OBJECTIVES: Diagnosis of urinary tract infection (UTI) in young children is difficult.  Prompt treatment could alleviate short-term symptoms and prevent serious long-term sequelae, but over-treatment will increase antibiotic resistance. We evaluated the cost-effectiveness of a UTI risk score based on signs, symptoms and dipstick test findings compared to clinical judgement in guiding urine sampling and antibiotic treatment. METHODS: We developed a risk score based on urine samples collected (by clean catch or nappy pad) in a multicentre diagnostic cohort study (DUTY) of 7,163 children <5 years presenting to primary care. The diagnostic value of symptoms, signs and dipstick test results were evaluated against a reference standard of urine culture results from a research laboratory. We constructed decision-analytical models comparing the cost-effectiveness of 3 DUTY risk score thresholds (high sensitivity, high specificity or intermediate) versus clinical judgement in younger (nappy pad) and older (clean catch) children. We explored the role of the dipstick in guiding diagnosis. We considered health service costs and patient utilities during the initial diagnosis, acute illness and long-term sequelae. RESULTS: The 'high specificity' DUTY threshold resulted in fewer urine samples than clinical judgement (4.8% versus 9.2%) with similar sensitivity (58.6% versus 57.1%) and higher specificity (96.1% versus 91.4%).  The difference in short-term net benefits between DUTY thresholds was small (range £1088 'high sensitivity' to £1091 'high specificity'). In younger children (nappy pads) the distinction in cost-effectiveness between the DUTY risk score and clinical judgement was not clear-cut.  Dipstick tests could potentially expedite therapy in higher risk children. CONCLUSIONS: Clinicians can reduce prescriptions and provide more cost effective care by using the DUTY risk score. Clean catch samples should be obtained whenever practical. Low UTI prevalence, imperfect NHS laboratory tests and an uncertain link between UTI and long-term sequelae mean that conservative sampling strategies may be most appropriate.