International Journal of Infectious Diseases 30 (2015) 122–124
Contents lists available at ScienceDirect
International Journal of Infectious Diseases journal homepage: www.elsevier.com/locate/ijid
Short Communication
The impact of temperature and humidity measures on influenza A (H7N9) outbreaks—evidence from China Yi Zhang a, Cindy Feng b, Chunna Ma a, Peng Yang a, Song Tang c, Abby Lau d, Wenjie Sun e,*, Quanyi Wang a,* a
Beijing Centre for Disease Prevention and Control (CDC), No. 16 He Pingli Middle St, Dongcheng District, Beijing 100013, China School of Public Health and Western College of Veterinary Medicine, University of Saskatchewan, Saskatoon, Canada School of Environment and Sustainability, University of Saskatchewan, Saskatoon, Canada d Tulane Infectious Disease Department, Tulane University, New Orleans, Louisiana, USA e School of Public Health and Tropical Medicine, Tulane University, 1440 Canal Street, Suite 2100, New Orleans, LA 70112, USA b c
A R T I C L E I N F O
Article history: Received 12 September 2014 Received in revised form 29 October 2014 Accepted 12 November 2014 Corresponding Editor: Eskild Petersen, Aarhus, Denmark. Keywords: Influenza A H7N9 Temperature Humidity Zero truncation Penalized spline function
S U M M A R Y
Objectives: To examine the non-linear effects of meteorological factors on the incidence of influenza A H7N9 and to determine what meteorological measure, and on which day preceding symptom onset, has the most significant effect on H7N9 infection. Methods: We applied a zero truncated Poisson regression model incorporating smoothed spline functions to assess the non-linear effect of temperature (maximum, minimum, and daily difference) and relative humidity on H7N9 human case numbers occurring in China from February 19, 2013 to February 18, 2014, adjusting for the effects of age and gender. Results: Both daily minimum and daily maximum temperature contributed significantly to human infection with the influenza A H7N9 virus. Models incorporating the non-linear effect of minimum or maximum temperature on day 13 prior to disease onset were found to have the best predictive ability. For minimum temperature, high risk was found to range from approximately 5 to 9 8C and moderate risk from 10 to 0 8C; temperatures of >9 8C represented a low risk. For maximum temperature, high risk was found to range from approximately 13 to 18 8C and moderate risk from 0 to 4 8C; temperatures of >18 8C represented a low risk. Relative humidity was not significantly associated with the incidence of infection. The incidence of H7N9 was higher for males compared to females (p < 0.01) and it peaked at around 60–70 years of age. Conclusions: We provide direct evidence to support the role of climate conditions in the spread of H7N9 and thereby address a critical question fundamental to our understanding of the epidemiology and evolution of H7N9. These findings could be used to inform targeted surveillance and control efforts aimed at reducing the future spread of H7N9. ß 2014 The Authors. Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/3.0/).
1. Introduction In addition to biological and ecological factors, climate factors influence the emergence of infectious diseases.1 Previous studies
* Corresponding authors. Tel./fax.: +86 10 6440 7109; Tel.: +1 504 988 4223. E-mail addresses: [email protected] (W. Sun), [email protected] (Q. Wang).
have shown that low temperature, low humidity, and the daily temperature difference contribute to the increased risk of seasonal influenza during the winter months.2–4 Although interest in assessing the impact of meteorological factors on influenza A H7N9 has increased, few studies have examined their non-linear effects on the incidence of H7N9. In addition, no research has been conducted to establish the temperature measure that is the best predictor of the incidence of H7N9 or to determine on which day
http://dx.doi.org/10.1016/j.ijid.2014.11.010 1201-9712/ß 2014 The Authors. Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).
Y. Zhang et al. / International Journal of Infectious Diseases 30 (2015) 122–124
123
preceding disease onset the meteorological factors have the most significant effect. The aim of this study was to fill these gaps. 2. Methods 2.1. Data sources Information was obtained from a database that included all cases of influenza A H7N9 virus infection reported by the health bureau of each province, starting from the first confirmed case in China on February 19, 2013 until February 18, 2014. The date of illness onset was acquired from the websites of the local health bureaus. Daily meteorological data, including daily maximum and minimum temperatures and relative humidity, spanning the 2 weeks prior to symptom onset for each laboratory-confirmed case were obtained through the China Weather Network outdoor weather reports. 2.2. Statistical analysis The number of laboratory-confirmed H7N9 cases was calculated for each unique combination of date of disease onset, age group, and gender, for which the value zero cannot occur. A zerotruncated Poisson regression model was therefore applied to quantify the non-linear effect of temperature, humidity, and age group on the number of H7N9 cases.5 Age was categorized into nine groups: 0–9, 10–19, 20–29, 30–39, 40–49, 50–59, 60–69, 70– 79, and 80 years. The zero-truncated Poisson model is:
Y ti j Zero truncated Poisson mti j
logðmti j Þ ¼ a þ s0 ageftig þ bgender ft jg þ s1 ðtem pti jflg Þ þ s2 ðhumidti jflg Þ where t is the date of disease onset, i indicates the age group (i = 1, 2,. . .9), and j denotes the gender group. Ytij and mtij are the observed and expected number of reported cases of H7N9 on day t for age i and gender group j, respectively. b is the regression coefficient for gender. temptij(l) was calculated by averaging the temperature measures for those cases l days preceding the date of disease onset on day t from age group i and gender group j (l = 1, 2,. . .14). humidtij(l) was calculated in the same manner. s0(), s1(), and s2() are non-parametric smooth functions of age, temperature, and humidity, respectively (see Supplementary Material A for further details). The smoothness of the spline functions depends on the degree of freedom (df), which was determined by minimizing the Akaike information criterion (AIC).6 The model with a lower AIC score is preferred as it achieves a more optimal combination of goodness of fit and parsimony. In this analysis, the minimum AIC was achieved when df = 3 for all of the spline terms. The statistical analysis was conducted using the ‘countreg’ package in R (version 2.14). 3. Results A total of 363 cases of H7N9 occurred during the study period, distributed across 14 provinces of mainland China. Figure 1 shows that models using the minimum temperature on day 13 preceding the date of disease onset have the best predictive ability, followed by models using maximum temperature as the temperature covariate. Figure 2 shows that for minimum temperature on day 13 preceding the date of disease onset, high risk ranges from approximately 5 to 9 8C and moderate risk from 10 to 0 8C;
Figure 1. AIC scores for a series of models using daily maximum and minimum temperature and the temperature difference spanning the 2 weeks prior to H7N9 disease onset. The solid line corresponds to the model: logðmti j Þ ¼ a þ s0 ðagei Þ þ b gender j þ s1 ðmaximum tem pti jðlÞ Þ þ þs2 ðhumidti jðlÞ Þ; The dashed line corresponds to the model: logðmti j Þ ¼ a þ s0 ðagei Þ þ b gender j þ s1 ðminimum tem pti jðlÞ Þ þ þs2 ðhumidti jðlÞ Þ; The dotted-dashed line corresponds to the model: logðmti j Þa þ s0 ðagei Þ þ b gender j þ s1 ðdaily tem p di f ferenceti jðlÞ Þ þ þs2 ðhumidti jðlÞ Þ; where l = 1, 2,. . .14.
temperatures of >9 8C represent a low risk. If based on the maximum temperature, high risk ranges from approximately 13 to 18 8C and moderate risk from 0 to 4 8C; temperatures of >18 8C represent a low risk. Neither daily temperature difference nor relative humidity was significantly associated with the incidence of infection. Figure S1 in the Supplementary Material presents the non-linear effect of age on the incidence of H7N9; this shows that the effect of age decreased as age increased up to 20–30 years of age and then peaked at around 60–70 years of age. The incidence of H7N9 was found to be higher for males compared to females (p < 0.01).
4. Discussion The most salient finding of our study is that the temperature on day 13 prior to disease onset, as covariates, has the best predictive ability regarding the incidence of H7N9. Of note, a previous study showed that the estimated mean incubation period was 3.1 days.6 However, the incubation period and the lag time related to temperature considered in the present study are two different concepts. The incubation period is defined as the time elapsed between exposure to a pathogenic organism and the first appearance of signs and symptoms of infection; our study aimed to investigate the relationship between H7N9 incidence and climate conditions. We speculate that the temperature on day 13 prior to illness onset affects virus activity and human immunity, and consequently increases vulnerability and susceptibility to infection for those who are exposed to poultry markets. Our results also demonstrated that temperature has a nonlinear effect on the incidence of H7N9, which provides a piece of critical evidence consistent with other studies.7 To be more specific, a high risk of H7N9 infection occurred within a certain temperature range, while both higher and lower temperatures can lead to a decline in the risk of infection. This can be explained by changes in virus activity and human susceptibility under different climate conditions.
124
Y. Zhang et al. / International Journal of Infectious Diseases 30 (2015) 122–124
Figure 2. The predicted logarithm of the number of H7N9-infected cases by minimum temperature at day 13 prior to disease onset and humidity (left panel) and the predicted logarithm of the number of H7N9-infected cases by maximum temperature at day 13 prior to disease onset and humidity (right panel).
No significant correlation was found between relative humidity and the incidence of H7N9 infection in the present study. While humidity is usually regarded as a factor impacting on the transmission of diseases that are transmitted by droplets or aerosols,8 the main routes of H7N9 transmission may be different from those of seasonal influenza.9 However, our study did not consider the impact of poultry market closures, which effectively reduced the numbers of cases of H7N9.10,11 The analysis of age showed that people aged between 20 and 30 years are at the lowest risk of becoming infected, while those aged between 60 and 70 years are at the highest risk of becoming infected; each age group increase from 30–40 to 40–50 to 50–60 years has a significant impact on the increase in the log mean of the infection incidence. This is consistent with, and even adds more to, the findings from a previous study that showed exposure to poultry markets to be associated with H7N9 virus infection.12 To be specific, exposure to the poultry market is likely to vary with age. For example, children and the workingage population are less likely to visit poultry markets when compared to those aged 60–70 years who are usually responsible for grocery shopping and are therefore more predisposed to infection. Finally, we also found that the incidence was higher in males than in females and this can be explained by greater exposure to poultry for males compared to females.13 In conclusion, a flexible statistical model that models the incidence of influenza A H7N9 was constructed in this study. It provided important public health information for the prediction of H7N9 outbreaks. This will help health policymakers to determine requirements for medical supplies and personnel in order to further curb the spread of H7N9. Acknowledgements This work was supported by the Beijing Municipal Science and Technology Commission (No. Z131100005613048), the Capital Health Research and Development of Special (2014-1-1011), and the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant for Cindy Feng. The funders had no role in
the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Conflict of interest: There are no conflicts of interest to declare.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijid.2014.11.010. References 1. Patz JA, Epstein PR, Burke TA, Balbus JM. Global climate change and emerging infectious diseases. JAMA 1996;275:217–23. 2. Tsuchihashi Y, Yorifuji T, Takao S, Suzuki E, Mori S, Doi H, et al. Environmental factors and seasonal influenza onset in Okayama city, Japan: case-crossover study. Acta Med Okayama 2011;65:97–103. 3. 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;8:e1000316. 4. Jaakkola K, Saukkoriipi A, Jokelainen J, Juvonen R, Kauppila J, Vainio O, et al. Decline in temperature and humidity increases the occurrence of influenza in cold climate. Environ Health 2014;13:22. 5. Long JS. Regression models for categorical and limited dependent variables. Thousand Oaks, CA: Sage Publications; 1997. 6. Akaike H. A new look at the statistical model identification. IEEE Transactions on Automatic Control 1974;19:716–23. 7. Zhang Z, Xia Y, Lu Y, Yang J, Zhang L, Su H, et al. Prediction of H7N9 epidemic in China. Chin Med J (Engl) 2014;127:254–60. 8. Koep TH, Enders FT, Pierret C, Ekker SC, Krageschmidt D, Neff KL, et al. Predictors of indoor absolute humidity and estimated effects on influenza virus survival in grade schools. BMC Infect Dis 2013;13:71. 9. Chan PK, Mok HY, Lee TC, Chu IM, Lam WY, Sung JJ. Seasonal influenza activity in Hong Kong and its association with meteorological variations. J Med Virol 2009;81:1797–806. 10. Yu H, Wu JT, Cowling BJ, Liao Q, Fang VJ, Zhou S, et al. Effect of closure of live poultry markets on poultry-to-person transmission of avian influenza A H7N9 virus: an ecological study. Lancet 2014;383:541–8. 11. Xu J, Lu S, Wang H, Chen C. Reducing exposure to avian influenza H7N9. Lancet 2013;381:1815–6. 12. Liu B, Havers F, Chen E, Yuan Z, Yuan H, Ou J, et al. Risk factors for influenza A(H7N9) disease—China, 2013. Clin Infect Dis 2014;59:787–94. 13. Cowling BJ, Jin L, Lau EH, Liao Q, Wu P, Jiang H, et al. Comparative epidemiology of human infections with avian influenza A H7N9 and H5N1 viruses in China: a population-based study of laboratory-confirmed cases. Lancet 2013;382:129–37.
Contents lists available at ScienceDirect
International Journal of Infectious Diseases journal homepage: www.elsevier.com/locate/ijid
Short Communication
The impact of temperature and humidity measures on influenza A (H7N9) outbreaks—evidence from China Yi Zhang a, Cindy Feng b, Chunna Ma a, Peng Yang a, Song Tang c, Abby Lau d, Wenjie Sun e,*, Quanyi Wang a,* a
Beijing Centre for Disease Prevention and Control (CDC), No. 16 He Pingli Middle St, Dongcheng District, Beijing 100013, China School of Public Health and Western College of Veterinary Medicine, University of Saskatchewan, Saskatoon, Canada School of Environment and Sustainability, University of Saskatchewan, Saskatoon, Canada d Tulane Infectious Disease Department, Tulane University, New Orleans, Louisiana, USA e School of Public Health and Tropical Medicine, Tulane University, 1440 Canal Street, Suite 2100, New Orleans, LA 70112, USA b c
A R T I C L E I N F O
Article history: Received 12 September 2014 Received in revised form 29 October 2014 Accepted 12 November 2014 Corresponding Editor: Eskild Petersen, Aarhus, Denmark. Keywords: Influenza A H7N9 Temperature Humidity Zero truncation Penalized spline function
S U M M A R Y
Objectives: To examine the non-linear effects of meteorological factors on the incidence of influenza A H7N9 and to determine what meteorological measure, and on which day preceding symptom onset, has the most significant effect on H7N9 infection. Methods: We applied a zero truncated Poisson regression model incorporating smoothed spline functions to assess the non-linear effect of temperature (maximum, minimum, and daily difference) and relative humidity on H7N9 human case numbers occurring in China from February 19, 2013 to February 18, 2014, adjusting for the effects of age and gender. Results: Both daily minimum and daily maximum temperature contributed significantly to human infection with the influenza A H7N9 virus. Models incorporating the non-linear effect of minimum or maximum temperature on day 13 prior to disease onset were found to have the best predictive ability. For minimum temperature, high risk was found to range from approximately 5 to 9 8C and moderate risk from 10 to 0 8C; temperatures of >9 8C represented a low risk. For maximum temperature, high risk was found to range from approximately 13 to 18 8C and moderate risk from 0 to 4 8C; temperatures of >18 8C represented a low risk. Relative humidity was not significantly associated with the incidence of infection. The incidence of H7N9 was higher for males compared to females (p < 0.01) and it peaked at around 60–70 years of age. Conclusions: We provide direct evidence to support the role of climate conditions in the spread of H7N9 and thereby address a critical question fundamental to our understanding of the epidemiology and evolution of H7N9. These findings could be used to inform targeted surveillance and control efforts aimed at reducing the future spread of H7N9. ß 2014 The Authors. Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/3.0/).
1. Introduction In addition to biological and ecological factors, climate factors influence the emergence of infectious diseases.1 Previous studies
* Corresponding authors. Tel./fax.: +86 10 6440 7109; Tel.: +1 504 988 4223. E-mail addresses: [email protected] (W. Sun), [email protected] (Q. Wang).
have shown that low temperature, low humidity, and the daily temperature difference contribute to the increased risk of seasonal influenza during the winter months.2–4 Although interest in assessing the impact of meteorological factors on influenza A H7N9 has increased, few studies have examined their non-linear effects on the incidence of H7N9. In addition, no research has been conducted to establish the temperature measure that is the best predictor of the incidence of H7N9 or to determine on which day
http://dx.doi.org/10.1016/j.ijid.2014.11.010 1201-9712/ß 2014 The Authors. Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).
Y. Zhang et al. / International Journal of Infectious Diseases 30 (2015) 122–124
123
preceding disease onset the meteorological factors have the most significant effect. The aim of this study was to fill these gaps. 2. Methods 2.1. Data sources Information was obtained from a database that included all cases of influenza A H7N9 virus infection reported by the health bureau of each province, starting from the first confirmed case in China on February 19, 2013 until February 18, 2014. The date of illness onset was acquired from the websites of the local health bureaus. Daily meteorological data, including daily maximum and minimum temperatures and relative humidity, spanning the 2 weeks prior to symptom onset for each laboratory-confirmed case were obtained through the China Weather Network outdoor weather reports. 2.2. Statistical analysis The number of laboratory-confirmed H7N9 cases was calculated for each unique combination of date of disease onset, age group, and gender, for which the value zero cannot occur. A zerotruncated Poisson regression model was therefore applied to quantify the non-linear effect of temperature, humidity, and age group on the number of H7N9 cases.5 Age was categorized into nine groups: 0–9, 10–19, 20–29, 30–39, 40–49, 50–59, 60–69, 70– 79, and 80 years. The zero-truncated Poisson model is:
Y ti j Zero truncated Poisson mti j
logðmti j Þ ¼ a þ s0 ageftig þ bgender ft jg þ s1 ðtem pti jflg Þ þ s2 ðhumidti jflg Þ where t is the date of disease onset, i indicates the age group (i = 1, 2,. . .9), and j denotes the gender group. Ytij and mtij are the observed and expected number of reported cases of H7N9 on day t for age i and gender group j, respectively. b is the regression coefficient for gender. temptij(l) was calculated by averaging the temperature measures for those cases l days preceding the date of disease onset on day t from age group i and gender group j (l = 1, 2,. . .14). humidtij(l) was calculated in the same manner. s0(), s1(), and s2() are non-parametric smooth functions of age, temperature, and humidity, respectively (see Supplementary Material A for further details). The smoothness of the spline functions depends on the degree of freedom (df), which was determined by minimizing the Akaike information criterion (AIC).6 The model with a lower AIC score is preferred as it achieves a more optimal combination of goodness of fit and parsimony. In this analysis, the minimum AIC was achieved when df = 3 for all of the spline terms. The statistical analysis was conducted using the ‘countreg’ package in R (version 2.14). 3. Results A total of 363 cases of H7N9 occurred during the study period, distributed across 14 provinces of mainland China. Figure 1 shows that models using the minimum temperature on day 13 preceding the date of disease onset have the best predictive ability, followed by models using maximum temperature as the temperature covariate. Figure 2 shows that for minimum temperature on day 13 preceding the date of disease onset, high risk ranges from approximately 5 to 9 8C and moderate risk from 10 to 0 8C;
Figure 1. AIC scores for a series of models using daily maximum and minimum temperature and the temperature difference spanning the 2 weeks prior to H7N9 disease onset. The solid line corresponds to the model: logðmti j Þ ¼ a þ s0 ðagei Þ þ b gender j þ s1 ðmaximum tem pti jðlÞ Þ þ þs2 ðhumidti jðlÞ Þ; The dashed line corresponds to the model: logðmti j Þ ¼ a þ s0 ðagei Þ þ b gender j þ s1 ðminimum tem pti jðlÞ Þ þ þs2 ðhumidti jðlÞ Þ; The dotted-dashed line corresponds to the model: logðmti j Þa þ s0 ðagei Þ þ b gender j þ s1 ðdaily tem p di f ferenceti jðlÞ Þ þ þs2 ðhumidti jðlÞ Þ; where l = 1, 2,. . .14.
temperatures of >9 8C represent a low risk. If based on the maximum temperature, high risk ranges from approximately 13 to 18 8C and moderate risk from 0 to 4 8C; temperatures of >18 8C represent a low risk. Neither daily temperature difference nor relative humidity was significantly associated with the incidence of infection. Figure S1 in the Supplementary Material presents the non-linear effect of age on the incidence of H7N9; this shows that the effect of age decreased as age increased up to 20–30 years of age and then peaked at around 60–70 years of age. The incidence of H7N9 was found to be higher for males compared to females (p < 0.01).
4. Discussion The most salient finding of our study is that the temperature on day 13 prior to disease onset, as covariates, has the best predictive ability regarding the incidence of H7N9. Of note, a previous study showed that the estimated mean incubation period was 3.1 days.6 However, the incubation period and the lag time related to temperature considered in the present study are two different concepts. The incubation period is defined as the time elapsed between exposure to a pathogenic organism and the first appearance of signs and symptoms of infection; our study aimed to investigate the relationship between H7N9 incidence and climate conditions. We speculate that the temperature on day 13 prior to illness onset affects virus activity and human immunity, and consequently increases vulnerability and susceptibility to infection for those who are exposed to poultry markets. Our results also demonstrated that temperature has a nonlinear effect on the incidence of H7N9, which provides a piece of critical evidence consistent with other studies.7 To be more specific, a high risk of H7N9 infection occurred within a certain temperature range, while both higher and lower temperatures can lead to a decline in the risk of infection. This can be explained by changes in virus activity and human susceptibility under different climate conditions.
124
Y. Zhang et al. / International Journal of Infectious Diseases 30 (2015) 122–124
Figure 2. The predicted logarithm of the number of H7N9-infected cases by minimum temperature at day 13 prior to disease onset and humidity (left panel) and the predicted logarithm of the number of H7N9-infected cases by maximum temperature at day 13 prior to disease onset and humidity (right panel).
No significant correlation was found between relative humidity and the incidence of H7N9 infection in the present study. While humidity is usually regarded as a factor impacting on the transmission of diseases that are transmitted by droplets or aerosols,8 the main routes of H7N9 transmission may be different from those of seasonal influenza.9 However, our study did not consider the impact of poultry market closures, which effectively reduced the numbers of cases of H7N9.10,11 The analysis of age showed that people aged between 20 and 30 years are at the lowest risk of becoming infected, while those aged between 60 and 70 years are at the highest risk of becoming infected; each age group increase from 30–40 to 40–50 to 50–60 years has a significant impact on the increase in the log mean of the infection incidence. This is consistent with, and even adds more to, the findings from a previous study that showed exposure to poultry markets to be associated with H7N9 virus infection.12 To be specific, exposure to the poultry market is likely to vary with age. For example, children and the workingage population are less likely to visit poultry markets when compared to those aged 60–70 years who are usually responsible for grocery shopping and are therefore more predisposed to infection. Finally, we also found that the incidence was higher in males than in females and this can be explained by greater exposure to poultry for males compared to females.13 In conclusion, a flexible statistical model that models the incidence of influenza A H7N9 was constructed in this study. It provided important public health information for the prediction of H7N9 outbreaks. This will help health policymakers to determine requirements for medical supplies and personnel in order to further curb the spread of H7N9. Acknowledgements This work was supported by the Beijing Municipal Science and Technology Commission (No. Z131100005613048), the Capital Health Research and Development of Special (2014-1-1011), and the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant for Cindy Feng. The funders had no role in
the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Conflict of interest: There are no conflicts of interest to declare.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijid.2014.11.010. References 1. Patz JA, Epstein PR, Burke TA, Balbus JM. Global climate change and emerging infectious diseases. JAMA 1996;275:217–23. 2. Tsuchihashi Y, Yorifuji T, Takao S, Suzuki E, Mori S, Doi H, et al. Environmental factors and seasonal influenza onset in Okayama city, Japan: case-crossover study. Acta Med Okayama 2011;65:97–103. 3. 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;8:e1000316. 4. Jaakkola K, Saukkoriipi A, Jokelainen J, Juvonen R, Kauppila J, Vainio O, et al. Decline in temperature and humidity increases the occurrence of influenza in cold climate. Environ Health 2014;13:22. 5. Long JS. Regression models for categorical and limited dependent variables. Thousand Oaks, CA: Sage Publications; 1997. 6. Akaike H. A new look at the statistical model identification. IEEE Transactions on Automatic Control 1974;19:716–23. 7. Zhang Z, Xia Y, Lu Y, Yang J, Zhang L, Su H, et al. Prediction of H7N9 epidemic in China. Chin Med J (Engl) 2014;127:254–60. 8. Koep TH, Enders FT, Pierret C, Ekker SC, Krageschmidt D, Neff KL, et al. Predictors of indoor absolute humidity and estimated effects on influenza virus survival in grade schools. BMC Infect Dis 2013;13:71. 9. Chan PK, Mok HY, Lee TC, Chu IM, Lam WY, Sung JJ. Seasonal influenza activity in Hong Kong and its association with meteorological variations. J Med Virol 2009;81:1797–806. 10. Yu H, Wu JT, Cowling BJ, Liao Q, Fang VJ, Zhou S, et al. Effect of closure of live poultry markets on poultry-to-person transmission of avian influenza A H7N9 virus: an ecological study. Lancet 2014;383:541–8. 11. Xu J, Lu S, Wang H, Chen C. Reducing exposure to avian influenza H7N9. Lancet 2013;381:1815–6. 12. Liu B, Havers F, Chen E, Yuan Z, Yuan H, Ou J, et al. Risk factors for influenza A(H7N9) disease—China, 2013. Clin Infect Dis 2014;59:787–94. 13. Cowling BJ, Jin L, Lau EH, Liao Q, Wu P, Jiang H, et al. Comparative epidemiology of human infections with avian influenza A H7N9 and H5N1 viruses in China: a population-based study of laboratory-confirmed cases. Lancet 2013;382:129–37.
