National disease burdens are often not estimated at all or are estimated using inaccurate methods, partly because the data sources for assessing disease burden-nationally representative household surveys, demographic surveillance sites, and routine health information systems-each have their limitations. An important step forward would be a more consistent quantification of the population at risk of malaria. This is most likely to be achieved by delimiting the geographical distribution of malaria transmission using routinely collected data on confirmed cases of disease. However, before routinely collected data can be used to assess trends in the incidence of clinical cases and deaths, the incompleteness of reporting and variation in the utilization of the health system must be taken into account. In the future, sentinel surveillance from public and private health facilities, selected according to risk stratification, combined with occasional household surveys and other population-based methods of surveillance, may provide better assessments of malaria trends.