Spatial mapping of temporal risk characteristics to improve environmental health risk identification: A case study of a dengue epidemic in Taiwan
Introduction
Mapping uneven events, such as disease cases or pollutants, makes it possible for public administrators to discover the origin of pollution or source of epidemic outbreaks and generate more hypotheses for further investigations. More importantly, it may spatially identify the high-risk areas, which can be targeted for environmental hazards and public health prevention. Mapping has been made easier and become more wisely used with the development of geographical information systems (GIS). However, mapping provides only a visual display of the uneven cases, but cannot definitively confirm clustering of cases or spatial correlations. Spatial statistics is extensively used to find relationship of case rates and their geographical location (Croner et al., 2000). Point pattern analysis is a method often employed in the detection of clustering patterns. Since points may represent actual locations of uneven events, the method involves in the issue of privacy (Maheswan and Haining, 2004). Risk surface estimation, including kernel estimation and geostatistical methods, produces continuous surfaces of risk across the whole study areas and potentially offer more insight into the nature of the clusters (Bithell, 1990, Bithell, 1999, Diggle, 2000). However, mapping the spatial clusters of uneven events is a static snapshot, which ignores the temporal kinetics of these uneven events, and it is difficult to evaluate whether the hazards or epidemics have been broken out or kept under control by policymakers. Time-series statistical techniques, such as Box-Jenkins (seasonal) autoregressive integrated moving average (ARIMA) models, are used to forecast the outbreak of uneven cases or to estimate the expected incidence values (Zeger et al., 2004, Earnest et al., 2005). But these kinds of approaches could not provide clues for identification of spatial risk areas. This paper intends to propose a spatial–temporal risk model for mapping geographic distribution of uneven events with temporally defined indices to improve health risks identifications. The model focuses on three temporal risk characteristics across geographic space: (1) how often these uneven cases occur, (2) how long these cases persist and (3) how significant cases occurring in consecutive periods. This model will be applied to the dengue fever epidemic in Taiwan in 2002 as a case study, which is the worst epidemic in the last 60 years.
Dengue is a member of flavivirus with four known antigenically distinct serotypes and is vectored mainly by domestically adapted mosquito species, specifically Aedes aegypti. Environmental risk factors, including water storage containers, housing patterns and frequency of waste disposal, etc., reflect the condition of environmental health in certain area (Bohra and Andrianasolo, 2001). These risk factors could have potential to cause more mosquito breeding sites, which might lead to an outbreak of dengue epidemic. However, routinely monitoring environmental risk factors is not feasible for public health and environment agencies. Therefore, this study only used dengue case-incidence data, including onset date of a confirmed dengue case and its geographic location, collected by official surveillance systems to investigate whether the approach will improve the detection of environmental health risk areas and the evaluation of vector or mosquito control measures.
Section snippets
Study area and data
Kaohsiung City, located in southern Taiwan, is the second largest metropolitan city, the largest commercial harbor and an important land, marine and air transportation hub. Fig. 1 shows the location of Kaohsiung City and one of its satellite cities, Fengshan City. This two-city area was used as our study area because it was the major dengue epidemic foci in southern Taiwan. The study focused on dengue cases based on surveillance data, which reported 4790 confirmed dengue cases (DEN), including
Spatial pattern analysis of temporal risk indices
Since any epidemic is a dynamic process, it would be important to identify spatial risk areas, if any, through the use of different temporal characteristics. For example, some area might have longer epidemic duration and the others might have stronger intensity even though the duration is short. Once the risk areas at different epidemic periods are identified, the comparisons of case-incidence with its temporal–spatial dynamics can be helpful in identifying the possible causes or effective
Discussion
We have described a procedure for identifying spatial risk with different temporal characteristics of disease epidemics. The spatial risk maps with the three proposed temporal characteristics improve spatial clustering analysis, which focused mainly on case-incidence data obtained through passive surveillance. We are able to identify other case clusters when we factored in the temporal properties, such as case-incidence frequency and the number of cases that occurred within a certain time, or
Concluding remarks
Space and time are two important dimensions in describing epidemic dynamics and risk distribution. Methods based on either complicated statistical analysis or sophisticated surveillance systems are difficult to carry out in developing counties. Therefore, this paper attempted to use minimum data requirement and more straightforward statistical methods to capture major temporal characteristics of epidemic dynamic process, including frequency, duration and intensity. Different epidemic patterns
Acknowledgements
The authors sincerely thank the staff of Kaohsiung City and County Health Bureau and the Center for Disease Control in Taiwan (Taiwan-CDC) for their efforts in surveillance and vector control during the hard time of the dengue epidemic. The authors also thank anonymous referee's comments for clarifying the original manuscript. Neal Lin would like to express appreciation for the support from the Fulbright grant, sponsored by the Foundation for Scholarly Exchange in Taiwan.
References (20)
- et al.
Users' perspectives on epidemiological, GIS and point pattern approaches to analyzing environment and health data
Health Place
(2004) - et al.
Analysing spatially referenced public health data: a comparison of three methodological approaches
Health Place
(2001) Local indicators of spatial association—LISA
Geogr Anal
(1995)An application of density estimation to geographical epidemiology
Stat Med
(1990)Disease mapping using the relative risk function estimated from areal data
- et al.
Application of GIS in modeling of dengue risk based on sociocultural data: case of Jalore, Rajasthan, India
Dengue Bull
(2001) - et al.
Silent transmission of the dengue virus in southern Taiwan
Am J Trop Med Hyg
(1996) - et al.
Geographic information systems (GIS): new perspectives in understanding human health and environmental relationship
Stat Med
(2000) Overview of statistical methods for disease mapping and its relationship to cluster detection
- DGBAS, Population and Housing Census Taiwan-Fukien Area, Republic of China, http://www.dgbas.gov.tw/, Nov. 2005 [in...
Cited by (75)
A survey on the suitability of risk identification techniques in the current networked environment
2021, Journal of Network and Computer ApplicationsCitation Excerpt :In this decade, researchers also started to use interdisciplinary models to identify the sources of hazards in events that have occurred. For example (Wen et al., 2006), used spatial mapping to investigate the reasons for the spread of an epidemic. By combining space with time, they studied the origin of epidemic outbreaks by which high-risk areas were shortlisted as in the 2002 case of dengue in Taiwan.
Different responses of dengue to weather variability across climate zones in Queensland, Australia
2020, Environmental ResearchEpidemic features of seasonal influenza transmission among eight different climate zones in Gansu, China
2020, Environmental ResearchCitation Excerpt :The maximum size of cluster search windows in our model was set as 50% of the population size (Seixas et al., 2018; Zhang et al., 2019). It remains difficult to determine whether an outbreak of disease occurs and its temporal features (Wen et al., 2006). Incidence rate is generally used to measure the health burden of the disease (Dunn et al., 2001), but it can neither reflect whether the occurrence of the disease is temporally accidental or clustered, nor evaluate the change patterns between epidemics in time (Wen et al., 2006).
Social-ecological modelling of the spatial distribution of dengue fever and its temporal dynamics in Guayaquil, Ecuador for climate change adaption
2019, Ecological InformaticsCitation Excerpt :The prevention of dengue infections depends on interrupting human-vector contact by controlling mosquito habitats and the presence of all four mosquito life stages inside households and in household vicinities. These goals can be accomplished through a combination of environmental management, chemical control, and biological control (Ministerio de Salud de la Nación, 2013; Wen et al., 2006). Habitats are eliminated by emptying and cleaning containers as well as by using insecticides, biological agents, and applying adulticides (Ministerio de Salud de la Nación, 2013).
Using Google Trends and ambient temperature to predict seasonal influenza outbreaks
2018, Environment InternationalAscertaining the impact of public rapid transit system on spread of dengue in urban settings
2017, Science of the Total Environment