Malaria occurrences are not randomly distributed across space and time . This variation is likely influenced by the combined effect of factors that characterize individuals and areas . Findings of the study show that factors at both the community-level and individual level are important in understanding the variation of malaria risk among individuals. In addition, temporal changes in the community factors, e.g., seasonal variation and control intervention programmes, can have a major impact on individual malaria outcome.
The relationship of malaria incidence with monthly minimum temperature rather than rainfall may be unique to this wet, higher altitude area. While the optimal temperature for the development of both vector and parasite is between 25°C and 30°C, the lower temperature during dry season may jeopardize the transmission capability of mosquitoes [17–19]. Since P. vivax is more tolerant at a relatively low temperature condition, the effect of temperature would be expected to be less for P. vivax than P. falciparum, which is consistent with the study findings. This may provide opportunities for new strategies that target interventions to eliminate transmission during these periods when temperature favors success. It may also raise concerns about future failure of control programmes in such regions if warming climate occurs even without changes in rainfall.
Malaria control interventions were associated with a substantial impact on the inter-annual reduction in malaria risk in this population, after adjusting for the temperature effect. Routine active surveillance allows detection of asymptomatic cases, which are the important infectious reservoirs of malaria [20–22]. In addition, symptomatic patients can be detected and treated earlier during active surveillance than when passive surveillance is operating alone . The resulting of reduction in both the number of reservoir hosts and in an individual's infectious period could both decrease infection pressure in a hamlet.
The finding of the effectiveness of artemisinin-based combination therapy (ACT) on P. falciparum occurrence is consistent with research conducted in other areas [23–25]. Artemisinin has been shown to be effective in killing both the asexual stages and the immature sexual stages of P. falciparum parasites, which consequentially prevents onward transmission [26–29]. Although ACT was used only for treatment of P. falciparum, a substantial reduction in P. vivax incidence was also observed during the ACT use period; this could be possibly explained by the additional effect of artesunate in killing asexual P. vivax in patients who initially had mixed infection. Mixed infection is likely to be undetected using microscopy . In Thailand, using molecular techniques, mixed P. falciparum and P. vivax infections have been reported in about 60% of clinical malaria patients, and when it presents, P. falciparum generally dominates . Although ACT has no reported impact on eliminating P. vivax hypnozoites, the liver stage parasite that causes relapse, a clinical trial conducted in Thailand has shown that about 50% of P. vivax patients who received either 5- or 7-day artesunate treatment remained parasite-free at day 28 after the treatment . However, the effect of ACT on reducing subsequent P.vivax reappearance among submicroscopic mixed infection patients is still unclear.
The effect of age and gender on malaria incidence in low transmission settings varies across areas [21, 22, 32, 33]. Unlike areas with high malaria transmission, an increased risk of malaria attack among children in malaria low transmission areas could be easily attributable to their lack of development of immunity, since exposure and clinical episodes are uncommon in all age groups [1, 32]. The pattern of age-gender distribution of malaria risk may, in part, be due to behavioural exposure to vector or to other unmeasured variables, such as intrinsic differences in immune function or hormonal response among different age groups and genders [34, 35]. Future work to explore this pattern, including immunological tests, is needed in hypo-endemic areas.
In this study, the risk of malaria attack was not significantly different for individuals living at variable locations relative to a stream. Although an association between topographical characteristics and malaria incidence has been reported in many areas, these observations are generally where there are large variations in attributes of house locations [21, 36–38]. For example, an increased risk of malaria was found to be associated with houses located farther than 500 or 1,000 meters from a river [21, 36, 37]. However, in this study area, the mean distance from an individual house to a stream was about 200 meters with a range of 200 to 500 meters. This relatively small variation may mean that vector abundance is similar among houses. The hamlet's light forest cover area had a marginal effect on the risk of P. falciparum attack but no effect on the risk of P. vivax attack. There are differences in the preferred habitats of Anopheles vector species in Thailand. An. minimus, a predominant vector in Thailand, is commonly found close to stream margins in forested and foothill areas [39–41]. However, the association between hamlets' light forest area and malaria risk was still uncertain due to limited number of landscape data for comparison at hamlet-level. Future research with wider spatial and temporal coverage is needed to understand which environments might favor transmission.
Malaria transmission directly depends on the presence of a human reservoir as well as the level of vector-human contact [1, 42]. While the factors most commonly found to be associated with vector abundance were relatively stable across the area, changes in the number of infectious human reservoirs in the community could potentially affect transmission probabilities for individuals in the population [43, 44]. For high transmission areas, stable areas of higher and lower incidence have been associated with spatially identifiable transmission and control factors. However, in low transmission areas where the majority of cases are quickly treated, other factors, such as immigrant cases or individual risk behaviours might be more important. In this study, the addition of one human reservoir per 100 persons in a hamlet resulted in 1.14 and 1.34 times higher risk for contracting P. falciparum and P. vivax, respectively, among individuals who lived in the hamlet. Spatial variation in level of gametocyte carriage in the population may have a major role in defining risk for malaria transmission in low endemicity areas.
Gametocyte carriers are vital for sustaining the disease in a population. However, number of humans available to infect mosquitoes may not be a limiting factor in high endemic regions. Interventions that have a strong community-level effect, particularly early detection and treatment to decrease number and duration of gametocyte carriage may have a more powerful effect on reducing malaria transmission in hypoendemic regions [18, 23]. The study findings showed that people who lived far from the malaria clinic were more likely to have P. falciparum attacks than those who lived closer. Distance to the malaria clinic may reflect how quickly people receive treatment. Those who live farther may take longer time to seek treatment, which consequently increases the infectious duration. Although the Euclidean distance used in the study was not accounted for the actual route and terrain to the clinic, the pattern of main roads and terrain features of the area suggests that the Euclidean distance and via route distance are likely to be proportionately similar.
Results of this study may be subject to misclassification bias. The incidence of malaria in a previous month was used as a proxy for the level of infectious individuals in each hamlet. For malaria infection, especially P. falciparum, infected patients do not always represent an infectious population. While this would bias estimates of incidence for each hamlet, the OR's would be biased toward the null (underestimated) unless the rates of infectious to infected individuals varied systematically between hamlets, which could possibly occur if the pattern of treatment and clearance of parasites differed across hamlets. Misclassification of landscape features was also possible because of the limited resolution of LandSat images. Additionally, land use changes were occurring over this period in several of hamlets, which can have variable effects on vector dynamics. However, characterizing spatial attributes may be less critical for a region with widespread water and vector habitat availability.
Although the one-month lag time used for hamlet incidence was based on the time from parasite uptake from one person until clinical symptoms appear in another individual, this lag time did not account for different durations of infectiousness and the infectivity dynamic among individuals. An individual's infectious period is usually less than one week. However, some individuals may remain infectious up to a month, even after treatment [44, 45]. In addition, recrudescence may prolong the duration of infectiousness; and relapse may result in multiple episodes of infectious period. Individuals who had a long infectious period could continue to infect mosquitoes and be a source for the disease in their hamlets in the next two or three months. However, in this study area, short duration from clinical attack to treatment and the low number of asymptomatic cases found during active surveillance suggest that a prolong individual's duration of infectiousness may be less likely, with an exception in 2003 and 2004 when recrudescence was observed in about 10% to 15% of all P. falciparum malaria cases, due to mefloquine resistance. Additionally, previous higher incidence in a hamlet may simply be a proxy for a higher risk in location where malaria is generally found. Findings from the preliminary analysis showed that the locations with highest incidence moved over time in this study area. Further, the effect of hamlet incidence became non-significant in the sensitivity analysis using different lag times (zero and two months lag).
To study the epidemiology of malaria, one must consider the relationship among factors that characterize time, person, and place. In this study, potential malaria risk factors at temporal-level, individual-level, and community-level were simultaneously examined. The 8-year longitudinal data used in the study allowed us to determine the effect of temporal and spatial changes over time. Unlike ecological studies where inferences are drawn on the basis of aggregated outcomes for a group, the individually specified, hierarchical outcomes used in this study provide more valid inferences regarding malaria. Results of this study provide new insight into the malaria epidemiology in a low malaria transmission area, where malaria elimination is feasible. Moreover, similar patterns of P. falciparum and P. vivax risk factors found in the study suggest that a uniform intervention could have an effect on both P. falciparum and P. vivax incidence. However, further research is needed to investigate the effects of ACT on P. vivax among individuals with submicroscopic mixed infection.