Home Articles Spatial and temporal dynamics of Dengue Hemorrhagic Fever Epidemics (Nakhon Pathom province,...

Spatial and temporal dynamics of Dengue Hemorrhagic Fever Epidemics (Nakhon Pathom province, Thailand, 1997-2001)

Wutjanun Muttitanon1, 2,5*, Pongpan Kongthong1 , Chusak Kongkanon1,
Sutee Yoksan2, 4, Jean Paul Gonzalez2, 3 and Philippe Babazan2, 3

1Department of Geography, Faculty of Education, Ramkhamhaeng University, Bangkok 10110, Thailand
2RCEVD – Center for Vaccine Development, Institute of Science and Technology for Research and Development, Mahidol University, Nakhon Pathom 73170, Thailand
3 Institute de Recherche pour le Developpement (IRD)Ur034, France.
4CVD, Mahidol University, Nakhon Pathom 73170, Thailand
5AIT – Asian Institute of Technologies, Pathumthani 12120, Thailand.
Email: [email protected]

Acknowledgement
This publication was supported by Institut de Recherche pour le Developpement (IRD)Ur034, France, and by fellowship Center for Vaccine Development, Institute of Science and Technology for Research and Development, Mahidol University, Thailand. We thank Dr. Sutee Yoksan for instituting the reporting system and helping with the analysis of the data.

Abstract
Several environmental factors modulate the distribution of dengue fever (DF) such as rainfall when the transmission peaks occur during the higher rainfall months, temperatures that reduce the mosquito extrinsic cycle duration, the human population density in urban areas, and the vector density. In order to take into account which variable. A Geographic Information Systems (GIS) has been build to create links between georeferenced data including medical records, socio-economic data, and environmental data. When applied to an epidemic-retrospective analytical study of DHF epidemics in Nakhon Pathom province (1997-2001), the GIS allowed a mapping of spatial variations of DHF incidence, the recognition of different temporal incidence patterns and the quantification of the spread of the diseases between defined spatial units. It showed that the diffusion process of these epidemics was of a contagious type as the distance between epidemics zones was significant lower than the average distance between one sub district to every sub-districts. This result indicates that these epidemics were likely to be due to the spread of a new or rare virus serotype in areas with a limited immune protection against the dengue virus for serotype differential level of specific herd immunity. 20-epidemic months was defined from 56 months from 1997-2001. There are negative correlation between the density of population and the DHF incidence rate. The average distance of epidemic sub district are not significantly smaller than the whole average distance, there are the global distribution. There are contagious diffusion , the probability of this emergence significantly decreasing with the distance from formerly epidemic sub districts.

1. Introduction
Dengue hemorrhagic fever (DHF) is caused by the dengue virus, comprising 4 serotypes, belonging to the genus Flavivirus, family Flaviviridae. The principal vector of dengue viruses is the mosquito Aedes aegypti. Dengue hemorrhagic fever (DHF) is one of the most important public health problems for Thailand and many tropical countries around the world. No treatment or vaccine is available, and vector control is the only method to control the dengue. Dengue hemorrhagic fever (DHF) is a viral disease worldwide distributed among all tropical area. It is caused by the dengue (DEN) virus, (genus Flavivirus, family Flaviviridae) which present four antigenic forms or serotypes: Den-1, Den-2, Den-3 and Den-4. In Thailand DHF has been endemic since 1958, with a cumulated total of 1 369 542 DHF number of cases. Several epidemic manifestation have been observed at two to four years of intervals and remain of a main concern for the Public Health authorities (Ministry of Public Health). In most of the studied areas two or three serotypes have been found co-circulating.

The main vector is the mosquito Aedes aegypti. As relevant to the virus transmission is anthropophilic, as females bite mainly human and also lay egg in man-made containers: (jars, cans, used tires…). Its short flight range, less than a kilometer., contributes also to a limited local spread of the disease. Vector control strategies are mainly based on mosquito population control by eliminating potential breeding sites.

As a consequence of vector biological features and urban environmental structure, two type of virus transmission can be described: i) The contagious/continuous type, dealing with intra communities transmission, mainly dependent on the density of human population and houses, and the number of vector females likely to carry the virus from one house to another; ii) the contagious/discontinuous type, inter communities, mainly dependent on the traffic among villages and towns and the transportation of viremic hosts or infectious vectors.

The knowledge of the mechanism of the DHF inter communities spread during epidemic periods is of primary interest in order to appreciate the areas at risk for high level of transmission and also to have precise view on the distance from the origin of an epidemic at which preventive control measures should be applied.

Assuming the stability several years at the epidemic scale of geographical factors involved in dengue transmission (urbanization, cultural and social characteristics), the emergence of DHF epidemics in an endemic area is likely to be due either to favorable climatic conditions, or to the emergence of a new or rare virus serotype, or to a combination of the two phenomena. These variant origins of an epidemic can induce different patterns of diffusion of the disease. An epidemic due to favorable climatic factors may cover a large area leading to the possible emergence of epidemic transmission anywhere in the surveyed area (random distribution). The emergence of a new serotype is more likely to exhibit particular spatial characteristics. It may start at the place where this serotype first arrived and then propagates to places where the specific herd immunity (towards this serotype) is low enough and the density of mosquito high enough, to allow a high level of transmission.

Assuming the second hypothesis the spread of a new serotype is likely to follow the main model of diffusion described for the spread of other types of moving phenomena (Hagerstrand 1952 quoted in Meade et al. 1988, pp 255). According to this model applied to the diffusion of an infectious disease, the probability for an area to be reached by the disease will depend on the distance to the formerly contaminated areas and, clusters of epidemic areas will secondary appear.

To test the validity of this model, a study was developed, covering the 1997-1998 and 2000-2001 DHF epidemics in Nakhon Pathom province in order to monitor spread of significantly higher levels of incidence rate (epidemic) among sub districts.

2. Material and methods

2.1 Data
Data on DHF were kindly provided by the Ministry of Public Health, Demographic data by the Administrative department, Ministry of Interior; and geographic map by the Royal Thai Survey Department.


Figure 1 Total DHF incidence in Nakhon Pathom Province; district level, recorded from January 1997 to August 2001

2.2 Study area
Nakhon Pathom province, is located in the central part of Thailand encompassing the latitude of 130 38′ 45.6″ to 140 10′ 37.2″ and the longitude of 990 51′ 10.8″ to 1000 17′ 6″. It covers 2164.35 km2 and has a population of 774,276 inhabitants. There are seven districts and 106 sub districts, where the population density ranges from 316 to 630 inhabitants / sq km. The province health department has reported 14,079 DHF cases during 1983 – 1992-2001, and during the last ten years, two DHF epidemics have broken out, in 1997-1998 and 2000-2001.

2.3. Method of analysis
The study aimed to describe the spatio-temporal dynamics of event, or it is for a sub district reaching a significantly higher than expected (epidemic) level of transmission of DHF.

An epidemic in a sub district is defined as a monthly period of time when the whole province has a number of cases significantly higher than expected and, a sub-district exhibits a significantly higher number of cases than the other ones.

An epidemic month has been defined as a month with a significantly higher number of cases during which the incidence is greater than expected value to the average plus one standard deviation of the monthly incidence specific of a glen month (i.e. January, February,…) observed from 1983 to 2001 (Barbazan P. et al 2001).

During an epidemic month, an epidemic sub-district is a sub-district where the incidence rate (for 100 000 inhabitants) is superior to the average incidence plus one standard deviation observed among every sub-districts during that month.

In a contagious model applied to infectious diseases, the spatial entities closer to an infective one are more at risk. In the spread of the epidemic phenomena among epidemic sub-districts, the observed distance between sub-districts reached each month by the epidemic, meaning facing an epidemic level of transmission is likely be smaller than the average expected distance between the whole sub-districts. To calculate these distances we used the euclidian distance between the centroid of sub districts.

  • The expected distance as the average distance between epidemic sub districts and every other sub districts.
  • The observed distance as the average distance between epidemic sub districts and all the other epidemic sub districts, during the same month (study of cluster), or from one month to the next one (study of spread).

H0 (null hypothesis)= the observed distance between epidemic sub-districts is not different than of the expected distance between every sub districts

H1 = observed distance < expected distance

The Z test is used to compare the average distances.


Figure 2 Observed and expected distances correlating between each epidemic sub districts during one and sub district epidemic during the next month
The method is applied to the study of two phenomena, 1) the occurrence epidemic sub district of cluster during one month and, 2) the spread of the epidemic from one month to the next one.

A Cluster is a geographically bounded group of occurrences of sufficient size and concentration to be unlike to have occurred by chance(Knox 1989 quoted in Alexander and J.Cuzick 1997). Clusters identification: comparison of observed distance between epidemic sub-districts and expected distance between any sub-districts, during one epidemic month.

The spread of epidemic of the diffusion of: the observed distance between epidemic sub-districts during one epidemic month and epidemic sub-districts during the next epidemic month is compared to the expected distance between epidemic sub-districts during one epidemic month and every sub-districts during the next epidemic month.

The distance, at which the epidemic can spread from one month to the next one, was estimated using a 5 km buffer created around each epidemic sub district, 1 = less than 5 km.; 2 = 5 – 10 km.; 3 = 10 – 15 km.; 4 = 15 – 20 km. and 5 = more than 20 km. For each monthly epidemic sub district, every epidemic sub district observed during the next month fall into the five classes according to its distance to the last epidemic sub districts.

3. Result

3.1 DHF Incidence
The incidence rate of DHF ranged from 76 to 2 835 cases per 100 000 inhabitants.

3.2. Epidemic months
Twenty epidemic month have been identified from January 1997 to August 2001 (Table 1); the number of epidemic months in one sub district ranged from 0 month (in 29 sub districts) to 11 months (Figure 3).
Table 1 Epidemic months chorological distribution from January,1997 to August,2001 (Nakhon Pathom, Thailand)

Year
Year  Jan  Feb  Mar  Apr  May  Jun  Jul  Aug  Sep  Oct  Nov  Dec
1997  X  X  X  X  X
1998  X  X  X  X  X
1999
2000  X  X  X
2001  X  X  X  X  X  X  X
X = Epidemic month


Figure 3 DHF average incidence rate per 100, 000 inhabitants per sub district and number of epidemic-month from January 1997 to August 2001.
3.3 Epidemic sub districts
During the first epidemic (1997-1998) a total of 49 epidemic sub districts have been identified, 61 during the second epidemic (2000-2001) and 31 during the two epidemics. The probability for one sub district epidemic during the first outbreak to be epidemic during the second epidemic is not significantly different of a random distribution.

3.4 Clusters
In the comparison of observed and expected distance during one month the epidemic periods, the average observed distances were significant lower than the average expected distance (according to the H1 hypothesis), significant level 0.95. In 78.67% of the distances.

3.5 Spread
The number of epidemic sub district during among the decrease with the distance to formerly epidemic sub districts distance during continuous epidemics months (Figure 4). Relative Frequency = Frequency for an epidemic sub district / frequency for all sub districts (The method used is strong enough or as looking at the map it appeared that during several more than one area (group of sub districts) was epidemic S+)


Figure 4 Comparison of DHF cases observed frequency of the distribution of classes of distance between every sub districts between epidemic during one month and every sub district epidemic during the next month VS the frequency of distance of class of distance between every sub districts.
Relative Frequency (RF) = Frequency for an epidemic sub district / frequency for all sub districts

4. Discussion

4.1 Epidemic months
The method used for the definition of epidemic month in province has been described (Barbazan P. et al, 2001). It allows to precisely frame epidemics, defined as periods during which the incidence is significantly over the observed average.

4.2 Choice of sub district as a spatial unit for this study
But this method to identify epidemic months at the province scale could not be used directly at the sub districts because these data are available over a short period of time and because of the great variance of results, due to the very low values of incidence often recorded (during the study period a null monthly incidence was reported in 66% of the months / sub districts) followed by epidemic periods.

The average surface of sub district is 21.83 km2 (standard deviation = 11.51). Sub districts comprise 3 to 24 villages, but villages cannot be used as spatial unit as many students and pupils do not live in their village and the address of patients is often only consistent at the sub districts scale. We considered in this study that displacements inside the sub district are sufficient to allow a mixity of the population and to consider the sub district as a homogenous unit towards DHF transmission.
4.3 Validity of the epidemic pattern as a representation of the spread of the disease
The method to identify epidemic sub districts allows to identify an “epidemic” pattern, in any sub district, whatever is its density of population as the distribution of epidemic sub district was not correlated to the density of population (Pearson’s correlation = -0.24, P = 0.71). Moreover, we used the incidence rate for 100 000 inhabitants to reduce the bias from the size of the population in every the sub districts.

Duration of the epidemic in a sub district : the definition of an epidemic month in provinces is a minima definition : i.e. a province where incidence is statistically higher than expected, is supposed to face and abnormal (epidemic) phenomena. During that period of time the sub districts of this province where the incidence is statistically higher than in the other ones are supposed to be responsible of this abnormal phenomena.

4.4 Clusters
The study describes the emergence of clusters of cases in sub districts during specific periods of time (epidemic month in the province) and their monthly spread over the province.

Clusters of epidemic sub districts maybe due to a geographic heterogeneity (density of urbanization, road network). Meanwhile, considering the whole studied period, the relative average distance between every sub districts having been epidemic at least during one month (more than 67% of the sub districts: average distance of 77 sub district are 28.36 km.) was not significantly smaller than the global average distance between every sub districts (average distance of 106 sub districts are 21.83 km.); meaning that their global distribution is not related to geographical factors.

The spread of epidemic sub districts follows the Hagerstrand’s model that has been used to describe many types of phenomena, such as the waves of innovation. Lost their energy with distance from the source of the innovation (Gould P.G. 1969) or the spread of new ideas (Hagerstrand 1952). In public health research it has been applied to the infectious influenza (Cliff et al 1986). Applied to the DHF epidemic in Nakhon Pathom, it means that during the epidemic periods each epidemic sub districts is the origin of the emergence of an epidemic in other sub districts during the next month, the probability of this emergence significantly decreasing with the distance from formerly epidemic sub districts. This model is the contagious type and maybe opposed to the random or homogenous model. In these two models occurrence of epidemic is due to an overall phenomena (for example and increase in temperature) and could be observed in any sub districts. In this case, the distribution of epidemic sub districts would not show any significant trend.

In the figure 2 observed distances are smaller than expected ones, but exhibit similar monthly variations. This is mainly because of the border effect, the propagation in neighboring provinces being not taken in account. During the months where epidemic sub districts are located on the periphery of the province, the average distance to other sub districts is larger than during the months where epidemic sub districts are located near the center of the province as directly neighboring sub districts (epidemic of not) are comparatively less numerous.

4.5 Origin of epidemics
The initial date and place of the emergence of the epidemic in the province cannot be identified, as the incidence progressively increases from the endemic pattern to the epidemic one. Meanwhile, the contagious distribution and spread of epidemic sub districts strongly suggests that the epidemic due to the emergence of a new or rare serotype in the epidemic sub districts.

DHF is endemic in Thailand and the different serotypes are largely distributed, at least two or three are generally found at the same time. Moreover, the information on serotypes is rare and limited. Therefore, the emergence of a DHF epidemic cannot be measured by using only the occurrence of the dengue infection or of a specific serotype, and indirect methods are necessary, such as the identification of epidemic months used here.

Inside human community the spread of DHF viruses from one house to neighboring houses maybe due to displacement of vectors or of hosts. On another hand, because of the relatively short range of flight of vectors females the spread of viruses among communities separated by several km cannot be due to the active dispersal of mosquitoes. Moreover, despite the transport of mosquito by cars has been described (Kuno 1995) , infected hosts are more like to be at the origin of the spread of viruses among communities.

According to Hagerstrand’s model, the probability to be reached by a new virus is inversely correlated to the distance between communities and positively correlated to the intensity of the communication between people, density of traffic, and road network. 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.

Areas not used by human to travel or limiting their displacements such as mountains, sea, boundaries act as barrier. This approach of the displacement of epidemics is likely to contribute to the delineation of areas at risk during epidemics, and to allow to public health to focus vector control activities in selected areas.

Reference

  • Barbazan, P., Sutee Yoksan, and Jean-Paul Gonzalez . 2001. Dengue Hemorrhagic Fever epidemiology in Thailand: Description and forecast of epidemics. Microbes and Infection. 1,4.
  • Cliff, A. D., P. Haggett, J.K. ord. 1986. Spatial aspects of influenza epidemic. London : Pion Limited.
  • Cuzick , J., and P.Elliott. 1997. Small-area studies : purpose and methods. Chapter 2 of Geographical & Environmental Epidemiology, eds. P.Elliott, J.Cuzick, E.English and R.Stern. 2d ed, New York : Oxford University Press.
  • D. Jay L., 1995. Probability and Statistics for Engineering and the Sciences. 4th ed, California : Duxbury Press.
  • Diggle, PJ. 1983. Statistical analysis of spatial point patterns .London: Academic Press.
  • English, D. 1997 . Geographical epidemiology and ecological studies. Chapter 1 of Geographical & Environmental Epidemiology, eds. P.Elliott, J.Cuzick, E.English and R.Stern. 2nd ed, New York : Oxford University Press .
  • Gubler, D.J. 1989. Surveillance for dengue and dengue hemorrhagic fever. Bulletin of PAHO 23(4) :397-404.
  • ……………. 1997. Dengue and dengue hemorrhagic fever: its history and resurgence as a global public health problem. Chapter 1 of Dengue and Dengue Hemorrhagic Fever, eds. D.J.Gubler and G.Kuno . Cambridge : The University Press.
  • Hagerstrand, T. 1952 . The propagation of innovation waves. Luud, Sweden : Gleerup.
  • Hay, S.I. 2000. An Overview of Remote Sensing and Geodesy for Epidemiology and Public Health Application. Remote Sensing and Geographical Information Systems in Epidemiology, eds. S.I.Hay, S.E.Randolph and D.J.Rogers. Great Britain : Academic Press.
  • Kuno, Goro .1995. Review of the factor modulating dengue transmission. Epidemiologic Reviews 12,2 : 321-335.
  • ……………..1997. Factors influencing the transmission of dengue viruses. Chapter 4 of Dengue and Dengue Hemorrhagic Fever, eds. D.J.Gubler and G.Kuno . Cambridge : The University Press.
  • Mayer, Jonathan D. 1983. The role of spatial analysis and geographic data in the detection of disease causation. Soc Sci Med. 17,16 : 1213-1221.
  • Meade, Milinda S., John W.Florin, and Wilbert M. Gesler. 1988. Medical Geography. London : The Guilford Press.
  • Morrison, Amy C., Arthur Getis, Marilyn Santiago, Jose G. Rigau-Perez, and Paul Reiter. 1998. Exploratory space-time analysis of reported dengue cases during an outbreak in Florida, Puerto Rico , 1991-1992. American Society of Tropical Medicine and Hygiene 58,3:287-298.