Novel approaches to risk stratification to support malaria elimination: an example from Cambodia

Background: Accurate malaria stratification is essential for effective targeting of interventions but represents a particular challenge in pre-elimination settings. In these settings transmission is typically sufficiently low and spatially heterogeneous to warrant a need for estimates of malaria risk at sub-district or village level but is also likely to be sufficiently high to render the type of decision support systems appropriate to the final stages of malaria elimination impractical. In such a scenario it is arguably more feasible to strengthen existing passive malaria surveillance systems so that routinely generated case data can provide an effective basis for stratifying malaria risk. This paper explores the utility of routine malaria surveillance data for the stratification of malaria risk in Cambodia, where the target is malaria elimination by 2025. Methods: A malaria information system (MIS) was developed to generate timely, routine data on temporal and spatial variations in malaria cases reported through public health facilities and village malaria workers (VMWs). The MIS was implemented across all malaria endemic districts in the country during 2010-11. In 2012 MIS data were extracted and assessed on the basis of coverage and completeness. Village-level incidence estimates for 2011 were generated using predefined data inclusion criteria. Results: In 2011, the MIS covered 681 health facilities and 1,489 VMW villages; the overall completeness of monthly reporting was 82% and 97% for health facilities and VMWs respectively. Using these data it was possible to estimate malaria incidence for 89% of villages covered by the MIS. The resulting stratification highlights the highly heterogeneous nature of malaria transmission in Cambodia and underlines the importance of village-level data for effective targeting of interventions, including VMWs. Challenges associated with implementing the MIS and the implications of these for developing viable and sustainable MIS in Cambodia and elsewhere are discussed. Conclusions: This study demonstrates the operational feasibility of introducing a system to routinely generate village level malaria case data in Cambodia. Although resulting incidence estimates are subject to various limitations and biases the data provide an objective, repeatable basis for a dynamic system of stratification which is appropriate for guiding the transition between malaria pre-elimination and elimination phases.