|Year : 2017 | Volume
| Issue : 2 | Page : 74-80
Spatiotemporal clustering of dengue cases in Thiruvananthapuram district, Kerala
Joanna Sara Valson1, Biju Soman2
1 PhD Scholar, Achutha Menon Centre for Health Science Studies, Sree Chitra Tirunal Institute for Medical Sciences and Technology, Thiruvananthapuram, Kerala, India
2 Additional Professor, Achutha Menon Centre for Health Science Studies, Sree Chitra Tirunal Institute for Medical Sciences and Technology, Thiruvananthapuram, Kerala, India
|Date of Web Publication||2-Jun-2017|
Achutha Menon Centre for Health Science Studies, Sree Chitra Tirunal Institute for Medical Sciences and Technology, Thiruvananthapuram - 695 011, Kerala
Source of Support: None, Conflict of Interest: None
| Abstract|| |
Background: Dengue cases are increasing in Kerala since 2010. Information on clustering of cases across locations and time periods is vital for disease surveillance and timely control. Objectives: The objective is to study spatiotemporal clustering of dengue cases and their climatic and physioenvironmental correlates in Thiruvananthapuram district during 2010–2014. Methods: Health department data on reported cases of dengue were obtained from January 2011 to June 2014. Cases were individually geocoded, using Google Earth. Moran's I index was estimated to analyze spatial autocorrelation using GeoDa software. Space–time clustering across 178 geo-divisions within the district was analyzed using SaTScan software. Correlation analysis was done for space–time clustering with climatic variables. Results: Definite spatial and temporal trends were found on analysis of a total of 8279 dengue cases. Significant spatial autocorrelation (Moran's I = 0.32, P< 0.01) and space–time clusters with very high log-likelihood ratios (P < 0.01) were found across geo-divisions. Pallichal panchayat was the most likely cluster in every year. The monthly incidence of dengue cases showed a significant positive association (P < 0.05) with a 2-month lag of mean minimum temperature (ρ = 0.39), 1-month lag of rainfall (ρ = 0.33), and 1-month lag of humidity (ρ = 0.38). Dengue occurrences showed an inverse association (P < 0.01) with mean maximum temperatures of the respective months (ρ= -0.48). Conclusion: Spatial analysis using epidemiological tools reveals spatial and temporal clustering of dengue cases within the district and their association with climatic parameters. This information can be used in controlling outbreaks in the future. This work upholds scope and feasibility of geospatial research in public health in India.
Keywords: Climate, clustering, dengue, geospatial, Kerala
|How to cite this article:|
Valson JS, Soman B. Spatiotemporal clustering of dengue cases in Thiruvananthapuram district, Kerala. Indian J Public Health 2017;61:74-80
|How to cite this URL:|
Valson JS, Soman B. Spatiotemporal clustering of dengue cases in Thiruvananthapuram district, Kerala. Indian J Public Health [serial online] 2017 [cited 2022 Aug 18];61:74-80. Available from: https://www.ijph.in/text.asp?2017/61/2/74/207409
| Introduction|| |
Dengue fever has emerged as a rapidly rising infectious disease in tropical countries. Asian and Latin American countries are at a high risk for severe dengue. Epidemiologists have recognized the importance of the spatial component in analyzing disease occurrence. Thiruvananthapuram district of Kerala has been endemic to dengue fever and has witnessed a substantial rise in the incidence of dengue cases since 2010. Control of dengue fever has remained a huge public health problem in Thiruvananthapuram, adding to the intermittent rains and monsoon season for the most part of the year. Precipitation, rainfall, and mean ambient temperature have been associated with the incidence of dengue fever.,, Geospatial techniques to analyze distribution and incidence of dengue fever in this district had not been attempted. Hence, this study aims to use geospatial techniques to find occurrence and clustering of dengue fever in the district.
| Materials and Methods|| |
Thiruvananthapuram, the capital city of Kerala, is located at the southernmost border of India. It is located between North latitudes 8°17′ and 8°54′ and East longitudes 76°41′ and 77°17′. The administrative divisions include 73 grama panchayats, 4 municipalities, and 1 corporation. The study area was demarcated into 178 geo-divisions (73 grama panchayats, 4 municipalities, and 100 corporation wards) that were the basic geographic units for analysis. The climate is mostly hot tropical, with the monsoon season extending from June to November.
Ethical clearance was obtained from the Institutional Ethics Committee of the author's institution.
Dengue case data from January 2011 to June 2014
Dengue case data from January 2011 to June 2014 were obtained from the Directorate of Health Services (Kerala) and verified with the data from the Public Health Laboratory, Thiruvananthapuram. Dengue being a notifiable disease, all the laboratory-confirmed cases, including those from the private sector, were reported to the health services.
Meteorological data from January 2011 to June 2014
Meteorological data from January 2011 to June 2014 were obtained from the Indian Meteorological Department. Monthly mean maximum temperature, precipitation, and relative humidity were captured from four substations in the district.
Spatial data, including spatial file of the rural and urban divisions of the district, were obtained from geospatial resources at Achutha Menon Centre for Health Science Studies.
Mapping of dengue fever cases
Mapping of dengue fever cases was done by geo-coding addresses of reported dengue cases using Google Earth.
Software used for analysis
SaTScan software (http://www.satscan.org/) (version 9.3.1) was used for space–time statistic test. R software (http://www.r-project.org/) (version 3.0.2; R development Core Team) and SPSS (IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp) were used for data analysis. GeoDa (http://geodacenter.asu.edu/) was used for spatial autocorrelation analysis., QGIS version 2.0.1 (http://www.qgis.org/en/site/) was used for geospatial mapping and analysis.
Spatial analysis was done panchayat- and block-wise for the study period. Monthly occurrences were analyzed. Choropleth maps were created in the QGIS software package for the number of cases reported in each panchayat and for the dengue fever occurrence per population density in each administrative division.
Spatial-autocorrelation analysis was done using GeoDa software, both global and local measures were analyzed. The global Moran's I index is a significance test for clustering and local Moran's I is a test for clusters. The global Moran's I is visualized by means of a Moran scatter-plot in which the slope of the regression line corresponds to Moran's I. Global Moran's I was estimated by testing a null hypothesis that there is a homogeneous distribution of dengue fever cases in the whole area of investigation. The Moran's I value ranges between +1 and −1. A value close to zero would indicate a spatially random pattern. A negative value would indicate negative spatial autocorrelation whereas a positive value would indicate a positive spatial autocorrelation. Analogous local measures are called local indicator of spatial association (LISA). The local analysis of LISA is visualized in the form of significance and cluster maps. The maps depict locations with significant local Moran's I statistics (LISA significance maps) and classify those locations by the type of associations (LISA cluster maps). Here, local Moran's I for each year was calculated by means of a neighborhood matrix, based on the criterion of “common border” (areas considered as neighbors). The Moran's I significance level was estimated using Monte-Carlo permutation test, which tests a null hypothesis that the dengue fever cases are randomly distributed. The number of permutations was selected to be 999 so as to reject the null hypothesis at a significance level of 0.001.
The local Moran's I estimated would also produce cluster maps. The spatial clusters are categorized as high-high, low-low, high-low, and low-high. A high-high cluster would mean higher incidence in the neighboring regions, while a low-low cluster would mean lower incidence in the neighboring regions. Both high-low and low-high clusters were considered outliers. This analysis was done for each year from 2011 to 2014.
Space–time clustering was analyzed for notified dengue cases yearly. Kulldorff space-scan statistic was calculated using SaTScan software., This statistic is defined by a cylindrical window with a circular geographical base and with a height corresponding to time, which is then moved in space and time so that we can obtain an infinite number of overlapping cylinders of different sizes and shapes, covering the entire study region, where each cylinder reflects a possible cluster. For each cylindrical window, the scan statistic tests the null hypothesis against the alternative hypothesis that there is an elevated risk of dengue within a window, compared to outside the window. Potential clusters are detected by calculating a maximum likelihood ratio for each cylindrical window. The window with the maximum likelihood ratio will be considered the most likely cluster. A large number of random replications of the dataset were done under the null hypothesis to obtain a P value through Monte-Carlo hypothesis testing. The maximum circle radius of the cylindrical window to detect disease clusters was set at 1 km. The maximum temporal (to detect clusters both in space and time) window was set at 50% of the study period (each year from 2011 to 2014). Primary and secondary space–time clusters were detected based on the log-likelihood ratio. (The primary cluster is the one with the maximum likelihood and that which is least likely to have occurred by chance. Secondary clusters will almost be identical with the most likely cluster and have almost a high likelihood value.) The significance (P value) of the clusters was set to 0.001 (999 permutations of Monte-Carlo simulation test which compares the rank of the maximum likelihood from the real data set with the maximum likelihoods from the random datasets) so as to reject the null hypothesis stating no space–time clustering at 0.05 level.
The geo-division-based space–time clusters were demonstrated using discrete Poisson regression model using the number of dengue cases in each geocode division and their respective populations at risk (SaTScan User Guide version 9.4, http://www.satscan.org). The input files were case, coordinate, and population files. Case file has information on reported dengue cases in each geo-division; the coordinate file has the geocordinates (latitude and longitude) of the centroid (geographical center) of each geo-division; and the population file has information on the population of each geo-division, to estimate the background population at risk. Year-wise analysis was done from 2011 to 2014.
The relation between clustering of dengue cases and climatic factors was determined using Spearman's rank correlation analysis.
Data storage and monitoring
The data were collected and stored with password encryption after receipt from the Directorate of Health Services. There was no sharing of data with anyone other than the authors themselves. Identifiers were removed, and anonymous data were used for analysis.
| Results|| |
Geographic information system mapping of dengue fever cases
All the 8279 cases were geo-tagged using Google Earth. Point maps of reported cases in each year from January 2011-June 2014 were created using QGIS software which showed higher occurrence in the corporation area (the urban region of Thiruvananthapuram district). Not many reported cases were found in the eastern areas. The eastern areas are highlands and are sparsely populated. Relatively higher number of dengue cases were reported during monsoon months. The Choropleth maps of the occurrence of dengue fever showed a similar pattern [Figure 1]. It revealed higher occurrence in the corporation area and the coastal regions, while fewer occurrences toward the northeastern regions.
|Figure 1: Choropleth maps showing distribution of dengue fever cases in (a) 2011, (b) 2012 (c) 2013 and (d) 2014.|
Click here to view
Spatial distribution of dengue fever
The block-wise distribution showed that the Thiruvananthapuram corporation area accounted for the highest number of cases each year, followed by Nemom block. Among the municipalities, the Nedumangadu municipality had the highest number of reported cases. Among the block panchayats, Pallichal had the highest number of dengue fever cases reported in the years 2011–2013, while in the year 2014, Poundukadavu had the highest number of dengue fever cases reported from January 2011-June 2014. As Thiruvananthapuram corporation had very high number of cases, the corporation ward-wise distribution was used for further analysis. Thus, a total of 178 geo-divisions (73 grama panchayats, 100 corporation wards, and 4 municipalities) were the basic geographic units of analysis.
Spatial autocorrelation of dengue fever cases
The total cases reported for the 4-year period were aggregated by month and panchayats/municipality/corporation areas. The units of analysis were 178 geo-divisions. Moran's I scatter plot for the study period (January 2011–June 2014) was plotted using spatial correlation analysis. The global Moran's I showed significant autocorrelation for the 4 years (significance level <0.01). Cluster maps were obtained for hotspots (high-high clusters). The clusters were located in the geo-division every year. There appeared to be a repetition of the hotspot every year (from 2011 to 2013).
Case-wise space–time clusters were analyzed using the space–time permutation model. The clusters are depicted in [Figure 2] However, further analysis could not be pursued at the individual case level because of the nonavailability of micro-level climatic data. Year-wise space–time clusters were then analyzed using SaTScan software. A maximum spatial cluster size of 50% of the population at risk and a circle of radius 1 km were selected for analysis. There were 45 space–time clusters in 2011, 78 space–time clusters in the year 2012, 97 space–time clusters in the year 2013, and 6 space–time clusters in 2014. The primary cluster and five secondary clusters in each year are listed in [Table 1]. Pallichal panchayat had the highest number of cases reported each year from 2011 to 2013, which were much higher than the expected cases. A very high log-likelihood ratio was also observed in this panchayat, hence causing the same to be the most likely cluster. Nevertheless, Karakulam and Vilappil were observed to be secondary clusters every year.
|Figure 2: Space–time clusters of dengue fever (case wise) in (a) 2011, (b) 2012, (c) 2013 and (d) 2014.|
Click here to view
|Table 1: Primary and secondary space.time clusters in each year (2011-2014)|
Click here to view
Relationship with climatic variables
Pearson's correlation analysis was performed to find relation between the month-wise number of dengue fever cases with the meteorological variables (mean minimum temperature, mean maximum temperature, average humidity, and average rainfall). The results are summarized in [Table 2]. The correlation analysis showed that the monthly mean minimum temperature was highly correlated with the number of dengue fever cases for each year with a lag of 2-month period. The monthly mean maximum temperature showed a correlation for dengue fever cases in the same month. Monthly average humidity and average rainfall were correlated with the dengue fever cases with a lag of 1 month (significance <0.05).
|Table 2: Spearman's rho (ρ) of monthly dengue cases with mean temperature, average rainfall, and humidity|
Click here to view
| Discussion|| |
This study explored the distribution of dengue fever cases in Thiruvananthapuram district using spatial and spatiotemporal analysis. The findings demonstrated that dengue fever occurrences are nonrandom in nature and that there are significant clusters of cases across Panchayats. It aimed at providing useful information for the health system to improve surveillance measures.
Point pattern analysis of diseases is found to be effective in disease surveillance and control, complementary to other methodological approaches. Spatial analysis of vector-borne diseases has been evidenced to be important tool for finding continuous risk surfaces and also to reveal heterogeneous patterns of risk at finer scales. Here, routine surveillance data were used for spatial analysis. Data scrutiny showed improvement in the quality of routine surveillance data over 2011–2014.
Analysis of spatial clusters of dengue fever using spatial autocorrelation has been done across countries to detect hotspots of occurrence. In Ecuador, spatial analysis of dengue fever from 2005 to 2009 found an autocorrelation of 0.37. While in Guangdong province of China, an autocorrelation of 0.24 was evidenced from 2005 to 2011., They also analyzed spatial autocorrelation across the years and found significant spatial clustering for the years 2005–2006 and 2009–2011. Jeefoo et al. also reported spatial autocorrelation of dengue fever cases in Thailand, significant for all the years from 1999 to 2007. Our study has found significant spatial clusters in all the years from January 2011-June 2014.
A systematic review on the dengue fever incidence in Asia-Pacific region has stated that every 2 years at least two more countries were added to the dengue fever-affected zone from 1955 to 2004., The reviewers have attributed economic growth without proper planned urbanization to be the possible cause. Incidentally, Thiruvananthapuram also has been undergoing rapid urbanization, especially in the corporation area. It has been recorded that the urban population in Thiruvananthapuram is higher than state average.
There seems to be an increasing trend of number of space–time clusters in Thiruvananthapuram district from 2011 to 2013. This could be because of the epidemiological trend of the disease and better surveillance measures. The Integrated Disease Surveillance Project (IDSP) at Thiruvananthapuram under the National Vector-borne Disease Control Programme was well established by 2012. However, it has been recorded that there are inherent weaknesses in the system by the way of frequent turnover and lack of motivation among the staff. Analysis of spatiotemporal trend of dengue fever from 2000 to 2009 in Bangladesh showed a decreasing trend despite the absence of a routine dengue vector control program.
Although there are routine dengue vector control measures in place, it is necessary to look at various other contributing factors to curb the occurrence of dengue fever. Recent years have witnessed the difficulties in urban waste management too.,,, The solid waste management was a cause of concern in this district since the closure of Vilappilsala plant; the plant could not handle large quantities of wastes that were brought from urban areas and was finally closed down due to strike from local residents. People were left with no choice, and the habit of indiscriminate dumping of wastes and half-burnt rubbishes around the premises has made favorable conditions for mosquito breeding in the district, especially in its urban areas.,
A review in 2011 has stated that climatic factors including rainfall, humidity, and temperature are closely linked to mosquito density. While relative humidity impacts the flight behavior of the mosquitoes, warmer temperature affects development and cooler temperature affects reproduction rates of mosquitoes. Banu in 2013 has reported that climatic variables can forecast dengue fever outbreaks within a period of 1–5 months. In our study, monthly average rainfall and humidity were associated with the dengue fever cases with a 1-month lag while mean minimum temperature was associated with dengue fever occurrence with a lag of 2 months. These findings were concurrent with the study from Bangladesh that rainfall and humidity were found to be significantly associated with dengue fever incidence with highest effects in the 2-month lag period. The study from Bangladesh also reported that a 2-month lag in rainfall and 1-month lag in temperature were found to be explanatory to the relationship between meteorological variables and dengue fever incidence. Jeefoo et al. have also reported a very high correlation with rainfall and relative humidity of 1 month before dengue occurrence. This study has indicated a similar relationship between climatic variables and dengue fever occurrence in Thiruvananthapuram district. On the contrary, an inverse association of monthly mean maximum temperature was found with dengue fever occurrence. Similar finding was reported in China in the year 2013.
We presume the study is special because we had used routine health data, adding much value to the public health surveillance. This was the first attempt to spatially map the dengue fever cases in Thiruvananthapuram district over years enabling us to analyze temporal trends. Finally, it was an attempt to incorporate technological advancements in routine public health management.
One potential limitation of this analysis could be the underreporting of cases, especially from private health institutions. However, that should not be a major concern here as the data were more than sufficient for trend and cluster analysis. The nonavailability of micro-level, climatic data, and the lack of social, cultural, behavioral or comorbid characteristics of the individual patients have limited the scope of analysis.
In spite of the above-mentioned limitations, the study implies that geospatial analysis of the routine public health data on reported dengue fever cases could be used to map spatiotemporal clustering of cases. It is also noted that the surveillance has improved across the years. This could be a valuable input for resource allocation and control measures. Use of the IDSP module in the DHIS2 (District Health Information System Software, Version 2, UiO, Oslo, HISP) will further refine the data, and geo-tagging of cases would be easier in the future. It is unfortunate that the meteorological data from around ten automated weather stations in the district are not being archived, which could have provided more robust climatic data for analysis. We hope this study could be showcased to convince the authorities on the utility of such microclimate data for public health surveillance.
Future research could include analysis of space–time clusters based on individual cases with climatic and other physiographic features. Household surveys in selected locations in the cluster and noncluster localities should reveal the social and behavioral patterns that favor mosquito breeding, which could in turn be addressed in our intervention strategies.
| Conclusion|| |
The study reveals that dengue fever in Thiruvananthapuram district occurs in a nonrandom manner. Spatial and space–time clusters of dengue fever occur every year. The significant correlation of dengue cases with climatic variables and entomological surveillance data show that timely geospatial analysis of the available routine health data could be useful in the prediction of potential outbreaks. Hence, we conclude that geospatial analysis health data should be a routine public health activity, even in resource-poor settings like India, which will help in the control and containment of infectious diseases such as dengue fever.
The authors acknowledge the support provided by the Directorate of Health Services, Kerala, and the Meteorological Centre, Trivandrum, for sharing data.
Financial support and sponsorship
This study was financially supported by the Kerala State Council for Science, Technology and Environment (No. 97/SPS/2014/CSTE).
Conflicts of interest
There are no conflicts of interest.
| References|| |
Kumar NP, Jayakumar PR, George K, Kamaraj T, Krishnamoorthy K, Sabesan S, et al.
Genetic characterization of dengue viruses prevalent in Kerala state, India. J Med Microbiol 2013;62:545-52.
Promprou S, Jaroensutasinee M, Jaroensutasinee K. Climatic factors affecting dengue haemorrhagic fever incidence in Southern Thailand. Dengue Bull 2005;29:41.
Castillo KC, Körbl B, Stewart A, Gonzalez JF, Ponce F. Application of spatial analysis to the examination of dengue fever in Guayaquil, Ecuador. Procedia Environ Sci 2011;7:188-93.
Jeefoo P, Tripathi NK, Souris M. Spatio-temporal diffusion pattern and hotspot detection of dengue in Chachoengsao Province, Thailand. Int J Environ Res Public Health 2011;8:51-74.
Kulldorff M, Heffernan R, Hartman J, Assunção R, Mostashari F. A space-time permutation scan statistic for disease outbreak detection. PLoS Med 2005;2:e59.
Anselin L. Exploring spatial data with GeoDaTM: A workbook. Urbana 2004;51:61801.
Er AC, Rosli MH, Asmahani A, Mohamad Naim MR, Harsuzilawati M. Spatial mapping of dengue incidence: A case study in Hulu Langat district, Selangor, Malaysia. Int J Hum Soc Sci 2010;5:410-4.
Kulldorff M. A spatial scan statistic. Commun Stat – Theory Methods 1997;26:1481-96.
Dunn CE, Kingham SP, Rowlingson B, Bhopal RS, Cockings S, Foy CJ, et al.
Analysing spatially referenced public health data: A comparison of three methodological approaches. Health Place 2001;7:1-12.
Eisen L, Lozano-Fuentes S. Use of mapping and spatial and space-time modeling approaches in operational control of aedes aegypti and dengue. PLoS Negl Trop Dis 2009;3:e411. Available from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2668799/
. [Last cited on 2013 Dec 02].
Fan J, Lin H, Wang C, Bai L, Yang S, Chu C, et al
. Identifying the high-risk areas and associated meteorological factors of dengue transmission in Guangdong Province, China from 2005 to 2011. Epidemiol Infect 2014;142:634-43.
Banu S. Examining the impact of climate change on dengue transmission in the Asia-Pacific region(PhD). Australia: Queensland University of Technology; 2013. Available from: http://eprints.qut.edu.au/66387/. [Last cited on 2017 May 09].
Li Z, Yin W, Clements A, Williams G, Lai S, Zhou H, et al.
Spatiotemporal analysis of indigenous and imported dengue fever cases in Guangdong Province, China. BMC Infect Dis 2012;12:132.
Banu S, Hu W, Hurst C, Guo Y, Islam MZ, Tong S. Space-time clusters of dengue fever in Bangladesh. Trop Med Int Health 2012;17:1086-91.
Varma RA. Technology Options for Treatment of Municipal Solid Waste with Special Reference to Kerala. Vol. 31. Suchitwa Mission Local Self Government Department, Government Kerala Thiruvananthapuram-695; 2009. Available from: https://www.sanitation.kerala.gov.in/pdf/workshop/techno_2.pdf
. [Last cited on 2014 Mar 22].
Vij D. Urbanization and solid waste management in India: Present practices and future challenges. Procedia Soc Behav Sci 2012;37:437-47.
Khormi HM, Kumar L. Examples of using spatial information technologies for mapping and modelling mosquito-borne diseases based on environmental, climatic and socio-economic factors and different spatial statistics, temporal risk indices and spatial analysis: A review. J Food Agric Environ 2011;9:41-9.
Chen SC, Liao CM, Chio CP, Chou HH, You SH, Cheng YH. Lagged temperature effect with mosquito transmission potential explains dengue variability in Southern Taiwan: Insights from a statistical analysis. Sci Total Environ 2010;408:4069-75.
[Figure 1], [Figure 2]
[Table 1], [Table 2]