Advertisement

Indoor relative humidity shapes influenza seasonality in temperate and subtropical climates in China

Open AccessPublished:November 21, 2022DOI:https://doi.org/10.1016/j.ijid.2022.11.023

      Highlights

      • Indoor relative humidity (RH) displays a seasonal pattern.
      • Outdoor RH remains high around the year in subtropical regions.
      • Absolute humidity has stronger an indoor-to-outdoor correlation than temperature and RH.
      • Indoor RH is a better indicator of influenza seasonality than outdoor RH.

      Abstract

      Objectives

      The aim of this study was to explore whether indoor or outdoor relative humidity (RH) modulates the influenza epidemic transmission in temperate and subtropical climates.

      Methods

      In this study, the daily temperature and RH in 1558 households from March 2017 to January 2019 in five cities across both temperate and subtropical regions in China were collected. City-level outdoor temperature and RH from 2013 to 2019 were collected from the weather stations. We first estimated the effective reproduction number (Rt) of influenza and then used time-series analyses to explore the relationship between indoor/outdoor RH/absolute humidity and the Rt of influenza. Furthermore, we expanded the measured 1-year indoor temperature and the RH data into 5 years and used the same method to examine the relationship between indoor/outdoor RH and the Rt of influenza.

      Results

      Indoor RH displayed a seasonal pattern, with highs during the summer months and lows during the winter months, whereas outdoor RH fluctuated with no consistent pattern in subtropical regions. The Rt of influenza followed a U-shaped relationship with indoor RH in both temperate and subtropical regions, whereas a U-shaped relationship was not observed between outdoor RH and Rt. In addition, indoor RH may be a better indicator for Rt of influenza than indoor absolute humidity.

      Conclusion

      The findings indicated that indoor RH may be the driver of influenza seasonality in both temperate and subtropical locations in China.

      Graphical Abstract

      Keywords

      1. Introduction

      In modern society, individuals spend approximately 80-90% of their time in enclosed buildings (
      • Graham SE
      • McCurdy T.
      Developing meaningful cohorts for human exposure models.
      ;
      • Klepeis NE
      • Nelson WC
      • Ott WR
      • Robinson JP
      • Tsang AM
      • Switzer P
      • et al.
      The National Human Activity Pattern Survey (NHAPS): a resource for assessing exposure to environmental pollutants.
      ). The transmission of respiratory viruses is much more common indoors than outdoors (
      • Bulfone TC
      • Malekinejad M
      • Rutherford GW
      • Razani N.
      Outdoor transmission of SARS-CoV-2 and other respiratory viruses: a systematic review.
      ). Owing to the use of air conditioning and heating systems, indoor and outdoor temperatures and relative humidity (RH) differ (
      • Pan J
      • Tang J
      • Caniza M
      • Heraud JM
      • Koay E
      • Lee HK
      • et al.
      Correlating indoor and outdoor temperature and humidity in a sample of buildings in tropical climates.
      ). Thus, it is reasonable to assume that the indoor ambient environment may significantly modulate influenza transmission compared with the outdoor ambient environment.
      Laboratory experiments examining influenza transmission in guinea pig and mouse hosts suggest that the spread of the influenza virus is dependent on both ambient RH and temperature (
      • Lowen AC
      • Mubareka S
      • Steel J
      • Palese P.
      Influenza virus transmission is dependent on relative humidity and temperature.
      ;
      • Schulman JL
      • Kilbourne ED.
      Experimental transmission of influenza virus infection in Mice. II. Some factors affecting the incidence of transmitted infection.
      ). Studies have observed a bimodal relationship between RH and influenza spread; that is, influenza spread is greatest at low RH (less than 30%), minimal at middle RH (approximately 50%), and moderate at high RH (greater than 80%). At the population level, numerous modeling studies have explored the associations between meteorological conditions and influenza transmission (
      • Ali ST
      • Cowling BJ
      • Wong JY
      • Chen D
      • Shan S
      • Lau E
      • et al.
      Influenza seasonality and its environmental driving factors in mainland China and Hong Kong.
      ;
      • Davey ML
      • Reid D.
      Relationship of air temperature to outbreaks of influenza.
      ;
      • Hemmes JH
      • Winkler KC
      • Kool SM.
      Virus survival as a seasonal factor in influenza and polimyelitis.
      ;
      • Hope-Simpson RE.
      The role of season in the epidemiology of influenza.
      ;
      • Moura FE
      • Perdigão AC
      • Siqueira MM.
      Seasonality of influenza in the tropics: a distinct pattern in northeastern Brazil.
      ;
      • Shaman J
      • Pitzer VE
      • Viboud C
      • Grenfell BT
      • Lipsitch M.
      Absolute humidity and the seasonal onset of influenza in the continental United States.
      ), and several climatic factors have long been speculated to be causally linked to influenza activity, including cold temperature (
      • Davey ML
      • Reid D.
      Relationship of air temperature to outbreaks of influenza.
      ), low humidity (
      • Hemmes JH
      • Winkler KC
      • Kool SM.
      Virus survival as a seasonal factor in influenza and polimyelitis.
      ), minimal solar radiation (
      • Hope-Simpson RE.
      The role of season in the epidemiology of influenza.
      ), and the rainy season (
      • Moura FE
      • Perdigão AC
      • Siqueira MM.
      Seasonality of influenza in the tropics: a distinct pattern in northeastern Brazil.
      ). Among these, humidity is of great research interest. Recently, the most accepted hypothesis explaining influenza seasonality, that is, the association between RH/absolute humidity (AH) and influenza transmission, has attempted to explain influenza transmission only in temperate regions. In temperate regions, influenza activity always peaks in winter, which could be explained by the low outdoor humidity. However, less attention has been given to influenza transmission in tropical and subtropical locations. This is mainly due to the more complicated seasonality of influenza in tropical and subtropical locations. In tropical and subtropical locations, there is either significant year-round influenza activity or two distinct influenza seasons (
      • Tamerius J
      • Nelson MI
      • Zhou SZ
      • Viboud C
      • Miller MA
      • Alonso WJ.
      Global influenza seasonality: reconciling patterns across temperate and tropical regions.
      ). In winter, outdoor humidity remains high; therefore, outdoor RH was not found to be associated with influenza outbreaks in the human population (
      • Schulman JL
      • Kilbourne ED.
      Experimental transmission of influenza virus infection in Mice. II. Some factors affecting the incidence of transmitted infection.
      ;
      • Tamerius JD
      • Shaman J
      • Alonso WJ
      • Bloom-Feshbach K
      • Uejio CK
      • Comrie A
      • et al.
      Environmental predictors of seasonal influenza epidemics across temperate and tropical climates.
      ). However, indoor RH, which is generally low in heated buildings during winter, was not evaluated. In addition, in an analysis of human population-level influenza surveillance data of the United States,
      • Shaman J
      • Pitzer VE
      • Viboud C
      • Grenfell BT
      • Lipsitch M.
      Absolute humidity and the seasonal onset of influenza in the continental United States.
      showed that AH controls influenza transmission more significantly than RH. However, temperature and AH are strongly correlated; therefore, the confounding effect of temperature cannot be excluded (
      • Tamerius J
      • Nelson MI
      • Zhou SZ
      • Viboud C
      • Miller MA
      • Alonso WJ.
      Global influenza seasonality: reconciling patterns across temperate and tropical regions.
      ).
      Northern China has a temperate climate, where influenza epidemics are concentrated in the winter-spring months, whereas Southern China, which has subtropical and tropical climates, has a semiannual cyclic pattern, with clear peaks in both summer and winter (
      • Shu YL
      • Fang LQ
      • de Vlas SJ
      • Gao Y
      • Richardus JH
      • Cao WC.
      Dual seasonal patterns for influenza.
      ). The influenza data of China are suitable for studying the impact of climate on influenza epidemics in both temperate and subtropical regions within similar socioeconomic settings. In this study, the relationships of influenza surveillance data with indoor/outdoor temperature and RH in five cities in China, two in temperate regions and three in subtropical locations, were analyzed by using regression models. The objective of this study was to determine whether influenza activity was better correlated with indoor RH or with outdoor RH.

      2. Methods

      2.1 Indoor and outdoor temperature and humidity measurements

      Beijing, Tianjin, Shanghai, Chongqing, and Hong Kong were chosen as the study regions because city-level influenza surveillance data were available for only these municipalities in mainland China. Beijing and Tianjin are located in temperate regions, whereas Shanghai, Chongqing, and Hong Kong are located in subtropical regions (Figure 1). All five cities have relatively high urbanization levels, with the highest at 100% in Hong Kong and the lowest at 64.1% in Chongqing (Table 1). Daily indoor temperature (°C) and RH (%) in four typical cities in mainland China were measured in 1577 households between March 2017 and July 2018. These records were measured using low-cost sensors, as described in our previous studies (
      • Ben Y
      • Ma F
      • Wang H
      • Hassan MA
      • Yevheniia R
      • Fan W
      • et al.
      A spatio-temporally weighted hybrid model to improve estimates of personal PM2.5 exposure: incorporating big data from multiple data sources.
      ;
      • Dong Z
      • Wang H
      • Yin P
      • Wang L
      • Chen R
      • Fan W
      • et al.
      Time-weighted average of fine particulate matter exposure and cause-specific mortality in China: a nationwide analysis.
      ). In total, 927 sensors were placed in Beijing, 567 in Shanghai, 50 in Tianjin, and 33 in Chongqing (Table 1). The daily indoor temperatures and RH in Hong Kong were obtained from the study by
      • Pan J
      • Tang J
      • Caniza M
      • Heraud JM
      • Koay E
      • Lee HK
      • et al.
      Correlating indoor and outdoor temperature and humidity in a sample of buildings in tropical climates.
      , in which they measured the indoor temperature and RH in a household from November 2017 to January 2019. The measurements in Beijing, Shanghai, Tianjin, Chongqing, and Hong Kong covered 494, 461, 379, 343, and 435 days, respectively.
      Figure 1
      Figure 1Location of study regions of Beijing, Tianjin, Chongqing, Shanghai, and Hong Kong in China.
      Table 1Background characteristics of the five cities involved in this study.
      CityBeijingTianjinShanghaiChongqingHong Kong
      No. of loggers
      Number of loggers used to monitor indoor temperature and relative humidity.
      92750567335
      Five loggers installed in one household.
      Urbanization level in 201786.5%82.9%87.7%64.1%100%
      Population size in 2017 (Million)21.715.624.230.87.5
      Population size in 2018.
      Latitude39.939.231.329.622.2
      Longitude116.4117.2121.5106.6114.1
      Climatic zoneTemperateTemperateSubtropicalSubtropicalSubtropical
      a Number of loggers used to monitor indoor temperature and relative humidity.
      b Five loggers installed in one household.
      c Population size in 2018.
      The city-level daily outdoor temperature and RH from 2013 to 2019 were obtained from the China Meteorological Administration (http://data.cma.cn, accessed on May 2, 2019) and the Hong Kong Observatory. City-level daily meteorological indicators were calculated as averages of the data from all weather stations in each city. AH values (g/m3) were computed from the temperature (°C) and RH (%) according to Eq. (1) (
      • Pan J
      • Tang J
      • Caniza M
      • Heraud JM
      • Koay E
      • Lee HK
      • et al.
      Correlating indoor and outdoor temperature and humidity in a sample of buildings in tropical climates.
      ).
      AH=1322.7×exp(17.625×TT+243.04)×(RH100)T+273.15
      (1)


      2.2 Influenza surveillance date

      Weekly reports of influenza cases from 2013 to 2019 were obtained from the Chinese Centre for Disease Control and Prevention and the Centre for Health Protection of the Hong Kong Special Administrative Region based on the influenza sentinel surveillance network. The weekly incidence rate was then interpolated to the daily incidence rate using splines (
      • Ali ST
      • Cowling BJ
      • Lau EHY
      • Fang VJ
      • Leung GM.
      Mitigation of influenza B epidemic with school closures, Hong Kong, 2018.
      ). Then, the data were converted to rates per 100,000 individuals, and all cities used the same indicator for describing influenza activities (see supplementary materials for more details).

      2.3 Estimation of the daily reproductive number (Rt)

      The effective reproductive number, Rt, which represents the mean number of secondary infections caused by a primary infection at time t (
      • Lei H
      • Xu M
      • Wang X
      • Xie Y
      • Du X
      • Chen T
      • et al.
      Nonpharmaceutical interventions used to control COVID-19 reduced seasonal influenza transmission in China.
      ), was used to measure influenza transmission in this study. The daily effective reproductive number at various time points was estimated using the instantaneous reproduction number method as implemented in the R package ‘EpiEstim’ (
      • Cori A
      • Ferguson NM
      • Fraser C
      • Cauchemez S.
      A new framework and software to estimate time-varying reproduction numbers during epidemics.
      ). We assumed that the serial interval for influenza followed a gamma distribution, with a mean of 2.6 days and an SD of 1.5 days (
      • Wallinga J
      • Lipsitch M.
      How generation intervals shape the relationship between growth rates and reproductive numbers.
      ).

      2.4 Time-series analysis of meteorological factors and reproductive number

      The distributed lag nonlinear models (DLNMs) proposed by
      • Gasparrini A.
      Distributed lag linear and non-linear models in R: the package dlnm.
      is a flexible model that estimates the nonlinearity and distributed lag effects of exposure-response relationships simultaneously, especially the effects of meteorological factors on health. The statistical model used in our study is defined by Eq. (2),
      Log(E(Rt))=α+cb(RH)+ns(Temperature,df)+βDOW+ns(t,df)
      (2)


      where t stands for time, α indicates the intercept, cb() represents the cross-basis matrix of RH and temperature, β is the coefficient of the day of week (DOW), ns() represents cubic spline function, and df is the degrees of freedom. Considering the overdispersion of influenza activity, we used a quasi-Poisson function to construct the model. The cross-basis matrix of RH was included to explore the cumulative and delayed effects of daily indoor and outdoor RH. According to the influenza incubation period studied in the previous literature (
      • Virlogeux V
      • Li M
      • Tsang TK
      • Feng L
      • Fang VJ
      • Jiang H
      • et al.
      Estimating the distribution of the incubation periods of human avian influenza A(H7N9) virus infections.
      ), we selected 7 days as the maximum number of lag days in our final model. The cross-basis was a natural cubic spline for RH, with the knot placed at 50% and the df defined as 3. The df was set between 4 and 7 for the sensitivity analysis, and the results showed that the conclusion was stable (Supplementary Material Figure S6-9). Temperature was included to control its confounding effect, and the df was defined as 3. T referred to seasonality and long-term trends in influenza, which was controlled using a cubic spline function with 2 df per year (
      • Qi L
      • Liu T
      • Gao Y
      • Tian D
      • Tang W
      • Li Q
      • et al.
      Effect of meteorological factors on the activity of influenza in Chongqing, China, 2012–2019.
      ). DOW was incorporated into the final model to control the effect of potential confounding factors. We used 1-10 df per year for time in the sensitivity analysis and found that the shape of the RH-influenza relationship remains stable with the change in the number of df and chose 2 df per year in the final models. When the parameters in the model were fixed, the cumulative effect under a specific delay using relative risk (RR) was calculated. The RR was calculated with corresponding 95% confidence interval (CI) relative to the reference levels, which was defined as the RH value corresponding to the RR minimum value. The distributed lag model in the DLNM package in R 3.63 (The R Project for Statistical Computing, Guangzhou, China) was used in this study.
      Given that indoor temperature and RH data in five cities were measured for about a year, it may not be enough to give statistically significant results. To further verify the stability of the results in the study, we further took the 1-year daily indoor temperature and RH data of five cities as a set of cycles and expanded them into 5 years. After expansion, the research time for Beijing, Tianjin, Shanghai, and Chongqing was from January 2013 to July 2018, and the research time for Hong Kong was from December 2013 to January 2019. After the same data preprocessing, the extended data were subjected to the same statistical analysis.
      To compare the impact of indoor and outdoor AH and RH on influenza transmission, same as Eq. (2), we established DLNMs for AH as well; the equation is as follows:
      Log(E(Rt))=α+cb(AH)+βDOW+ns(t,df)
      (3)


      The temperature was not included in this model because the temperature was highly correlated with AH (
      • Shimmei K
      • Nakamura T
      • Ng CFS
      • Hashizume M
      • Murakami Y
      • Maruyama A
      • et al.
      Association between seasonal influenza and absolute humidity: time-series analysis with daily surveillance data in Japan.
      ).

      2.5 Statistical analysis

      We calculated Pearson correlation coefficients (r) and 95% CI between the daily average of indoor and outdoor meteorological factors and compared the indoor and outdoor conditions using a t-test at a significance level of 0.05. All statistical analyses were carried out in R 3.63 (The R Project for Statistical Computing, Guangzhou, China).
      In this study, the study period was divided into cold and warm/hot days based on the 25th quartile of the combined outdoor temperatures in the five cities, that is, 12°C. Therefore, the cold days in Beijing, Tianjin, Shanghai, Chongqing, and Hong Kong were 173, 150, 130, 87, and 10 days, respectively. The warm/hot days in Beijing, Tianjin, Shanghai, Chongqing, and Hong Kong were 330, 254, 368, 271, and 425 days, respectively.

      3. Results

      3.1 Temperature, humidity, and influenza activities

      Both indoor and outdoor temperatures and AH displayed seasonal patterns in Beijing and Shanghai (Figure 2a, 2c), with highs during the summer months and lows during the winter months. The indoor temperature had a much lower variability than the outdoor temperature, and the indoor temperature was approximately 20-30°C in both Beijing and Shanghai year-around. In contrast, the outdoor RH fluctuated with no consistent pattern, especially in Shanghai, Chongqing, and Hong Kong (Figure 2b, Supplementary Material Figure S1b), where the outdoor RH remained relatively high year-round. However, indoor RH followed a seasonal pattern (Figure 2b, Supplementary Material Figure S1b), with highs during the summer months and lows during the winter months. The means and SDs of the measured indoor daily temperatures and RH in different households in the four cities are presented in Supplementary Material, Figure S2. During the early phase of indoor temperature and RH measurement, only a few sensors worked in each city; therefore, the variability between the different indoor environments was low (Supplementary Material Figure S2). Generally, compared with the variability of indoor RH, the variability of indoor temperature is relatively lower.
      Figure 2
      Figure 2Daily indoor and outdoor meteorological data from 2017 to 2018; (a) temperature (°C), (b) RH (%), and (c) AH (g/m3) in Beijing in the temperate regions, and Shanghai in the subtropical regions. Data in Tianjin, Chongqing and Hong Kong are presented in Supplementary Material Figure S1.
      Abbreviations: RH, relative humidity; AH, absolute humidity.
      Influenza activity in temperate and subtropical regions also displayed different patterns. In the temperate regions, Beijing and Tianjin, influenza epidemics peaked during the winter-spring months. There was a relatively small wave in the summer months, but the peak was much lower than that in the winter-spring months (Figure 3a, Supplementary Material Figure S3a), whereas in the subtropical regions, Shanghai, Chongqing, and Hong Kong, in some years, there were influenza activities in the summer months, and the intensity was similar to that in the winter-spring waves. During the 5-year study period, there were three summer influenza waves in Shanghai, Chongqing, and Hong Kong in 2014, 2015, and 2017, respectively (Figure 3b, Supplementary Material Figure S3b, Figure S3c).
      Figure 3
      Figure 3Influenza activities from 2013 to 2018 in (a) Beijing in the temperate regions, and (b) Shanghai in the subtropical regions. Data in Tianjin, Chongqing and Hong Kong are presented in Supplementary Material Figure S2.

      3.2 Correlation between indoor and outdoor temperatures and humidity

      Generally, except for indoor and outdoor temperatures during cold days in Beijing (P-value = 0.33), Tianjin (P-value = 0.06), and Hong Kong (P-value = 0.61), all other indoor-outdoor correlations were statistically significant (P <0.05) (Supplementary Material Figure S4). AH exhibited the strongest indoor-to-outdoor correlation, and all the Pearson correlation coefficients were >0.8 (Supplementary Material Figure S4). On combining the data from the five cities for the analysis, the study found a significant deviation from linearity in the relationship between indoor and outdoor temperatures, RH, and AH (Figure 4b, c, d). On cold days, the median outdoor temperature was significantly lower than the indoor temperature, and the outdoor temperature distribution was more dispersed. On warm/hot days, the median difference between the indoor and outdoor temperatures was small, and the distribution of outdoor temperatures was more scattered and uniform than that of indoor temperatures (Figure 4a). The indoor temperature on cold days was similar to that on warm/hot days. The goodness of fit between the indoor and outdoor data was better for AH (R2 = 0.80 and 0.83) than with RH (R2 = 0.64 and 0.70) and temperature (R2 = 0.69) (Figure 4b, c, d). In addition, there was a stronger correlation between indoor RH and indoor/outdoor AH on cold days (r = 0.85 and 0.93, R2 = 0.72 and 0.86) than that on warm/hot days (r = 0.76 and 0.67, R2 = 0.58 and 0.44; Figure 4e, f).
      Figure 4
      Figure 4The associations of indoor and outdoor meteorological data. (a) violin plot of the combined indoor and outdoor temperature and RH during cold days and warm/hot days respectively; the associations of indoor and outdoor (b) temperature, (c) RH, (d) AH, during cold and warm/hot days respectively; the associations of indoor RH and indoor/outdoor AH during (e) cold and (f) warm/hot days respectively. Black markers represent the cold days and blue markers represent the warm/hot season.
      Abbreviations: RH, relative humidity; AH, absolute humidity.

      3.3 Correlation between indoor/outdoor RH/AH and influenza transmission

      Supplementary Material Figure S5 presented the relationships between indoor and outdoor RH in different climate regions and Rt of influenza transmission in different lag days. Previous studies have found the lag between humidity and influenza transmission to be approximately 1-2 days (
      • Dai Q
      • Ma W
      • Huang H
      • Xu K
      • Qi X
      • Yu H
      • et al.
      The effect of ambient temperature on the activity of influenza and influenza like illness in Jiangsu Province.
      ). To facilitate comparison, the lag time was all set to 1 day in this study to analyze the overall cumulative RR change of RH/AH and Rt indoors and outdoors. In temperate regions, both indoor and outdoor RH represented a U-shaped relationship with influenza Rt, but the relationship was not statistically significant for both indoor and outdoor RH at low RH (Figure 5a, 5b). In subtropical regions, there was a U-shaped relationship between influenza Rt and indoor RH (Figure 5c), whereas the relationship between influenza Rt and outdoor RH increased monotonically (Figure 5d) and was statistically significant only at high RH. As for AH, no matter whether it is in a temperate region, subtropical region, or temperate and subtropical region, the relationships between indoor/outdoor AH and Rt were also not statistically significant in both low and high AH (Figure 6).
      Figure 5
      Figure 5Association between RH and influenza transmissibility reproduction number, using the measured indoor data. (a, c, e), indoor RH; (b, d, f), outdoor RH; (a, b), in temperate regions; (c, d), in subtropical regions; (e, f), in combined regions; light shaded areas represent the 95% confidence intervals.
      Abbreviation: RH, relative humidity.
      Figure 6
      Figure 6Association between AH and influenza transmissibility reproduction number, using the measured indoor data. (a, c, e), indoor AH; (b, d, f), outdoor AH; (a, b), in temperate regions; (c, d), in subtropical regions; (e, f), in combined regions; light shaded areas represent the 95% confidence intervals.
      Abbreviation: AH, absolute humidity.
      One potential reason for the nonsignificant relationship between indoor or outdoor RH and influenza Rt (Figure 5e and 5f) may be that the data size we used in this study was relatively small, that is, there was only approximate 1-year data. After expanding the indoor data into 5 years, we found that there was statistically significant U-shaped relationship between indoor RH and Rt of influenza (Figure 7a), i.e., influenza transmissibility was greatest at low RH, minimal at middle RH, and moderate at high RH. The findings coincide with the results from laboratory experiments (
      • Lowen AC
      • Mubareka S
      • Steel J
      • Palese P.
      Influenza virus transmission is dependent on relative humidity and temperature.
      ;
      • Schulman JL
      • Kilbourne ED.
      Experimental transmission of influenza virus infection in Mice. II. Some factors affecting the incidence of transmitted infection.
      ). The relationship between outdoor RH and Rt of influenza transmission was still monotonically increasing (Figure 7b), which reflects the stability of the results.
      Figure 7
      Figure 7Association between RH and influenza transmissibility reproduction number; (a), indoor RH; (b), outdoor RH; (a, b), using the expanded 5-year indoor data; light shaded areas represent the 95% confidence intervals.
      Abbreviation: RH, relative humidity.

      4. Discussion

      In this study, according to the time-series analyses of the daily indoor and outdoor temperature and RH in five cities across both temperate and subtropical regions in China, we found that indoor RH had a more obvious seasonal pattern than outdoor RH, with highs during the summer months and lows during the winter months, especially in the subtropical locations. AH exhibited the strongest indoor-to-outdoor correlation, and the correlation between indoor and outdoor temperatures was not even statistically significant in the cold season. These findings are consistent with previous studies (
      • Pan J
      • Tang J
      • Caniza M
      • Heraud JM
      • Koay E
      • Lee HK
      • et al.
      Correlating indoor and outdoor temperature and humidity in a sample of buildings in tropical climates.
      ). This is likely due to the use of heating systems in temperate regions but not in subtropical regions during the cool season and the use of air conditioners in both temperate and subtropical regions during the hot season. Influenza epidemics peaked in the winter-spring months in the temperate regions; however, in subtropical regions, there were two distinct waves, both in summer and winter-spring months, which corresponds to previous studies (
      • Lau EH
      • Cowling BJ
      • Ho LM
      • Leung GM.
      Optimizing use of multistream influenza sentinel surveillance data.
      ;
      • Tamerius J
      • Nelson MI
      • Zhou SZ
      • Viboud C
      • Miller MA
      • Alonso WJ.
      Global influenza seasonality: reconciling patterns across temperate and tropical regions.
      ,
      • Tamerius JD
      • Shaman J
      • Alonso WJ
      • Bloom-Feshbach K
      • Uejio CK
      • Comrie A
      • et al.
      Environmental predictors of seasonal influenza epidemics across temperate and tropical climates.
      ;
      • Tang JW
      • Lai FY
      • Nymadawa P
      • Deng YM
      • Ratnamohan M
      • Petric M
      • et al.
      Comparison of the incidence of influenza in relation to climate factors during 2000–2007 in five countries.
      ). The DLNMs indicated that Rt had a consistent U-shaped relationship with indoor RH in both temperate and subtropical regions, whereas Rt did not have a U-shaped relationship with outdoor RH. However, regardless of the region, the relationship between AH and Rt was not statistically significant. All these findings indicated that indoor RH may be a better indicator of influenza seasonality. The U-shaped form indicates that influenza transmissibility is higher during periods of low or high indoor RH but lower in periods of moderate RH, which is consistent with experimental studies (
      • Lowen AC
      • Steel J
      • Mubareka S
      • Palese P.
      High temperature (30 degrees C) blocks aerosol but not contact transmission of influenza virus.
      ;
      • Lowen A
      • Palese P.
      Transmission of influenza virus in temperate zones is predominantly by aerosol, in the tropics by contact: a hypothesis.
      ;
      • Moriyama M
      • Hugentobler WJ
      • Iwasaki A.
      Seasonality of respiratory viral infections.
      ;
      • Shaman J
      • Pitzer VE
      • Viboud C
      • Grenfell BT
      • Lipsitch M.
      Absolute humidity and the seasonal onset of influenza in the continental United States.
      ). At high RH, evaporation happens slowly, and the solute concentration remains at levels that are harmless to the influenza virus, and at low RH, all water is evaporated quickly, and the solutes crystallize; thus, the vitality of influenza virus is retained. However, at medium RH, evaporation causes the increase in salt concentration and leads to virus inactivation (
      • Yang W
      • Elankumaran S
      • Marr LC.
      Relationship between humidity and influenza A viability in droplets and implications for influenza's seasonality.
      ). A study found that the reason for the U-shaped relationship between the virus survival rate and the RH may be that the impact of antiviral proteins on viral particles reaches the maximum at medium RH because their contact time and overlapping area are the largest when RH is at a moderate level, leading to the lowest viral activity (
      • Kong ZM
      • Sandhu HS
      • Qiu L
      • Wu J
      • Tian WJ
      • Chi XJ
      • et al.
      Virus dynamics and decay in evaporating human saliva droplets on fomites.
      ). Low RH also makes the droplets suspend longer in the air; thus, sedimentation is slower due to more intense evaporation (
      • Marr LC
      • Tang JW
      • Van Mullekom J
      • Lakdawala SS.
      Mechanistic insights into the effect of humidity on airborne influenza virus survival, transmission and incidence.
      ). These may be some of the reasons why RH has a U-shaped relationship with Rt.
      Because influenza transmission in temperate and subtropical regions shows different patterns, some studies have attempted to explain influenza transmission using outdoor AH (
      • Deyle ER
      • Maher MC
      • Hernandez RD
      • Basu S
      • Sugihara G.
      Global environmental drivers of influenza.
      ;
      • McDevitt J
      • Rudnick S
      • First M
      • Spengler J.
      Role of absolute humidity in the inactivation of influenza viruses on stainless steel surfaces at elevated temperatures.
      ;
      • Shaman J
      • Kohn M.
      Absolute humidity modulates influenza survival, transmission, and seasonality.
      ). For example,
      • Shaman J
      • Kohn M.
      Absolute humidity modulates influenza survival, transmission, and seasonality.
      found that AH could explain 50% of influenza transmission, whereas RH could only explain 12%. However, temperature and AH are strongly correlated, which makes it difficult to exclude a confounding effect (
      • Tamerius J
      • Nelson MI
      • Zhou SZ
      • Viboud C
      • Miller MA
      • Alonso WJ.
      Global influenza seasonality: reconciling patterns across temperate and tropical regions.
      ). In addition, from the mechanism of droplet evaporation and the physical and chemical characteristics of droplets, RH is more likely to modulate influenza survival and transmission than AH (
      • Marr LC
      • Tang JW
      • Van Mullekom J
      • Lakdawala SS.
      Mechanistic insights into the effect of humidity on airborne influenza virus survival, transmission and incidence.
      ).
      • Zhao Y
      • Aarnink AJ
      • Dijkman R
      • Fabri T
      • de Jong MC
      • Groot Koerkamp PW
      Effects of temperature, relative humidity, absolute humidity, and evaporation potential on survival of airborne Gumboro vaccine virus.
      studied the survival of aerosolized Gumboro virus and found that a large part of the effect of AH on viral survival was attributable to temperature. They found that temperature and RH were the best predictors of virus survival and that survival was not significantly affected by AH. Because the Gumboro virus is nonenveloped, the relationship between viability and RH is likely to differ from that of the influenza virus (
      • Tang JW.
      The effect of environmental parameters on the survival of airborne infectious agents.
      ); however, the relationship between temperature, AH, and RH is universally applicable. In addition, the 1-year results of our study indicate that the correlation between RH and Rt is more statistically significant than that between AH and Rt, which supports that indoor RH may be a better indicator to predict the transmission of influenza. Outdoor AH might be a good proxy for indoor AH in both temperate and subtropical regions, whereas outdoor RH is not a good proxy for indoor RH in subtropical regions, especially during hot/warm days. Furthermore, our research also found that the correlation between indoor RH and indoor/outdoor AH is strong on cold days. These could partly explain why outdoor AH was a better indicator of influenza transmission than outdoor RH in other studies (
      • Deyle ER
      • Maher MC
      • Hernandez RD
      • Basu S
      • Sugihara G.
      Global environmental drivers of influenza.
      ;
      • McDevitt J
      • Rudnick S
      • First M
      • Spengler J.
      Role of absolute humidity in the inactivation of influenza viruses on stainless steel surfaces at elevated temperatures.
      ;
      • Shaman J
      • Kohn M.
      Absolute humidity modulates influenza survival, transmission, and seasonality.
      ). In addition, the study of indoor and outdoor environmental conditions also had other implications. For example, the indoor RH can be used as an indicator for wooden home fire risk (
      • Log T.
      Indoor relative humidity as a fire risk indicator.
      ); indoor environment can also be used as an indicator to study the protection of ancient buildings and the health and comfort of indoor individuals (
      • Marcu F
      • Hodor N
      • Indrie L
      • Dejeu P
      • Ilieș M
      • Albu A
      • et al.
      Microbiological, health and comfort aspects of indoor air quality in a Romanian historical wooden church.
      ).
      This study had several limitations. First, the data time span used in this study was insufficient, and the number of sensors measuring indoor environmental factors in Tianjin, Chongqing, and Hong Kong was small, which may cause the poor robustness of the results by city level. Thus, we merged the data and expanded the measured indoor environment data into 5-year data; this may not be suitable, but using the expanded data also produce similar results, which may partly suggest that the results from 1-year data were robust. The findings also indicated that the transmission of influenza virus in the next few days may be predicted based on the information of weather forecast. In the future, it is necessary to measure indoor temperature and RH for a much longer period to further confirm our results and explore the relationship between indoor and outdoor environments to preliminarily establish an influenza transmission prediction system based on the indoor environment. Second, due to the lack of related data, other factors that may affect influenza seasonality, such as ventilation rate, cannot be studied in this study. In the future, relevant data are needed to further verify the findings of this study. Third, only indoor temperature and RH in household environments were measured; however, indoor temperature and RH in other indoor spaces, such as schools and workplaces, were not measured, but these indoor spaces have been proven to be the primary contexts of influenza transmission (
      • Ferguson NM
      • Cummings DA
      • Cauchemez S
      • Fraser C
      • Riley S
      • Meeyai A
      • et al.
      Strategies for containing an emerging influenza pandemic in Southeast Asia.
      ;
      • Goldstein E
      • Lipsitch M
      • Cevik M.
      On the Effect of age on the transmission of SARS-CoV-2 in households, schools, and the community.
      ). Thus, measurements of temperature and RH in these indoor environments, especially in schools, should be performed in the future. Finally, in this study, only the indoor temperatures and RH in temperate and subtropical locations were measured, and the indoor temperature and RH in tropical regions, where influenza activities showed more diverse seasonality, were not measured and analyzed. Future studies in tropical regions are required to further explore the relationship between indoor RH and influenza transmissibility.
      In conclusion, we identified that indoor RH is a better indicator of influenza seasonality than outdoor RH in both temperate and subtropical locations. Accordingly, an influenza early warning system based on indoor environmental factors should be developed and implemented to enable a timely response to surges in influenza activities. In addition, although this paper studied only the influenza virus, the results may also be applicable to other respiratory infections, such as SARS-CoV-2, because the transmission of SARS-CoV-2 is also much more common indoors than outdoors.

      Declaration of competing interest

      The authors have no competing interests to declare.

      Funding

      This study was supported by grants from the National Natural Science Foundation of China (Grant No. 82003509 to HL, Grant No. 82173577 to SY), the Natural Science Foundation of Zhejiang Province (Grant No. LQ20H260009 to HL), and Fundamental Research Funds for the Central Universities (to HL).

      Ethical approval

      The approval was not required.

      Acknowledgments

      The authors would like to thank Prof. Linsey C. Marr for providing access to the indoor temperature and relative humidity data in Hong Kong from their extensive work in
      • Pan J
      • Tang J
      • Caniza M
      • Heraud JM
      • Koay E
      • Lee HK
      • et al.
      Correlating indoor and outdoor temperature and humidity in a sample of buildings in tropical climates.
      .

      Author contributions

      HL and YS conceived, designed, and supervised the study. HL and MY analyzed the data. XD, ZD, TC, LY, and DW collected the data. HL cleaned data. HL and MY wrote the drafts of the manuscript. KH, ZD, NZ, SY, and YL commented on and revised drafts of the manuscript. All authors read and approved the final report.

      References

        • Ali ST
        • Cowling BJ
        • Lau EHY
        • Fang VJ
        • Leung GM.
        Mitigation of influenza B epidemic with school closures, Hong Kong, 2018.
        Emerg Infect Dis. 2018; 24: 2071-2073
        • Ali ST
        • Cowling BJ
        • Wong JY
        • Chen D
        • Shan S
        • Lau E
        • et al.
        Influenza seasonality and its environmental driving factors in mainland China and Hong Kong.
        Sci Total Environ. 2022; 818151724
        • Ben Y
        • Ma F
        • Wang H
        • Hassan MA
        • Yevheniia R
        • Fan W
        • et al.
        A spatio-temporally weighted hybrid model to improve estimates of personal PM2.5 exposure: incorporating big data from multiple data sources.
        Environ Pollut. 2019; 253: 403-411
        • Bulfone TC
        • Malekinejad M
        • Rutherford GW
        • Razani N.
        Outdoor transmission of SARS-CoV-2 and other respiratory viruses: a systematic review.
        J Infect Dis. 2021; 223: 550-561
        • Cori A
        • Ferguson NM
        • Fraser C
        • Cauchemez S.
        A new framework and software to estimate time-varying reproduction numbers during epidemics.
        Am J Epidemiol. 2013; 178: 1505-1512
        • Dai Q
        • Ma W
        • Huang H
        • Xu K
        • Qi X
        • Yu H
        • et al.
        The effect of ambient temperature on the activity of influenza and influenza like illness in Jiangsu Province.
        China. Sci Total Environ. 2018; 645: 684-691
        • Davey ML
        • Reid D.
        Relationship of air temperature to outbreaks of influenza.
        Br J Prev Soc Med. 1972; 26: 28-32
        • Deyle ER
        • Maher MC
        • Hernandez RD
        • Basu S
        • Sugihara G.
        Global environmental drivers of influenza.
        Proc Natl Acad Sci U S A. 2016; 113: 13081-13086
        • Dong Z
        • Wang H
        • Yin P
        • Wang L
        • Chen R
        • Fan W
        • et al.
        Time-weighted average of fine particulate matter exposure and cause-specific mortality in China: a nationwide analysis.
        Lancet Planet Health. 2020; 4: e343-e351
        • Ferguson NM
        • Cummings DA
        • Cauchemez S
        • Fraser C
        • Riley S
        • Meeyai A
        • et al.
        Strategies for containing an emerging influenza pandemic in Southeast Asia.
        Nature. 2005; 437: 209-214
        • Gasparrini A.
        Distributed lag linear and non-linear models in R: the package dlnm.
        J Stat Softw. 2011; 43: 1-20
        • Goldstein E
        • Lipsitch M
        • Cevik M.
        On the Effect of age on the transmission of SARS-CoV-2 in households, schools, and the community.
        J Infect Dis. 2021; 223: 362-369
        • Graham SE
        • McCurdy T.
        Developing meaningful cohorts for human exposure models.
        J Expo Anal Environ Epidemiol. 2004; 14: 23-43
        • Hemmes JH
        • Winkler KC
        • Kool SM.
        Virus survival as a seasonal factor in influenza and polimyelitis.
        Nature. 1960; 188: 430-431
        • Hope-Simpson RE.
        The role of season in the epidemiology of influenza.
        J Hyg (Lond). 1981; 86: 35-47
        • Klepeis NE
        • Nelson WC
        • Ott WR
        • Robinson JP
        • Tsang AM
        • Switzer P
        • et al.
        The National Human Activity Pattern Survey (NHAPS): a resource for assessing exposure to environmental pollutants.
        J Expo Anal Environ Epidemiol. 2001; 11: 231-252
        • Kong ZM
        • Sandhu HS
        • Qiu L
        • Wu J
        • Tian WJ
        • Chi XJ
        • et al.
        Virus dynamics and decay in evaporating human saliva droplets on fomites.
        Environ Sci Technol Forthcoming. 2022;
        • Lau EH
        • Cowling BJ
        • Ho LM
        • Leung GM.
        Optimizing use of multistream influenza sentinel surveillance data.
        Emerg Infect Dis. 2008; 14: 1154-1157
        • Lei H
        • Xu M
        • Wang X
        • Xie Y
        • Du X
        • Chen T
        • et al.
        Nonpharmaceutical interventions used to control COVID-19 reduced seasonal influenza transmission in China.
        J Infect Dis. 2020; 222: 1780-1783
        • Log T.
        Indoor relative humidity as a fire risk indicator.
        Build Environ. 2017; 111: 238-248
        • Lowen A
        • Palese P.
        Transmission of influenza virus in temperate zones is predominantly by aerosol, in the tropics by contact: a hypothesis.
        PLoS Curr. 2009; 1: RRN1002
        • Lowen AC
        • Mubareka S
        • Steel J
        • Palese P.
        Influenza virus transmission is dependent on relative humidity and temperature.
        PLoS Pathog. 2007; 3: 1470-1476
        • Lowen AC
        • Steel J
        • Mubareka S
        • Palese P.
        High temperature (30 degrees C) blocks aerosol but not contact transmission of influenza virus.
        J Virol. 2008; 82: 5650-5652
        • Marcu F
        • Hodor N
        • Indrie L
        • Dejeu P
        • Ilieș M
        • Albu A
        • et al.
        Microbiological, health and comfort aspects of indoor air quality in a Romanian historical wooden church.
        Int J Environ Res Public Health. 2021; 18: 9908
        • Marr LC
        • Tang JW
        • Van Mullekom J
        • Lakdawala SS.
        Mechanistic insights into the effect of humidity on airborne influenza virus survival, transmission and incidence.
        J R Soc Interface. 2019; 1620180298
        • McDevitt J
        • Rudnick S
        • First M
        • Spengler J.
        Role of absolute humidity in the inactivation of influenza viruses on stainless steel surfaces at elevated temperatures.
        Appl Environ Microbiol. 2010; 76: 3943-3947
        • Moriyama M
        • Hugentobler WJ
        • Iwasaki A.
        Seasonality of respiratory viral infections.
        Annu Rev Virol. 2020; 7: 83-101
        • Moura FE
        • Perdigão AC
        • Siqueira MM.
        Seasonality of influenza in the tropics: a distinct pattern in northeastern Brazil.
        Am J Trop Med Hyg. 2009; 81: 180-183
        • Pan J
        • Tang J
        • Caniza M
        • Heraud JM
        • Koay E
        • Lee HK
        • et al.
        Correlating indoor and outdoor temperature and humidity in a sample of buildings in tropical climates.
        Indoor Air. 2021; 31: 2281-2295
        • Qi L
        • Liu T
        • Gao Y
        • Tian D
        • Tang W
        • Li Q
        • et al.
        Effect of meteorological factors on the activity of influenza in Chongqing, China, 2012–2019.
        PLoS One. 2021; 16e0246023
        • Schulman JL
        • Kilbourne ED.
        Experimental transmission of influenza virus infection in Mice. II. Some factors affecting the incidence of transmitted infection.
        J Exp Med. 1963; 118: 267-275
        • Shaman J
        • Kohn M.
        Absolute humidity modulates influenza survival, transmission, and seasonality.
        Proc Natl Acad Sci U S A. 2009; 106: 3243-3248
        • Shaman J
        • Pitzer VE
        • Viboud C
        • Grenfell BT
        • Lipsitch M.
        Absolute humidity and the seasonal onset of influenza in the continental United States.
        PLoS Biol. 2010; 8e1000316
        • Shimmei K
        • Nakamura T
        • Ng CFS
        • Hashizume M
        • Murakami Y
        • Maruyama A
        • et al.
        Association between seasonal influenza and absolute humidity: time-series analysis with daily surveillance data in Japan.
        Sci Rep. 2020; 10: 7764
        • Shu YL
        • Fang LQ
        • de Vlas SJ
        • Gao Y
        • Richardus JH
        • Cao WC.
        Dual seasonal patterns for influenza.
        China. Emerg Infect Dis. 2010; 16: 725-726
        • Tamerius J
        • Nelson MI
        • Zhou SZ
        • Viboud C
        • Miller MA
        • Alonso WJ.
        Global influenza seasonality: reconciling patterns across temperate and tropical regions.
        Environ Health Perspect. 2011; 119: 439-445
        • Tamerius JD
        • Shaman J
        • Alonso WJ
        • Bloom-Feshbach K
        • Uejio CK
        • Comrie A
        • et al.
        Environmental predictors of seasonal influenza epidemics across temperate and tropical climates.
        PLoS Pathog. 2013; 9e1003194
        • Tang JW.
        The effect of environmental parameters on the survival of airborne infectious agents.
        J R Soc Interface. 2009; 6: S737-S746
        • Tang JW
        • Lai FY
        • Nymadawa P
        • Deng YM
        • Ratnamohan M
        • Petric M
        • et al.
        Comparison of the incidence of influenza in relation to climate factors during 2000–2007 in five countries.
        J Med Virol. 2010; 82: 1958-1965
        • Virlogeux V
        • Li M
        • Tsang TK
        • Feng L
        • Fang VJ
        • Jiang H
        • et al.
        Estimating the distribution of the incubation periods of human avian influenza A(H7N9) virus infections.
        Am J Epidemiol. 2015; 182: 723-729
        • Wallinga J
        • Lipsitch M.
        How generation intervals shape the relationship between growth rates and reproductive numbers.
        Proc Biol Sci. 2007; 274: 599-604
        • Yang W
        • Elankumaran S
        • Marr LC.
        Relationship between humidity and influenza A viability in droplets and implications for influenza's seasonality.
        PLoS One. 2012; 7: e46789
        • Zhao Y
        • Aarnink AJ
        • Dijkman R
        • Fabri T
        • de Jong MC
        • Groot Koerkamp PW
        Effects of temperature, relative humidity, absolute humidity, and evaporation potential on survival of airborne Gumboro vaccine virus.
        Appl Environ Microbiol. 2012; 78: 1048-1054