Do et al.

BMC Public Health 2014, 14:1078

http://www.biomedcentral.com/1471-2458/14/1078

RESEARCH ARTICLE Open Access

Climatic-driven seasonality of emerging dengue

fever in Hanoi, Vietnam
Thi Thanh Toan Do1*, Pim Martens2, Ngoc Hoat Luu1, Pamela Wright3 and Marc Choisy4,5

Abstract
Background: Dengue fever (DF) has been emerging in Hanoi over the last decade. Both DF epidemiology and
climate in Hanoi are strongly seasonal. This study aims at characterizing the seasonality of DF in Hanoi and its links
to climatic variables as DF incidence increases from year to year.
Methods: Clinical suspected cases of DF from the 14 central districts of Hanoi were obtained from the Ministry of
Health over a 8-year period (2002–2009). Wavelet decompositions were used to characterize the main periodic
cycles of DF and climatic variables as well as the mean phase angles of these cycles. Cross-wavelet spectra between
DF and each climatic variables were also computed. DF reproductive ratio was calculated from Soper’s formula and
smoothed to highlight both its long-term trend and seasonality.
Results: Temperature, rainfall, and vapor pressure show strong seasonality. DF and relative humidity show both
strong seasonality and a sub-annual periodicity. DF reproductive ratio is increasing through time and displays two
clear peaks per year, reflecting the sub-annual periodicity of DF incidence. Temperature, rainfall and vapor pressure
lead DF incidence by a lag of 8–10 weeks, constant through time. Relative humidity leads DF by a constant lag of
18 weeks for the annual cycle and a lag decreasing from 14 to 5 weeks for the sub-annual cycle.
Conclusion: Results are interpreted in terms of mosquito population dynamics and immunological interactions
between the different dengue serotypes in the human compartment. Given its important population size, its strong
seasonality and its dengue emergence, Hanoi offers an ideal natural experiment to test hypotheses on dengue
serotypes interactions, knowledge of prime importance for vaccine development.
Keywords: Dengue fever, Seasonality, Emergence, Climatic factors, Hanoi, Vietnam

Background understood. The virus is transmitted by bites of female

Dengue Fever (DF) has recently been recognized by WHO Aedes aegypti or albopictus mosquitoes in the inter-tropical
as the fastest spreading tropical disease across all conti- regions of the world. In absence of vaccine (under develop-
nents. Bhatt et al. [1] estimated that the global number of ment), mosquito control is the only available method of
new infections per year (390 millions, 95% confidence inter- prophylaxy.
val: 284–528) is largely underestimated: only 96 millions In Vietnam, dengue is recognized as a major cause of
(95% confidence interval: 67–136) being declared yearly. mortality and morbidity and ranks amongst the top ten
DF is one of the many symptoms (ranging from mild fever communicable diseases in terms of overall health burden
to hemorrhagic fever and shock syndrome) caused by one [2]. All four dengue virus serotypes have been found
of the four serotypes of dengue virus (Flaviridae family). circulating in Vietnam with the dominant one varying
Even though recovery from dengue confers life-long over time. Reports from the National Institute of
immunity against the infecting serotype, immunological Hygiene and Epidemiology show that DENV-1 and
interactions between the different serotypes are not fully DENV-2 have been the predominant circulating viruses
almost every year. DENV-3 emerged in the late 1990s
* Correspondence: [email protected] and was responsible for the large outbreak of 1998,
1
Biostatistics and Medical Informatics Department, Institute of Training for whereas DENV-4 was also detected between 1999 and
Preventive Medicine and Public Health, Hanoi Medical University, Hanoi,
Vietnam 2003 [3]. Dengue transmission occurs throughout the
Full list of author information is available at the end of the article

© 2014 Do et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain
Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article,
unless otherwise stated.
Do et al. BMC Public Health 2014, 14:1078 Page 2 of 10
http://www.biomedcentral.com/1471-2458/14/1078

year in Vietnam, with peaks in the numbers of cases (72% risk of dengue to be associated with high temperature, high
of total cases) reported between June and November [4]. relative humidity and rainfall, but inversely associated with
There are regional variations in the seasonality of dengue duration of sunshine [22]. Findings from other studies in
epidemiology in Vietnam. In the Northern and Central many other parts of the world also show climatic variables
Highland regions, dengue notifications are low during the to have an effect on dengue transmission. Studies in
winter time from December to March, while in southern Thailand [11], Barbados [12], Taiwan [16], Guangzhou,
regions, dengue transmission occurs throughout the year, China [18], the French West Indies [21] and in Colombia
even if it sharply increases during the rainy season from [26] showed a positive correlation between dengue inci-
July to September. Given that dengue is a vector-born dence and minimum and maximum temperatures, precipi-
disease and that vector population dynamics are strongly tation and relative and absolute humidities. However,
dependent on climatic factors, the diversity of climates in depending on the approach of analysis and the areas, these
Vietnam may explain the observed diversity of dengue correlations were more or less strong. In Barbados, the
epidemiological dynamics. strongest correlation was found at a lag of 6, 12 and
Hanoi, the capital of Vietnam, is located in the North 16 weeks for vapour pressure and minimum and maximum
of the country and known as a low transmission setting temperatures respectively, whereas in Taiwan the highest
of DF [5]. Hanoi experiences annual seasonal dengue correlations were found with maximum temperature at a
outbreaks with the pinnacle of epidemics usually falling in lag of 5 weeks and with total precipitation at a lag of
September/October and ending in November/December. 7 weeks. In Colombia, Eastin et al. [26]‘s results suggest that
Over the last decade, the number of DF cases has been DF cases increase 2 to 5 weeks after the daily temperature
increasing from year to year, reaching a peak in 2009. range remains for an extended period within the
According to the Ministry of Health’s statistics, the temperature range optimal for vector survival and disease
outbreak in Hanoi in 2009 is the most important outbreak transmission. Nagao et al. [11] in Thailand and Yi et al. [27]
of the last decade, with 384 notified cases per 100,000 in Guangdong, China, demonstrated that the distributions
individuals. Interestingly, 2009 was also the year El Niño of Aedes species and dengue cases were positively associ-
increased actively [6,7]. There are only few studies ated with high absolute humidity, which itself increases
published on dengue epidemiology in Hanoi that are with high temperature and rainfall. In San Juan, Puerto
based on the public health surveillance data routinely Rico, Schreiber [9] used a water budgeting technique and
collected through the Ministry of Health’s notifiable dis- showed that high levels of dengue are associated with
eases surveillance program. Toan et al. [8] show that there reduced actual evapotranspiration, minimum temperature
are spatio-temporal clusters of DF limited to a radius of and with high levels of precipitation. In Taiwan, using auto-
1,000 m and a duration of 29 days. This study also demon- regressive integrated moving average models, Wu et al.
strates that most of the DF cases occur between June and [20], found a negative association of dengue incidence with
November, during which the rainfall and temperatures are temperature and relative humidity. Finally, in the city of
highest. Cuong et al. [5] use wavelet analysis to relate Noumea (New Caledonia), Descloux et al. [24] recently
dengue incidence to climatic variables and suggest that all documented a high seasonality of dengue incidence, with
the tested local climatic variables (total rainfall; mean wind an epidemic peak (March-April) lagging the warmest
velocity; mean, maximum and minimum temperatures; temperature by 1 to 2 months and in phase with maximum
relative humidity) are significantly associated with dengue precipitations, relative humidity and entomological indices.
incidence around the annual periodicity: on average, In the present study, we consider vapor pressure and
dengue incidence follows the seasonal peak of rainfall and relative humidities, temperature and rainfall in order to
mean temperature with a lag of 1 to 2 months. identify which of these variables are most critical for the
Other studies have been carried out on the correlation onset of dengue epidemics. Compared to Cuong et al.
between climate and DF in other parts of Vietnam as well [5], who also investigated the links between dengue and
as in other parts of the world, using a wide spectrum of climatic variables in Hanoi, Vietnam, we here consider
mathematical and statistical modeling methods [6,9-25]. In vapor pressure in addition to relative humidity. Vapor
Vietnam, most of the studies have been carried out in the pressure is a measure of absolute humidity and this
south and the center of the country, and showed significant climatic variable is often neglected in the studies investi-
associations between climatic variables and dengue inci- gating the links between climate and disease transmis-
dence. A wavelet analysis of monthly dengue cases from sion, even though it has been proved to play a role more
the province of Binh Thuan has shown a non-stationary important than relative humidity for the transmission of
relationship between El Niño Southern Oscillation indice some diseases such as influenza (e.g. [28]). A second
and dengue incidence in the 2–3 year periodic band [6]. difference with Cuong et al. [5] is that we here work on
Meanwhile, a correlation study carried out on monthly weekly incidence aggregates instead of monthly aggre-
dengue cases from the province of Daklak has found the gates. With this finer temporal resolution we aim at
Do et al. BMC Public Health 2014, 14:1078 Page 3 of 10
http://www.biomedcentral.com/1471-2458/14/1078

investigating intra-annual patterns of seasonality. We first that in the above formula temperature T is expressed in °C
quantify the synchrony between weekly dengue incidence instead of K as in Shaman and Kohn [28].
and the four climatic variables between 2002 and 2009. Climatic variables were aggregated by week using sums
We then estimate the values of reproductive ratio of for rainfall and mean values for all the other variables.
dengue fever through time and characterize the trend and
seasonality of reproductive ratio. We finally discuss the Wavelet analysis
results in light of immunological and entomological fac- Epidemiological data can be substantially non-stationary
tors specific to dengue epidemiology in Hanoi, Vietnam. [31-33] as is the case for dengue in Hanoi where it is
emerging (see in particular the increase in mean and
Methods amplitude over time on Figure 1B). Here we performed
Setting: Hanoi (1,760 km2) is located in the Red River Delta, wavelet decomposition, a time-series statistical analysis
in the centre of North Vietnam (21o 2 ’ N, 105° 51 ’ E). The allowing to efficiently deal with non-stationary data. Spe-
city experiences the typical climate of northern Vietnam, cifically, we used the Morlet wavelet [34], classically used
where summers are hot and humid, and winters are, by in ecology with a non-dimensional frequency ω0 = 6. An
national standards, relatively cold and dry. advantage of using the Morlet wavelet is that it is a com-
plex wavelet, allowing to quantify the phase and thus
Data collection calculate time lags between different time series.
The data used in this study have been previously published Coherence based on wavelets allows to perform similar
by Cuong et al. [5] and Toan et al. [8]. Clinical suspected analysis as cross-correlation but for potentially non-
cases of DF in old Hanoi (14 districts of Hanoi before mer- stationary signals. Wavelet coherences were calculated to
ging with Ha Tay in 2008) are reported to the surveillance examine the association between two time series, both in
system of Hanoi Center for Preventive Health. The criteria time and frequency. Coherence spectra allow to investi-
for notification of DF disease are based on the guidelines gate whether different periodic modes of two time series
of the Ministry of Health (2006) on surveillance, diagnosis tend to oscillate simultaneously and, if yes, to identify the
and treatment of dengue, in which suspected dengue cases periodicity around which this association takes place.
are based on acute febrile illness (≥38°C) of 2–7 days Significance levels were calculated by a Chi-square test
duration with at least two of the following non-specific assuming that the wavelet coefficients are normally dis-
manifestations of dengue fever: headache, retro-orbital tributed as described in Torrence and Compo [35]. The
pain, myalgia, arthralgia, rash, hemorrhagic manifestations,
and leucopenia [29]. The data analyzed here include all

reproductive ratio
0.0 1.0 2.0
reported cases from January 2002 to December 2009, (A)
aggregated by week.
Daily weather data for Hanoi from 2002 to 2009 were
4.00 1.00 0.25 0.0 60 15 30

provided by the National Centre for Hydrometeorological

incidence

(B)
Forecasting. They include the records of mean, maximum,
and minimum temperatures (T, in°C), rainfall (in mm)
(C)
6
and relative humidity (in %).
period (year)

log2(power)
2

Vapor pressure
−2

We used vapor pressure (VP, in mb) as a proxy of absolute

−6

humidity. VP was calculated from relative humidity (RH,

in %) and temperature T using the Clausius–Clapeyron 2003 2004 2005 2006 2007 2008 2009
time (year)
formula [28,30]:
Figure 1 The reproductive ratio, time series and wavelet power
spectrum of DF in Hanoi (14 districts) from 2002 to 2009.
(A) The reproductive ratio was estimated from equation 2 (see text)
V P ¼ V P 0    and smoothed by lowest regressions with smoothing factors equal
L 1 1 RH
 exp −T 0 − : ð1Þ to 0.05 (blue) and 0.90 (red). The shaded areas around the lines
Rv T 0 T þ T 0 100 represent the 95% confidence intervals calculated assuming a
normal distribution of errors. (B) Time series of the square-root
transformed weekly DF incidence. (C) Wavelet power spectrum of
where L = 2,257 J/g is the latent heat of evaporation the square-root transformed weekly DF incidence. The black contour
for water, Rv = 416.5 J/(kg K) is the gas constant for lines show the regions of power significant at the alpha-risk of 0.05.
water vapor, T0 = 273.15 K and VP0 = 6.11 mb is the The paled region of the spectrum delineates the cone of influence
due to the zero-padding of the time series. The power increases
vapor pressure at which water would change phase be-
from dark blue to dark red.
tween vapor and liquid if the temperature was T0. Note
Do et al. BMC Public Health 2014, 14:1078 Page 4 of 10
http://www.biomedcentral.com/1471-2458/14/1078

detailed theory for wavelet analysis has been described reproductive ratio (susceptible depletion and external
elsewhere [4]. Before wavelet decomposition, time series forcing), we assume that susceptible depletion is negli-
were square-root transformed in order to mitigate the gible before any external forcing such as climatic drivers
weights of high values. They were also zero-padded to the on the mosquito population dynamics.
next power of 2 of their length (i.e. 512), in order to Time series of the calculated seasonal factor k were
minimize edge effects [35]. smoothed by lowess regression with smoothing factors
equal to 0.05 and 0.90 in order to reveal its seasonality
Reproductive ratio and its long-term trend respectively. Confidence intervals
The reproductive ratio R is the expected number of in- were calculated by assuming normal distribution of errors.
fections caused by one infected individual. It is maximal Given the uncertainty on the infection generation length
at the start of an epidemic when the host population is and the assumption made on the number of susceptibles,
fully susceptible and then decreases over the course of estimates of the reproductive ratio will be treated with
the epidemics. It reaches the equilibrium value of 1 at caution: only their trend and seasonality will be inter-
epidemic peak and decreases below 1 after it. The initial preted, not their absolute values, on which we will have
and maximum value of the reproductive ratio is called limited confidence.
the basic reproductive ratio R0 and is the classical All analyses were conducted in R (R Core Team [39])
epidemiological statistics as its value relative to 1 and wavelet analyses were performed using the “biwavelet”
informs about the potential for an epidemic to occur. R package [40].
We followed Soper [36] as reported in Keeling and
Rohani [37] and approximated the reproductive ratio R Results
by: Dengue incidence and its reproductive ratio in Hanoi
  From 2002 to 2009, 23,195 DF cases were reported in
C tþ1 α Hanoi with the average annual incidence rate of 69.22/
R≈ ð2Þ
Ct 100,000. Overall, the incidence of DF increased over the
8 years of the study, with a sharp increase during the
where Ct+1 and Ct are the numbers of cases at times t + 1 period 2005–2009. This period of increasing incidence is
and t respectively and α is a parameter that expresses the visible on the wavelet spectrum of Figure 1C. The highest
infection generation length (i.e. the sum of the infectious peak of 3,697 cases was recorded in September 2009. Over
and latent periods) in the same units as the data time steps the 5 years from 2005 to 2009, annual (1 year) and sub-
(here 1 week). The infection generation length for dengue annual (6 months) periodicities were significant. These two
being approximately 2 weeks [38], we set this parameter α periodicities correspond to a slow increase of DF incidence
to the value of 2. The other part of Soper [36]’s formula ex- from the beginning of the year to weeks 22–24 (June),
presses the reproductive ratio as a function of the number followed by a rapid increase of incidence until weeks
of susceptibles in the population: 44–46 (November) which ends by a sharp decrease in inci-
X tþ1 dence at the end of the rainy season.
R≈k ð3Þ Figure 1A shows a long-term increase in the repro-
X
ductive ratio (orange curve) as well as a non-stationary
where X* and Xt+1 are the numbers of susceptibles be- sub-annual periodicity with two peaks of this reproduct-
fore the epidemics and at time t + 1, respectively and k is ive ratio per year (blue curve). These peaks are of
a parameter reflecting some potential external forcings roughly equal magnitude from 2005 to 2008, the second
(such as climatic ones). This latter part of the Soper peak is substantially higher than the first one in 2004
[36]‘s equation shows that variations in the reproductive and 2009, and the number of cases are too low before
ratio are due either to variations in the number of sus- 2004 for any clear pattern to be visible (large confidence
ceptibles in the population, or to some external forcings intervals on Figure 1B).
directly affecting the transmissibility of the disease (in
our case climatic factors acting on the vector population Meteorological variables in Hanoi
dynamics and density). Given that dengue is emerging in The wavelet power spectra of temperature, precipitation,
Hanoi and that the basic reproductive ratio of dengue is and humidities (relative humidity and vapor pressure) in
generally low [38], we expect the depletion of susceptible Hanoi during the study period are shown in Figure 2.
in the population to be very slow. We expect it to be Temperature, precipitation and vapor pressure in Hanoi
even slower given the fact that dengue can actually be show significant annual periodicities that are constant
caused by four different serotypes with no permanent through time, whereas relative humidity shows both
cross-immunity between them. Hence, among the two annual and sub-annual periodicities, as observed on the
above-cited factors that can affect the seasonality of the DF incidence time series (Figure 1B).
Do et al. BMC Public Health 2014, 14:1078 Page 5 of 10
http://www.biomedcentral.com/1471-2458/14/1078

of 576.6 mm brought by the typhoon Maysak (Center for

mean temperature (°C)

Excellence in Disaster Management and Humanitarian

Assistance, [41]). The relative humidity in Hanoi is quite
10 20 30

high (78.62 ± 4.28%), and is usually higher in February and

March (cool but very rainy: 84.12 ± 2.56), and August and
September (drier but very hot: 79.81 ± 3.54), than the rest

period (year)
of the year (76.80 ± 3.99).

4.00 0.25
Coherences between meteorological variables and DF
rainfall (mm)

incidence in Hanoi
0 300

Results of wavelet coherences between DF incidence and

climate variables are shown in Figure 4. Significant

period (year)
coherences were observed between DF incidence and

4.00 0.25
temperature, precipitation and humidity for the annual
relative humidity (%) absol. humidity (mg/L)

periodicity from approximately 2005 to 2009 (Figure 4A-D).

Moreover a weaker, but still significant association between
10 20 30

relative humidity and DF incidence was also seen for the

sub-annual periodicity from 2006 to 2009 (Figure 4D).
In analyzing the phase difference at the annual cycle
period (year)

between DF incidence and the climatic variables, we

4.00 0.25

found that dengue incidence was consistently trailing

temperature, rainfall, vapor pressure and relative humid-
ities with a delay of 9.37 ± 0.02 (standard error), 8.71 ±
60 80

0.02, 10.29 ± 0.03, and 18.05 ± 0.24 weeks respectively

(see Figure 4). Interestingly, when looking at the statis-
tical association between dengue incidence and relative
period (year)
4.00 0.25

humidity for the sub-annual cycle, it appears that the

time delay of dengue incidence compared to relative hu-
midity decreases from 14.30 to 5.27 weeks over the
2003 2005 2007 2009 8 years of the study.
time (year)
Figure 2 Time series and wavelet power spectra of mean Discussion and conclusion
temperature, cumulative rainfall and mean absolute and Using monthly aggregated data, Cuong et al. [5] showed
relative humidities in Hanoi, from 2002 to 2009. The black
a clear annual cycle for dengue transmission in Hanoi
contour lines show the regions of power significant at the alpha-risk
of 0.05. The paled region of the spectrum delineates the cone of from 1998 to 2009. In the present study, using weekly
influence due to the zero-padding of the time series. The power data allowed us to further characterize a sub-annual peri-
increases from dark blue to dark red. odicity, in addition to the annual one. This sub-annual
periodicity is reflected in the DF incidence time series, by
a slow increase of incidence from the beginning of the
The cross-correlation coefficients among four climatic year to the weeks 22–24 (June), followed by a faster in-
variables in Hanoi are presented in Figure 3. It shows that crease of incidence until weeks 44–46 (November), which
mean temperature, rainfall and vapor pressure are much ends by a sharp decrease of incidence at the end of the
correlated and in phase, whereas there is a significant lag rainy season. A potential drawback of working on weekly
between these three variables and relative humidity. instead of monthly data is that it decreases the incidence
Over the 8 years of the study, the average temperature values and thus increases the noise. However, given that
in Hanoi was found to be lowest in January and February we can still detect clear periodicities in our wavelet spec-
(16.07 ± 1.65°C standard deviation) and highest in June tra, this does not seem to affect our analysis too much.
and July (29.80 ± 0.47°C). Over the studied period, the When characterizing the reproductive ratio throughout
coolest temperature was observed in week 5 (February) of the studied period, it displays, most of the years, two peaks
2008 and the warmest was observed in week 27 (July) of per year, which is in accordance with the sub-annual peri-
2009 (10.11°C and 32.84°C respectively). July to September odicity of DF incidence. Among the climatic variables that
was the time of the year receiving the majority of the an- we investigated (temperature, rainfall, relative humidities
nual rainfall (271.9 ± 111.9 mm in average over one week). and vapor pressure), all of them expectedly displayed
Week 44 (November) in 2008 recorded a rainfall extreme strong annual periodicities with temperature, rainfall and
Do et al. BMC Public Health 2014, 14:1078 Page 6 of 10
http://www.biomedcentral.com/1471-2458/14/1078

mean temperature
cross−correlation coefficient
0.0 0.2 0.4

rainfall
−0.4

−20 −10 0 10 20 1.0 rainfall

cross−correlation coefficient
0.0 0.2 0.4

vapor pressure
0.5
0.0
−0.5
−0.4

−20 −10 0 10 20 −20 −10 0 10 20

vapor pressure
cross−correlation coefficient

0.3

relative humidity
0.3

0.2

0.1
0.1

0.0

−0.1
−0.3 −0.1

−0.2

−0.3

−20 −10 0 10 20 −20 −10 0 10 20 −20 −10 0 10 20

lag (bi−weeks) lag (bi−weeks) lag (bi−weeks)
Figure 3 Cross-correlation between the 4 climatic variables: mean temperature, rainfall, vapor pressure and relative humidity.
Horizontal blue dotted lines materialize the significativity thresholds at p = 0.05.

vapor pressure leading DF incidence by a constant delay of extended periods of either very cool or very hot tempe-
8 to 10 weeks. In addition to this strong annual periodicity, ratures. Likewise, a study in Taiwan found three turning
relative humidity displays a sub-annual periodicity, as ob- points of DF that occurred around early August, late
served on DF incidence. The annual periodicity of relative August/early September, and late October/early November.
humidity leads the annual periodicity of DF incidence by a The first two turning points were shown to relate
constant delay of 18 weeks whereas the sub-annual period- with two typhoons around early to mid August in
icity of relative humidity leads the sub-annual periodicity Taiwan that witnessed a sharp drop in temperature
of DF by a delay that decreases from 14.30 weeks in 2002 and substantial rainfall after it [16]. Similarly, other
to 5.27 weeks in 2009 at an almost constant rate of studies in Thailand and Singapore also revealed sig-
1.13 week per year. These results are in general agreement nificant associations between climatic variables and
with the findings of other studies that climatic factors play dengue incidence ([13,14,42]; Tipayamongkholgul [43,44]).
a role in the transmission cycles of DF. Interestingly, these For example, Tipayamongkholgul [43] conducted a study
two incidence peaks per year that we observed in Hanoi in the Gulf of Thailand and showed that the monthly aver-
with periods of low incidence occurring in January and age local relative humidity in the previous 3–6 months
February (the coldest months in Hanoi) and in June and was negatively associated with epidemics of dengue and
July (the warmest months in Hanoi) are in accordance with incidence of dengue cases. Woongkon et al. [44] in Chiang
Eastin et al. [26]‘s observation in Columbia where they Rai, Thailand, showed that all climatic factors including
noted a significant decreases of DF cases soon after minimum, maximum temperature, minimum and average
Do et al. BMC Public Health 2014, 14:1078 Page 7 of 10
http://www.biomedcentral.com/1471-2458/14/1078

(A)

4.00 0.50 0.06 4.00 0.50 0.06 4.00 0.50 0.06 4.00 0.50 0.06

phase (radians)

phase (weeks)
π
period (years)

0 20 40
−π 0
(B) 2003 2005 2007 2009 2003 2005 2007 2009

phase (radians)

phase (weeks)
π
period (years)

0 20 40
−π 0
(C) 2003 2005 2007 2009 2003 2005 2007 2009

phase (radians)

phase (weeks)
π
period (years)

0 20 40
(D) 2003 2005 2007 2009 −π 0 2003 2005 2007 2009
phase (radians)

phase (weeks)
π
period (years)

0 10 20
−π 0

2003 2005 2007 2009 2003 2005 2007 2009

phase (radians)

time (years)

phase (weeks)
π

0 20 40
−π 0

2003 2005 2007 2009

time (years)
Figure 4 Cross-wavelet power spectra between DF and mean temperature (A), rainfall (B) and absolute (C) and relative (D) humidities
in Hanoi from 2002 to 2009 (left column). The right column shows the phase angles of the climatic variables (blue, left y-axis) and DF (red, left
y-axis), as well as their difference (black, right y-axis). These phase angles are calculated on signals that have been filtered around the period of
maximal power in the spectra of the left column, i.e. annual periodicity for all the climatic variables, as well as also the semi-annual periodicity for
the relative humidity. In spectra of the left column, the black contour lines show the regions of power significant at the alpha-risk of 0.05, the
paled region of the spectrum delineates the cone of influence due to the zero-padding of the time series, and the power increases from dark
blue to dark red.

relative humidity, evaporation, wind speed and rainfall larvae or eggs [45]. In the spring, when weather condi-
lead increasing DF incidence by 0–2 months. tions become favorable again, eggs hatch and adults
In our study, DF incidence is characterized by 2 quasi- emerge, probably causing the first peak on the repro-
cycles with a periodicity of 6 months with the first one ductive ratio and the consequent DF incidence increase.
showing a slow and constant increase and the second The second peak on the reproductive ratio could be due
one showing a marked epidemic. The sharp rupture to the second mosquito generation of the year (issued
between these phases can be explained by the fact that from the first one), hence its potential to be higher than
the reproductive ratio is not constant throughout the the first one and even partially conceal it. This second
year but actually exhibits two peaks per year, with the peak of higher magnitude would be the cause of the epi-
second peak at least as high as the first one. Given that demic peak observed on DF incidence during the second
dengue is vector-born, the factor limiting its transmis- half of the year. This epidemic peak would thus be due
sion is either due to the mosquito population (mostly its more to an increase of the number of infected people
population size), or the human population (mostly its than to an increase in the mosquito population size and
proportion of susceptibles). Winter climatic conditions the dengue reproductive ratio. Indeed, dengue epidemic
in Hanoi are not favorable to adult mosquitoes and most peak appears even when the second peak on the repro-
of the mosquito population survive the winter either as ductive ratio is not higher than the first one. Such a
Do et al. BMC Public Health 2014, 14:1078 Page 8 of 10
http://www.biomedcentral.com/1471-2458/14/1078

hypothesis to explain the mechanism of dengue epidemi- mechanism could also explain the number of detected
ology in Hanoi can be tested by collecting entomological cases from year to year, the earlier detection of the epi-
data (larvae and adult densities estimates) in Hanoi all demic, and thus the shift in the sub-annual periodicity of
year-round and translating it into mathematical equa- the DF incidence mentioned above. Such a hypothesis
tions. This would allow to check whether a model based could be tested by collecting immunological data from the
on this hypothesis can generate epidemiological patterns human population of Hanoi (by an aged-stratified sero-
that are in accordance with the ones observed on DF prevalence survey for example) and investigating whether a
incidence data. mathematical model built on this hypothesis does generate
The 2 above-mentioned 6-month cycles observed on the trend in DF incidence mean and timing that we observe
time series of DF incidence translate into the sub-annual on the data.
periodicity that we have characterized in addition to the Both DF incidence and relative humidity exhibit con-
annual periodicity, with the annual periodicity of DF spicuous annual and sub-annual periodicities and these pe-
incidence mostly accounting for the high epidemic peak riodicities happen to be strongly correlated. However, we
of the second half of the year, and the sub-annual peri- warn against over-interpretation of such correlations in
odicity mostly accounting for the slow and constant term of biological causation. One reason for such a caution
increase of DF incidence of the first half of the year. An in particular is that relative humidity is a variable that
interesting result of our analysis is that relative humidity depends on both absolute humidity and temperature (the
also shows these two annual and sub-annual periodic- former being naturally strongly influenced by rainfalls). In
ities and that the sub-annual periodicity of relative case where absolute humidity (or rainfalls) and temperature
humidity leads the sub-annual periodicity of DF incidence are not perfectly correlated (which is most likely the case),
by a lag that decreases from 14.30 weeks in 2002 to we do expect that relative humidity exhibits annual and bi-
5.27 weeks in 2009. As interpreted above, this sub-annual annual periodicities, as the resultant of two periodic signals
periodicity reflects the first peak of the reproductive ratio that are not perfectly in phase. Thus, instead of looking for
that we interpreted in the paragraph above as the first mos- a mechanistic link between DF incidence and relative
quito hatching of the year. Explaining the observed shift in humidity, it may be more relevant to look for two links: (i)
the timing of this first peak by a shift in mosquito hatching one between DF incidence and absolute humidity and (ii)
is biologically unrealistic. Alternatively, we propose that this one between DF incidence and temperature, possibly
shift is due to (i) the building-up of the human population accounting for a possible interaction between the two
immunity from year to year and (ii) the interactions climatic variables. This particular point will be the topic of
between dengue serotypes (antibody-dependent enhance- a subsequent study.
ment, ADE) as explained below. In conclusion, our analysis on the links between climatic
Most of primary dengue infections are asymptomatic variables and DF incidence in Hanoi raises a number of
[1]. Before the emergence of dengue in Hanoi (in 2002), questions of general interest on the relationships between
most of the human population may have been susceptible climate and infectious diseases epidemiology. Because of its
to the 4 dengue serotypes and hence most of the dengue highly seasonal climate (and thus potentially highly seasonal
cases may have been primary infections, most likely dengue transmission too), its important population size and
asymptomatic and thus unnoticed by the surveillance density, and its dengue epidemiological transition (current
system. As the disease progressively emerges in Hanoi, emergence), Hanoi appears as the ideal set-up to test
population immunity to different dengue serotypes hypotheses about interaction between serotypes. This is an
increases, thus increasing the number of secondary infec- issue both under-understood and potentially of high rele-
tions relative to primary ones, and thus increasing the vance for vaccine development. Further investigations on
number of symptomatic detected cases. Expected conse- dengue in Hanoi call for additional entomological and im-
quences of this mechanism is not only an increase in the munological data, as well as for theoretical developments.
number of detected cases from year to year (as visible
through the upward trend of DF incidence), but also an
Competing interests
earlier detection of the epidemics. The latter would The authors declare that they have no competing interests.
explain this observed shift in the sub-annual periodicity of
the DF incidence. This mechanism can potentially be rein-
forced by some ADE-related mechanisms. Indeed, poten- Authors’ contributions
DTT: Designed the study, developed the outline, and contributed to the
tial epidemiological consequences of the ADE hypothesis analysis, writing and revision of the manuscript. PM: Developed the outline,
that have been proposed in the literature are that it contributed to writing the manuscript. LNH: Revised the outline, contributed
increases the susceptibility to secondary infections and/or to writing the manuscript. PW: Revised the outline, contributed to writing
the manuscript. MC: Helped developing the outline and writing the
the transmissibility from individuals suffering from sec- manuscript, contributed to the analysis. All authors read and approved the
ondary infections (see for example [46]). Thus, such a final manuscript.
Do et al. BMC Public Health 2014, 14:1078 Page 9 of 10
http://www.biomedcentral.com/1471-2458/14/1078

Acknowledgements 18. Lu L, Lin H, Tian L, Yang W, Sun J, Liu Q: Time series analysis of dengue
This study was funded by the Netherlands Higher Education (NPT) project fever and weather in Guangzhou, China. BMC Public Health 2009, 9:395.
on “Strengthening teaching and research capacity in preventive medicine in 19. Hu W, Clements A, William G, Tong S: Dengue fever and El Nino/Southern
Vietnam”. MC is supported by the “Biodiversity and Infectious Diseases in Oscillation in Queensland, Australia: a time series predictive model.
Southeast Asia” CNRS-funded GDRI. Occup Environ Med 2010, 67(5):307–311.
20. Wu P-C, Guo H-R, Lung SC, Lin CY, HJa S: Weather as an effective predictor
Author details for occurrence of dengue fever in Taiwan. Acta Trop 2007, 103:50–57.
1
Biostatistics and Medical Informatics Department, Institute of Training for 21. Gharbi M, Quenel P, Gustave J, Cassadou S, La Ruche G, Girdary L, Marrama
Preventive Medicine and Public Health, Hanoi Medical University, Hanoi, L: Time series analysis of dengue incidence in Guadeloupe, French West
Vietnam. 2International Centre for Integrated assessment and Sustainable Indies: forecasting models using climate variables as predictors.
development, Maastricht University, Maastricht, The Netherlands. 3The BMC Infect Dis 2011, 9(11):166.
Medical Committee Netherlands-Vietnam, Amsterdam, The Netherlands. 22. Pham HV, Doan HT, Thao TTP, Minh NNT: Ecological factors associated
4
MIVEGEC (IRD 224-CNRS 5290-Université Montpellier 1 et 2), Centre IRD, with dengue fever in a Central Highlands province, Vietnam. BMC Infect
Montpellier, France. 5Oxford University Clinical Research Unit, Hanoi, Vietnam. Dis 2011, 11:172.
23. Pinto E, Coelho M, Oliver L, Massad E: The influence of climate variables
Received: 14 May 2014 Accepted: 29 September 2014 on dengue in Singapore. Int J Environ Health Res 2011, 21(6):415–426.
Published: 16 October 2014 24. Descloux E, Mangeas M, Menkes CE, Lengaigne M, Leroy A, Tehei T,
Guillaumot L, Teurlai M, Gourinat AC, Benzler J, Pfannstiel A, Grangeon JP,
Degallier N, Lamballerie X: Climate-based models for understanding and
References
forecasting dengue epidemics. PLoS Negl Trop Dis 2012, 6(2):e1470.
1. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL, Drake JM,
25. Hsieh YH, de Arazoza H, Lounes R: Temporal trends and regional variability
Brownstein JS, Hoen AG, Sankoh O, Myers MF, George DB, Jaenisch T,
of 2001–2002 multiwave DENV-3 epidemic in Havana City: did Hurricane
William Wint GR, Simmons CP, Scott TW, Farrar JJ, Hay SI: The global
Michelle contribute to its severity? Trop Med Int Health 2013, 18(7):830–838.
distribution and burden of dengue. Nature 2013, 496:504–507.
26. Eastin MD, Delmelle E, Casas I, Wexler J, Self C: Intra- and inter-seasonal
2. Tien TKN, Ha DQ, Hien TK, Quang LC: Predictive indicators for forecasting
autoregressive prediction of dengue outbreaks using local weather and
epidemic of dengue/dengue haemorrhagic fever through
regional climate for a tropical environment in Colombia. Am J Trop Med
epidemiological, virological and entomological surveillance. Dengue Bull
Hyg 2014, Available online at http://www.ajtmh.org/content/early/recent
1999, 23:34–39.
since 23 June 2014.
3. NIHE: Final Report on evaluation of communicable diseases surveillance system
in Vietnam 2008. Hanoi: National Institute of Hygiene and Epidemiology; 27. Yi B, Zhang Z, Xu D, Xi Y, Fu J, Luo J, Yuan M, Liu S, Zuo Z: Relationship of
2009. dengue fever epidemic to Aedes density changed by climate factors in
4. Ha DQ, Ninh TU: Virological surveillance of dengue haemorrhagic fever in Guangdong Province. Wei Sheng Yan Jiu 2003, 32(2):152–154.
Vietnam, 1987–1999. Dengue Bull 2000, 24:18–23. 28. Shaman J, Kohn M: Absolute humidity modulates influenza survival,
5. Cuong HQ, Hien NT, Duong TN, Phong TV, Cam NN, Farrar J, Nam VS, Thai transmission, and seasonality. Proc Natl Acad Sci U S A 2009, 106(9):3243–3248.
KTD, Horby P: Quantifying the emergence of dengue in Hanoi, Vietnam: 29. Vietnamese Ministry of Health: Guidelines for Surveillance, Prevention and
1998–2009. PLoS Negl Trop Dis 2011, 5(9):e1322. Control of Dengue and Dengue Haemorrhagic Fever. Circula 2006, 1266/
6. Thai KTD, Cazelles B, Nguyen NV, Vo LT, Boni MF, Farrar J, Simmons CP, 2006/QD-BYT. Available online at http://thuvienphapluat.vn/archive/Quyet-
Rogier van Doorn H, de Vries PJ: Dengue Dynamics in Binh Thuan dinh-1266-QD-BYT-sua-doi-huong-dan-giam-sat-va-phong-chong-benh-
Province, Southern Vietnam: Periodicity, Synchronicity and Climate vb109491.aspx. Medicine Publising House.
Variability. PLoS Negl Trop Dis 2010, 4(7):e747. 30. Wallace JM, Hobbs PV: Atmospheric Science, An Introductory Survey. Academic
7. Vietnamese Ministry of Health: Guidance on declaration, communication, New York. 2nd edition. 2006.
and reporting infectious deseases. Circula 2010, 48/2010/TT-BYT. Available 31. Cazelles B, Chavez M, McMichael AJ, Hales S: Nonstationary influence of El
online at http://vanban.chinhphu.vn/portal/page/portal/chinhphu/ Niño on the synchronous dengue epidemics in Thailand. PLoS Med 2005,
hethongvanban?class_id=1&mode=detail&document_id=99852. 2:313–318.
8. Toan DTT, Hu W, Thai PQ, Hoat LN, Wright P, Martens P: Hot spot detection 32. Cazelles B, Hales S: Infectious Diseases, Climate Influences, and
and spatio-temporal dispersion of dengue fever in Hanoi, Vietnam. Nonstationarity. PLoS Med 2006, 3(8):328.
Global Health Action 2013, 6:18632. 33. Simões TC, Codeço CT, Nobre AA, Eiras AE: Modeling the Non-Stationary
9. Schreiber KV: An investigation of relationships between climate and Climate Dependent Temporal Dynamics of Aedes aegypti. PLoS One
dengue using a water budgeting technique. Int J Biometeorol 2001, 2013, 8(8):e64773.
45:81–89. 34. Cazelles B, Chavez M, Constantin de Magny G, Guegan JF, Hales S:
10. Hales S, de Wet N, Maindonald J, Woodward A: Potential effect of Time-dependent spectral analysis of epidemiological time-series with
population and climate changes on global distribution of dengue fever: wavelets. J R Soc Interface 2007, 4(15):625–636.
an empirical model. Lancet 2002, 360(9336):830–834. 35. Torrence C, Compo GP: A practical guide to wavelet analysis. Bull Am
11. Nagao Y, Tharava U, Chinumsup P, Tawatsin A, Chansang C, Campbell- Meteorol Soc 1998, 79:61–78.
Lendrum D: Climatic and social risk factors for Aedes infestation in rural 36. Soper HE: The Interpretation of Periodicity in Disease Prevalence. Roy Stat
Thailand. Trop Med Int Health 2003, 8(7):650–659. Soc A 1929, 92:34–61.
12. Depradine C, Lovell E: Climatological variables and the incidence of 37. Keeling MJ, Rohani P: Modeling Infectious Diseases in Humans and
Dengue fever in Barbados. Int J Environ Health Res 2004, 14(6):429–441. Animals. Princeton University Press 2008, 5:155–189.
13. Promprou S, Jaroensutasinee M, Jaroensutasinee K: Climatic factors 38. Andraud M, Hens N, Marais C, Beutels P: Dynamic epidemiological models
affecting dengue haemorrhagic fever incidence in Southern Thailand. for dengue transmission: a systematic review of structural approaches.
Dengue Bulletin 2005, 29:41–48. PLoS One 2012, 7(11):e49085.
14. Promprou S, Jaroensutasinee M, Jaroensutasinee K: Impact of Climatic 39. RCoreTeam: R: A Language and Environment for Statistical Computing R.
Factors on Dengue Haemorrhagic Fever Incidence in Southern Thailand. Vienna, Austria: Foundation for Statistical Computing; 2012.
Walailak J Sci & Tech 2005, 2(1):59–70. 40. Gouhier T: Biwavelet: Conduct univariate and bivariate wavelet analyses. 2013.
15. Chowell G, Torre CA, Munayco_Escate C, Suarez-Ognio L, Lopez-Cruz R, (Version 0.14). Available from http://biwavelet.r-forge.r-project.org.
Hyman JM, Castillo-Chavez C: Spatial and temporal dynamics of dengue 41. Center for Excellence in Disaster Management and Humanitarian
fever in Peru: 1994–2006. Epidemiol Infect 2008, 136(12):1667–1677. Assisstance: Dengue fever cases rise after floods plague Vietnam. 2008.
16. Hsieh YH, Chen CWS: Turning points, reproduction number, and impact Available online at http://reliefweb.int/report/viet-nam/dengue-fever-cases-
of climatological events for multi-wave dengue outbreaks. Trop Med Int rise-after-floods-plague-vietnam since 28 Nov 2008.
Health 2009, 16:1–11. 42. Thu HM, Aye KM, Thein S: The effect of temperature and humidity on
17. Johansson MA, Dominici F, Glass GE: Local and global effects of climate dengue virus propagation in Aedes aegypti mosquitoes. Southeast Asian
on dengue transmission in Puerto Rico. PLoS Negl Trop Dis 2009, 3(2):e382. J Trop Med Public Health 1998, 29:280–284.
Do et al. BMC Public Health 2014, 14:1078 Page 10 of 10
http://www.biomedcentral.com/1471-2458/14/1078

43. Tipayamongkholgul M, Fang C-T, Klinchan S, Liu CM, King CC: Effects of the
El Niño-Southern Oscillation on dengue epidemics in Thailand,
1996–2005. BMC Public Health 2009, 9:422.
44. Wongkoon S, Jaroensutasinee M, Jaroensutasinee K: Climatic variability and
dengue virus transmission in Chiang Rai, Thailand. Biomedica 2011,
27(19):5–13.
45. Higa Y, Yen NT, Kawada H, Son TH, Hoa NT, Takagi M: Geographic
Distribution of Aedes aegypti and Aedes albopicuts Collected from Used
Tires in Vietnam. Am Mosq Control Assoc 2010, 26(1):1–9.
46. Recker M, Blyuss KB, Simmons CP, Hien TT, Wills B, Farrar J, Gupta S:
Immunological serotype interactions and their effect on the
epidemiological pattern of dengue. Proc R Soc B 2009, 276(1667):2541–2548.

doi:10.1186/1471-2458-14-1078
Cite this article as: Do et al.: Climatic-driven seasonality of emerging
dengue fever in Hanoi, Vietnam. BMC Public Health 2014 14:1078.

Submit your next manuscript to BioMed Central

and take full advantage of:

• Convenient online submission

• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution

Submit your manuscript at

www.biomedcentral.com/submit