From 2011 to 2021, there has been an overall reduction in the number of malaria cases reported in Si Sa Ket and Ubon Ratchathani provinces in Thailand, despite a large outbreak in Ubon during this period. An increasing proportion of those cases were classed as indigenous to the patient’s village of residence. There are likely to be multiple factors contributing to this pattern, including changes to surveillance strategies and dynamic human processes. The introduction of the 1-3-7 surveillance strategy and updated online reporting dashboard in 2017(9) coincided with a gradual increase in the proportion of cases classified such that from 2017, at least 50% of reported cases were classified each month, increasing to more than 75% in 2021. The increase in the number and proportion of indigenous malaria cases in both provinces may therefore represent improved reporting rather than a true increase in malaria transmission within villages.
The receptivity of an area to malaria is affected by human behaviours. Changes in human processes have presented challenges to control strategies in the past and the possibility that they may do so in the future should be prepared for. Human processes in areas covered by plantations and tropical forests may lead to malaria importation and subsequently local transmission(20). Community members in Si Sa Ket and Ubon participate in different agricultural activities in rice fields(21), rubber plantations(21, 22), cassava plantations(22), and forest-going activities(21). Mobile and migrant populations (MMPs) also pose major challenges to the malaria prevention and control programme in the country(7). Strategies targeting forest-goers such as portable ITNs and asymptomatic testing have variable uptake(21).
There was a large peak in unclassified cases between 2014 and 2016 in Ubon: the number of reported malaria cases rose from 1,081 in 2013 to 8,321 in 2014, an increase of 670%. This outbreak, confined largely to Buntharik, Na Chaluai, and Nam Yuen districts in Ubon which share a forest border with Lao PDR to the east, was anecdotally related to an increase in the price of rosewood (Dalbergia cochinchinensis Pierre ex Lannes), which led to an increase in forest-going activity to harvest the wood(19, 23). An entomological study of anopheles mosquitos in 8 village sentinel sites at the time found no Plasmodium species, although few primary vectors were collected(19). This implies that infection was occurring outside of the villages and within the forest. Controlling transmission around cases imported from the forest and other regions is important: the location of stable hot spots for indigenous cases and all cases in our data are very similar. This suggests that indigenous cases may be largely introduced cases fed by non-indigenous infections, and so curbing local transmission also relies on reducing importation, which was high during the 2014 outbreak in Ubon. Were a similar situation to arise in the future, this would present a challenge for surveillance efforts: forest-goers with fever may be hesitant to present to healthcare facilities or disclose their movements if they have been engaging in potentially illegal forest-going activity.
There were few indigenous village hot spots compared to subdistrict indigenous hot spots, which included cases acquired in any area within the subdistrict of residence. During the outbreak in the southeast of Ubon, there were high-confidence subdistrict hot spots but no village hot spots. This is likely due to low rates of case classification. No indigenous cases were reported in some months of 2013 to 2015 in Ubon, but there were high numbers of unclassified cases: 56.1% of all cases in 2014 were unclassified. This improved as case numbers dropped. A likely explanation for this transient reduction in reporting quality is that healthcare workers were overwhelmed with cases and therefore unable to complete the epidemiological surveillance forms. Individual village/clusters risk stratification data suggests that indigenous malaria transmission was ongoing in villages or clusters of houses in villages at the time. Prior to 2016, an A1 village/cluster was defined as a “perennial transmission village or hamlet where indigenous cases are reported at least 6 months out of the year”, while an A2 village/cluster was where “indigenous cases are reported fewer than 6 months out of the year”(24). If a village/cluster had not reported indigenous transmission for at least 3 years, but primary vectors were found or conditions were felt to be favourable for breeding, it was classed as B1(24). In our risk stratification dataset, in Ubon, 14 villages/clusters were classed as A1 in 2013, 15 in 2014, and 6 in 2015. There were 29 A2 villages/clusters in 2013, 33 in 2014, and 126 in 2015. This increase in the number of high-risk villages/clusters implies that the presence of indigenous transmission was being acknowledged despite only two village indigenous cases having been reported in the province in 2014, one of which was within the study districts. Accounting for this, and the high proportion of unclassified cases, we assume that the lack of village hot spots compared to subdistrict hot spots is due to low rates of classification rather than a true absence of village indigenous malaria transmission. We assume that the missing, unclassified data is balanced and representative of the other proportions.
The majority of infections over the study period were P. vivax, although there were high P. falciparum case numbers during the 2014 outbreak. Since 2016, there has been a marked reduction in the proportion of P. falciparum infections and hot spots. In contrast, while P. vivax hot spots have become less numerous over time, they persisted throughout the study period in parts of both provinces. The proportions of each species during the outbreak years were similar to reports across the border in Lao PDR in 2013 to 2016 in a study area including Moonlapamok and Sukhuma districts, which border Na Chaluai and Buntharik districts in Ubon(25). It is important to be aware of the species mixes, as different Plasmodium species present specific elimination challenges. Antimalarial drug resistance is a particular concern in the GMS where strains of artemisinin and ACT-resistant P. falciparum circulate. Chloroquine resistance is also present in P. vivax(26) which presents an additional challenge to elimination efforts due to the potential for relapse from the dormant liver stages(27). While radical cure with primaquine can help to prevent this recurrence, there is a risk of haemolysis in people with G6PD deficiency, which is estimated to affect 8–24% of the population in Northeast Thailand(28). The recent introduction of point-of-care G6PD testing has increased access to primaquine, but a significant proportion of the population cannot safely take the full dose.
The use of malaria terminology can vary between programs(29), but applying flexible definitions allows examination of data on different levels depending on intervention scale. The WHO defines indigenous malaria as a “case contracted locally with no evidence of importation and no direct link to transmission from an imported case”(29). However, a specific definition of local is not given, and use of the term varies(30). The time frame within which someone is considered at highest risk of having an imported case following travel to an endemic area also varies, from 10 days to 3 months(31). The Thai DVBD defines an indigenous case as “a patient who contracted malaria in the village where the patient lived during infection period”(9) and states no specific time frame. A malaria infection acquired anywhere else is classed as imported. In this analysis, we stratified malaria cases in two ways: firstly, locally acquired was defined as acquired within the village of residence only, per the DVBD definition; and secondly, where locally acquired was defined as acquired within the village or the subdistrict of residence. These stratifications reflect both the level of interventions, which is tailored to individual villages/clusters(1), and the high level of mobility in at-risk forest-going populations(32), which often includes moving beyond the village. There were high-confidence indigenous hot spots in 2013 and 2014 in the northeast of Ubon which were not identified in the village-level analysis, indicating high levels of transmission within the subdistrict but outside of the home village.
The key hot spots of indigenous transmission identified in this analysis appear to be stable, and these hot spots contribute high numbers of malaria cases. Stable hot spots have been found in previous studies to be highly predictive of future malaria risk(33, 34), which supports the spatial targeting of receptive villages/clusters with interventions. There were high-confidence hot spots of indigenous malaria transmission at the village and subdistrict level in the south of Si Sa Ket and the southeast of Ubon for much of the study period. The statistical significance of these hot spots remained high despite declining case numbers, implying that they were very stable(34). Targeting malaria foci in elimination and post-elimination settings has been recommended by the WHO(35). However, there is mixed evidence regarding whether geographical malaria hot spots should always be targeted for interventions in the near-elimination setting(36). A study in the Western Kenyan Highlands found that targeted interventions had a modest impact on parasite prevalence, and this did not affect prevalence in areas adjacent to the hot spots(37). Stresman et al(36) recommend that where the bulk of transmission occurs away from settlements, targeting of behavioural traits rather than geographic locations would be advantageous. This is consistent with the current approach taken by the DVBD, which specifies that villages/clusters which are not active foci (i.e. which do not have ongoing indigenous transmission) should implement educational strategies for night-time forest-goers(1), thus aiming to reduce the importation of cases. The findings of this study support the country’s current focus on targeting high-risk groups as well as the spatial targeting of villages/clusters that are active foci. They also highlight the need to prioritise malaria service provision in hot spots which are homes to farmers, forest-going populations, and MMPs at risk of contracting malaria.
Strengths and limitations
Here we analysed high-quality malaria surveillance data over a dynamic 10-year period, including a cross-border malaria outbreak and changes to the national malaria strategy. We used manually geolocated village coordinates for the spatial analyses. We considered the factors contributing to the decline in overall cases and increase in indigenous cases, and how these are relevant to future control efforts.
Important limitations include a lack of data on the prevalence of asymptomatic infection from active surveillance, which is currently offered to close contacts of malaria cases, and before and during the rainy season in active foci(1). Many cases were not classified by likely origin between 2012 and 2016, but from 2017 onwards there was a much greater proportion of cases classified. Comparisons between earlier and later data should therefore be made with caution due to differing levels of completeness. There was also the potential for inaccurate geolocation of cases, for example in the case of recrudescence/relapse of previous infection, or inaccurate reporting of travel history to those performing case investigations. It was also not possible to differentiate potential recurrences of P. vivax from new infections.
Village location data used in this study was collated from different sources as described in the Methods. Due to the quantity of villages with no location data and the amount of time required to locate them, we decided to include only the villages in the districts that comprised 95% of the cases in each province in village-level analyses. Compiling a list of village location data of quality that is interoperable with other existing datasets, such as malaria surveillance data and population counts, would be of great usefulness for future disease mapping efforts. Ideally, such a list should be centrally managed by the government as a single high-quality master list for all purposes, including those beyond health.
Future work should utilise information sharing with programs in Lao PDR and Cambodia to allow better understanding of cross-border epidemiology and work towards the shared goal of elimination across the GMS. Furthermore, now that case numbers are low enough to permit it, spatiotemporal modelling techniques could be used to form likely links between cases and determine whether locally transmitted cases are largely indigenous or introduced by importation(8, 38).