A retrospective study based on paediatric ward registers was conducted in the Ugandan districts of Hoima and Kabale, which are areas of stable and unstable malaria transmission, respectively. The records of Hoima hospital from 1990 to 2001 and of Kabale hospital from 1994 to 2000 were reviewed and the initial diagnoses for all young children (i.e. those aged <5 years) were noted. Admissions for malaria and for anaemia were significantly more common among the young children admitted to Hoima hospital than among those admitted to Kabale hospital (P<0.0001 for each). Over the study periods, there were significant linear increases in the numbers of young children admitted with malaria or anaemia, at both Hoima hospital (with χ² values of 25.6 and 191.5, respectively; P<0.0001 for each) and at Kabale hospital (with χ² values of 31.6 and 29.0, respectively; P<0.0001 for each). Anaemia was not an important cause of mortality at Kabale hospital during the period reviewed. As in other sites in the East African highlands, the increasing malaria-related morbidity and mortality at Kabale hospital between 1994 and 2000 could be explained by the general increase in air temperatures over the same period. This increase may have made the local climate more conducive to mosquito survival and to parasite development in the vector, leading to increases in the intensity of transmission. At Hoima hospital, however, the increasing numbers of admissions for anaemia or malaria between 1990 and 2001 seem more likely to be the result of increased resistance to chloroquine (CQ) and sulfadoxine-pyrimethamine (SP) in the parasites and to changes in treatment-seeking behaviour. With the recent change in the national drug policy, from the use of CQ alone as the first-line treatment of uncomplicated malaria to the use of a combination of CQ with SP, and the abolition of user charges at government health facilities, a reversal of these worrying trends might be anticipated. Although it may be not be appropriate to extrapolate the conclusions of studies based on hospital records to the communities at risk of malaria, such conclusions do allow the health services to monitor general trends in the morbidity and mortality associated with malaria and anaemia.