In countries where malaria is endemic, routine blood slide examinations remain the major source of data for the public health surveillance system. This approach has become inadequate, however, as the public health emphasis has changed from surveillance of laboratory-confirmed malaria infections to the early detection and treatment of the disease. As a result, it has been advocated that the information collected about malaria be changed radically and should include the monitoring of morbidity and mortality, clinical practice and quality of care. To improve the early diagnosis and prompt treatment (EDPT) of malaria patients, three malaria case definitions (MCDs) were developed, with treatment and reporting guidelines, and used in all static health facilities of Cox's Bazar district, Bangladesh (population 1.5 million). The three MCDs were: uncomplicated malaria (UM); treatment failure malaria (TFM); and severe malaria (SM). The number of malaria deaths was also reported. This paper reviews the rationale and need for MCDs in malaria control programmes and presents an analysis of the integrated surveillance information collected during the three-year period, 1995-97. The combined analysis of slide-based and clinical data and their related indicators shows that blood slide analysis is no longer used to document fever episodes but to support EDPT, with priority given to SM and TFM patients. Data indicate a decrease in the overall positive predictive value of the three MCDs as malaria prevalence decreases. Hence the data quantify the extent to which the mainly clinical diagnosis of UM leads to over-diagnosis and over-treatment in changing epidemiological conditions. Also the new surveillance data show: a halving in the case fatality rate among SM cases (from 6% to 3.1%) attributable to improved quality of care, and a stable proportion of TFM cases (around 7%) against a defined population denominator. Changes implemented in the EDPT of malaria patients and in the surveillance system were based on existing staff capacity and routine reporting structures.