Incremental diagnostic accuracy of dipstick urinalysis for the diagnosis of UTI compared with clinical evaluation alone in young children presenting with a febrile illness

Jonathan Craig1,2, Dr Mike Jones1,2, Dr Miriam Codarini1,2, Dr Petra Macaskill1
1. School of Public Health, University of Sydney
2. Centre for Kidney Research, The Childrens' Hospital at Westmead, Australia

Background: UTI occurs in about 10% of children. Typically the diagnosis is made and treatment commenced before the reference standard test, urine culture, is available. 70+ studies have evaluated the accuracy of dipstick for diagnosis of UTI in children but none have evaluated the incremental benefit of dipstick over clinical tests alone.

Methods: Since July 2004 a prospective cohort study of children under 5 years presenting with a febrile illness to the Emergency Department of The Children's Hospital, Westmead, was established (FEVER). We compared the accuracy of clinical assessment alone (routinely collected clinical history and physical findings), urinalysis alone (dipstick positive for leucocyte esterase and/or nitrites), physician judgment on probability of urinary tract infection and a combination of these tests, with the reference standard (positive urine culture using standard microbiological criteria). ROC methods accounted for multiple test thresholds with accuracy summarized as area under the curve (AUC).

Results: Complete data are available on 12,489 children (92.3% of those eligible). UTI was diagnosed in 375 (3%), of whom 65.1% were given antibiotics. AUC was 0.8435 for urinalysis combined with clinical assessment, 0.8090 for urinalysis alone, 0.7465 for clinical assessment alone, and 0.7066 for subjective judgment of probability of UTI (with maximum sensitivity of 50% only).

Conclusions: Signs and symptoms elicited by clinicians are moderately accurate for UTI, especially when combined explicitly and statistically and not just implicitly by clinicians who tend to discount the sensitivity of their judgment. Adding urinalysis results substantially improves accuracy and should reduce delay in antibiotic use.

Presented at the Australian & New Zealand Society of Nephrology [ANZSN]. 42nd Annual Meeting. Melbourne , 16-18 Aug 2006