Background: It is not clear which young children presenting acutely unwell to primary care should be investigated for urinary tract infection (UTI) and whether or not dipstick testing should be used to inform antibiotic treatment.
Objectives: To develop algorithms to accurately identify pre-school children in whom urine should be obtained; assess whether or not dipstick urinalysis provides additional diagnostic information; and model algorithm cost-effectiveness.
Design: Multicentre, prospective diagnostic cohort study.
Setting and participants: Children < 5 years old presenting to primary care with an acute illness and/or new urinary symptoms.
Methods: One hundred and seven clinical characteristics (index tests) were recorded from the child's past medical history, symptoms, physical examination signs and urine dipstick test. Prior to dipstick results clinician opinion of UTI likelihood ('clinical diagnosis') and urine sampling and treatment intentions ('clinical judgement') were recorded. All index tests were measured blind to the reference standard, defined as a pure or predominant uropathogen cultured at ≥ 10(5) colony-forming units (CFU)/ml in a single research laboratory. Urine was collected by clean catch (preferred) or nappy pad. Index tests were sequentially evaluated in two groups, stratified by urine collection method: parent-reported symptoms with clinician-reported signs, and urine dipstick results. Diagnostic accuracy was quantified using area under receiver operating characteristic curve (AUROC) with 95% confidence interval (CI) and bootstrap-validated AUROC, and compared with the 'clinician diagnosis' AUROC. Decision-analytic models were used to identify optimal urine sampling strategy compared with 'clinical judgement'.
Results: A total of 7163 children were recruited, of whom 50% were female and 49% were < 2 years old. Culture results were available for 5017 (70%); 2740 children provided clean-catch samples, 94% of whom were ≥ 2 years old, with 2.2% meeting the UTI definition. Among these, 'clinical diagnosis' correctly identified 46.6% of positive cultures, with 94.7% specificity and an AUROC of 0.77 (95% CI 0.71 to 0.83). Four symptoms, three signs and three dipstick results were independently associated with UTI with an AUROC (95% CI; bootstrap-validated AUROC) of 0.89 (0.85 to 0.95; validated 0.88) for symptoms and signs, increasing to 0.93 (0.90 to 0.97; validated 0.90) with dipstick results. Nappy pad samples were provided from the other 2277 children, of whom 82% were < 2 years old and 1.3% met the UTI definition. 'Clinical diagnosis' correctly identified 13.3% positive cultures, with 98.5% specificity and an AUROC of 0.63 (95% CI 0.53 to 0.72). Four symptoms and two dipstick results were independently associated with UTI, with an AUROC of 0.81 (0.72 to 0.90; validated 0.78) for symptoms, increasing to 0.87 (0.80 to 0.94; validated 0.82) with the dipstick findings. A high specificity threshold for the clean-catch model was more accurate and less costly than, and as effective as, clinical judgement. The additional diagnostic utility of dipstick testing was offset by its costs. The cost-effectiveness of the nappy pad model was not clear-cut.
Conclusions: Clinicians should prioritise the use of clean-catch sampling as symptoms and signs can cost-effectively improve the identification of UTI in young children where clean catch is possible. Dipstick testing can improve targeting of antibiotic treatment, but at a higher cost than waiting for a laboratory result. Future research is needed to distinguish pathogens from contaminants, assess the impact of the clean-catch algorithm on patient outcomes, and the cost-effectiveness of presumptive versus dipstick versus laboratory-guided antibiotic treatment.
Funding: The National Institute for Health Research Health Technology Assessment programme.