Background
The annual global burden of dengue is estimated at 390 million infections, of which 96 million present clinically [
1]. Four closely related RNA viruses in the family
Flaviviridae (DENV1 to DENV4) are responsible for dengue disease. They are transmitted by
Aedes (primarily subgenus
Stegomyia) mosquitoes, particularly
Aedes aegypti (L.) and
Aedes albopictus (Skuse) [
2]. Dengue has developed from a sporadically occurring disease to a major and re-emerging global public health problem over recent decades causing substantial economic disruption and social burden in endemic areas in Asia, Africa, and the Americas. There is no effective treatment for dengue and vaccination, so far, offers only incomplete protection [
3,
4]. Therefore, vector control remains the most important means of prevention [
5]. Effective vaccine or not, vector control will remain the cornerstone of dengue control for years to come [
3].
Due to increasing incidence and rapid geographical expansion, dengue is the most common vector-borne disease in Thailand [
6]. From 2000 to 2011, the number of reported cases varied from 20,000 to 140,000 cases each year [
7]. Both
Ae. aegypti and
Ae. albopictus are common species and widely distributed in Thailand [
8]. All four serotypes co-circulated in each of the major outbreaks that occurred in 1958, 1987, 1998, 2001, 2013, and 2015 [
9‐
14]. The highest incidence typically occurs in 13–24 year-old age group with case clustering seen predominately in urban areas [
15]. Males represent the majority of reported dengue cases in several Asian countries [
16]. A study in Singapore showed that men were more exposed to infected mosquitoes than women, during daytime hours, at the workplace or while travelling to and from work. A forceful public health policy in Singapore [
17] has greatly reduced the number of mosquitoes in and around homes, potentially rendering the larger male labor force more exposed to mosquito bites during working hours [
16,
18]. Other causes for these apparent gender differences could be different health seeking behaviors or male-female differences in disease severity [
19]. In the Lao People’s Democratic Republic male-female ratios in dengue cases varied between years and provinces [
16]. We are not aware of similar spatio-temporal or socioeconomic differences in Thailand.
Thailand has adapted the dengue control strategy of the World Health Organization (WHO) [
2], which consists of three main pillars: 1) patients diagnosed with dengue are required to avoid mosquito bites to prevent dengue transmission; 2) active community case detection of cases which do not result in clinical consultation; and 3) vector control, consisting of environmental management, source reduction, and chemical interventions using insecticide fogging against adult vectors and larvicides to control immature stages in containers [
20]. Follow-up interventions are conducted by health officers or village health volunteers [
20]. To determine the most appropriate and feasible intervention or combination of interventions, health officers need to consider local environmental, resource, and contextual factors that may influence effectiveness [
21].
Climate variables are predictors of dengue infection [
4,
22,
23]. Seasonal variation in climate shows a strong relationship with
Ae. aegypti abundance and historical dengue incidence [
24]. Temperature affects population biology of
Aedes mosquitoes [
25]. Higher temperatures increase larval development [
26] and rates of multiple feeding, but reduce mosquito size [
27]. The extrinsic incubation period declines as temperature rises, thus increasing the proportion of infected vectors, and enhancing the transmission potential of the vector [
27‐
29].
As ambient temperature increases, so does dengue epidemic potential, peaking at around 29 °C and then decreases [
29]. In subtropical and tropical regions such as Thailand, with mean diel temperatures of 26 °C (20 °C ≤ T ≤ 32 °C), an increase in diurnal temperature range can enhance transmission [
29]. An analysis of data from Thailand (1978–1997) showed the incidence of dengue hemorrhagic fever (DHF) was negatively associated with higher rainfall in the southern region of the country, but positively associated with elevated ambient temperatures in the central and northern regions [
30]. Another study using provincial monthly dengue data from 1983 to 2001 concluded that the relationships between weather variables and dengue transmission are very complex in Thailand [
31]. The study found that transmission occurs within a specific temperature range, but that changes in humidity within this range can amplify the transmission potential with 80% of dengue cases occurring at a mean temperature of between 27.0 and 29.5 °C and a mean relative humidity of > 75%. They further found that large epidemics begin earlier, develop faster and can be predicted at a defined onset time. Non-linear modeling of more than 30 years (1982–2013) of monthly data by province in Thailand showed that inter-annual variations in rainfall and temperature with a lag time of one month can improve the explanation of dengue relative risk compared to a seasonal-spatial model [
32]. The relationship between rainfall and dengue is complex, as it may create abundant breeding sites for the vector [
33], but can also flush out sites if rain is too intense [
33,
34]. Because household water storage may increase in the dry season, the resulting breeding habitats may weaken, or even reverse, the positive association between dengue and rainfall [
35‐
39].
Spatio-temporal analysis can detect clusters of dengue disease and is useful for a better understanding of the dynamics of disease dispersion. Analysis of spatial and temporal variations is also useful in identifying high-risk locations and times of higher transmission risk, which are important for disease surveillance and control [
15,
40].
The above-mentioned research on climate and dengue focused on larger spatiotemporal scales, such as monthly dengue surveillance and climate records at the provincial level [
31,
41,
42]. The current study is novel because it uses data on the lowest administrative level, the sub-district, in one province to understand fine-scale spatial dengue-climate relationships. This is useful for developing more reliable prediction models for future projections applied in early warning and response systems, thus ultimately improving timely control interventions.
We analyzed data on reported dengue cases in Khon Kaen Province, northeastern Thailand collected between 2006 and 2016 to 1) describe demographic characteristics and seasonal variations of dengue cases; 2) determine the potential impact of climate variability on dengue incidence; and 3) identify clusters of dengue cases at the sub-district level.
Discussion
The majority (~ 90%) of patients were below the age of 30 years. The trend during the study period showed that the proportion of dengue cases younger than 15 years declined from almost 80% in 2006 to below 50% in 2016. Dengue fever is generally more common in younger age groups [
53], although there is evidence showing increasing incidence of more severe disease and outcomes among older age groups [
54]. Our observations are also consistent with a population age shift, potentially influenced by demographic changes, such as the birth and death rates that show decreasing trends during 2011 and 2015 [
55]. Thailand, in general, is undergoing a demographic transition where the proportion older adults are gradually increasing with an increase in median age of the general population. A higher proportion of adults will also increase the number of immune individuals (those with previous exposure to dengue virus) in the population, which might theoretically decrease the risk of dengue infection in younger people by providing alternative blood sources for infectious mosquitoes [
56]. This age shift has also been observed in other Asian countries with a higher frequency of dengue cases among people 15 years of age and older [
16]. Increases in disease incidence in older age groups may be explained by an increase in secondary infections and changes in circulating dengue virus serotypes [
57], which have been shown to be important risk factors for severe clinical presentations [
58‐
62].
There were clear seasonal patterns of dengue incidence in Khon Kaen Province during the study period. Dengue occurs throughout the rainy season, with 73% of cases reported between May and September. Although maximum temperature was associated with higher incidence (Table
2), the model with meteorological covariates had similar performance (in terms of the WAIC) to a non-mechanistic model, which simply fitted a sinusoidal pattern with a period of 12 months. In our study, a 1 cm increase in monthly rainfall was associated with a 0.4% increase in dengue incidence. In Timor Leste, results from similar modeling analyses showed a far larger effect: a 47% increase in incidence per 1 mm increase in annual rainfall [
63]. Different climate patterns between Timor Leste and Thailand might explain these differences. Rainfall can affect the availability of mosquito larval habitats [
34]. During rainy and dry periods of the year, permanent water containers are common in and around households; some located in toilet or bathroom spaces providing continuous year round mosquito production [
35‐
39,
64]. Large water storage jars and tanks are the most commonly used containers in Thailand [
64]. A study correlating rainfall and clinical dengue cases in Thailand from 2002 to 2003 also found that the dengue incidence was closely related with rainfall [
65].
Temperature is another primary environmental risk factor for dengue transmission. Sea surface temperature (SST) changes, generally related to periodic El Niño Southern Oscillation effects, and air temperature, having more direct short-term effects, have both been shown to influence dengue incidence [
63,
66]. Dengue incidence increased by 19.4% with a 1 °C increase in SST and 2.6% with a 1 °C increase in weekly maximum temperature in the Texas-Mexico border region [
66]. Another study found that a 1 °C monthly increase in mean ambient temperature, dengue incidence increased by 0.7% [
63]. In our study, the rate ratio for maximum temperature was 1.055 per °C, within the range from 30.7 °C to 44.9 °C. Higher temperatures enhance viral replication in the vector mosquito in a shorter amount of time and thus increase transmission potential of dengue viruses. A study of the extrinsic incubation period (EIP) of dengue serotype 2 in
Aedes albopictus found that the virus remained in the midgut at 18 °C but could disseminate and invade the salivary glands at temperatures between 23 °C and 32 °C [
67], thereby showing higher temperatures produce a shorter EIP and greater transmission potential. The strong and consistent relationships between climate, particularly rainfall and temperature, and the number of dengue cases have been used to develop prediction models to implement more timely dengue control measures [
68,
69]. Relationships between dengue transmission and climatic variables have been examined in numerous studies, as shown above, but the question remains how to use such relationships in predicting impending outbreaks and applying effective interventions in time to avert them. User-friendly tools, such as the operational guide on Early Warning and Response System developed with support from the WHO/TDR and the European Union [
70], are needed and will be tested in forthcoming work in Khon Kaen Province.
The highest dengue incidence seen in this study occurred in two areas of the province: around Khon Kaen Mueang District in the northeast, and in Manchakhiri and Khokphochai districts in the southwest. Mueang District includes the provincial capital and has the highest human population density, and in general, more conducive to dengue transmission. Manchakhiri and Khokphochai districts have lower population densities, but are, from our observations, seemingly similar to other districts in the province, i.e. vector species are present, larval habitats are plentiful, with a susceptible human population; therefore there must be other yet unexplored factors that support high dengue transmission in these two districts.
Although dengue incidence is influenced by rainfall and temperature, in our data there is no apparent spatial clustering of cases associated with the spatial variability in these environmental parameters. Rather, other factors such as urbanization are likely causes of the observed clustering effect [
71]. However, population density, which was included in the regression model as a measure of urbanization, was not independently associated with dengue incidence. The residual spatial variation visible in Fig.
4b suggests that variables beyond those included in the spatial regression model are needed to explain differences in incidence between urban and rural subdistricts. Moreover, hotspots in more rural areas of southwestern Khon Kaen Province, further corroborate the influence of factors other than urbanization driving transmission. We do not know of any specific reasons for why these rural areas should have elevated dengue prevalence. One speculation could be that the lakes and swamps that are common in this area may provide suitable humidity for mosquitoes to thrive, but this was not studied here. Large changes in population size over time will affect outcomes. However, during 2000 and 2015, the average annual population growth rate in Thailand was less than 0.5% [
72], which might not have affected the results substantially. Rural-urban migration is common in Thailand, with people drawn by, for example, better education, job opportunities, health facilities, standard of living, and wages [
73]. Human movement is also an important factor in the dynamics of dengue transmission [
74]. Adults are more likely to have greater mobility than younger age groups; therefore, to understand the circulation of the virus information on recent travel history and working conditions (location, time of work, etc.) is required. Elsewhere in Thailand, greater vulnerability to dengue infection has been observed in villages situated closer to urban centers [
75]. Such neighboring effects are related to similarities in human behavior, development infrastructure, and ecological surroundings. Moreover, similar lifestyles and social interactions between neighboring areas are evident between villages that share social and religious centers such as schools, temples, mosques and community halls [
75]. Hence, the results presented here are generalizable to most of northern Thailand, Laos, and Cambodia, and potentially Vietnam and Myanmar as well, under similar epidemiological settings.
Data collected from national surveillance systems come with inherent limitations, including underreporting and misreporting of symptomatic cases as well as the absence of subclinical and asymptomatic infections [
76]. Moreover, dengue cases are seldom laboratory confirmed or identified to serotype. Another limitation of this study is inaccuracy, albeit minor, of the population denominators within sub-districts, as these were taken as fixed values from a single census (2010). Lastly, the possibility of travel-related infections was not determined in this study, which would provide potential misclassification bias. Nationally, the importance of travel-related dengue would vary by locality based on mobility. Obviously, we cannot exclude the possibility that some dengue infections were acquired outside the study area, thus potentially affecting the analysis and conclusions. However, if the general travel patterns had not changed significantly over the 11-year observation period, the dengue disease trends reported in this study would remain valid.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.