The World Health Organization (WHO) clinical case definition for paediatric AIDS was tested during a 1-month period on 221 consecutive hospitalized children in Kigali, Rwanda. Relevant clinical features not included in the WHO case definition were also evaluated. Thirty-four out of the 221 children (15.4%) were HIV seropositive. Although the specificity of the WHO case definition was high (92%), the sensitivity and the positive predictive value (PPV) were low (41 and 48%, respectively). The following individual signs had a PPV at least equal to the complete WHO case definition: chronic diarrhoea (47%), respiratory distress secondary to lower respiratory tract infection (50%), oral candidiasis (53%), parotitis (67%), generalized lymphadenopathy (88%), and herpes zoster infection (100%). When logistic regression analysis was done on the nine variables included in the WHO case definition, confirmed maternal infection was the best predictive variable for HIV seropositivity in children (P less than 10(-5). We further excluded the serological status of the mother from the analysis and performed a stepwise logistic regression analysis on the 18 clinical signs and symptoms for which information had been collected. Those signs and symptoms contributing the most to the regression were: respiratory distress, chronic diarrhoea and generalized lymphadenopathy. Based on these findings, we propose a simplified clinical case definition for paediatric AIDS in Africa with better sensitivity, specificity and PPV than the WHO case definition. Further work is needed using this approach to develop case definitions useful for epidemiological surveillance and for case management.