A decline in malaria transmission is evident in malaria endemic areas of sub-Saharan Africa and is likely to
reduce the proportion of fevers due to malaria. Fever has been used as a predictor of malaria, however, the
proportion of fevers due to malaria vary with prevalence such that low malaria infection prevalence might alter the
accuracy of fever as a marker of malaria. This study examined the diagnostic accuracy and proportion of fevers
attributable to malaria among under-fives in a cross-sectional survey carried out in Bagamoyo district, Tanzania from
April–May 2012 during peak malaria transmission. Consecutive under-fives with and without history of fever were
recruited; for each, fever was measured by digital thermometer, and two Giemsa stained thick and thin blood films
taken for parasite count and species identification. Accuracy of fever for prediction of malaria was assessed by
performance indices, microscopy as gold standard. Proportion of fevers attributable to malaria was computed by the
odds ratio technique at 0.05 significance level.{Formatting Citation}
Only 98 out of 925 (10.6%) under-fives had parasitaemia. Among under-fives with a history of fever, the fraction
attributable to malaria was 71.4% [95%CI: 54.8–81.9]; in those with measured fever ≥ 37.5˚C, the fraction was
74.3% [95%CI: 61.8–82.7]. In bivariate and multivariate analyses, at 1001-10000 parasites/µl the attributable
fraction was 66%, and 93% for parasitaemia>10000/µl. Fever was more likely to be due to malaria among
infants<12 months than subsequent months. Despite the recorded decline in malaria infection prevalence, fever is
highly likely to be due to malaria among under-fives with fever and malaria infection in peripheral blood. This
observation highlights the need to scale up and maintain parasitological confirmation of malaria; and to look for other
causes of fever