NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Hay AD, Birnie K, Busby J, et al.; on behalf of the DUTY team. The Diagnosis of Urinary Tract infection in Young children (DUTY): a diagnostic prospective observational study to derive and validate a clinical algorithm for the diagnosis of urinary tract infection in children presenting to primary care with an acute illness. Southampton (UK): NIHR Journals Library; 2016 Jul. (Health Technology Assessment, No. 20.51.)

Cover of The Diagnosis of Urinary Tract infection in Young children (DUTY): a diagnostic prospective observational study to derive and validate a clinical algorithm for the diagnosis of urinary tract infection in children presenting to primary care with an acute illness

The Diagnosis of Urinary Tract infection in Young children (DUTY): a diagnostic prospective observational study to derive and validate a clinical algorithm for the diagnosis of urinary tract infection in children presenting to primary care with an acute illness.

Show details

Chapter 4Microbiological diagnosis of urinary tract infection by NHS and research laboratories

Introduction

Laboratory diagnosis is based on colony counts following culture, which reflect the concentration of organisms in urine and hence the likelihood that the bacteria grown arise from a UTI rather than contamination. UTI is typically caused by a single organism that is present in a high concentration, usually ≥ 105 CFU/ml.81 However, laboratory guidelines differ regarding the nature and extent of bacterial growth required to confirm UTI.82,83 NICE guidelines do not provide a definitive threshold for diagnosing UTI on culture but provide advice about the level of bacterial growth in relation to symptoms and signs.2 Based on paediatric data, others have since proposed higher thresholds, for example ≥ 106 CFU/ml.80 It is not surprising, then, that laboratory guidelines differ regarding the nature and extent of bacterial growth required to confirm UTI according to the patient’s age, symptoms and urine collection method.82,83 Although NHS laboratories in the UK follow the UK Standards for Microbiological Investigation for examination of urine, application of the method varies between laboratories.

For this reason, and when sufficient urine was available, we sent a fraction of the urine to a single research laboratory where it was to be processed by a small number of technicians and according to a SOP. Although we wanted the algorithm results to be generalisable to current NHS practice, we did not know whether the NHS or research laboratory was providing the more reliable and accurate results.

Therefore, the aim of this chapter is to report the comparison of reliability and accuracy of UTI laboratory diagnosis between routine NHS laboratories and a single research laboratory. We investigated associations of pre-specified symptoms and signs related to UTI, and urine dipstick test results identified from the literature, with different laboratory definitions of urine culture positivity.

Methods

Urine collection and sample processing

Urine samples were obtained by clean catch, where possible, for children who were toilet trained or for whom the parent was happy to attempt such collection. A full description of the urine collection methods is given in Chapter 2.

NHS laboratories processed urine samples using their local SOPs and recorded data according to local reporting procedures, including, where possible, quantifying bacterial growth (as < 103; 103 to < 105; or ≥ 105 CFU/ml), purity of growth (pure/predominant; mixed growth two species; mixed growth > 2 species), organism speciation for up to two species, and microscopy for white and red cells. The research laboratory quantified absolute colony counts (range 101–1010 CFU/ml) for all organisms present and established species ID for organisms present at ≥ 103 CFU/ml.

Statistical analysis

Results reported by the NHS and research laboratories were initially classified based on extent and purity of growth and whether or not the species grown was a uropathogen, defined as a member of the Enterobacteriaceae group. For NHS laboratory results, samples for which the laboratory reported pure/predominant growth of a uropathogen at ≥ 105 CFU/ml were considered UTI positive. For research laboratory results, samples were considered UTI positive if there was growth of ≥ 105 CFU/ml of a single uropathogen (‘pure growth’) or growth of ≥ 105 CFU/ml of a uropathogen with ≥ 3 log10 difference between the growth of this and the next species (‘predominant growth’). Agreement between laboratories was assessed using kappa statistics, with analyses additionally stratified by urine collection method (clean catch or nappy pad) and by age (0 to < 2, 2 to < 3 and 3 to < 5 years).

Analyses were restricted to samples with a result from both the NHS and research laboratories. Samples classified as UTI positive by both NHS and research laboratories results were denoted ‘Agree UTI positive (step 1)’. Where there was disagreement between the NHS and research laboratories, we considered whether or not the combined evidence was consistent with a UTI. When the result classified as negative in one of the laboratories met either of the two conditions (1) growth of ≥ 105 CFU/ml of a uropathogen with lesser growth of at most one other species, or (2) pure/predominant growth of 103 to 105 CFU/ml of a uropathogen, samples were denoted ‘Agree UTI positive (step 2)’.

A priori (before examining their associations with different definitions of microbiological positivity from different laboratories), we selected a small number of variables (symptoms, signs and dipstick test results) reported in the literature to be clearly related to presence of a UTI.116 Those thought suitable for all ages and collection methods were urinary symptoms (pain/crying when passing urine, passing urine more often, changes in urine appearance); temperature ≥ 39 °C, and nitrite- or leucocyte-positive results from urine dipstick tests. Additional symptoms and signs thought to be relevant mainly for older children were daytime or bed wetting when previously dry, and a history of UTI. Most parent-reported symptoms were recorded using categories ‘no’, ‘slight’, ‘moderate’ or ‘severe problem’; we decided a priori (based only on inspection of symptom frequencies) to dichotomise these into ‘no or slight problem’ and ‘moderate or severe problem’. Responses of ‘not known’ were coded with ‘no or slight problem’. A small number of samples for which there were missing data on most or all urine symptoms (five samples), or for which urine dipstick tests were not available (12 samples), or there was missing information on prior infection (three samples) were excluded. Missing data on temperature (204 children) were coded as < 39 °C. The remaining, sporadic missing values (on five children) were coded as ‘no or slight problem’.

We used separate logistic regression models to quantify associations of the selected variables with UTI positivity in the NHS and research laboratories, and for the different outcomes ‘agree UTI’, ‘disagree (NHS positive, research negative)’ and ‘disagree (NHS negative, research positive)’, all compared with ‘agree UTI negative’. We plotted receiver operating characteristic (ROC) curves and AUROCs to quantify diagnostic utility. The maximum value of the AUROC is 1 (perfect prediction) while a value of 0.5 corresponds to no association with any predictor. The symptoms ‘day or bed wetting when previously dry’ and ‘history of UTI’ were recorded in too few of the children who provided nappy pad samples to permit their associations with microbiology results to be examined. For these children, therefore, we examined associations of the remaining six signs, symptoms and dipstick results with microbiology results. For children who provided a clean-catch sample, we fitted two sets of logistic regression models, one including all eight symptoms and dipstick results (the ‘eight variable model’) and the other, for comparability, including the same six symptoms and results as for children providing nappy pad samples (the ‘six variable model’).

In sensitivity analyses we (1) stratified by age (< 3 and ≥ 3 years); (2) allowed for ‘not known’ categories in the questionnaire responses for variables for which such responses occurred sufficiently frequently; (3) stratified according to whether samples were coded as ‘agree UTI positive’ at step 1 or step 2; (4) stratified by whether the sample was collected at the surgery or at home; (5) stratified by time between recruitment and laboratory sample receipt (< 24 hours and ≥ 24 hours); (6) stratified NHS laboratory results according to extent of pure/predominant growth (≥ 105, ≥ 103 to < 105 CFU/ml); (7) stratified research laboratory results according to extent of pure/predominant growth (≥ 107; ≥ 106 to < 107; ≥ 105 to < 106; ≥ 104 to < 105; and ≥ 103 to < 104 CFU/ml); (8) stratified NHS and research laboratory results according to whether the WBC count was < 30 or ≥ 30/mm3; (9) stratified research laboratory results according to whether growth was pure or predominant, and (10) as the research laboratory received only the urine available after the priority fraction was placed in the NHS collection sample, we stratified by research laboratory urine volume (using the cut-point of median urine Monovette weight for clean-catch and nappy pad samples). All analyses were carried out using Stata version 12 (StataCorp LP, College Station, TX, USA).

Results

Of the 7163 children included in the study, 6241 provided a urine sample using the clean-catch or nappy pad collection methods, for whom 5945 and 5071 culture results were available from the NHS and research laboratories, respectively. A total of 4828 children had results from both laboratories, of whom 4808 had information available on candidate predictors. [The starting point for the number of urine results in this chapter was determined by the use of the individuals with cultured specimens in both laboratories (see box B in Figure 4).] The detail of the urine samples in this chapter is shown in Figure 8. Most children (4543, 94.5%) were recruited from GP surgeries (Table 23).

FIGURE 8. Flow of participants in microbiological diagnosis of UTI in young children.

FIGURE 8

Flow of participants in microbiological diagnosis of UTI in young children. a, See Figure 4 for how this number was reached.

TABLE 23

TABLE 23

Characteristics of children and urine samples collected via clean catch or nappy pads, for the 4808 children with both NHS laboratory and research laboratory results

There were approximately equal numbers of boys and girls. Of 2884 children aged < 3 years, urine samples for 758 (26.3%) were collected using clean catch. By contrast, 1861 (96.7%) of 1924 children aged 3–5 years provided a clean-catch sample. Among children aged < 3 years, samples were obtained in the surgery in 1470 cases (51.0%), compared with 1477 (76.8%) among children aged 3–5 years. Parents reported the following symptoms as a moderate or severe problem: pain or crying when passing urine in 217 (4.5%) children; passing urine more often in 484 (10.1%); day or bed wetting when previously dry in 209 (4.3%); and a change in urine appearance in 523 (10.9%). The symptoms that were most commonly reported as ‘not known’ among children aged < 3 years were pain or crying when passing urine (980, 34.0%) and passing urine more often (1038, 40.0%). A history of UTI was reported in 221 (4.6%) children, 140 of whom were aged ≥ 3 years. Only 185 (3.8%) children had a temperature ≥ 39 °C. Dipstick urine results were positive for nitrites in 416 (8.7%) and for leucocytes in 670 (13.9%) children. Both nitrite (12.9% compared with 2.2%) and leucocyte (16.0% compared with 10.8%) positivity were more common in children aged < 3 years than in children aged ≥ 3 years. Monovette weights (grams) were available for all but 22 of the research laboratory urine samples. Median weight was 16.23 g (IQR 13.67–17.03 g) for clean-catch samples, and 11.08 g (IQR 9.53–13.26 g) for nappy pad samples.

A total of 251 (5.2%) and 88 (1.8%) samples were classified as positive according to the NHS and research laboratory result, respectively. The causative organism distributions were similar between laboratories: in the NHS, E. coli 71.7%, unidentified coliforms 19.5%, other coliforms 2.8%, Proteus spp. 6.0%; and in the research laboratory, E. coli 84.1%, other coliform (Klebsiella spp., Enterobacter spp., Serratia spp., Citrobacter spp., Morganella spp.) 10.2%, Proteus spp. 5.7%. NHS laboratory positivity was more common (6.6%) in children aged < 3 years than in those aged 3–5 years (3.2%). By contrast, rates of research laboratory positivity were similar in these age groups (1.8% and 1.9%, respectively). Only 64 (1.3%) samples were positive in both laboratories and coded as ‘agree UTI positive (step 1)’. In 187 (3.9%) the NHS laboratory result was positive but the research laboratory result negative, while in 24 (0.5%) the research laboratory result was positive but the NHS laboratory result negative (Table 24).

TABLE 24

TABLE 24

Extent of agreement between NHS and research laboratory result

Overall agreement between the NHS and research laboratories was moderate (kappa = 0.36; 95% CI 0.29 to 0.43; see Table 24). Agreement was better for clean-catch samples (0.54; 95% CI 0.45 to 0.63) than for nappy pads (0.20; 95% CI 0.12 to 0.28). For children aged ≥ 3 years, too few nappy pad samples were available to allow assessment of reliability. For clean-catch samples, reliability was similar in children aged ≥ 3 years (0.55; 95% CI 0.43 to 0.67) and < 3 years (0.52; 95% CI 0.37 to 0.67), which was better than for nappy pad samples in children aged < 3 years (0.20; 95% CI 0.12 to 0.28). Similar patterns were seen when comparisons were further stratified into age groups < 2 and ≥ 2 to < 3 years. It thus appeared that lower reliability was attributable to nappy pad samples rather than to child’s age.

There was little evidence that passing urine more often or day/bed wetting when previously dry were associated with UTI positivity (Table 25). Associations of pain or crying when passing urine, and dipstick nitrite and leucocyte positivity, were markedly stronger in clean-catch than nappy pad samples and with research laboratory than with NHS laboratory positivity. The association with temperature ≥ 39 °C appeared stronger in clean-catch than nappy pad samples, though CIs were wide. Associations with change in urine appearance and history of prior UTI did not differ markedly between NHS and research laboratory results.

TABLE 25

TABLE 25

Associations of signs, symptoms and urine dipstick tests with separate NHS and research laboratory results, and ‘agree UTI positive’

For both clean-catch and nappy pad samples, values of the AUROC were consistently lower for NHS than research laboratory positivity (see Table 25 and Figure 9). For clean-catch samples (eight-variable model) the AUROC was 0.75 (95% CI 0.70 to 0.81) for NHS laboratory positivity and 0.87 (0.80 to 0.93) for research laboratory positivity. Values of the AUROC were similar in the six-variable models. Values of the AUROC were lower in nappy pad than clean-catch samples: 0.65 (95% CI 0.61 to 0.70) for NHS laboratory positivity and 0.79 (95% CI 0.70 to 0.88) for research laboratory positivity.

FIGURE 9. Receiver operating characteristic curves of symptoms, signs and urine dipstick tests with different definitions of urine culture positivity.

FIGURE 9

Receiver operating characteristic curves of symptoms, signs and urine dipstick tests with different definitions of urine culture positivity. CC, clean catch; NP, nappy pad.

Of the 211 samples that were coded as positive in one laboratory only, 38 (10 from clean-catch samples and 28 from nappy pad samples) were coded as ‘agree UTI positive (step 2)’ when the NHS and research laboratory results were considered together. Thus, a total of 102 (2.1%) samples were considered as ‘agree UTI’ on the basis of both the NHS and research laboratory results. Strong associations of pain/crying passing urine and nitrite and leucocyte dipstick positivity with an overall ‘agree UTI’ microbiology result in clean-catch samples were lower or absent in nappy-pad samples (see Table 25). For the six-variable model, the AUROC was higher in clean-catch (0.88; 95% CI 0.82 to 0.94) than nappy pad samples (0.74; 95% CI 0.67 to 0.81) (see Table 25 and Figure 9).

Values of the AUROC were much lower for samples classified as ‘disagree (NHS positive, research negative)’ (six-variable model 0.62; 95% CI 0.54 to 0.69 for clean-catch and 0.62, 95% CI 0.57 to 0.68 for nappy pad samples). Only a small number (11 clean catch, 7 nappy pad) of samples were classified as ‘disagree (NHS negative, research positive)’ and values of the AUROC were also lower for these than for samples classified as ‘agree UTI’. Values of the AUROC in sensitivity analyses conducted using the six-variable model are shown in Table 26.

TABLE 26

TABLE 26

Areas under the ROC curve (95% CI) from sensitivity analyses using six-variable logistic regression models

Results for clean-catch samples using the eight-variable model and six-variable model were similar, and are available from the authors on request. For clean-catch samples, the values of the AUROC were similar for children aged < 3 and 3 to < 5 years, for both NHS laboratory and research laboratory positivity. Allowing for ‘not known’ responses for children aged < 3 years made little difference to the AUROC, for either NHS or research laboratory positivity. The AUROC was markedly higher for samples that were positive in both NHS and research laboratories than in those coded as ‘agree UTI’ after considering both results, for both clean-catch and nappy pad samples. For the research laboratory, but not for NHS laboratories, values of the AUROC were higher for samples collected in surgery than those collected at home. Values of the AUROC were similar in samples received by both laboratories within 24 hours and samples received after 24 hours except for nappy pad samples in the research laboratory, where the diagnostic accuracy appeared higher for samples received within 24 hours. For both NHS and research laboratory positivity, the AUROC increased with increasing threshold of pure/predominant growth count. For research laboratory positivity, values of the AUROC were markedly lower for pure/predominant growth < 105 CFU/ml. Values of the AUROC were markedly higher in samples with WBC count ≥ 30/mm3, except for research laboratory positivity in nappy pad samples. There was little evidence that values of the AUROC were higher for research laboratory positivity with pure than with predominant growth. There was little difference in AUROCs when stratifying by research laboratory urine volume (Monovette weight) in clean-catch and nappy pad samples.

Discussion

Summary of findings

Based on a large, unselected cohort of children presenting to primary care in England and Wales with acute illness, the reliability of microbiological diagnosis of UTI in routine NHS laboratories and a research laboratory was lower than expected and worse for urine samples collected using nappy pads than for clean-catch samples. Associations of microbiological positivity with pre-specified symptoms, signs and urine dipstick test results were lower for NHS laboratories than the research laboratory, and for nappy pad samples than clean-catch samples. Urines giving a ‘positive’ result in a NHS laboratory but not in the research laboratory had only modest associations with the preselected symptoms, signs and dipstick test results. These findings did not appear to be attributable to the younger age of the children providing nappy pad samples. Discrimination improved with increasing bacteriuria concentration and with the presence of WBCs in the urine sample (pyuria). Results of urine microbiology should, therefore, not be regarded as dichotomous result, but rather as a continuous phenomenon to be interpreted in the clinical context, with UTI possible even when bacterial concentrations are between 103 and 104 CFU/ml, and becoming increasingly probable with higher concentrations of pure or predominant bacterial growth in the presence of pyuria.

Results in context with other studies

To our knowledge, this is the largest and most generalisable primary care-based study comparing the diagnostic performance of NHS laboratories with a research laboratory, and using nappy pad and clean-catch collection methods. However, the number of UTI positive samples was relatively small, especially for clean-catch samples in younger children and for the research laboratory. We do not know which samples were sent to NHS laboratories in containers with boric acid: all samples were sent via the routine mechanisms for that laboratory. All research laboratory samples were sent in containers with boric acid by Royal Mail post, introducing longer delays (which were not associated with UTI status) before processing than with NHS laboratories. In current UK primary care, nappy pad and bag samples are often the only feasible methods for obtaining urine samples from young children: there is usually insufficient space or staffing for parents and children to wait to provide clean-catch specimens. Moreover, most primary care clinicians (other than those with specialist paediatric or emergency department training) are not trained in SPA or catheter techniques, and neither of these are acceptable to parents. Nappy pads have been shown to be acceptable to parents63 and endorsed by NICE.2

In the early development of microbiology, laboratory diagnosis of UTIs required isolation of the same organism from repeated urine samples. It was recognised in the 1940s that high urine bacterial counts were related to UTI, and subsequently proposed that a single sample with a high count could support a laboratory diagnosis of UTI and allow early treatment.117 Early proposed thresholds to define positivity were derived from detailed investigation of fresh urine samples and ranged from ≥ 103 CFU/ml118 to ≥ 3 × 103 CFU/ml119 and ≥ 105 CFU/ml.120 Recently, it was suggested that a threshold of ≥ 106 CFU/ml would be more appropriate.80 Current laboratory guidelines differ with respect to recommended thresholds. The UK Standards for Microbiological Investigations do not have specific paediatric guidance and suggest a threshold of a ‘single organism ≥ 104 CFU/ml indicating UTI’, but other thresholds are also discussed.82 European paediatric guidelines suggest a threshold of ≥ 104 CFU/ml if symptoms are present and ≥ 105 CFU/ml if no symptoms are present for midstream specimens, and lower thresholds for specimens collected by bladder catheterisation or SPA.83 US guidance suggests that clinicians require both urinalysis evidence of infection and a threshold of ≥ 5 × 104 CFU/ml.121

Microbiological examination of urine requires quantification of bacteria and the ability to differentiate mixed from pure growths. The pour-plate method proved too labour-intensive given the large numbers of urine samples submitted to routine microbiology laboratories in the UK:82 in 2012, 663,355 samples (12,689 from children aged < 5 years) were submitted in Wales alone, equating to some 12.1 million samples annually (250,000 from children aged < 5 years) in England and Wales. More rapid methods using calibrated loops, filter paper strips or multipoint methods to deliver a standard inoculum were developed in response to the need for rapid throughput.122124 All were validated against viable counts performed by pour plates or the method of Miles and Misra.125 The Standards for Microbiological Investigation followed by most UK laboratories provide options for urine culture using these methods to inoculate CLED or chromogenic agar, but they have not previously been calibrated against clinical symptoms. Difficulties in defining mixed growths and achieving accurate bacterial counts may be due to the small volumes of urine inoculated onto small areas of agar.80,122 Spiral plating, which was used by the research laboratory in this study and involves a much larger inoculum (50 µl) over an entire 9-cm agar plate, is a more accurate method of quantifying bacterial counts and allows easy differentiation of mixed cultures.126 Enhanced diagnostic performance might also be achieved through improvements in procedures for sample collection and transport.

NHS laboratory UTI positivity was consistently less strongly associated with urinary symptoms and dipstick results than research laboratory UTI positivity. A substantial number of UTI positive NHS laboratory samples [128 (5.8%) nappy pad and 59 (2.3%) clean catch] were negative in the research laboratory, and associations of the NHS laboratory positive results with symptoms, signs and dipstick results were modest. These findings suggest that the diagnostic performance of current routine NHS laboratory testing is suboptimal and may lead to overtreatment and unnecessary investigations.

Clinical and microbiological implications

Even for samples processed in the research laboratory, the diagnostic utility of microbiology based on nappy pad samples was less than for clean-catch samples. The prevalence of UTI positivity diagnosed in the research laboratory was lower for nappy pad (1.3%) than for clean-catch (2.3%) samples, suggesting that UTIs are missed in nappy pad samples because of contamination. Therefore, primary care clinicians should try to obtain clean-catch samples in even very young children in whom they suspect a UTI,127 for example by providing time and space to support urine collection. If an algorithm based on parent-reported symptoms can provide earlier ID of the children at greatest risk of UTI, then parents could be advised to obtain a urine sample prior to attending primary care.

In adult medicine, results from urine microbiology can be interpreted in the clinical context of the patient’s presentation. However, in young children the significant difficulties in obtaining uncontaminated samples, together with the non-specific nature of the presenting symptoms, mean that there is greater reliance on the laboratory result. More detailed routine microbiological examination of paediatric urine samples would have resource implications that could better be justified if urines were selected for testing through an algorithm that increased the prior probability of positivity. Our results suggest that NHS laboratories should distinguish primary care paediatric (age < 5 years) samples from adult samples and consider reporting these in more detail, and that national procedures should be correspondingly updated.

Adoption of the more accurate but labour-intensive research laboratory methods would not be appropriate to use for all urines processed in NHS laboratories but, in many laboratories, extra or augmented methods are used for ‘special urines’, usually received in small numbers, to enhance reporting. It seems reasonable that research laboratory methods be implemented in NHS laboratories for the processing of paediatric urines if associated with enhanced diagnostic performance.

Implications for the DUTY study

For the purposes of the DUTY study, samples will be considered UTI positive (called ‘the reference standard’ in Chapter 5) if there is growth in the research laboratory of ≥ 105 CFU/ml of a single uropathogen (‘pure growth’) or growth of ≥ 105 CFU/ml of a uropathogen with ≥ 3 log10 difference between the growth of this and the next species (‘predominant growth’).

Image 08-66-01-fig4
Copyright © Queen’s Printer and Controller of HMSO 2016. This work was produced by Hay et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK373515

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (18M)

Other titles in this collection

Recent Activity

  • Microbiological diagnosis of urinary tract infection by NHS and research laborat...
    Microbiological diagnosis of urinary tract infection by NHS and research laboratories - The Diagnosis of Urinary Tract infection in Young children (DUTY): a diagnostic prospective observational study to derive and validate a clinical algorithm for the diagnosis of urinary tract infection in children presenting to primary care with an acute illness

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...