The results of the present study revealed that the Q
1 and Q
2 of temperatures for Wuhan and Hong Kong (Figs. 4a and a’) were essentially same as for Japan. In our preceding study, bimodal cycles of reported chickenpox incidences that were clearly observed in areas of northern Japan such as Hokkaido (latitude 43°N) disappeared at lower latitudes, and unimodal cycles appeared in Okinawa Prefecture, the most southern prefecture in Japan (latitude 26°N). This transition of patterns of reported chickenpox incidences in Japan was considered to be temperature-dependent. Thus, it is reasonable that temporal patterns of chickenpox incidences observed for Wuhan and Hong Kong (Fig. 2) were dominated by temperature as well.
The trends of Q
1 and Q
2 of temperature for Wuhan and Hong Kong (Fig. 4a and a’) support the findings reported by Shoji et al. , who showed that the reported cases of chickenpox increased at 5 °C–20 °C (i.e., the temperature range at which the chickenpox virus is activated) and decreased at temperatures lower than 5 °C and higher than 20 °C. In Wuhan, where the temperature falls below 5 °C in winter and exceeds 20 °C in summer, the occurrence of epidemics is bimodal (Fig. 2a). In contrast, the occurrence of epidemics is expected to be unimodal in Hong Kong, where the temperature rarely falls below 5 °C in winter; however, it was actually bimodal (Fig. 2b). This bimodal cycle of chickenpox epidemics in Hong Kong may be related to the fact that the values of Q
2 for Hong Kong are larger than those for Japanese prefectures with the same mean temperature (Fig. 4a’). One study revealed a spring peak in a bimodal pattern of chickenpox cases related to spring vacation . With respect to Wuhan and Hong Kong, school children do not have spring vacation; thus, it is unlikely that the occurrences of bimodal reported chickenpox incidences in Wuhan and Hong Kong are related to the school calendar. Rather, the bimodal pattern likely results from the mean temperature, as shown in Fig. 4a’.
The increase in the magnitude of Q
2 for Wuhan and Hong Kong (Fig. 4a’) may be explained in terms of (i) the age distribution of reported chickenpox cases, as well as (ii) the effect of rainfall.
(i) Age distribution of reported chickenpox cases: For Hong Kong, more than 90% of the notifications were regarding children aged <18 years, with 29.1% of notified pediatric cases of chickenpox receiving treatment at public or private hospitals . For Wuhan and Hong Kong, the proportion of reported cases among school children was high relative to Japan. Specifically, the 5–19 year age group accounted for 60%–70% of all cases from 2007 to 2015 for Wuhan and the 6–17 year age group accounted for 52.4% of all cases for Hong Kong , while pre-school children (0–4 years) accounted for 78% of the cases reported during 2009–2011 for Japan . For Wuhan and Hong Kong, it can be assumed that school children tend to stay indoors during the day, which in turn increased domestic transmission. Given the limited indoor space in schools in Wuhan and Hong Kong, school children may have a greater opportunity for close contact, facilitating chickenpox transmission. In addition, the high population densities of Wuhan (1248 persons/km2) and Hong Kong (6622 persons/km2) may make it easier for chickenpox to spread from one person to another.
(ii) Effect of rainfall: It has been reported that rainfall was positively associated with chickenpox notifications in Hong Kong . Indeed, the annual mean summation of rainfall in Hong Kong (2364 mm) is larger than that of Wuhan (1798 mm) and 43 of the 47 prefectures in Japan (Fig. 4c’). During the hot and wet season with high temperature and heavy rain, children tend to stay indoors with air-conditioning, which in turn increases domestic transmission. Given the limited indoor space in Hong Kong and Wuhan, children may have a greater opportunity for close contacts with one another, facilitating chickenpox transmission. However, the exact mechanism for the association between rainfall and chickenpox notifications remains unclear.
Critselis et al. examined the influence of meteorological conditions on chickenpox in Greece; before introduction of the vaccine, the authors found that the occurrence of hospitalized chickenpox cases was positively associated with wind speeds of 2.7–3.5 m/s . We confirmed that for Wuhan, the Q
1 and Q
2 values were, respectively, 0.13 and 0.63 with an annual mean wind speed of 1.8 m/s; for Japan’s 47 prefectures, the mean wind speed appeared to display a randomly scattered pattern (Additional file 1).
As with Japan, the Q
1 and Q
2 values for Wuhan and Hong Kong are evidently dominated by temperature ; the meteorological data appear in Table 1. In Table 1c, there is clearly large variance in the daily rainfall data for Wuhan and Hong Kong. The amount of rainfall depends on the amount of water vapor in the atmosphere, which influences relative humidity . The variation in the relative humidity for Wuhan and Hong Kong (Table 1b) was relatively small compared with that for rainfall (Table 1
c). This finding is the result of relative humidity being constrained by the amount of saturated water vapor, which is dependent on air temperature . Thus, it is reasonable to infer that the unimodal and bimodal cycles observed in temporal variations of the reported chickenpox incidence were dominated by temperature.
There was a large difference between vaccine coverage in Wuhan (10%) and Hong Kong (57.6%); however, the temporal patterns of the incidence data for the two areas indicated approximately the same pattern, i.e., the bimodal cycle (Fig. 2
a, b). Thus, the vaccination coverage rate may not have affected the temporal patterns of the incidence data during the period of the present study (2008–15). In a previous investigation, we confirmed that the bimodal cycle of chickenpox epidemics observed in Japan was independent of the influence of the vaccination coverage rate (20%–40%) . If the vaccination coverage rate exceeded 80% in Wuhan, Hong Kong, and Japan, the seasonal peak superposed on a 1-year cycle would diminish; that phenomenon has been observed in the United States .
It is possible that the epidemiology of chickenpox in subtropical Wuhan and Hong Kong differs from that in temperate Japan. In a previous study, we found that preschool children (aged ≤4 years) in Japan accounted for over 70% of all chickenpox cases in 2000–11 . In Wuhan, chickenpox typically occurs at a later age, with many cases among schoolchildren and adolescents (5–19 years). The relatively low number of reported chickenpox cases among preschool children (≤4 years) in Wuhan may be the result of the reduced chickenpox virus transmission reported for tropical areas . Garnett et al.  proposed that in tropical regions, the transmission potential of the chickenpox virus could be adversely affected by a combination of high ambient temperatures and humidity. For example, outbreaks of chickenpox in India appear to be more common in the cooler than in the warmer months of the year . By contrast, in equatorial regions, such as Singapore, the incidence does not seem to vary according to the time of year .
Recently, Rice  interpreted the seasonality of the reported incidence of chickenpox in tropical regions with respect to levels of ultraviolet radiation and air pollution. Accordingly, the effect of meteorological factors on chickenpox incidence may differ from one country to another in different climate regions. To determine the underlying causes of chickenpox virus transmission in each climate region and effectively utilize the obtained results for health services, it is necessary to conduct a systematic study; the investigation should quantify the impact of meteorological factors on the chickenpox incidence in various countries in each climate region. Toward that end, the susceptible-exposed-infected-recovered (SEIR) model, which is a well-known mathematical model of infectious disease epidemics, may be appropriate; it has been shown to be effective with measles . Thus, we would expect that a theoretical procedure, such as the SEIR model, could contribute to future investigations into the meteorological factors related to chickenpox transmission .
It is important to assess the sensitivity and representativeness of incidence data, as was performed by Souty et al.  for chickenpox in France. However, the CISDCP in Wuhan and CHP in Hong Kong have not announced the relevant information (such as age, location, and number of reported severe cases of chickenpox) that would allow us to determine the sensitivity and representativeness of the data used in the present study. There should be increased awareness of the information needed to determine the sensitivity and representativeness of the data in the surveillance system; that would allow a more accurate estimate of the burden of chickenpox and help prevent it in Wuhan and Hong Kong.
It should be noted that this study was limited in that, for Hong Kong, we used monthly data for chickenpox, while we used weekly data for Wuhan, because monthly measures are the minimum unit of measurement released by the Hong Kong Observatory website. We investigated the Q1 and Q2 values for the monthly data for Wuhan, by converting the weekly data for Wuhan into monthly data to enable comparisons with the results shown in Fig. 4. Using the monthly data, we confirmed that the Q1 and Q2 values (0.14 and 0.68, respectively) were essentially consistent with the weekly data (0.13 and 0.65, respectively). Thus, we concluded that the effect on the results of using monthly data compared with weekly data for Wuhan was negligible (Additional file 2). Further studies using weekly data for Hong Kong should be conducted to address the issue of the relationship between climate patterns and chickenpox epidemiology.