BACKGROUND:

Identification of children who need antimalarial treatment is difficult in settings where confirmatory laboratory testing is not available, as in much of sub-Saharan Africa. The current national policy in Malawi is to treat all children with fever, usually defined as the mother's report of fever in the child, for presumed malaria. To assess this policy and to find out whether a better clinical case definition could be devised, we studied acutely ill children presenting to two hospital outpatient departments in Malawi.

METHODS:

The parent or guardian of each enrolled child (n = 1124) was asked a standard series of questions about the symptoms and duration of the child's illness. Each child was examined, axillary and rectal temperatures and blood haemoglobin concentrations were measured, and a giemsastained thick smear was examined for malaria parasites. Logistic regression procedures were used to identify clinical predictors of parasitaemia.

FINDINGS:

High temperature (37.7 degrees C or above), nailbed pallor, enlarged spleen, and being seen at one of the clinics rather than the other were associated with an increased risk of malaria parasitaemia in univariate analyses. A revised malaria case definition of rectal temperature of 37.7 degrees C or higher, splenomegaly, or nailbed pallor was 85% sensitive in identifying parasitaemic children and 41% specific; the corresponding sensitivity and specificity for the nationally recommended definition that equates mother's history of fever with malaria were 93% and 21%. The revised case definition had 89% sensitivity in identifying parasitaemic children with haemoglobin concentration below 80 g/L and 89% sensitivity in identifying children with parasite density greater than 10,000/microL, characteristics that indicate a clear need for antimalarial treatment.

INTERPRETATION:

These results suggest that better clinical definitions are feasible, that splenomegaly and pallor are helpful in identifying children with malaria, and that much overtreatment of children without parasitaemia could be avoided.