OBJECTIVES--To assess whether ultrasonography alone is adequate for routine screening of childhood urinary infection, whether clinical features determine the need for further investigations, and which investigations are most appropriate. DESIGN--Prospective survey of children with proved urinary infection and a preinvestigation record of clinical features. Ultrasonography and intravenous urography were routine, with choice of further studies determined by ultrasonographic findings. SETTING--A children's hospital and two district general hospitals in Mersey region. MAIN OUTCOME MEASURES--Sensitivity and specificity of ultrasonography both generally and in relation to clinical features. Accuracy of intravenous urography compared with radioisotope examinations. RESULTS--Specificity of ultrasonography was good (99% (95% confidence interval 96% to 100%)) but sensitivity modest (43% (32% to 55%)), principally with respect to detecting vesicoureteric reflux and renal scarring. Among older children (aged 2-10 years) with positive ultrasound results and fever or vomiting the sensitivity in detecting reflux (with and without renal scarring) was 78% (62% to 89%) and the specificity 69% (60% to 78%); in detecting renal scarring (with and without reflux) the sensitivity was 100% (80% to 100%) and specificity 65% (56% to 74%). Renal scarring and obstructive uropathies were better assessed by radioisotope examinations than by intravenous urography. CONCLUSIONS--Ultrasonography alone is inadequate for routine screening of childhood urinary infection. Though further investigations remain advisable in infants, in older children they can be restricted to a minority who have positive ultrasound examinations or have had fever or vomiting. Radioisotope examinations largely eliminate the need for intravenous urography.