth wave have spread to a much broader areas of the province (Figure 3).  at home or around the house, and occupational exposure across the three waves (Table 2). Among rural cases, the exposure to the sick or dead poultry was highest in the fifth wave and an increasing trend was observed (3.70% for the third, 10.53% for the fourth, and 20.97% for the fifth wave). A similar trend was not observed among urban cases (Table 2). Consistently, among cases from LPMs closure urban area, non-LPMs closure urban area, and non-LPM closure rural area, over half of them had LPMs exposure across the three waves ( Figure 4).

| Exposure pattern
When we conducted analyses by defining the rural/urban cases according to the location of residence, the results were consistent to the above analyses except that an increasing trend of exposure to the sick or dead poultry was not observed among rural cases (P = .098) ( Table 3).

| Key timeline
We

Closure of LPMs was conducted to block the transmission of H7N9
and has been considered as the most effective method for restricting the epidemic to date in China. 3,11,17 Zhejiang Province closed all LPMs in central urban areas concerning the high incidence of H7N9 cases in July 2014, prior to the third wave. Previous studies reported that the number of H7N9 cases in the third wave was significantly lower than that in the first and second wave. 13 The T A B L E 2 Exposure pattern of laboratory-confirmed cases of A(H7N9) virus infection between rural and urban areas in Zhejiang province, July 2014-June 2017 (rural/urban), and the exposure patterns remained relatively unchanged, with the exception in the proportion of cases exposed to sick or dead poultry among rural cases being increasing, raising concerns over the control of sick or dead poultry in rural areas. Further, the significantly shortened period between the illness onset and the start of antiviral treatment indicated a better preparation and enhanced awareness of the H7N9 epidemic among health workers in general.
The sudden increase in the number of H7N9 cases during the fifth wave raises global concern. 8,9,12 The national study showed that the main epidemiological characteristics remained unchanged. 9,10 Another study conducted in Jiangsu province, the most impacted province in China in the fifth wave, also reached similar conclusions. 18 Our study also demonstrated that the demographic characteristics and exposure patterns of H7N9 cases had no significant changes during the most recent waves. However, there was a wide geographic distribution of cases as well as new occurring districts/counties in the fifth epidemic comparing with the third and fourth waves, which was also consistent to the above-mentioned studies. 9,10,18 Moreover, despite the wide geographic coverage of cases in the fifth wave, the larger proportion of them actually occurred in the non-LPMs closure areas. This reassured the effectiveness of the LPMs permanent closure approach in the central urban areas. The potential cause for a sudden increase in H7N9 cases in the fifth outbreak was still unclear, but our study indicated that if we had managed to expand the LPMs closure areas, the total number of H7N9 cases in the fifth wave could have been reduced.
Although the government has permanently closed LPMs in the main urban areas within Zhejiang province, new H7N9 cases were still reported from those places. Due to the long-established Chinese tradition of consuming live poultry, we inferred that illegal live poultry trading in these areas might have existed, as suggested by other studies as well. 16,19 In addition, massive closures of large LPMs were only implemented in central urban areas, which means many LPMs were still operating in other urban/suburban areas, and residents living in central urban areas could still access those operating LPMs through transportation and be exposed to live poultry, as suggested by our study that over half of cases from non-LPMs closure urban area had LPMs exposure across the three waves. These findings suggest that in the urban areas where LPMs were required to close, additional efforts However, our study showed that although rural cases still account for majority of the total count, there is not enough evidence to suggest an increasing trend during the latest three waves. Recently, a national study showed that more cases shifted from urban locations to semi-urban and rural areas. 10 However,as the definition for rural and urban cases differed in our study, we were not able to make a direct comparison. Nevertheless, previous studies consistently highlighted that strict control measures should be taken in rural areas as well. 10  Secondly, the limited sample size of H7N9 cases could have restricted our statistical power to detect true differences in epidemiological characteristics across the three waves, especially when subgroup analyses were conducted. Thirdly, A(H7N9) virus infection among the sick or dead poultry was not studied. Therefore, although we found an increasing trend of human cases exposed to sick or dead poultry in rural areas from the analysis, it was not possible to establish a direct association between human cases and H7N9 infections in poultry. Additionally, to better reflect the sources of infection and follow the geographic changes in the H7N9 epidemic more closely, unlike many of the other studies, 5,9,13,20 we defined the urban or rural cases by subjects' exposure addresses instead of their current residential addresses. Some of the results may need to be interpreted cautiously, as the exposure addresses were still not necessarily the actual sources of infection, with potential impacts on the calculation of the proportions of the rural and urban cases.
Finally, the geography of LPM closure or often blurred distinction between rural and urban areas could have resulted in additional bias that is not fully addressed in the current analyses.

| CONCLUSIONS
Few epidemiological characteristics of the H7N9 had changed since the closure of the LPMs in the main urban areas in Zhejiang province. The sudden increase in the number of H7N9 cases in the fifth wave was mainly attributed to the excessive cases reported from areas where LPMs were not permanently closed. The sizable number of rural cases and the increasing proportion of rural cases exposed to the sick or dead poultry across the three waves suggested the importance of strengthened surveillance in the rural area.