Hot spot detection and spatio-temporal dispersion of dengue fever in Hanoi, Vietnam.

0.65 MB·English

by  Toan, Do Thi Thanh

(cid:2)PUBLIC HEALTH IN VIETNAM: HERE’S THE DATA, WHERE’S THE ACTION?
Hot spot detection and
spatio-temporal dispersion of dengue
fever in Hanoi, Vietnam
Do Thi Thanh Toan1*, Wenbiao Hu2, Pham Quang Thai3,
Luu Ngoc Hoat1, Pamela Wright4 and Pim Martens5
1Institute ofTraining forPreventive Medicine and Public Health, Hanoi MedicalUniversity, Hanoi,
Vietnam;2SchoolofPopulation Health, TheUniversity ofQueensland, Queensland, Australia;
3NationalInstituteofHygieneandEpidemiologyofVietnam,Hanoi,Vietnam;4TheMedicalCommittee
Netherlands(cid:2)Vietnam,Amsterdam,theNetherlands;5InternationalCentreforIntegratedassessment
and Sustainabledevelopment, Maastricht University, Maastricht,the Netherlands
Introduction: Dengue fever (DF) in Vietnam remains a serious emerging arboviral disease, which generates
significant concerns among international health authorities. Incidence rates of DF have increased
significantly during the last few years in many provinces and cities, especially Hanoi. The purpose of this
studywastodetectDFhotspotsandidentifythediseasedynamicsdispersionofDFovertheperiodbetween
2004and2009in Hanoi, Vietnam.
Methods:DailydataonDFcasesandpopulationdataforeachpostcodeareaofHanoibetweenJanuary1998
and December 2009 were obtained from the Hanoi Center for Preventive Health and the General Statistic
OfficeofVietnam.Moran’sIstatisticwasusedtoassessthespatialautocorrelationofreportedDF.Spatial
scanstatistics and logistic regressionwereused toidentify space(cid:2)timeclusters anddispersion of DF.
Results: The study revealed a clear trend of geographic expansion of DF transmission in Hanoi through
thestudyperiods(OR1.17,95%CI1.02(cid:2)1.34).Thespatialscanstatisticsshowedthat6/14(42.9%)districtsin
Hanoi had significant cluster patterns, which lasted 29 days andwere limited to a radius of 1,000 m. The
studyalsodemonstratedthatmostDFcasesoccurredbetweenJuneandNovember,duringwhichtherainfall
andtemperaturesarehighest.
Conclusions:ThereisevidencefortheexistenceofstatisticallysignificantclustersofDFinHanoi,andthat
thegeographicaldistributionofDFhasexpandedoverrecentyears.Thisfindingprovidesafoundationfor
furtherinvestigationintothesocialandenvironmentalfactorsresponsibleforchangingdiseasepatterns,and
provides datatoinform program planning forDFcontrol.
Keywords: denguefever;hotspots;dynamicdispersion;Hanoi;Vietnam
Received:25 April 2012; Revised:21 December 2012; Accepted:21 December 2012; Published:24 January 2013
Dengue fever (DF), a mosquito transmitted viral thelevelofrisk,riskfactors,andthepopulationexposed.
infection, is a serious public health concern In recent years, the development of geographic informa-
worldwide, particularly in developing countries. tion systems (GIS) has provided a supportive spatial
Annually, the number of DF cases has been estimated analyticaltoolthathasenabledepidemiologiststoinclude
to range from 50 to 100 million cases worldwide; of moresimplyaspatialcomponentinepidemiologicstudies
which,upto500,000casesresultindenguehaemorrhagic (6).Inthefieldofinfectiousandvector-bornediseasessuch
fever (1, 2). The infection remains a major threat to the asmalariaorDF,GIShavebeenwidelyusedfordisease
community well-being because it is associated with an mappingofdifferentpathologies,inanalysisofspaceand
increased riskof premature mortality and incurs signifi- space(cid:2)timedistributionofdiseasedata,inidentifyingrisk
canthealthcarecosttosociety(3).InVietnam,anumber factors,andinmappingriskareas(5,7).Moreover,totest
ofrecentoutbreaksofDFhavealsogeneratedsignificant whetheranyclusterscanbedetectedorifthepointprocess
concerns among international health authorities (4, 5). is purely randomly distributed, temporal, spatial, and
As the risk of DF varies with space and time, it is space(cid:2)time scan statistics (SaTScan) have recently come
importanttohavepreciseknowledgeoftheregionsatrisk, intocommonuse(8,9).TheadvantagesofusingSaTScan
GlobHealthAction2013.#2013DoThiThanhToanetal.ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommonsAttribution- 1
Noncommercial3.0UnportedLicense(http://creativecommons.org/licenses/by-nc/3.0/),permittingallnon-commercialuse,distribution,andreproduction
inanymedium,providedtheoriginalworkisproperlycited. Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
DoThiThanhToanetal.
Summary of Policy Recommendations
For Hanoi Preventive Medicine Center:
. HotspotsanalysisforDFshouldbewidelyusedinDFsurveillancesinceitcanhelpreallocatetheresourcetodeal
with the outbreak more effectively.
For Preventive Medicine System:
. Case-basedsurveillancesystemshoulduseGPSdatatotrackthediseaseoutbreakandtheeffectofintervention.
are that it can adjust for confounding variables, it can year (1,682 mm rainfall/ year). The winter months
reducepre-selectionbiasasitsearchesforclusterswithout from November to March are relatively dry, although
specifyingtheirsizeorlocation,itgivesasinglep-valueas springthenoftenbringslightrains.Theminimumwinter
the likelihood-ratio-based test takes account of multiple temperatureinHanoicandipaslowas6(cid:2)78C(43(cid:2)458F)
testing,andfinally,itcanbeappliedtoawholeregionto not including the wind chill, while summer can get as
detectsignificantclustersinthatregion(10). hot as 38(cid:2)408C (100(cid:2)1048F). The period from May to
WithincreasingconcernaboutthethreatposedbyDFin September is suitable for the development of the mos-
Hanoi, a clearer picture of the epidemiology and impor- quito vector: high temperatures and high precipitation
tantriskfactorswasneeded.Thisstudyaimedtofillthat favour increased rates of mosquito development and
need. Using spatial scan statistics and GIS, we investi- a decreased length of reproductive cycle, as well as
gatedthespatialdistributionofconfirmedcasesofDFand providing more sites for egg deposition and larva
investigatedtheareasofhighriskwithinall14districtsof development (12).
Hanoi. In addition, we have used GIS and spatial scan With the above characteristics, Hanoi provides very
statistics to detect the hot spots and identify the disease favourable conditions for the existence, circulation and
dynamicsdispersionofDFoverperiodbetween2004and development of infectious diseases such as DF and
2009inHanoi. dengue haemorrhagic fever.
Methods
Datasources
Daily data on DF cases between January 1998 and
Studysite
December 2009 in old Hanoi were obtained from the
HanoiislocatedinthenorthofVietnam,inthelowlying
Hanoi Center for Preventive Health. Data included
anddenselypopulatedRedRiverdelta.Hanoi,beforeits
the onset date and place of onset of the notified cases
merging with part of neighbouring provinces in 2008
ofDFinfection, age,sex,occupationof thepatients and
(known as ‘old Hanoi’), had 14 districts divided into
laboratorytestresults.ThecriteriafornotificationofDF
229postcodeareas.In2009,theoldcityhadapopulation
disease are based on guidelines from the Ministry of
of 3.5 million; the population density was quite high, at
Health, 1999, on surveillance, diagnosis, and treatment
1,943people/km2(11).Inthisstudy,thegeographicareaof
‘oldHanoi’wasselectedasastudysitebecauseitmadeit
possible to use consistent data for all of the study time
periods(2004(cid:2)2009)(Figure1).
ThereisarapidpopulationgrowthinHanoiduetothe
Soc Son
influx of workers from rural areas. Previous studies in
Vietnamshowedthatpeoplelivinginpoorerareasandold
tenementhousing,includinghousesfortransientworkers,
generallytendtohaveunhygienicconditionswithalackof
watersupplyandabsenceofwindowscreens,bothofwhich Dong anh
may promote Aedes mosquito proliferation and contact.
Therapidmovementofpopulationbetweencommunities Tu Liem Tay Ho Long Bien
Gia Lam
may have promoted the largest outbreak of dengue in Cau Giay Hoan Kiem
Dong Da
Hanoi:in2009,thecityrecordedalmost8,000DFpatients, Thanh XuanHai Ba Trung
representinga15-foldincreasefromthepreviousyear. Hoang Mai
Hanoi experiences the typical climate of northern Thanh Tri
Vietnam,wheresummersarehotandhumid,andwinters
02.5 5 10 15 20
Kilometers
are relatively cool and dry. The summer months from
May to September receive the majorityof rainfall in the Fig. 1. Locationof the studyarea(cid:2)Hanoi, Vietnam.
2
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
Detectionanddispersionofdenguefever
Table1. Descriptivestatisticsof monthly numbersof postcode areaswithnotified denguecases
Period Mean SD Minimum Q1a Median Q3b Maximum
2004(cid:2)2005 41.00 32.36 5 15.75 36 59 95
2006(cid:2)2007 49.33 32.47 8 24.25 47 72.25 104
2008(cid:2)2009 79.33 45.53 17 36.75 74 122.5 146
aQ1,firstquartilevalue.
bQ3,thirdquartilevalue.
ofdengue,inwhichindividualssuspectedtohavedengue Moran’s I statistic in the program ArcGIS 9.2. Spatial
are those who have acute febrile illness (]388C) of 2(cid:2)7 autocorrelationwasconsideredsignificantifthepB0.05.
days duration with two or more of the following non- Moran’sIrangesfrom (cid:3)1to1andcanbeinterpretedas
specific manifestations of DF: headache, retro-orbital follows: avalue close to 0 indicates spatial randomness,
pain, myalgia, arthralgia, rash, haemorrhagic manifesta- while a positive value indicates positive spatial autocor-
tions,andleucopenia(11).Thetotalnumberofcaseswas relation and a negative value indicates negative spatial
25,483 after excluded caseswhere the residential address autocorrelation.
wasunspecified(theyears1999(cid:2)2002havenoaddresses).
The locations of patient’s residence were taken from a Cluster analysis
Garmin GPS60 (Garmin Corporation, Taipei County, Scan statisticswere used to detect and evaluate the clus-
Taiwan) global positioning system while doing a field ters of cases in either a purely temporal, purely spatial,
survey.Atotalof51.4%(13,092cases)oftheDFrecords or space(cid:2)time setting. This was achieved by gradually
are available as case event data. scanning awindow across time and/or space, noting the
The population data of each postcode area for number of observed and expected observations inside
2004and2009wereobtaineddirectlyfromthedecennial the window at each location. For each location and size
national Population and Housing Census, conducted of the scanning window, the alternative hypothesis was
by the General Statistic Office of Vietnam (GSO; http:// that there was an elevated risk within the window
www.gso.gov.vn). comparedtooutside.
In SaTScan software, the scanning window was an
Dataanalysis interval (in time), a circle or an ellipse (in space), or a
Descriptive analysis cylinder with a circular or elliptic base (in space(cid:2)time).
Descriptive statistics of numbers of Hanoi areas with Multiple different window sizes were used. The window
notifiedDFcasesinthreeperiods(2004(cid:2)2005,2006(cid:2)2007, withthemaximumlikelihoodwasthemostlikelycluster,
and2008(cid:2)2009),incidencerateofDFintheseperiods,or that is, the cluster least likely to be due to chance.
choroplethmapswereanalysed to describe the dynamics A p-value was assigned to this cluster. The standard
ofthedisease.ThemonthlydistributionofDFaccording purely spatial scan statistic imposed a circular window
toseasonalitywasalsoperformedusingboxplots. on the map. The window was in turn centred on each
of several possible grid points positioned throughout
Spatial autocorrelation analysis the study region. For each grid point, the radius of the
The spatial autocorrelation of the expected incidence window varied continuously in size from zero to some
rates of DF in three different periodswas assessed using upperlimitspecifiedbytheuser.Inthisway,thecircular
4000 160 er
v
er of dengue fever12233505050000000000 1116802400000 PA with dengue feinfection
mb1000 40 of
Nu 500 20 ber
m
0 0 u
1 4 8 12 4 8 12 4 8 12 4 8 12 4 8 12 4 8 12 N
2004 2005 2006 2007 2008 2009
Fig. 2. Numbers of dengue cases (_____) and postcode areas with dengue notifications (-----) between January 2004 and
December 2009inHanoi.
3
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
DoThiThanhToanetal.
Winter Spring number of overlapping cylinders of different sizes and
150 150
shapes were obtained, which jointly cover the entire
100 100 studied region, where each cylinder reflected a possible
50 50 cluster. The temporal scan statistic used a window that
moved in one dimension, time, defined in the same way
0 0
astheheightofthecylinderusedbythespace(cid:2)timescan
Autumn Summer
150 150 statistic. This meant that it was flexible in both start
100 100 andenddate.Forpurelyspatialandspace(cid:2)timeanalyses,
50 50 SaTScanalsoidentifiedsecondaryclustersinthedataset
0 0 in addition to the most likely cluster and lined them up
2004–2005 2006–2007 2008–2009 2004–2005 2006–2007 2008–2009
by their likelihood ratio test statistic. For purely tem-
Fig. 3. Boxplots of the seasonal distribution of numbers poralanalyses,onlythe mostlikelycluster wasreported.
of postcode areaswith dengue infection in three periods in No geographic overlap was used as a default setting, so
Hanoi. The boxplots display the values of the 25th, 50th, secondary clusters would not overlap the most signifi-
and75thpercentiles.
cantcluster.Inordertoscanfromsmalltolargeclusters,
the maximum cluster size was set to 50% of the total
window was flexible both in its location and size, while
population at risk. To ensure sufficient statistical power,
each circle was a candidate cluster. The space(cid:2)time
the numberof Monte Carlo replicationswas set to 999.
scan statistic was defined by a cylindrical window with
acircular(orelliptic)geographicbaseandwithitsheight Dynamic dispersion of DF
corresponding to the time. The base was defined exactly In this study, we try to identify whether changes in DF
as for the purely spatial scan statistic, while the height varied with latitude and longitude of villages centroids
reflected the time period of potential clusters. The cylin- in the three periods. Logistic regression models can be
dricalwindowwasthenmovedinspaceandtime,sothat constructed with the dichotomous outcome variable
for each possible geographical location and size, it also defined as whether or not an increase of DF occurred
visited each possible time period. In effect, an infinite in each villagebetweenthe three periods. Longitude and
± ± ±
Decimal Degrees
00.020.04 0.08 0.12 0.16
0.00 0.00 – 25.00 25.00 – 50.00 >50.00
Fig. 4. Mapshowing rawdengueincidence rates in threeperiods.
4
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
Detectionanddispersionofdenguefever
Table2. Spatialautocorrelation analysisfordenguein 2887.6/100 000 and kept increasing, from 97 to 132 and
Hanoi, 2004(cid:2)2009 subsequently 160 postcode areas in the three periods.
Period Annualincidence Moran’sI E(I) p SpatialautocorrelationofDF
A significant positive spatial autocorrelation of dengue
2004(cid:2)2005 226.32 0.19 (cid:3)0.011 0.001
incidence for all three periods is presented in Table 2,
2006(cid:2)2007 1049.78 0.32 (cid:3)0.007 0.001
where Moran’s I index was 0.19 (expected Moran’s
2008(cid:2)2009 3774.49 0.22 (cid:3)0.006 0.001
I(cid:4)(cid:3)0.011, pB0.001) during 2004(cid:2)2005 (expected
Moran’s I(cid:4)(cid:3)0.007, p(cid:4)0.001), 0.32 during 2006(cid:2)2007,
latitude of village centroids were entered as explanatory and 0.22 (expected Moran’s I(cid:4)(cid:3)0.006, p(cid:4)0.001)
variables. Spatial dispersions can be expressed in terms during2008(cid:2)2009.Thismeansthatvillagesclosertogether
ofoddsratios(OR)forlongitudeandlatitude,with95% tend to have more similar baseline incidence rates than
confidence interval (CI). thosefurtherapart.
Purelytemporalclustering
Results
The results of the purely temporal clustering analysis in
each year also indicate the seasonal tendency of dengue
Descriptiveanalysis
transmission. Table 3shows the temporal clusters of DF
Monthlyaveragenumbersofpostcodeareaswithnotified
cases in the study area from 2004 to 2009. There were
denguecasesinHanoiforthreeperiods2004(cid:2)2005,2006(cid:2)
significantly temporal clusters in all years (range of RR
2007, and 2008(cid:2)2009 are summarised in Table 1. It was
from1.34(cid:2)14.26,p(cid:4)0.001).ThehighestRRwasfoundin
shownthatthereisacleartrendofgeographicexpansion
2005 (RR: 14.26, p(cid:4)0.001), which suggested there were
of dengue transmission in Hanoi through periods (with
strongesttemporalclustersin2005.Thetemporalclusters
mean equal to 41.0, 49.33, and 79.33, respectively).
of DF in Hanoi often covered 6 months (June to
Figure 2 presents a striking variation in the monthly
December).
numbers of dengue cases and monthly numbers of post-
codeareaswith dengue from2004 to2009. A largepeak
of dengue incidence occurred in October and November Purelyspatialclustering
2009(3,696 cases and2,698 cases, respectively).Peaks in Analysisofpurelyspatialclusteringofdenguecasesfrom
incident cases coincided with high monthly numbers of 2004 to 2009, with the maximum spatial cluster size of
postcodeareaswithDFcases. 50% of the total population, identified the most likely
Boxplots of the monthly numbers of postcode areas clusterforeachofthe6years.However,onlyin2004and
with dengue are shown in Fig. 3. The results indicated 2005didtheresultshowarandomdistributionofdengue
a strongly seasonal pattern (with a peak in autumn and in space of Dong Da, Hai Ba Trung, Thanh Xuan and
early winter) and suggested that there was an upward HoangMaidistricts(Table4).Fortheyears2006(cid:2)2009,it
trend of dengue incidence from 2004 to 2009. wasfoundthattheriskisthesameinsideandoutsidethe
Figure 4 shows the geographic distribution of the raw cluster since p(cid:2)0.05.
incidenceofnotifieddenguecasesinHanoiinthreetime
(cid:2)
periods. There was an expansion of postcode areaswith Spacetimeclustering
dengue to the west-northern wards in Hanoi between The space(cid:2)time clustering analysis of the dengue data
2004 and 2009. Dengue incidence ranged from 4.55 to from 2004 to 2009 was also tested. Figure 4 and Table 5
Table3. The clusters ofdenguecases detected usingthe purelytemporal analysis
Year Clustertimeframe Totaldaysofcluster Obsa Expb Relativerisk LLRc p
2004 4/8(cid:2)1/12 118 216 160.30 1.90 17.68 0.001
2005 5/6(cid:2)1/12 177 246 167.56 14.26 74.0 0.001
2006 5/6(cid:2)1/12 177 1,679 1,556.38 2.30 45.39 0.001
2007 5/7(cid:2)31/12 177 1,051 937.56 1.88 33.36 0.001
2008 4/8(cid:2)16/12 133 992 886.41 1.60 23.84 0.001
2009 20/6(cid:2)16/12 179 7,725 7,634.84 1.34 12.64 0.001
aThenumberofobservedcasesinacluster.
bThenumberofexpectedcasesinacluster.
cLoglikelihoodratio.
5
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
DoThiThanhToanetal.
Table4. The clusters of denguecases detectedusingthe purelyspatial analysis
Year Location Obsa Expb Relativerisk LLRc p
2004 A ThanhTri 175 150.08 1.33 3.56 0.932
2005 A HaiBaTrung,HoangMai 65 37.10 2.01 10.4 0.002
B ThanhXuan,DongDa 128 96.19 1.66 8.18 0.017
2006 A DongDa,TayHo,LongBien 833 806.05 1.06 0.81 1.000
2007 A DongDa,BaDinh 658 609.28 1.17 3.89 1.000
2008 A TuLiem,CauGiay 238 215.30 1.13 1.41 1.000
2009 A HoanKiem,HaiBaTrung 2,346 2323.1 1.02 0.16 1.000
aThenumberofobservedcasesinacluster.
bThenumberofexpectedcasesinacluster.
cLoglikelihoodratio.
A:Mostlikelycluster;B:Secondarycluster.
illustrate the clusters in all districts of Hanoi at a 5% of postcode centroids in the three periods (Table 6). The
significant level (pB0.05) in this period. results suggest that changes in DF were significantly
The results reveal a high significance of space(cid:2)time associated with latitude (OR 1.17, 95% CI 1.02(cid:2)1.34)
association with DF transmission. It is revealed that six betweentheperiods2008(cid:2)2009and2004(cid:2)2005.However,
out of the 14 districts of Hanoi had significant cluster there was no association between DF and longitude in
patterns, in which Dong Da, Hoang Mai, and Thanh any period.
Xuanhavethehighestnumberofspace(cid:2)timeclusters.The
most likely cluster was found to differ during all three
Discussion
year periods. In 2004(cid:2)2005, only a few clusters were
The results of this study revealed significant spatio-
found, distributed over a large distance and time. In
temporal variation in the distribution of DF in Hanoi,
2006(cid:2)2007, the most likelyclusteroccurred in Dong Da,
Vietnam. Previous studies showed that, in Vietnam, the
with 149 cases and within 14 days in November. In
dengue epidemic often had a cycle of 3(cid:2)5 years andwas
August 2008, the most likely cluster was reported in
expectedtoreacha10-yearpeak(5,13,14).Thesepeaks
Hoang Mai and was limited to 250 m (RR(cid:4)5.58, p(cid:4)
werein1987,1998and2009,asalsorecordedinthisstudy,
0.001).InNovember2009,thehighestnumberofdengue
showinganupwardtrendofdenguecaseswiththelargest
cases(553)wasagainfoundinDongDaandHoanKiem,
outbreaks in 2009. Most of the dengue cases occurred
within the radius of 1,560 m. The RRwithin the most
between June and November, when the rainfall and
likelyclusterwas6.41(p(cid:4)0.001).Thesecondaryclusters
temperature are highest of the year. This time period
reportedinHoangMai,TayHo,andHaiBaTrungwere
appeared in our analysis of purely temporal clustering,
also limited at 1,000 m andwithin 29 days (Fig. 5).
showing high-risk months for dengue eachyear between
2004and2009.Theresultsareconsistentwiththoseofa
Dynamic dispersion of DF study conducted in a Central Highlands province of
Logistic regression models were constructed to identify Vietnam,whichfoundthatdenguewasmostprevalentin
whetherchangesinDFvariedwithlatitudeandlongitude thewetseason(11).Asothertropicalcountries,Vietnam’s
climateisfavourableforthetransmissionofDF.Awarm
Table5. SaTScan statistics forspace(cid:2)time clusterswith
temperature is crucial to the mosquito’s life and gono-
significantly higher incidence inHanoi from 2004to2009
trophic cycle, and to virus replication. In addition,
(most likelycluster)
stagnant water and higher humidity could augment the
epidemicduringarainyseason(9,12,15(cid:2)19).
Radius Relative Resultsfromspace(cid:2)timeclustering identifiedthehigh-
Location (km) Timeframe risk p riskareasoverthelargerregionandovertheyears.Using
the maximum spatial cluster size of 50% of the total
HoangMai 2.50 4/8(cid:2)19/7/2004 2.03 0.001
population, and the maximum temporal cluster size of
ThanhXuan 0.68 17/11(cid:2)31/12/2005 2.66 0.001
50% of the total population, we identified six among
DongDa 0.64 2(cid:2)16/11/2006 3.51 0.001
14 districts of Hanoi as having significant cluster pat-
DongDa 0.99 17/11(cid:2)16/12/2007 3.38 0.001
terns within a period of 29 days and limited at 1,000 m
HoangMai 0.29 19/8(cid:2)2/10/2008 5.58 0.001
on average. The areas recording the highest numbers of
HoanKiem, 1.56 2(cid:2)16/11/2009 6.41 0.001
space(cid:2)time clusters were Dong Da, Hoang Mai, and
HaiBaTrung
Thanh Xuan, with some expansion to the north-western
6
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
Detectionanddispersionofdenguefever
Hanoi, 2004 Hanoi, 2005
80 80
70 70
60 60
y) 50 y) 50
a a
e (D 40 e (D 40
m 30 m 30
Ti Ti
20 20
10 10
0 0
0.0 1.0 2.0 3.0 0.00 0.50 1.00 1.50 2.00
Space (Km) Space (Km)
Hanoi, 2006 Hanoi, 2007
180 180
160 160
140 140
Day)110200 ay) 110200
me ( 80 e (D 80
Ti 60 Tim 60
40 40
20 20
0 0
0.00 1.00 2.00 3.00 4.00 0.00 2.00 4.00 6.00
Space (Km) Space (Km)
Hanoi, 2008 Hanoi, 2009
180 180
160 160
140 140
120 120
ay)100 ay) 100
D D
e ( 80 e ( 80
m m
Ti 60 Ti 60
40 40
20 20
0 0
0.00 2.00 4.00 6.00 0.00 0.50 1.00 1.50 2.00
Space (Km) Space (Km)
Hoan Kiem Hai Ba Trung Thanh Xuan Dong Da Hoang Mai Tu Liem
Fig. 5. Scatterplotof significantspace(cid:2)time clustersin Hanoi from2004to 2009.
wards of Hanoi, where a higher population is concen- that DF had high spatial autocorrelation in three differ-
trated. Similar results were reported in Malaysia, where enttime periods. Thepatternswerecloselyrelatedtothe
geographical weighted regression analysis revealed that topographyoftheenvironment,inthatthevillagescloser
the spatial distribution of DF was closely related to together tended to have more similar baseline incidence
population distribution (9). Moreover, the result of their rates.
space(cid:2)time permutation scan statistics showed that most To the best ofour knowledge, this is the first study to
of the clusters were in medium or high population apply a spatial scan technique, using SaTScan software,
areas. Spatial clustering of disease is almost inevitable, toinvestigatethetemporalandgeographicalclusteringof
sincehumanpopulationsgenerallyliveinspatialclusters DF disease in Vietnam, in this case in Hanoi. The study
rather than random distributions. provides useful information on the prevailing epidemio-
Finally, spatial autocorrelation and logistic regression logicalsituationofDFinHanoicity.Thisnewknowledge
analysis are valuable tools for studying how spatial about the presence of hotspots of DF in the city can
patternschangeovertime(20(cid:2)26).Inthisstudy,wefound helpHanoiPreventiveMedicineCentertointensifytheir
7
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
DoThiThanhToanetal.
Table6. Changesofdenguefeverinlatitudeandlongitude, References
Hanoi, 2004(cid:2)2009
1. HoppMJ,FoleyJA.Global-scalerelationshipsbetweenclimate
andthedenguefevervector,Aedesaegypti.ClimChang2001;
Latitude Longitude
48:441(cid:2)63.
2. Honorio NA, Nogueira RMR, Codeco CT, Carvalho MS,
Changeinperiods OR 95%CI OR 95%CI
Cruz OG, Magalha˜es MAFM, et al. Spatial evaluation and
modelingofdengueseroprevalenceandvectordensityinRiode
Period3(cid:2)Period1 1.17 1.02(cid:2)1.34 1.06 0.93(cid:2)1.34
Janeiro,Brazil.PLOSNeglTropDis2009;3:e545.
Period2(cid:2)Period1 0.94 0.85(cid:2)1.03 0.98 0.87(cid:2)1.11 3. ShepardDS,SuayaJA,HalsteadSB,NathanMB,GublerDJ,
Period3(cid:2)Period2 1.09 0.99(cid:2)1.22 0.98 0.89(cid:2)1.06 Mahoney RT, et al. Cost-effectiveness of a pediatric dengue
vaccine.Vaccine2004;22:1275(cid:2)80.
Note: Period 1: 2004(cid:2)2005; Period 2: 2006(cid:2)2007; Period 3: 4. ThaiKTD,CazellesB,NguyenNV,VoLT,BoniMF,FarrarJ,
2008(cid:2)2009. et al. Dengue dynamics in Binh Thuan province, Southern
Vietnam:periodicity,synchronicityandclimatevariability.Negl
TropDisJ2010;4:e747.
remedialmeasuresintheidentifiedareasofhighDFpre-
5. NguyenLAP,ClementsACA,JefferyJAL,YenNT,NamVS,
valenceandchalkoutfuturestrategiesformoreeffective VaughanG,etal.AbundanceandprevalenceofAedesaegypti
DFcontrol. immatures and relationships with household water storage in
The study does have some potential limitations. The ruralareasinsouthernVietnam.IntHealth2011;3:115(cid:2)25.
6. Lawson AB, Williams FLR. An introductoryguide to disease
diseasecasedatawerenotsurvey-basedbutusedsentinel
mapping.Chichester,UK:JohnWiley&Son;2001.
surveillance data, which only records patients presenting
7. Hu W, Clements A, Williams G, Tong S. Spatial analysis of
at hospitals. The study analysed statistically significant notified dengue fever infections. Epidemiol Infect 2011; 139:
clustersofDFinHanoibutdidnotexaminetheircauses. 391(cid:2)9.
Futureresearchshouldfocusontheeffectofvarioussocio- 8. Jeefoo P, Tripathi NK, Souris M. Spatio-temporal diffusion
pattern and hotspot detection of dengue in Chachoengsao
economic and environmental factors that could affect
province, Thailand. Int J Environ Res Public Health 2011; 8:
diseasetransmission.
51(cid:2)74.
9. Seng SB, Chong AK, Moore A. Geostatistical modelling,
analysis and mapping of epidemiology of dengue fever in
Conclusion and policy implications
Johor state, Malaysia; 2005. Available from: http://www.
Thestudyhasshown thepresenceof long-termhotspots
business.otago.ac.nz/sirc/conferences/2005/13_su.pdf [cited 15
ofDFoccurrence,whichwashighestinDongDa,Hoang January2012].
Mai, and Thanh Xuan districts of Hanoi. We have also 10. Kulldorff M. (2006) SaTScan user guide (http://www.satscan.
shown the expansion of geographic distribution of DF org)v7.0,19(cid:2)20.
11. Ministry of Health. Guidelines on the surveillance, diagnosis,
over recent years. The results demonstrate the necessity
andtreatmentofdenguehemorrhagicfever.MedicalPublishing
to further improve our understanding of the impact of
House.Hanoi,Vietnam;1999.
socio-environmental change and ecosystem stress on the 12. Martens P, Huynen MMTE. Will global climate change
transmissionofDF.Thestudyhasillustratedhow,using reduce thermal stress in the Netherlands? Epidemiology 2001;
existing health data, spatial scan statistic and GIS can 12:753(cid:2)4.
13. Pham HV, Doan HTM, Phan TTT, Minh NNT. Ecological
providepublichealthofficialswithnecessaryinformation
factors associated with dengue fever in a central highlands
about the prevalence of statistically significant hotspots
province,Vietnam.BMCInfectDis2011;11:172.
of DF in the city, thus enabling them to chalkout more 14. Cuong HQ, Hien NT, Duong TN, Phong TV, Cam NN,
effectivestrategiestocontainthisscourge.Moreover,hot FarrarJ,etal.QuantifyingtheemergenceofdengueinHanoi,
spot analysis using GIS should be widely used in DF Vietnam:1998(cid:2)2009.PloSNeglTropDis2011;5:e132.
15. Martens P. Health transitions in a globalising world: towards
surveillance since it can help reallocate the resource to
morediseaseorsustainedhealth?Futures2002;37:635(cid:2)48.
deal with the outbreak more effectively. This effort will
16. Martens P, Huynen MMTE. A future without health? Health
contribute to dengue control strategy. dimensioninglobalscenariostudies.BullWorldHealthOrgan
2003;81:896(cid:2)901.
17. Martens P. Integrated assessment models. In: Valkering P,
Acknowledgements AmelungB,VanderBruggeR,RotmansR,eds.Morepuzzle-
solvingforpolicy:integratedassessmentfromtheorytopractice.
MaastrichtUniversity.Maastricht:ICIS;2006.
We thank Dr. Nguyen Nhat Cam, Vice-Director of Hanoi Pre-
18. CazellesB,ChavezM,McMichaelAJ,HalesS.Nonstationary
ventive Medicine Center and his staff for assistance in collecting
influence of El Nino on the synchronous dengue epidemics in
surveillancedata.
Thailand.PloSMed2005;2:e106.
19. AliM,WagatsumaY,EmchM,BreimanRF.Useofgeographic
Conflict of interest and funding informationsystemfordefiningspatialriskfordenguetransmis-
sion in Bangladesh: role for aedes albopictus in an urban
This study was funded by the Netherlands Higher Edu-
outbreak.AmJTropMedHyg2003;69:634(cid:2)40.
cation (NPT) project on ‘Strengthening teaching and
20. Almeida AS, Medronho RA, Valencia LIO. Spatial analysis
researchcapacityinpreventivemedicineinVietnam’. of dengue and the socio economic context of the city of
8
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)
Detectionanddispersionofdenguefever
RiodeJaneiro(SoutheasternBrazil).RevSaudePublica2009; 26. Hsieh YH, Chen CWS. Turning points, reproduction number,
43. and impact of climatological events for multi wave dengue
21. OoiE-E,GublerDJ.DengueinSoutheastAsia:epidemiological outbreaks.TropMedIntHealth2009;14:628(cid:2)38.
characteristics and strategic challenges in disease prevention.
CadSaudePublica2008;25(Suppl1):S115(cid:2)S24.
*DoThiThanhToan
22. NishiuraH.Mathematicalandstatisticalanalysesofthespread
BiostatisticsandMedicalInformaticsDepartment
ofdengue.DengueBull2006;30.
InstituteofTrainingforPreventiveMedicineand
23. LuL,LinH,TianL,YangW,SunJ,LiuQ,etal.Timeseries
PublicHealth
analysis of dengue fever and weather in Guangzhou, China.
HanoiMedicalUniversity
BMCPublicHealth2009;9:395.
Vietnam
24. BarbazanP,GuiserixM,BoonyuanW,TuntaprasartW,Pontier
Tel:[phone]
D,GonzalezJ-P,etal.Modellingtheeffectoftemperatureon
Email:[email protected]
transmissionofdengue.MedVetEntomol2010;24:66(cid:2)73.
25. Wu PC, Guo HR, Lung SC, Lin CY, Su HJ. Weather as an
effective predictor for occurrence of dengue fever in Taiwan.
ActaTrop2007;103:50(cid:2)7.
9
Citation:GlobHealthAction2013,6:18632-http://dx.doi.org/10.3402/gha.v6i0.18632
(pagenumbernotforcitationpurpose)