Background
Epidemiological data on malaria over time are essential to planning of malaria control and elimination. In Cambodia, the control and elimination of malaria has garnered much attention due the recent discovery of artemisinin resistance in Pailin Province on the Thai border [
1]. This area has historically been the source of resistance to other anti-malarial drugs and if artemisinin resistance were also to spread it would be a major threat to malaria control and elimination efforts worldwide [
2,
3]. Resistance has since been identified in elsewhere in Cambodia and parts of Thailand [
4], Myanmar [
5] and Vietnam [
6]. Mathematical modelling has been employed to predict the most effective interventions to try to eliminate artemisinin-resistant malaria parasites [
7,
8]. The modelling predictions are heavily dependent on accurate information on baseline malaria epidemiology.
Since 1992, the Department of Planning and Health Information has used a standardized system for collecting monthly data on malaria cases across the whole of Cambodia into a centralized database, the Health Information System (HIS). This includes clinical cases in each province who attend public sector health facilities. Since 2004, the National Centre for Parasitology, Entomology and Malaria Control (CNM) has rolled out a system of Village Malaria Workers (VMWs) across the areas of the country with highest malaria prevalence. These provide diagnosis and treatment at the village level and report their activities through the Malaria Information System (MIS) on cases diagnosed and treated to CNM. Between these two sources, HIS and MIS, a detailed picture of malaria in Cambodia over time can be gleaned.
This study analysed data collected through HIS and MIS from 2004 to 2013 to produce a picture of the spatial and temporal epidemiology of malaria in Cambodia.
Discussion
The overall number of cases of P. falciparum and P. vivax malaria attending public sector facilities in Cambodia has fallen markedly by 59% since 2011. This coincided with a rapid scale-up of distribution of insecticide treated bed nets and coverage of villages in high-risk areas by village malaria workers. P. falciparum remained relatively stable from 2004–2011 whilst numbers of P. vivax cases increased despite substantial distribution of ACT and the roll-out of RDTs during that period. The reasons for this delayed fall in cases are unclear but further examination is warranted to derive lessons for malaria control programmes elsewhere. The fall in percent of tests for malaria which were positive since 2011 with similar annual numbers tested throughout suggests the decline in incidence of clinical malaria was real and not an artefact due to changing surveillance efforts. Additionally, it appears that more effort has been put into testing to detect the same number of cases, particularly in the low season. This may be masking an earlier actual decline in P. falciparum in 2010 and 2011.
The marked increase in
P. vivax in this study began a year too early for it to be related to the decrease in
P. falciparum[
11,
12]. Alternatively, it could be a result of improved diagnosis with introduction of a HRP-2/pLDH RDT (CareStart™) in 2009 in place of a Pf only HRP2 RDT (Paracheck Pf®), improved microscopic differentiation of species on diagnosis and better sensitivity of testing with development of the malaria control programme and improvements in training of microscopists from 2009 [
13]. Other differences between
P. falciparum and
P. vivax included different timing of seasonal peaks in cases and differences in the geographical distribution in numbers of cases over time. A greater relative decline in cases of
P. falciparum than
P. vivax and differences in the periodicity were also seen in Peru, particularly in mountain and jungle regions with some influence of precipitation and temperature [
14]. In Cambodia, further investigation of the influence of climatic factors and malaria control measures on the two species over space and time would be informative.
The decrease in mortality found in this study is likely to be due to improvement in healthcare provision during the study period. The difference in mortality in VMW and non-VMW areas in the HIS data suggest that VMWs may have contributed to this decreased mortality, presumably through earlier effective diagnosis, treatment and onward referral at a village level.
The roll-out of VMWs across remote and high transmission areas of Cambodia has resulted in detection of a large previously unrecorded burden of disease. With the scale-up of VMWs to over 40% of the villages in the higher transmission areas of Cambodia in the past five years, the total number of detected cases of both species nationally has almost doubled. Numbers of people reported through HIS did not fall during the roll-out of VWMs. This suggests that many of those people now presenting to VMWs previously sought care in the private sector and that the cohort who previously accessed public sector healthcare services for diagnosis and treatment of malaria continue to do so. If VMWs were rolled out further, additional increases in the detected caseload would thus be expected. VMWs have the advantage of being accessible to villagers who may not otherwise attend public sector facilities with their malaria. In the absence of VMWs, a high proportion of individuals are thought to purchase anti-malarials from the private sector. This was estimated at 87% in a survey in 2002 [
15] and 54% in 2011 [
16]. Despite the intensified malaria control efforts in and around Pailin, 43% were estimated to have received treatment from the private sector in 2011 [
16]. It is likely that many of these will choose instead the free testing and treatment available from the VMWs as had been shown previously [
15] and in a recent evaluation by CNM [
17]. As data were not available on attendees at private pharmacies and clinics during the time of the study, the total burden of malaria in Cambodia remains unclear although it is evident from these data that it is grossly underestimated. This was suggested in a previous study, which showed that this underestimate is most prominent in remote areas of the country [
18]. A conservative estimate for total malaria burden in Cambodia from the present study would be over 100,000 cases per year from 2004–2011, falling to 70,000 in 2012 and 42,000 in 2013. These figures do not include those who sought treatment in the private sector (conservative estimates being over 50%) and it is not clear how many of these would attend a VMW if available. In 2013, VMWs were present in 44% of the villages in ODs with high malaria prevalence. If malaria prevalence across all villages in these ODs with VMWs were similar, then the actual number of cases could be 1/0.44 = 2.3 times the current number detected by VWMs i.e. 47,000. This would give a total of around 68,000 malaria cases per year in Cambodia in 2013, an API of 4.6 per 1000. Similarly for 2012, the estimated API using this method would have been 7.5 per 1000. This is 2.8 times higher than the estimate from WHO for that year of 2.7/1000 [
19]. In 2015, the private sector will no longer be permitted to diagnose or treat malaria. A more accurate picture of malaria burden in Cambodia can be expected as a result. By providing free testing and treatment, VMWs could be expected to increase the number of people being treated for malaria. This should lead to a decrease in transmission overall in areas with VMWs. This study confirmed this with a greater decrease in cases of
P. falciparum in ODs with VMWs compared to ODs with no VMWs. Malaria mortality was also lower in ODs with VMWs highlighting the benefits of early effective diagnosis and treatment (EDAT).
During the study period, the geographical distribution of malaria in Cambodia has changed. This is likely to be due to a combination of factors including changes in malaria control activities, migration and deforestation. In some areas, particularly in parts of the west of Cambodia,
P. falciparum decreased. Pailin Province had the greatest decrease in
P. falciparum cases from 2007 onwards. It is not clear which factors caused this decrease although it coincided with the beginning of a period of intense focus on Pailin as the location of newly discovered artemisinin resistance [
20]. This resulted in intensification of malaria control activities in the area including increased availability of ACT and intensified active surveillance, such as focused and mass screening and treatment (FSAT and MSAT). It was also around this time that long-lasting insecticide treated bed nets were introduced [
21]. Additionally, extensive deforestation has occurred in Pailin during this period, resulting in very little forest cover in 2013. Before 2007, Pailin was the OD with the highest malaria transmission in Cambodia (API/1000 = 244 in 2004). That the surrounding Province of Battambang experienced a coincident decline in malaria, whilst other parts of Cambodia did not, may reflect the decreasing influence of the Pailin malaria ‘hot spot’ on adjacent areas. As of 2012, every village in Pailin has a VMW and this is likely to result in both improved collection of malaria data and a further decrease in the burden of disease through EDAT and directly-observed therapy (DOT). It is perhaps reassuring that the large decrease in malaria burden in Pailin was achieved despite the presence of reduced parasite clearance rates with artemisinins. However, reductions in disease burden were much less in Pursat Province where artemisinin resistance has also been identified [
22] and bed net coverage has been high. It is possible that this difference between Pursat and Pailin is due to there having been far fewer VMWs and less deforestation in Pursat.
The northeast of Cambodia differed from the rest of the country with an increase in
P. falciparum from 2007–2010 and, despite recent decreases in disease burden, is now the area with highest transmission intensity. There was also an even larger proportional increase in
P. vivax than that seen nationally since 2009. Reasons for this are not clear. The Northeast is a relatively remote area with remaining high forest cover and some parts have had relatively lower levels of malaria control activity than elsewhere in Cambodia. Although VMWs have been present in the Northeast since 2003, there have been relatively few in 2 of the 4 provinces (Steung Treng and Kratie). Additionally few ITNs had been distributed in Steung Treng to the end of 2012, although the rest of the Northeast had a high proportion of households with ITN. The main focus for malaria control recently has been on Pailin and other areas with or suspected to have artemisinin resistance [
23]. Changing seasonal migration between the Northeast and the highly populated south of the country is one possible factor worth exploring further [
24]. The contribution of ongoing deforestation in parts of Cambodia to the reduction in malaria also needs to be investigated further.
In 2013, the Provinces bordering Laos, Vietnam and Thailand in the north and northeast of the country had the highest transmission intensity for both species. These borders are frequented by migrant and mobile populations who have proven difficult to target for surveillance and malaria control activities. A recent study by CNM found 14% of individuals crossing the Cambodia-Laos border at one location in Steung Treng to be positive for malaria by polymerase chain reaction (PCR) and two thirds of these were asymptomatic (unpublished data). This raises the possibility that importation of malaria from neighbouring countries is contributing to the higher incidence of infection found in the border areas of Cambodia. Migrant populations more often delay seeking appropriate diagnosis and treatment and are a potential source and means of dispersal of anti-malarial drug resistance, which is frequently first identified in border areas [
25,
26].
This study had several limitations. This study did not measure transmission intensity directly. Instead, API was used as a surrogate measure. Although the majority of cases are likely to have been infected in the same district as where they were diagnosed, it should be noted that for some individuals transmission may have occurred elsewhere. Malaria surveillance data are notoriously prone to collection bias. Self-treated and privately treated cases including those treated by private pharmacies and traditional healers were not captured by these data and migrant and mobile populations were probably underrepresented. Although it is likely that many of these underrepresented cases chose to attend VMWs where available, the true numbers of such cases are not known. Remote populations are also thought to have been underrepresented by the HIS data [
18]; however, most of these communities are now included in the VMW project. Changes in efficiency of the systems used to collect the data are difficult to quantify and may introduce biases. The strength of these data is that several trends are apparent over time in the two independent datasets. During the study period, there was widespread deforestation, and the contribution of this to the reduction in malaria, transmitted predominantly by forest-dwelling mosquitoes was not examined in this study.
This analysis highlights a number of areas requiring further detailed study. Further investigation of trends over time in individual ODs matched with details of malaria control interventions employed, changes in environmental determinants and investigation of the contribution of mobile and migrant populations would be highly informative for planning future malaria control activities. In particular detailed data on ACT use have been collected in large national surveys and collection of information on the behavior and malaria burden in mobile and migrant populations (MMPs) is ongoing and a National MMP Strategy has been developed [
27]. The VMW project collects more detailed data than we have had space to present here in this paper including age, gender, migration status, pregnancy status and breakdown by individual village. This growing dataset is an invaluable resource for the future for Cambodia. By combining these data with other studies currently being undertaken, for example, on population migration and the geographical distribution of artemisinin resistance, important new insights will be gained.
Competing interests
The authors declare that they have no competing interest.
Authors’ contributions
RJM designed the study, performed the analyses and wrote the report. CN, PL, TB and PN arranged collection of the data. PL, PN, SEC, AMD and LJW wrote the report. All authors approved the final version of the manuscript.