Dengue

Dengue Bulletin Volume 28 (2004)

 PDF Version

Spatial and Temporal Dynamics of Dengue Haemorrhagic Fever Epidemics,  Nakhon Pathom Province, Thailand, 1997-2001

Wutjanun Muttitanon*, Pongpan Kongthong**, Chusak Kongkanon**,
Sutee Yoksan***, Narong Nitatpattana***, Jean Paul Gonzalez^‡
and Philippe Barbazan^†#

*Asian Journal of Geoinformatics; Space Technology Application and Research Program,
Asian Institute of Technology. P.O. Box 4, Klong Luang, Pathumthani, Thailand
**Department of Geography, Faculty of Education, Ramkhamhaeng University, Bangkok 10110, Thailand
***Center for Vaccine Development (CVD), Institute of Science and Technology for Research and Development, Mahidol University, Nakhon Pathom 73170, Thailand
^†Research Center for Emerging Viral Diseases (RCEVD) – IRD – Center for Vaccine Development,
Institute of Science and Technology for Research and Development, Mahidol University,
Nakhon Pathom 73170, Thailand
^‡Institut de Recherche pour le Développement (IRD) Ur034, 213 rue La Fayette, 75480,
Paris cedex 10, France

Introduction

Dengue fever (DF) is a viral disease with a worldwide distribution in all tropical areas. It is caused by the dengue virus (genus Flavivirus, family Flaviviridae) which presents four antigenic forms or serotypes: DEN-1, DEN-2, DEN-3 and DEN-4. In Thailand, Dengue haemorrhagic fever (DHF) a severe form of dengue fever has been endemic since 1958, with a cumulative total of 1,369,542 cases till date^[1]. Epidemics occur with a periodicity of between two and four years; these epidemics are of significant concern for the public health authorities. In most of the areas where serotype identifications were performed, two or three serotypes were found to be co-circulating^[2].

The dengue virus is an arbovirus (arthropod-borne virus) transmitted by the mosquito Aedes aegypti (L.). Control of the spread of the disease focuses on vector control strategies based mainly on the elimination of potential breeding sites^[3]. A major attribute of the virus transmission is its anthropophilic behaviour, as females mainly bite humans and lay eggs in man-made containers near houses (for example, water jars, cans, used tyres). The short flight range of the vector, less than 1 km, contributes to the limited spread of the disease by an infected female. Most of the infections by dengue viruses are not severe and present asymptomatically, allowing infected patients to maintain normal activities.

Two types of viral spread can be described: (i) the diffusion of human infections as a function of the spatial distribution of houses and the limited flight range of infectious or infected Aedes aegypti females (intra-communal, contagious/ continuous); and (ii) inter-communal dispersion, largely a function of the stochastic movement of incubating/infectious humans and the transport via vehicles of virus-positive females^[4].

The understanding of the mechanism of the inter-community spread of DHF during epidemic periods is a primary factor likely to lead to an evaluation of the risk of virus transmission and disease dispersal^[5]. Moreover, it would provide some guidance on the distance from the spatial origin of an epidemic at which preventive control measures should be applied.

At a monthly time-scale, the main geographical factors involved in dengue transmission (urbanization, demography, cultural and social characteristics) are stable^[6]. A change in the pattern of monthly DHF transmission, such as the emergence of epidemics in an endemic area, should then rather be related to factors evolving with time: climate, density of vectors, emergence of a new or rare virus serotype, each type of factor inducing a specific pattern of diffusion of the disease^[7,8]. The emergence of a new serotype in a given population is likely to exhibit particular spatial characteristics. The outbreak would begin where the serotype first arrived and then move to places where a low specific herd immunity (towards this serotype) and a sufficient density of mosquito allow a high level of transmission. The spread of a new serotype is then likely to follow the main model of contagious diffusion described for the spread of other types of moving phenomena^[9]. Applied to the diffusion of an infectious disease, it means that the probability for an area to be reached by a contagious disease will be inversely correlated to the distance to the formerly contaminated areas, leading to clusters of epidemic areas.

In order to test the validity of this model in the frame of dengue dispersal, a study was conducted to describe the spread of significantly higher levels of incidence rate (of epidemic significance) among sub-districts. The study, done in a province of Thailand, covering the period 1997-2001, included two DHF epidemics.

Materials and methods

Data collection

Data on clinically diagnosed DHF cases were recorded at the Ministry of Public Health, the demographic data were provided by the Administrative Department of the Ministry of Interior, and the geographical maps by the Royal Thai Survey Department. DHF cases were defined according to WHO criteria^[10].

Population and study area

Nakhon Pathom province is a part of the central plain region in Thailand encompassing the latitude of 13° 38’45.6” N to 14° 10’37.2” N and the longitude of 99° 51’10.8” E to 100° 17’6” E. It covers 2,164 sq km, has a population of 774,276 inhabitants and includes 7 districts and 106 sub-districts (Figure 1a). The population density ranges from 153 to 623 inhabitants/sq km. The average surface area of sub-districts is 20.4 sq. km. The provincial health department reported 14,079 DHF cases during 1983-2001; two DHF epidemics occurred in 1997-1998 and 2000-2001 (Figure 2).

Figure 1. District scale approach: (i) Administrative limits of districts and sub-districts; density of population; main roads; (ii) Average incidence observed before the epidemics (cases/100,000 people), January 1992 – June 1997; (iii) Ratio of the incidence during the first three months of the DHF epidemic compared to the average incidence from
January 1992 – June 1997

Figure 2. Monthly DHF incidence in Nakhon Pathom province, Thailand, from January 1992 to August 2001

Method of analysis

The study aimed to describe the spatial-temporal dynamics at a monthly time-scale of a DHF epidemic among Nakhon Pathom’s 106 sub-districts considered as the spatial units. As a first step, epidemics were defined at the province level as periods of time (at least two consecutive months) when the incidence is higher than the average, plus one standard deviation of the monthly incidence of each month (i.e. January, February, etc.). The average was calculated over the entire 1983-2001 period^[11].

During these epidemic months (EMs), epidemic sub-districts (ESDs) were those where the monthly incidence was significantly higher than in other sub-districts. The threshold for a significantly higher incidence was leveled at the average monthly incidence (per 100,000 inhabitants) plus one standard deviation, observed among every sub-district during that EM.

In a contagious model for an infectious disease, the spatial entities close to an infected one were assumed to be more at risk to become infected than the distant ones. Applied to the diffusion of an epidemic phenomenon, it meant that the distance between the new epidemic sub-districts and the former ones (observed distance) should be shorter than the average (expected) distance between all the sub-districts. The distance between sub-districts was defined as the Euclidian distance between their centroids. The expected distance was the average distance between each ESD and every other sub-district. The observed distance was the average distance between each ESD and every other ESD, during the same month (cluster study), or from one month to the next (spread study).

H₀ (null hypothesis) = the average observed distance (between ESD) was not different from the average expected distances.

H₁ = average observed distance <average expected distance.

The Z test was used to compare the average distances.

The method was applied to the study of two phenomena: (i) the occurrence of clusters of ESD during one month; and (ii) the spread of the epidemic among sub-districts from one month to the next. A cluster is defined here as an aggregation of ESD (during one EM) of sufficient size and concentration to be unlikely to have occurred by chance, i.e. if the average distance (between these ESD) is shorter than the average distance between all sub-districts. The spread of the epidemic is based on the comparison of observed and expected distances during two consecutive EMs, i.e. the average distance between ESD during one epidemic month (EM_m) and ESD during the next epidemic month (EM_m+1), versus the average distance between ESD during EM_m and every sub-district during EM_m+1.

The (discrete) distance, at which an epidemic can spread in one month, was estimated by summing the number of ESD centroid during EM_m+1 observed inside circles centred on each ESD during EM_m and drowned at 5 km; 10 km; 15 km; 20 km and out of 20 km. This number is compared to the number of sub-districts centroids distributed in these surfaces to build a relative risk index.

Relative risk index =

Results

At the district scale, the DHF incidence was higher in the central-west part of the province. The epidemic broke out in the northern district with a medium density of population (Figures 1b and 1c). At the sub-district scale, the maximum DHF incidence rate reached 540 cases per 100,000 inhabitants in July 1997.

Nineteen EMs were identified in Nakhon Pathom province from January 1997 to August 2001 (Table); the number of EMs in one sub-district ranged from 0 month (in 27 sub-districts) to a maximum of 11 months.

Table. Chronological distribution of epidemic months from January 1997 to August 2001 in Nakhon Pathom province, Thailand

A total of 49 ESDs were identified during the first outbreak (1997-1998) and 61 during the second outbreak (2000-2001); 31 sub-districts were epidemic-affected during the two outbreaks. The probability of one sub-district being epidemic-affected during the first outbreak to be epidemic-affected during the second outbreak as well is not significantly different from a random distribution (P<0.05).

Cluster study: During EM, 78.67% of the average observed distances were
significantly lower (P<0.05) than the average expected distance, characterizing the occurrence of clusters of ESDs according to the H₁ hypothesis.

Spread study: The distance observed between ESDs during EM_m and ESDs during EM_m+1, was significantly smaller than the expected distance (Figure 3), characterizing the contagious spread of DHF among ESDs according to the H₁ hypothesis.

Figure 3. Study of the spread of DHF epidemic among sub-districts: observed distances are calculated between epidemic-affected sub-districts during one month and the epidemic-affected sub-districts during the next month; expected distances are calculated between all the sub-districts (see text for details)

As a consequence, the distribution of ESDs during EM_m+1 in surfaces drowned round ESD during EM_m showed a significant (P<0.05) aggregation within the first two circles (5 and 10 km).

Discussion

The method used for the identification of an epidemic month in the province^[11] allowed a precise framing of epidemics, defined as periods during which the incidence was significantly higher than the observed average over the complete time series of data (19 years). Meanwhile, this method could not be used directly at the sub-district scale because of the lack of long-time data series on the incidence at this scale. Moreover, the variance of the DHF incidence in most of the sub-districts was very high because of the low values of incidence often recorded (during the study period a null monthly incidence was reported in 66% of the 5,936 months X sub-districts). Similarly, the village scale (from 3 to 24 villages per sub-district) could not be used as the spatial unit as the addresses of patients were often consistent only at the sub-district scale and many students and pupils did not live in their village.

We assumed in this study that sub-districts could be considered as homogeneous small areas and that human displacements were sufficient to produce a homogenization of the population, allowing consideration of the sub-district as a unit towards DHF transmission. An ‘epidemic’ pattern can then be identified in any sub-district, whatever its density of population: the distribution of ESD was not correlated to the density of population (Pearson’s correlation = -0.24, P = 0.71). Moreover, we used the incidence rate per 100,000 inhabitants to reduce the bias related to the size of the population.

The geographical heterogeneity of the environment, e.g. the density of urbanization or the road network, could also be at the origin of clusters of ESDs. Meanwhile, after the two epidemics, ESDs were found to be uniformly distributed over the entire province, and the spatial distribution of all sub-districts having been epidemic-affected at least during one month (67% of the sub-districts) was not significantly different from the spatial distribution of all sub-districts (average distances not different, P=0.95). Meanwhile, the results implied a high degree of spatial auto-correlation, meaning that neighbouring sub-districts shared similar characteristics, such as the level of immunity for the different serotypes (due to a similar epidemiological history) or the density of the vector.

As shown in Figure 3, the observed distances are smaller than the expected ones, but exhibit similar monthly variations. This was mainly because of a border effect, the propagation in sub-districts located in neighbouring provinces not being taken into account. During the months where ESDs were located on the periphery of the province, the average distance to other sub-districts was larger than during the months where ESDs were located near the centre of the province, as several neighbouring sub-districts located in other provinces (epidemic or not) were ‘missing’ in the calculation. The absolute level of expected and observed distances was then directly dependent on the location of the ESD in the province.

The spread of the epidemic between sub-districts followed Hagerstrand’s model that has been used to describe many types of phenomena, such as the spread of new ideas^[9] or the waves of innovation which lose their ‘energy’ when the distance from the source increases^[12]. In public health research it has been applied to infectious influenza^[13]. Applied to the DHF epidemic in Nakhon Pathom, it means that during the epidemic periods the ESDs were the origin of the emergence of epidemics in neighbouring sub-districts during the next month. The probability of this emergence at m+1 significantly decreased with the distance from the former ESD. This model is of a contagious type and may be opposed to a random or homogeneous model. In the homogeneous models the occurrence of an epidemic could be due to a global phenomenon, such as an increase in temperature, which should have been observed in any sub-district, leading to a random distribution of ESD^[14] and an observed distance not different from the expected distance.

Inside human communities (villages) it has been shown that the spread of DHF viruses from one house to neighbouring houses due to the displacement of infected vectors or hosts follows a pattern similar to what we have described between sub-districts^[15]. Meanwhile, among communities separated by several kilometers, the spread of viruses cannot be due to the active dispersal of mosquitoes or to their transport by car, which is much more rare than the displacement of infected hosts. More than 80% of infections by dengue virus are unapparent or not severe, allowing healthy carriers to travel. The presence of sufficient densities of vectors in destination communities is also necessary to allow the transmission of the virus after it has been imported.

The contagious distribution and spread of the two DHF epidemics among the sub-districts strongly suggests that they were due to the emergence of a new or rare serotype. DHF is endemic in Thailand and the different serotypes are largely distributed, as at least two or three serotypes are generally found at the same time in the same area^[2,16]. Meanwhile, during epidemic periods the relative prevalence of the serotypes varies, as the 2000 epidemic in Bangkok that was due mainly to the serotypes DEN-1 and DEN-2 (each reaching 42% of total isolations), whereas the 1994 epidemic was due mainly to the rise in DEN-4 (36%). But as serology and isolation of viruses are rarely performed, the emergence of a DHF epidemic cannot be forecast by using these methods. Indirect methods, such as the statistical identification of epidemic months, are then necessary to identify early the emergence of DHF epidemics.

The epidemiology of DHF in Thailand is changing^[17]. This approach of the displacement of epidemics is likely to contribute to the localization of the origins of outbreaks and the delineation of areas at risk during epidemics, as well as to help public health authorities to focus vector control activities on selected areas.

Acknowledgements

The study and the preparation of this paper was supported by the Institut de Recherche pour le Développement (IRD)-Ur034, France, by a fellowship to the Center for Vaccine Development, Institute of Science and Technology for Research and Development, Mahidol University, Thailand, and by the Department for Technical and Economic Cooperation, Thailand. We thank Professor Natth Bramapavarati for his constant support.

References

1.      Annual epidemiological surveillance report, 1983-2002. Bangkok: Ministry of Public Health, Department of Disease Control, Bureau of Epidemiology, 2002.

2.      Vaughn DW, Green S, Kalayanarooj S, Innis BL, Nimmannitya S, Suntayakorn S, Rothman AL, Ennis FA and Nisalak A. Dengue in the early febrile phase: viremia and antibody responses. Journal of Infectious Diseases, 1997, 176: 322-330.

3.      World Health Organization. Dengue prevention and control. Executive Board, 109^th Session, 2001, EB109/16.

4.      Meade MS, Florin JW and Gesler WM. Medical Geography. London, UK: The Guilford Press, 1988.

5.      Cuzick J and Elliott P. Small-area studies: purpose and methods. Chapter 2 of Geographical & Environmental Epidemiology (Eds. P. Elliott, J. Cuzick, D. English and R. Stern). 2^nd ed. New York: Oxford University Press, 1997.

6.      Kuno G. Review of the factor modulating dengue transmission. Epidemiologic Reviews, 1995, 12(2): 321-335.

7.      Focks DA, Daniels E, Haile DG and Keesling JE. A simulation model of the epidemiology of urban dengue fever: literature analysis, model development, preliminary validation, and samples of simulation results. American Journal of Tropical Medicine and Hygiene, 1995, 53(5): 489-506.

8.      Mayer JD. The role of spatial analysis and geographic data in the detection of disease causation. Social Science and Medicine, 1983, 17(16): 1213-1221.

9.      Hagerstrand T. The propagation of innovation waves. Lund studies in Geography, Series B, 1952, 4.

10.   World Health Organization. Dengue haemorrhagic fever: diagnosis, treatment and control, 2^nd edition. Geneva: 1997.

11.   Barbazan P, Yoksan S and Gonzalez JP. Dengue Hemorrhagic Fever epidemiology in Thailand: Description and forecast of epidemics. Microbes and Infection, 2002, 1(4): 699-705.

12.   Gould PR. Spatial diffusion. Washington, US, Association of American Geographers, 1969.

13.   Cliff AD, Haggett P and Ord JK. Spatial aspects of influenza epidemic. London, Pion Limited. 1986.

14.   Jackson EK. Climate change and global infectious disease threats. The Medical Journal of Australia, 1995, 163(4): 570-573.

15.   Morrison AC, Getis A, Santiago M, Rigau-Perez JG and Reiter P. Exploratory space-time analysis of reported dengue cases during an outbreak in Florida, Puerto Rico, 1991-1992. American Journal of Tropical Medicine and Hygiene, 1998, 58(3): 287-298.

16.   Burke DS, Nisalak A, Johnson DE and Scott RM. A prospective study of dengue infections in Thailand. American Journal of Tropical Medicine and Hygiene, 1988, 38(1): 172-180.

17.   Chareonsook O, Foy HM, Teeraratkul A and Silarug N. Changing epidemiology of dengue hemorrhagic fever in Thailand. Epidemiology and Infection, 1999, 122: 161-166.