Isolation of Candida from blood is the gold standard for the diagnosis of candidaemia, but because of rather low sensitivity and difficulty in collection of large amounts of blood from children, alternative methods are required. We evaluated the reliability of candiduria as an indicator of systemic candidosis. Blood and suprapubic sample of urine were cultured on the same day from 91 children of paediatric intensive care unit who were clinically suspected to have disseminated candidosis. Candida was isolated from blood in 41.8 per cent patients, from urine in 50.5 per cent patients and both from blood and urine in 27.5 per cent patients. C tropicalis was the commonest yeast isolated from blood (55.3%) and urine (58.7%). The sensitivity and specificity of urine culture for predicting candidaemia were 65.8 and 60.4 per cent respectively, positive predictive value was 54.3 per cent and negative predictive value 71.1 per cent. Urine microscopy had comparable sensitivity (52.6%), specificity (75.5%), positive (60.6%) and negative (70.0%) predictive values. Therefore, candiduria is not a reliable indicator of candidaemia. However, the isolation of non-albicans Candida species from urine was better indicator for candidaemia compared to isolation of C. albicans as 59.5 per cent patients with non-albicans Candida species in urine had candidaemia compared to 33.3 per cent with C. albicans.