Background The accuracy of malaria diagnosis has received renewed interest in recent years due to changes in treatment policies in favour of relatively high-cost artemisinin-based combination therapies. The use of rapid diagnostic tests (RDTs) based on histidine-rich protein 2 (HRP2) synthesized by Plasmodium falciparum has been widely advocated to save costs and to minimize inappropriate treatment of non-malarial febrile illnesses. HRP2-based RDTs are highly sensitive and stable; however, their specificity is a cause for concern, particularly in areas of intense malaria transmission due to persistence of HRP2 antigens from previous infections. Methods In this study, 78,454 clinically diagnosed malaria patients were tested using HRP2-based RDTs over a period of approximately four years in four highland sites in Kenya and Uganda representing hypoendemic to mesoendemic settings. In addition, the utility of the tests was evaluated in comparison with expert microscopy for disease management in 2,241 subjects in two sites with different endemicity levels over four months. Results RDT positivity rates varied by season and year, indicating temporal changes in accuracy of clinical diagnosis. Compared to expert microscopy, the sensitivity, specificity, positive predictive value and negative predictive value of the RDTs in a hypoendemic site were 90.0%, 99.9%, 90.0% and 99.9%, respectively. Corresponding measures at a mesoendemic site were 91.0%, 65.0%, 71.6% and 88.1%. Although sensitivities at the two sites were broadly comparable, levels of specificity varied considerably between the sites as well as according to month of test, age of patient, and presence or absence of fever during consultation. Specificity was relatively high in older age groups and increased towards the end of the transmission season, indicating the role played by anti-HRP2 antibodies. Patients with high parasite densities were more likely to test positive with RDTs than those with low density infections. Conclusion RDTs may be effective when used in low endemicity situations, but high false positive error rates may occur in areas with moderately high transmission. Reports on specificity of RDTs and cost-effectiveness analyses on their use should be interpreted with caution as there may be wide variations in these measurements depending upon endemicity, season and the age group of patients studied.