http://orcid.org/0000-0003-3112-6370
Figures
Abstract
Influenza is associated with highly contagious respiratory infections. Previous research has found that influenza transmission is often associated with climate variables especially in temperate regions. This study was performed in order to fill the gap of knowledge regarding the relationship between incidence of influenza and three meteorological parameters (temperature, rainfall and humidity) in a tropical setting. This was a retrospective study performed in Yaoundé-Cameroon from January 2009 to November 2015. Weekly proportions of confirmed influenza cases from five sentinel sites were considered as dependent variables, whereas weekly values of mean temperature, average relative humidity and accumulated rainfall were considered as independent variables. A univariate linear regression model was used in determining associations between influenza activity and weather covariates. A time-series method was used to predict on future values of influenza activity. The data was divided into 2 parts; the first 71 months were used to calibrate the model, and the last 12 months to test for prediction. Overall, there were 1173 confirmed infections with influenza virus. Linear regression analysis showed that there was no statistically significant association observed between influenza activity and weather variables. Very weak relationships (-0.1 < r < 0.1) were observed. Three prediction models were obtained for the different viral types (overall positive, Influenza A and Influenza B). Model 1 (overall influenza) and model 2 (influenza A) fitted well during the estimation period; however, they did not succeed to make good forecasts for predictions. Accumulated rainfall was the only external covariate that enabled good fit of both models. Based on the stationary R2, 29.5% and 41.1% of the variation in the series can be explained by model 1 and 2, respectively. This study laid more emphasis on the fact that influenza in Cameroon is characterized by year-round activity. The meteorological variables selected in this study did not enable good forecast of future influenza activity and certainly acted as proxies to other factors not considered, such as, UV radiation, absolute humidity, air quality and wind.
Citation: Monamele GC, Vernet M-A, Nsaibirni RFJ, Bigna JJR, Kenmoe S, Njankouo MR, et al. (2017) Associations between meteorological parameters and influenza activity in a subtropical country: Case of five sentinel sites in Yaoundé-Cameroon. PLoS ONE 12(10): e0186914. https://doi.org/10.1371/journal.pone.0186914
Editor: Jeffrey Shaman, Columbia University, UNITED STATES
Received: May 5, 2017; Accepted: October 10, 2017; Published: October 31, 2017
Copyright: © 2017 Monamele et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This work was supported by the U.S. Department of Health and Human Services (DHHS) grant number 6 DESP060001-01-01 via the International Network of Pasteur Institutes to Richard Njouom.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Influenza is associated with highly contagious respiratory illness throughout the world. Illnesses can result in hospitalization and death mainly among high-risk groups (the very young, elderly or chronically ill). Annually, epidemics due to influenza result in approximately 3 to 5 million cases of severe illness, and about 250 000 to 500 000 deaths [1]. The resulting economic impact is large and includes both direct and indirect cost [2–4].
There are different patterns in the circulation of human influenza in tropical and subtropical regions [5]. All countries with winter seasons have a clear peak in influenza activity during the winter months [6]. However, in countries with tropical or subtropical climates, there is no distinct profile in the circulation of influenza, year-round activity can be observed with higher frequency during the rainy seasons [6]. Despite recognition of this phenomenon, mechanisms driving influenza seasonality are not well understood.
Several factors have been found to favour transmission of influenza virus transmission. These factors act through different mechanisms by providing appropriate conditions for: influenza virus survival and increased contact with infected individuals especially in circumstances where there is limited immunization in the population to provide herd immunity [7–9]. In this regard, the likelihood of an infection occurring can be affected by meteorological and environmental conditions [9]. Several laboratory-based studies regarding the role of environmental factors on the spread of influenza have been performed particularly associated to humidity and temperature [8, 10]. When the air becomes dry, ambient humidity decreases, causing respiratory particles to evaporate moderately. The lighter drops of respiratory particles can remain in the air for prolonged periods of time [8, 11]. With respect to temperature, Lowen et al. showed that transmission of influenza virus decreases with increasing temperature, then reaches a point of complete halt at 30°C [10]. Rainfall is another meteorological parameter that has been thought to be implicated in influenza transmission [12]. The probable hypothesis is that rainfall promotes indoor crowding and human-human contact leading to more contact or aerosol transmission [13].
There is paucity of data on the association between influenza and weather parameters in the tropics [14, 15], and to date, there has been no data reporting associations between meteorological variables and influenza activity in Cameroon. In 2010, Tchidjou et al. reported that rain, high relative humidity and low temperature were associated with increased frequency of acute respiratory infections in Cameroon [16]. Another study performed in Cameroon laid emphasis on the seasonality of influenza to be characterized by year-round circulation and semi-annual peaks in incidence [17]. This study was performed in order to fill the gap of knowledge regarding influenza seasonality. Furthermore, it is essential to understand the role of meteorological factors associated with influenza seasonality because it can help in making projections on the dynamics of infection with fluctuations in climatic factors.
Materials and methods
Study population and design
Located at the bottom of the Gulf of Guinea, Cameroon has 21,657,488 inhabitants and an area of 475,442 km2. Its capital, Yaoundé, has about 4 million inhabitants [18]. The climatic condition in Cameroon alters with altitude and locations. Yaoundé, which is located in the Centre region is endowed with 3° 52' N Latitude and 11° 32' E Longitude. The central region has an equatorial climate with moderate rainfall between March and June and intense precipitation between September and November.
This was a retrospective study performed in Yaoundé from January 2009 to November 2015. The study population comprised individuals with Influenza-like illness (ILI) or severe acute respiratory infection (SARI) based on the WHO criteria present in five sentinel clinics in Yaoundé. Nasopharyngeal swab was collected from all individuals that respected the WHO case definition for ILI and SARI throughout the study period. An ILI case was defined as an outpatient presenting sudden onset of fever (temperature >38°C), and cough or sore throat, with the onset of symptoms within the previous 5 days. For SARI, the case definition was similar to that of ILI and in addition required hospital admission. Of the five sentinel sites considered, one received both inpatients and outpatients (“Centre Hospitalier d’Essos”) whereas the four others received only outpatients (“Centre Médico-Social Ambassade de France”, “Centre Médical d’Arrondissement de Nkomo”, “Centre d'Animation Sociale et Sanitaire de Nkolndongo”, “Centre Médical d’Etoudi”). These sites were chosen based on their location (Yaoundé), the completeness of their influenza time series data, and the availability of meteorological parameters for this region. Fig 1 shows the location of the study sites.
1 = “Centre Médico-Social Ambassade de France” (3.85N, 11.51E), 2 = “Centre Hospitalier d’Essos” (3.87N, 11.53E), 3 = “Centre d'Animation Sociale et Sanitaire de Nkolndongo” (3.86N, 11.53E), 4 = “Centre Médical d’Etoudi” (3.92N, 11.52E), 5 = “Centre Médical d’Arrondissement de Nkomo” (3.84N, 11.55E).
Virological data
Naso-pharyngeal swabs were tested for influenza by real-time reverse transcription polymerase chain reaction (rRT-PCR) using the CDC protocol prior to RNA extraction with the QIAamp Viral RNA Mini Kit (QIAGEN, Hilden, Germany). RNA extraction was performed following the manufacturer’s instructions. These analyses were performed at the National Influenza Centre of Cameroon located in Centre Pasteur of Cameroon which performs influenza surveillance since November 2007 and reports to the World Health Organization (WHO) Global Influenza Surveillance and Response System. All samples were selected in the flu database based on the date of onset of symptoms from January 2009 to November 2015.
Meteorological data
Meteorological data considered for this study were: relative humidity, temperature and precipitation. These parameters were chosen because they have been reported in several studies to be implicated in influenza transmission [9, 13, 19]. This data was retrieved from the Department of Meteorology at the Ministry of Transport in Cameroon from two different sources. From 2009–2013, meteorological data were obtained from a ground station located at latitude: 3°50'N, longitude: 11°31'E and 760M altitude. Data from 2014 to 2015 were obtained from an automatic station located at latitude 3°8421'N, longitude 11°5242'E and 754M altitude, which records data at a frequency of 15 minutes. The automatic station gives more reliable results but was installed only in 2014. Data obtained from these two sources are specific to Yaoundé, where the five sentinel sites considered in this study are located.
Analysis
Influenza data was retrieved from the flu database (Access) as an Excel sheet. Individual data for influenza diagnosis was aggregated into daily counts then merged with meteorological data using R version 3.2. This data was further aggregated by week. Weekly proportions of positive laboratory-confirmed cases of influenza A and/or B were considered as dependent variables. As independent variables, weekly values of mean temperature, average relative humidity, and accumulated rainfall were included in the analysis.
A univariate linear regression model was used in determining associations between the different influenza groups (overall influenza, influenza A and influenza B) and weather covariates.
A time series method was as well applied to investigate associations between influenza and weather variability. The time-series data was divided into 2 parts; the first 71 months of data (from January 2009 to November 2014) were used to calibrate the time-series model, and the last 12 months of data (December 2014 to November 2015) were used to test the model prediction. In this study, we used the Statistical Package for Social Sciences (SPSS) Expert Modeller [20, 21], which makes use of the autoregressive integrated moving average model, ARIMA (p,d,q), where p represents the autoregressive (AR) order, d stands for differencing order and q for moving average (MA) order. Autoregressive orders specify which previous values from the series are used to predict current values. Differencing is necessary when trends are present and is used to remove their effect. Moving average orders specify how deviations from the series mean for previous values are used to predict current values. The Ljung–Box (modified Box–Pierce) test was used to determine if the model was correctly specified.
The ARIMA model generally can be written as:
Where, Zt is the influenza count at week t, Zt-p is the count at previous p week (AR), μ is the intercept, Xi,t-n is environmental variable i with lag n and εt-q is the process error at lag q (MA). The parameters Øp, βi, n, ϑq are estimated using least square method during fitting process [22]. Statistical analysis was performed with SPSS version 22.0.
Results
During the study period, there were 5216 naso-pharyngeal swabs collected from patients with ILI and SARI. Data regarding gender of the study population was available for 3781 participants whereas information on age distribution of the study population was available for 5136 individuals. Table 1 gives a summary of the socio-demographics of the study population.
Among the collected specimens, 1161 (22.2%) had confirmed infections with influenza A and B virus, that is, 767 (14.7%) cases of infection with influenza A virus, 382 (7.3%) with influenza B virus and 12 (0.2%) individuals with both infections. S1 Table gives detailed information on the raw data obtained during the study. Overall, there were 362 observations with weekly data for number of influenza cases and climatic data. Table 2 summarizes the characteristics of these variables during the study period.
Fig 2 shows the trends in influenza virus circulation and meteorological variables during the period of study. Throughout the study period, influenza activity displayed year round activity. The distribution of influenza A had a similar profile to that observed for overall influenza cases. Visually, there was no clear trend observed in the association between influenza cases and weather variables.
A corresponds to mean temperature, B to accumulated rainfall, C corresponds to average relative humidity, D to overall population, E to proportion of influenza cases, F to proportion of influenza A cases, and G to proportion of influenza B cases.
There was no statistically significant association (P>0.05) found between influenza virus and weather variables; very weak relationships (-0.1 < r < 0.1) were observed with univariate linear regression analysis (Fig 3). Less than 1% of variations in the meteorological factors could explain the influenza activity (r2 < 0.01)
An Expert Modeller was used to predict on future influenza activity. This Modeller automatically found the best-fitting model for each dependent series. The Expert Modeller selected, for inclusion in ARIMA models, independent variables that have a statistically significant relationship with the dependent series. Out of the three candidate predictors fitted in the model, the Expert Modeller identified one meteorological variable that was useful for forecasting for each dependent variable. Table 3 summarizes the model performances. For overall influenza cases, the ARIMA (3,0,1) was the selected model and gave significant forecast with accumulate rainfall at lag 0 to lag 2. ARIMA model (2,0,0) was observed for influenza A with similar lags while no predictor was observed for influenza B. Accumulated rainfall was the only covariate that enabled good fit of models. The Ljung–Box (modified Box–Pierce) test indicated that model 1 for overall influenza cases was correctly specified (p-value > 0.05) and that there were no outliers with the expert modeller. Meanwhile, a p-value of below 0.05 for model 2 and 3 indicate that the residuals are not independent. Based on the stationary R2, 29.5% and 41.1% of the variation in the series can be explained by model 1 and 2, respectively.
The graphical representation of the different models is shown in Fig 4. Model 1 (overall influenza cases) and model 2 (Influenza A cases) fitted well during the estimation period (January 2009—November 2014); however, they did not succeed to make good forecast for predictions. The estimation and prediction model for influenza A performed better than that for influenza B.
Red curves are the observed data; blue curves are the modelled results; and bold-blue curves are the prospectively estimated influenza activity using actual meteorological data and ARIMA models trained with influenza data from previous years.
Discussion
This study aimed to determine the relationship between influenza cases and weather variability. We found that influenza in Yaoundé is characterized by a year round activity as described by Heraud et al. in 2008 and 2009 respectively, who found influenza to circulate year-round or to have a seasonal peak (May through November) in Yaoundé-Cameroon [17]. These results are however different from that reported by Caini et al. who observed a heterogeneous pattern in several tropical countries [5].
The inverse relationship between temperature and influenza is a fact that has been long confirmed particularly in countries with temperate climates, where influenza peaks during the winter seasons [15]. Previous animal studies performed in guinea pigs showed that influenza virus is more stable in the cold with robust transmission at low temperatures due to increased virus half-life [23]. Though Cameroon is a subtropical country and has little variations in temperature across the year, a negative correlation was also observed in this study although this relationship was not significant.
Generally, data from different geographic locations show that rainfall affects influenza virus infection with increased activity depending on the region [19]. We found that the amount of rainfall is weakly related to the influenza activity. This is contradictory to reports by other studies, for example, in studies involving regions of low latitudes [15], Ivory coast [24], Thailand [25], Honduras and Nicaragua [13] where a significant correlation was observed with the number of influenza cases. A previous study reported a negative correlation between the number of influenza cases and rainfall though it was in cases of avian influenza [26]. Murray et al. reported that influenza A/H5N1 infections in Egypt were negatively correlated with precipitation between 2006 and 2008 [26]. The proportional correlation with rainfall can be supported by the hypothesis that rainfalls can be sufficiently intense to favour indoor gathering, thereby increasing aerosol and direct contact transmissions [24].
Relative humidity negatively correlated with the weekly influenza activity though the correlation was not significant and relatively low. Similarly, N'gattia et al. reported poor correlation between humidity and weekly influenza activity in Ivory Coast. It is worth noting that they observed moderately high ranges of relative humidity between 54–93% which is similar to that obtained in this study [24]. These results corroborate with past findings which reported that influenza virus circulates efficiently under dry and cold conditions [9, 17, 23].Several reports from studies performed in the tropics have however demonstrated a proportional association between influenza and humidity [14, 15, 27] which is contradictory to our findings. This can be explained by the fact that the primary mode of transmission during these humid conditions is probably by droplet and contact modes as opposed to aerosol transmission which is favoured by dry weather conditions [8].
Through the use of ARIMA models, we explored the relationship between meteorological variables and influenza activity and made forecasts with significant predictors. The Expert Modeller used in this analysis gave three models for the different viral groups (overall influenza, influenza A and influenza B). The ARIMA models did not enable predictions that were in agreement with the observed data though all models performed relatively well in the estimation period. Accumulated rainfall was the only external covariate that enabled good fit of models. Moreover, some synchronies were noted between rainfall and influenza activity as shown in Fig 2. Based on the stationary R2, 29.5% and 41.1% of the variation in the series could be explained by model 1 and 2, respectively. These results are the first of a kind that did not enable predictions to be made despite good fits of model [13, 25]. Association between the three weather variables and influenza may not involve a direct causal relationship and rather acted as proxies to other factors not considered in the model, such as, UV radiation, absolute humidity, air quality and wind/stability [8]. Moreover the irregular pattern observed in influenza distribution might have influenced the performance of the model. This raises more concern on the difficulty of performing timely vaccination campaigns in tropical settings. The mechanism driving influenza seasonality still remains uncertain.
Contrarily to other studies performed in developed regions, where lowering of the temperature and humidity by the air conditioning system was a commonly raised bias [13, 28], it was not an issue here. Outdoor temperature, and humidity measurements used in this study do not vary much from indoor measurements in our low income settings, where there is little availability to even electricity. Moreover, the time frame for this study was relatively long and enabled an evaluation of the different variations in influenza cases and weather parameters over time.
Although this study’s focus was to assess impacts of weather variability on influenza, there might be other environmental factors that could be associated and were not considered. For instance, absolute humidity and solar radiation have been indicated to be key environmental variables in influenza virus transmission [7, 8]. A major drawback to this study was the lack of rigorous statistical analysis such as deviance analysis that would have enabled a better understanding of the role of each weather variable in explaining the regression model. Moreover, the expert modeller used in this study is less efficient than the commonly used ARIMA model. Another drawback to this study was the lack of inclusion of other determinants capable of playing a role in the incidence of influenza such as: socio-economic elements, living environment, viral stability, host behaviour and biology.
More so, meteorological parameters were collected at a single point within the region. These point-parameters may therefore not reflect the reality of the climate conditions present in all the study sites due to possible variations in these factors from one location to another.
The fact that this study was performed in only one geographic location did not enable comparisons to be made with other regions within the same country in order to evaluate variations obtained with changes in climate and socio-economic factors.
Conclusion
This study laid more emphasis on the fact that influenza in Cameroon is characterized by year-round activity. The exploited meteorological parameters did not give significant forecasts on expected influenza cases during the prediction period probably due to several other environmental factors not included in the model that might have been implicated in the circulation of this virus as well as intrinsic factors specific to the individual or socio-economic elements. Future studies taking into consideration these factors are warranted.
Supporting information
S1 Table. Complete study data.
MT = Mean temperature, ARF = Accumulated Rainfall, ARH = Average Relative Humidity, Inf. = Influenza.
https://doi.org/10.1371/journal.pone.0186914.s001
(PDF)
Acknowledgments
We would like to acknowledge the focal points of the influenza sentinel sites in Yaounde for identification of cases and data collection. We are thankful to Matthieu Schoenhals for editing this manuscript. This work was supported by the U.S. Department of Health and Human Services (DHHS) grant number 6 DESP060001-01-01 via the International Network of Pasteur Institutes.
References
- 1.
WHO. Influenza (Seasonal). Fact sheet N°211 [Internet]. 2014 14 March 2016. Available from: http://www.who.int/mediacentre/factsheets/fs211/en/.
- 2. Molinari NA, Ortega-Sanchez IR, Messonnier ML, Thompson WW, Wortley PM, Weintraub E, et al. The annual impact of seasonal influenza in the US: measuring disease burden and costs. Vaccine. 2007;25(27):5086–96. pmid:17544181
- 3. Fischer WA, Gong M, Bhagwanjee S, Sevransky J. Global burden of Influenza: Contributions from Resource Limited and Low-Income Settings. Global heart. 2014;9(3):325–36. pmid:25667184
- 4. Emukule GO, Paget J, van der Velden K, Mott JA. Influenza-Associated Disease Burden in Kenya: A Systematic Review of Literature. PLoS One. 2015;10(9):e0138708. pmid:26398196
- 5. Caini S, Andrade W, Badur S, Balmaseda A, Barakat A, Bella A, et al. Temporal Patterns of Influenza A and B in Tropical and Temperate Countries: What Are the Lessons for Influenza Vaccination? PLoS ONE. 2016;11(3):e0152310. pmid:27031105
- 6.
WHO. Influenza seasonality: timing and formulation of vaccines. Bulletin of the World Health Organization [Internet]. 2014 14 March 2016; 92:[311 p.]. Available from: http://www.who.int/bulletin/volumes/92/5/14-139428/en/.
- 7. Tamerius J, Nelson MI, Zhou SZ, Viboud C, Miller MA, Alonso WJ. Global Influenza Seasonality: Reconciling Patterns across Temperate and Tropical Regions. Environmental Health Perspectives. 2011;119(4):439–45. pmid:21097384
- 8. Fuhrmann C. The Effects of Weather and Climate on the Seasonality of Influenza: What We Know and What We Need to Know. Geography Compass. 2010;4/7:718–30.
- 9. Soebiyanto RP, Gross D, Jorgensen P, Buda S, Bromberg M, Kaufman Z, et al. Associations between Meteorological Parameters and Influenza Activity in Berlin (Germany), Ljubljana (Slovenia), Castile and Leon (Spain) and Israeli Districts. PLoS One. 2015;10(8):e0134701. pmid:26309214
- 10. Lowen AC, Mubareka S, Steel J, Palese P. Influenza Virus Transmission Is Dependent on Relative Humidity and Temperature. PLoS Pathogens. 2007;3(10):e151.
- 11. Weber TP, Stilianakis NI. Inactivation of influenza A viruses in the environment and modes of transmission: a critical review. J Infect. 2008;57(5):361–73. pmid:18848358
- 12. Alonso WJ, Guillebaud J, Viboud C, Razanajatovo NH, Orelle A, Zhou SZ, et al. Influenza seasonality in Madagascar: the mysterious African free-runner. Influenza and other respiratory viruses. 2015;9(3):101–9. pmid:25711873
- 13. Soebiyanto RP, Clara WA, Jara J, Balmaseda A, Lara J, Lopez Moya M, et al. Associations between seasonal influenza and meteorological parameters in Costa Rica, Honduras and Nicaragua. Geospatial health. 2015;10(2):372. pmid:26618318
- 14. Imai C, Brooks WA, Chung Y, Goswami D, Anjali BA, Dewan A, et al. Tropical influenza and weather variability among children in an urban low-income population in Bangladesh. Global health action. 2014;7:24413.
- 15. 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;9(3):e1003194. pmid:23505366
- 16. Tchidjou HK, Vescio F, Boros S, Guemkam G, Minka E, Lobe M, et al. Seasonal pattern of hospitalization from acute respiratory infections in Yaounde, Cameroon. Journal of tropical pediatrics. 2010;56(5):317–20. pmid:20080936
- 17. Heraud JM, Njouom R, Rousset D, Kadjo H, Caro V, Ndiaye MN, et al. Spatiotemporal circulation of influenza viruses in 5 African countries during 2008–2009: a collaborative study of the Institut Pasteur International Network. The Journal of infectious diseases. 2012;206 Suppl 1:S5–13.
- 18.
BUCREP. Rapport National sur l'état de la Population. Regards sur le genre au Cameroun [Internet]. 2014 14 March 2016. Available from: http://www.bucrep.cm/index.php/fr/ressources-et-documentations/telechargement/category/60-rnep-2014?download=95:rapport-national-sur-l-etat-de-population-2014.
- 19. Pica N, Bouvier NM. Environmental Factors Affecting the Transmission of Respiratory Viruses. Curr Opin Virol. 2012;2(1):90–5. pmid:22440971
- 20.
IBM. IBM SPSS Forecasting 22. http://www.sussex.ac.uk/its/pdfs/SPSS_Forecasting_22.pdf.
- 21. Khaliq A, Batool SA, Chaudhry MN. Seasonality and trend analysis of tuberculosis in Lahore, Pakistan from 2006 to 2013. Journal of epidemiology and global health. 2015;5(4):397–403. pmid:26318884
- 22. Soebiyanto RP, Kiang RK. Modeling Influenza Transmission Using Environmental Parameters. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Science. 2010:330–4.
- 23. Lowen AC, Steel J. Roles of Humidity and Temperature in Shaping Influenza Seasonality. Journal of virology. 2014;88(14):7692–5. pmid:24789791
- 24. N'Gattia A K, Coulibaly D, Nzussouo NT, Kadjo HA, Cherif D, Traore Y, et al. Effects of climatological parameters in modeling and forecasting seasonal influenza transmission in Abidjan, Cote d'Ivoire. BMC public health. 2016;16:972. pmid:27624302
- 25. Chadsuthi S, Iamsirithaworn S, Triampo W, Modchang C. Modeling Seasonal Influenza Transmission and Its Association with Climate Factors in Thailand Using Time-Series and ARIMAX Analyses. Computational and mathematical methods in medicine. 2015;2015:436495. pmid:26664492
- 26. Murray EJ, Morse SS. Seasonal oscillation of human infection with influenza A/H5N1 in Egypt and Indonesia. PLoS One. 2011;6(9):e24042. pmid:21909409
- 27. Soebiyanto RP, Clara W, Jara J, Castillo L, Sorto OR, Marinero S, et al. The role of temperature and humidity on seasonal influenza in tropical areas: Guatemala, El Salvador and Panama, 2008–2013. PLoS One. 2014;9(6):e100659. pmid:24956184
- 28. 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(10):1797–806. pmid:19697414