The Thai-Cambodian border has long been considered to be an epicentre of emerging resistance of Plasmodium falciparum to anti-malarial drugs, including chloroquine, sulphadoxine-pyrimethamine and mefloquine [1]. Artemisinin-based combination therapy (ACT) was implemented as the first-line treatment for uncomplicated falciparum malaria in Cambodia in 2000. Artemisinin resistance, expressed as reduced parasite clearance rates in vivo, was first detected in Pailin province in 2007–2008 [2] and has recently been reported in neighbouring Pursat province [3]. In late 2008 Cambodia and Thailand embarked on a joint strategy for the containment of artemisinin-tolerant malaria parasites in Southeast Asia (ARCE), which involved the introduction or scaling up of a number of interventions in key districts on both sides of the Thai-Cambodia border [4]. Within this strategy the presence of P. falciparum parasitaemia 72 hours after the initiation of treatment with an ACT (i.e. at day-3) has been used as a standard in vivo indicator of the possible presence of drug resistance [5]. In areas considered to be at high risk of developing artemisinin resistance (and particularly areas bordering zones where resistance has already been detected), there is a recognized need to develop surveillance systems through which health care providers can identify and report delayed clearance at day-3, both as a means of monitoring geographical patterns of resistance and as a basis for facilitating appropriate response activities [4].

Within Cambodia the ARCE strategy has supported the National Centre for Parasitology, Entomology and Malaria Control (CNM) in developing new approaches to malaria surveillance and stratification. This has included the introduction of a new system of routine case reporting that supports incidence-based stratification of malaria at the village level. It has also involved testing systems for real-time, community-level reporting of cases at the point of care using village malaria workers (VMWs). Since its introduction in 2001, the Cambodia VMW network has been substantially scaled up [6,7] and today covers a total of 1,576 villages in 18 malaria-endemic provinces (unpublished CNM data). VMWs are primarily tasked with performing rapid diagnostic tests (RDTs) on individuals suspected of having malaria and with treating or referring test-positive cases according to defined guidelines. They are also encouraged to conduct active case detection, follow up malaria patients and disseminate information on malaria preventive measures.

Between July 2010 and September 2011 a new system using community-based VMWs to detect, treat and report day three-positive cases of P. falciparum was piloted in selected districts of western Cambodia by CNM with technical support from University Research Co. (URC), FHI 360 and the Malaria Consortium. The primary objectives of the pilot phase were: (a) to assess the feasibility and viability of community-based systems to capture and disseminate information on day three-positive P. falciparum cases; and, (b) to collect and analyse basic epidemiological data relating to day three-positive P. falciparum cases. A secondary objective was to test the viability of a SMS alert system for providing CNM and other key stakeholders with real-time notifications of day three-positive P. falciparum cases. This paper focuses on the effectiveness of surveillance activities introduced at the community level through the VMW system.

day-3 surveillance pilot activities have provided valuable data on rates of day-3 P. falciparum positivity in key areas of western Cambodia and also on the programmatic viability of community-based systems for detecting and reporting day-3 positive cases. The overall rate of day-3 positivity across all pilot sites was 21%, with the largest number of both day-0 and day-3 positive cases coming from a single site (Ta Sanh). This figure greatly exceeds not only the 3-5% background rate of day-3 positivity that might be expected in the absence of resistance to artemisinin, but also the 3-10% range which in the past has been seen as appropriate window for initiating containment activities [5]. Data from these community-based pilots are consistent with those published from therapeutic efficacy studies [8] which suggest that in key areas of western Cambodia rates of day-3 positivity far exceed the threshold figure of 10% which has been suggested as a possible indicator for the presence of drug resistance [4]. While attention has been previously focused on Pailin, data from these pilots indicate high levels of day-3 positivity are also characteristic of parts of neighbouring Battambang.

Community case management of malaria through village volunteers has been identified as a key mechanism for achieving targets relating to prompt access of appropriate malaria treatment [9] and as a prerequisite for malaria elimination [10]. The viability of using community-based volunteers to prescribe appropriate treatment on the basis of RDT diagnosis has been demonstrated in a variety of settings [11-22] and, as of 2010, 42 malaria-endemic countries have reported some degree of implementation of RDTs at community level [9]. Cambodia has a particularly long history of community-based diagnosis and treatment of malaria using VMWs [7] and there is good evidence that this has been an effective mechanism for improving access to appropriate diagnosis and treatment [23]. In areas of potentially emergent drug resistance the detection and follow-up of individuals with prolonged parasite clearance times would seem to be a logical and justifiable extension of the role of VMWs; however findings from this evaluation have also highlighted the non-trivial nature of implementing and sustaining day-3 surveillance activities. Effective day-3 surveillance is a highly intensive activity that introduces new roles and responsibilities (as well as significant extra work load) to VMWs and health facility staff. For the system to be effective patients testing positive for P. falciparum on day-0 need to be successfully traced and re-tested on day-3 and adherence to the relevant ACT regimen needs to be assured through the administration of DOT. In any event, not all of these elements were achieved consistently at all of the pilot sites.

Results presented in the previous section focused on key variations in protocols and experiences among different arms of the pilot; however a consistent finding from interviews conducted at all sites was the very high level of project acceptability and near universal support for the public health rationale of the project. There was clearly an appetite on the part of VMWs and health centre staff to improve the quality of care provided for malaria cases and day-3 surveillance was seen to contribute directly to this through its emphasis on treatment and monitoring of parasite clearance. Notably, however, these high levels of acceptability did not always translate into high levels of engagement. For most VMWs and facility staff, issues relating to existing workloads, financial incentives and availability of basic supplies created practical limits to the amount of time and effort they could justify in supporting the project.

A clear outcome from the pilot phase has been a strong evidence base to support the feasibility of surveillance predicated on blood slides obtained by VMWs. VMWs at all sites were willing and able to prepare blood films as part of their routine activity and with appropriate training VMWs with no relevant experience were capable of producing good quality smears by the end of the pilot. The amount of training input needed to achieve this should not be underestimated, however, and regular refresher training and supervision are required, particularly in very low transmission settings where VMWs have limited opportunity to practice these skills. This is consistent with experience in Vietnam, where village health workers are routinely expected to prepare blood films but lack of appropriate training has been identified as a key determinant of poor performance [24]. It is worth noting that within the URC pilot all refresher training was carried out within the existing system of supervision and support via routine monthly meetings.

Evidence from the evaluation suggests that effective DOT is achievable but requires a high level of motivation on the part of VMWs. There was considerable variability between (and sometimes within) pilot sites in terms of VMW provision of DOT. At Ta Sanh, VMWs routinely provided DOT and travelled to patients’ houses to do so. At all other sites arrangements were more variable, but in practice VMWs often opted to make appointments for patients to return to the VMW station for treatment. Some VMWs reported using accompanying friends and relatives to administer DOT. Others clearly had never attempted to provide DOT. Overall, the evidence points to the fact that comprehensive provision DOT cannot be achieved unless VMWs are provided with the resources (finance and transport) to allow them to follow up patients at home. It is unrealistic to expect all patients to re-visit VMWs on the second and third day to receive treatment. Indeed, instituting such a system would arguably be counter-productive and increase rates of non-compliance. Any system of DOT would also have to retain some flexibility and autonomy on the part of VMWs as to when and where DOT is achievable. Because under these circumstances the universal use of DOT cannot be assured, the validity of VMW-reported day-3 positivity rates should also be interpreted with caution.

On the evidence of process data and user feedback there was substantial variation between the sites in terms of the basic operational effectiveness of the pilot system. The most obvious distinction was between pilot activities carried out under the auspices of the non-governmental organisations (URC and FHI 360) and those supervised directly by CNM. In many ways the CNM system represented a ‘minimum’ model of implementation with comparatively low rates of remuneration for staff involved and with relatively modest inputs in terms of supervision. It is likely that the relatively smooth running of pilot activities under URC and FHI 360 can be attributed, at least in part, to strong supervisory support provided at the provincial level. In contrast, some of the problems experienced at Pramaoy could be linked to the absence of equivalent support mechanisms below the central level. Linked to this, it is also likely that pilot experiences were influenced strongly by the underlying characteristics of the VMW networks at each site. Most notably at Pramaoy, where the pilot was particularly ineffective, a large majority of VMWs complained about frequent (and sometimes chronic) stock-outs of ACT and RDTs, as well as problems with the supply of consumables for the day-3 pilot itself, both of which were symptomatic of weak supervision. For community volunteers the demotivating effect of such factors linked to their working environment have been clearly demonstrated in other settings [11,25].

Given the intensive nature of day-3 surveillance, the effectiveness of the system is linked directly to the performance of VMWs and key health centre staff (particularly laboratory staff), which itself will reflect a variety of individual-level and contextual factors [26,27]. Health worker motivation has been identified as an important direct and indirect determinant of performance and is therefore considered to be a key area for intervention. Previous reviews have highlighted the importance of financial incentives; however these alone may not be sufficient to motivate health workers [28-30], particularly in the case of community volunteers who are by definition non-salaried. Within the current evaluation none of the VMWs or health centre staff interviewed considered the payments provided through the pilot system to constitute an incentive. A number of VMWs pointed out that financial considerations were not a primary factor in their original decision to volunteer and most were keen to highlight the satisfaction they derived from serving their community and/or to their increased level of knowledge through training. At the same time it is clear that realistic payments for the extra tasks are a prerequisite for the operational viability of VMW-centred initiatives such as this. Although VMWs felt that payments were too low to fully compensate for the time and effort involved, they were generally considered sufficient to cover their basic costs. Most VMWs made it clear that they would be unable to continue following up malaria cases in the absence of these basic payments. A number of VMWs and health centre staff also stressed the importance of payments being linked explicitly to specific activities such as slide transport or patient follow-up. Within the CNM pilot it was evident that use of a flat-rate monthly payment to VMWs was not conducive to increasing levels of VMW engagement with pilot activities. Moreover, key aspects of VMWs’ practices within the URC and FHI 360 pilots could be linked to the payment structures used in each case. Under FHI 360, for example, VMWs were not paid to travel to patients’ houses to administer DOT and as a result patients were required to travel to the VMW’s house to receive treatment. This proved to be viable in Pailin but would not be recommended as a general model of implementation. Under URC, on the other hand, VMWs routinely carried out active follow-up of patients and a number stated explicitly that this would only be possible while they received financial compensation for doing so. At both sites compensation was contingent on a day-3 slide being submitted to the supervising health centre which no doubt contributed to very high rates of day-3 follow-up (as distinct from experiences at the CNM site).

Typically when asked what was good about the project, or what motivated them to be involved, VMWs focused on improved case management. However, many felt they were not receiving sufficient support through training, guidance or supervision to maximize their effectiveness in this area. It has been shown in other settings that provision of adequate training and appropriate job-aids are critical to health worker performance [11,20,31] and there is a need to examine areas of best practice from these pilots and similar studies to tailor these elements in future. As noted above, adequate supervision is also a key factor influencing health worker performance. This fact is widely recognized, although supervision systems for community volunteers are nevertheless typically weak [27]. On a positive note, experience from the current pilots indicates that many of the training and supervision inputs required to support day-3 surveillance activities can be achieved through the existing mechanism of VMW monthly meetings. In addition, relatively small changes in the way health centre staff provide feedback to VMWs may produce beneficial effects. Perhaps somewhat counter-intuitively the amount of effort required to implement and sustain day-3 surveillance activities is likely to be greater in low transmission settings than in areas where P. falciparum cases are relatively common.

Within the piloted systems much of the burden of activity rests with VMWs but the viability of the system also depends on effective engagement by key health centre staff, and most notably laboratory staff responsible for blood slide examination. Within this evaluation feedback from laboratory staff interviewed was quite variable, particularly regarding the importance (or otherwise) of financial incentives. However, there was a clear general consensus that the degree to which laboratories could support day-3 surveillance was limited inherently by pre-existing commitments to other projects and routine activities, and in many cases staff were reluctant to take on extra duties. It is clear that any attempt to scale up new surveillance activities will need to address this constraint and, where required, strengthen laboratory capacity. The issue of competing commitments is also highly relevant to VMWs who in addition to their own personal commitments are now required to treat acute respiratory infections and cases of diarrhoea [6] and in some cases may also be responsible for the delivery of other services, including maternal and child health and family planning. It has been suggested that in the future community volunteers may represent a suitable platform for community-based approaches for reducing malaria transmission, including active case detection, focused screening and treatment and targeted focal vector control [10]. However, the capacity of VMWs to support such activities requires careful review prior to developing new policy in this area.

It is worth noting that in the early planning stages of the day-3 surveillance pilots, the novel use of SMS messaging to provide rapid alerts of day-3 P. falciparum cases to CNM and other stakeholders was a prominent feature of planned activities. In practice, however, this technological element represented a very small component of the overall work flow and any problems users had with the system (e g, inexperience in sending texts) appeared to be entirely tractable. It is important to recognize that within the current day-3 framework information technology (IT) is used essentially to increase the effectiveness of surveillance data once they have been generated. IT does not make the gathering of these data any easier. In reality the principal impediments to achieving effective surveillance of day-3 P. falciparum cases relate to basic health system constraints (outlined above) and there are clear limits to the extent that technology can support this process. In the elimination literature there has been a tendency to see the strengthening of surveillance systems as a technical challenge involving the deployment of new means of managing and transmitting surveillance data [32-34]. Gains from these developments will only be realised, however, once wider health system constraints have been effectively resolved.

On balance, evidence from this evaluation suggests that introducing community-based systems for detecting and reporting day-3 positive cases of P. falciparum is viable, but that a number of prerequisites need to be successfully met before such a system can be considered to be practically feasible. At the most basic level pre-existing local VMW-health centre networks need to be effective, incorporating adequate supervision and resupply of RDTs, drugs, slides and other consumables. Relatively frequent training of VMWs is required to ensure that slide quality is adequate and to fully empower VMWs in the process of case management. Given the relatively intensive nature of day-3 surveillance activities it is essential that VMWs are directly compensated for travel associated with obtaining and delivering blood slides and administering DOT. Careful consideration needs to be given to existing time commitments of both VMWs and health centre staff and where necessary human and operational capacity needs to be strengthened before such add-on activities can be supported. This latter point is pertinent not only in the case of day-3 surveillance but also in the wider context of using community volunteers to support interventions aimed at drug resistance containment and/or malaria elimination.

Results from pilot activities indicate that an appropriately resourced and well-supported community-based system for day-3 surveillance is certainly capable of providing robust indicator data on day-3 positivity. However, on its own such a system cannot be considered comprehensive as it fails to capture P. falciparum cases presenting directly to government facilities or private providers. Creating a more comprehensive day-3 surveillance system would involve effective integration of community- and facility-based systems. A suitable blueprint for such a system is relatively easy to envisage. In practice, however, achieving the sort of close coordination of facility- and village-based activities required to make this work would be more challenging. The question of whether there is proper justification for this type of integrated system depends ultimately on the perceived purpose of day-3 surveillance. If it is primarily to serve as a platform for monitoring temporal changes in day-3 positivity rates then arguably an extensive sentinel site system, most likely consisting of inpatient facilities, would almost certainly be a more efficient and cost-effective approach to generating suitable datasets. If the purpose of day-3 surveillance is to identify as many P. falciparum infections in the community as possible, and use this information to clear these and nearby infections, other mechanisms (e.g. FSAT) may be more appropriate. If, however, the rationale for the day-3 system is to provide an effective alert system for identifying potential clusters of day-3 positives, the type of system piloted in this exercise will be potentially valuable, providing they are appropriately targeted geographically and linked explicitly to well-defined plans for response. Whatever option is selected for scale-up, this study underlines the importance of evaluation in assisting programmes and donors to make rational decisions.

ACT: Artemisinin-based combination therapy; ARCE: Strategy for the containment of artemisinin-tolerant malaria parasites in Southeast Asia; CNM: National centre for parasitology, entomology and malaria control; DOT: Directly observed therapy; IT: Information technology; RDT: Rapid diagnostic test; VMW: Village malaria workers.

The authors have declared that they have no competing interests.

JC, LDS, KST, SN, and SM conceived and designed the study. JC, LDS, TB, KST, and SN participated in the implementation and coordination of the study. JC performed the data analysis. All authors contributed to the writing of the manuscript and have approved the final version.