Introduction
Diagnosis of UTI is difficult to make and may be missed in up to 50% of children presenting to primary care.48,107 Establishing a diagnosis in pre- or early school-aged children is challenging, because a high proportion are preverbal, there is a lack of specific symptoms, and collection of uncontaminated urine samples is difficult, particularly in primary care settings where the provision of time and private space for young children is challenging.64 UK NICE guidelines suggest that samples should be collected using a method suitable for the age of the infant or child.2 ‘Clean-catch’ samples are preferred, but urine collection pads (nappy pads) are suggested if this is not possible. In the USA, SPA is the gold standard and has been used along with urethral catheterisation, but both are invasive and can be painful for infants, with limited success rate.102,120,146,147 US guidelines recognise ultrasonography increases success but at increased cost.
Laboratory diagnosis is based on colony counts after culture, using quantity and type of bacteria present in the urine sample as the main criteria. Culture methods differ between NHS laboratories but all are derivatives of the UK standards for microbiology investigations (B41),82 which describe criteria for significant bacteruria for children as ≥ 105 CFU/ml of a single species from a clean-catch specimen, 104–105 CFU/ml single species or ≥105 CFU/ml of any bacteria from a SPA. These guidelines also describe colony counts of ≤ 104 CFU/ml and ≥105 CFU/ml from bag urines as diagnostically useful for a negative result and as an indicator of a poor-quality sample, respectively.
It is recognised that difficulty in specimen collection and the interpretation of specimens potentially contaminated prior to culture, from skin, faeces and other sources, may contribute to the misdiagnosis of UTI.89 Overdiagnosing UTI can lead to unnecessary investigations and treatment, which entail risks of complications and psychological stress to the child and his or her family. Discriminating between contamination with faecal organisms and potential UTI pathogens in the laboratory is difficult, as the most common faecal organism and the most common pathogen causing UTI is E. coli, and high colony counts may be found in children without UTI.
Urine samples in the DUTY study were collected by either clean catch or nappy pad, with samples split between a local NHS laboratory and, for a more detailed examination, a research laboratory. The aim of this chapter is to compare the prevalence of contamination by urine collection method, based on the research laboratory result, and to identify the independent factors associated with contamination using the local laboratory results.
Methods
Participants, urine collection and microbiological methods
These have all been described in detail in previous chapters and our protocol paper.105 For a full description of the measures taken to minimise contamination of urine specimens, see Chapter 2.
Definition of contamination
Interpretation of laboratory results from microscopy and culture is not straightforward, with definitions, collection methods and criteria for contamination and UTI inextricably linked. The HPA in its 2012 guideline82 does not use differing cut levels for the diagnosis of UTI in children, but UTI criteria are dependent upon all patient information, with different criteria for various patient groups. Generally, UTI is suspected if culture counts of a pure or predominant growth of organism are ≥ 105 CFU/ml. Any counts below this with ≥ 2 organisms present may be indicative of contamination. In children, the guidelines suggest that cultures < 104 CFU/ml from bag urines may be regarded as negative for UTI and that cultures ≥ 105 CFU/ml may be contaminated and should be repeated with a reliable specimen. There is no mention of any use of nappy pads even though this has been established as a recommended collection method by NICE.2 Squamous epithelial cells present in the urine are a useful indicator of the degree of contamination, but no clear cut-off is recommended in the HPA guidelines.
The European Urinalysis Guidelines of 2000136 consider ‘most acute uncomplicated urinary tract infections result from one bacterial species’ and ‘the isolation of more than one organism from a single specimen of urine must be interpreted in the light of (potential) contamination’. The European Association of Urology Guidelines of 2009 regard colony counts of between 103 and 105 CFU/ml as UTI in symptomatic adults and ≥ 105 CFU/ml in asymptomatic adults.83 In children, the UTI cut-off criteria are based on collection method: any growth from SPA, 103 to 104 CFU/ml from catheter urine, ≥ 104 CFU/ml in children with UTI symptoms and ≥ 105 CFU/ml without UTI symptoms. Neither mention of contamination nor any mention of nappy pad as a collection method is made. A variety of different definitions of contamination have been proposed, differences being in the number of different organisms present and the quantity, derived from colony counts (Table 66).
TABLE 66
Comparing other research definitions of contamination (in date order)
For the purposes of this chapter we defined contamination using the research laboratory results, with a criterion of count ≥ 105 and > 2 organisms, equivalent to Jackson et al.’s94 definition of frankly contaminated. We also excluded any urines collected via bags (n = 73) and any urines with an unknown collection method (n = 17), leaving 5017 urines to analyse.
We also considered a number of other definitions of contamination, firstly equivalent to Feasey90 or Rao et al.91 and secondly equivalent to Bekeris et al.96 or to Jackson et al.93 of probable contamination or higher. Although we did not carry out extensive analyses using these definitions, we considered their effect on the numbers of contaminated urines by collection method. We selected these definitions from Table 66 to most closely reflect what we thought a NHS laboratory would report in practice and also which did not include too many borderline growths (lower than 104 CFU/ml) that could be a UTI.
Statistical analysis
All analyses used results from the research laboratory to define contamination. We compared contamination rates by urine collection method. To create a model showing factors associated with contamination, we selected a priori a number of variables. The variables were grouped in order to reduce the potential for spurious associations into those associated with laboratory and clinical findings, with faeces, skin variables and variables from our prediction rule (see Chapter 5). The laboratory and clinical findings included were dipstick tests results, leucocytes, pH, specific gravity, proteins, ketones, blood, nitrites, urine collection method, and the urine being smelly, dark or cloudy. In addition, we examined where the sample was collected, whether the nurse was an in-house or a DUTY nurse, and the time taken for the sample to reach the laboratory. The faecal variables were diarrhoea in the past 24 hours, the child taking laxatives and constipation in the last week. The skin variables were sex, circumcision, wearing nappies, dehydration, number of nappies used in the previous 24 hours, number of showers/baths in a week and nappy rash. The variables from our prediction rule (see Chapter 5) included were cough, global impression of child, pain/crying when passing urine, abdominal tenderness on examination, any acute abnormality on ear examination and history of UTI. We then examined the frequency of variable categories, blind to their associations with contamination, and merged the least frequent categories prior to analyses. Univariable associations with contamination were estimated using logistic regression. P-values were derived from a LR test of adding each single variable. In the case where the variable was ordinal a single heterogeneity p-value was reported and another for the trend. These univariable associations were stratified by urine collection method. We selected any variable for our multivariable model if it had a p-value < 0.01 considering either heterogeneity or trend p-values. These two multivariable models were then displayed using ORs and their 95% CIs. All analyses were completed using a complete case approach.
We carried out a number of comparisons: the number of contaminated urines in clean catch and nappy pads; and for the contaminated samples only, the numbers of urines with organisms (coagulase-negative staphylococci, E. coli, enterococci and staphs) present in clean catch and nappy pads. These comparisons included the risk difference and risk ratio along with their 95% CIs and a p-value derived from a chi-squared test.
Finally, we compared the number of squamous epithelial cells (using 10 or more squamous epithelial cells as a cut-point) in contaminated and uncontaminated samples; and the number of urines with organisms [E. coli, enterococci and other Enterobacteriaceae (e.g. Proteus)] present in contaminated and UTI-positive urines. We looked at both the risk difference and risk ratio with their 95% CIs and p-values derived from a chi-squared test. These analyses were stratified by collection method.
Results
Of the 7163 children in the study as a whole, we received urine from 6390. The local laboratories cultured 6079 samples and the research laboratory 5107 samples, with 4910 urines cultured at both laboratories (Figure 29). Thirty-three urines were not received at the research laboratory. Of the 5107 research laboratory results, 5017 were collected using either clean catch or nappy pads with 73 bag urines, and 17 with missing collection method were removed from the analysis due to the unsuitability of the collection method. This is why the numbers are not the same as in Chapter 5. Similar numbers of urines from each collection method were included: 55% clean catch and 45% nappy pad. Figure 30 shows the flow of participants in this chapter which relates only to urines received and processed at the research laboratory.
FIGURE 29
Effect of clinical suspicion of UTI on the number of days until symptoms improved among children with UTI.
FIGURE 30
Participant flow in contamination chapter (research laboratory).
Microbiological predictors of contamination
Table 67 shows that both clean-catch and nappy pad samples have increasing prevalence of contamination as the definition is relaxed according to the criteria outlined in Table 66 (although note that we restricted this to those definitions that could be reported by a NHS laboratory) and that, for all definitions used, the nappy pad contamination rate consistently exceeds that seen in the clean-catch samples.
TABLE 67
Numbers of contaminated urines by different definition (research laboratory)
Table 68 shows that there was a 1.8% chance that clean-catch urines were contaminated, compared with 12.2% in nappy pad urines. The risk was 10% (95% CI 8.9 to 11.8%) greater and the risk of contamination 6.7 times larger in nappy pad than in clean catch.
TABLE 68
Number of contaminated urines by collection method (clean catch/nappy pad)
The most dominant organisms found in contaminated urines from 10 research laboratories are shown in Tables 69–71. To verify which organisms were associated with contamination, the prevalence of all organisms present in the contaminated urines were compared with those present in urines from patients considered to have a UTI from the same study (see Chapter 4). Tables are divided by collection method of urine to evaluate any difference between clean catch and nappy pad.
TABLE 69
Organisms prevalent in clean-catch contaminated urines compared with UTI-positive urines
TABLE 71
Organisms prevalent in contaminated urines
E. coli was present in 85% out of 86.7% of clean-catch/nappy pad urines from patients diagnosed with a UTI, compared with 76% out of 84.1% in contaminated urines (see Tables 69 and 70). The numbers of urines with E. coli present were similar for UTI-positive and contaminated samples in both clean-catch and nappy pad urines (with p-values of 0.232 and 0.715, respectively). Proteus species was also present at similar levels in clean catch (p-value of 0.518) from patients diagnosed with a UTI and contaminated urines. However, looking at other Enterobacteriaceae and Enterococcus species showed evidence of a difference between UTI-positive and contaminated samples in both clean-catch and nappy pad samples, that is 6.7% of clean-catch and 13.3% of nappy pad urines from patients diagnosed with a UTI contained one of three other Enterobacteriaceae (Klebsiella, Enterobacter and Serratia species), compared with 52% and 41.1% in contaminated urines, respectively. Enterococci were present in 23.3% of clean-catch and 53.3% of nappy pad urines from patients diagnosed with a UTI, compared with 82% and 92.1% in contaminated urines, respectively.
TABLE 70
Organisms prevalent in nappy pad contaminated urines compared with UTI-positive urines
In order to determine if any particular organism was associated with contamination in nappy pad samples, the most dominant organisms present in the contaminated urines were compared according to collection method.
Nappy pad urines are 30.8% more likely to be contaminated with coagulase-negative staphylococci than clean catch (risk difference) or, equivalently, the risk was 1.8 times higher in nappy pad urine. In both nappy pad and clean-catch urines, contamination caused by E. coli was similar, with a p-value of 0.161. However, there is some evidence to suggest that enterococci are more likely to be present in contaminated nappy pad samples, 92.1%, than in contaminated clean-catch samples, 82% (p-value of 0.025) (see Table 71).
Squamous epithelial cells are routinely used in the laboratory as a guide to contamination. The numbers of contaminated and all other urines from each collection method to exhibit ≥ 10 squamous epithelial cells present are shown in Table 72.
TABLE 72
Numbers of urines from nappy pad and clean-catch collection methods with ≥ 10 SECs
In clean-catch urines, the number of urines with ≥ 10 squamous epithelial cells was significantly higher in contaminated (10%) than in non-contaminated urines (2.7%). However, in nappy pad urines the levels of squamous epithelial cells were similar for contaminated (10.5%) and non-contaminated (9%) urines.
Clinical predictors of contamination
Univariable associations with contamination were analysed in groups, laboratory and clinical (faecal and skin) variables, to link potential contamination sources and variables associated with the DUTY algorithm identified in Chapter 5.
Table 73 shows the variables associated with contamination in clean-catch samples. Other variables analysed, but not found to be associated, were all other dipstick results, urine colour, odour and opacity, presence of nurse during collection, urine transport time, laxative use, constipation, presence of diarrhoea, circumcision, dehydration, number of baths/showers per week, nappy rash (important as this was shown to be inversely associated with UTI in the nappy pad samples), presence of cough, global impression of child by clinician, pain/crying on passing urine, abdominal tenderness, ear abnormality and history of UTI.
TABLE 73
Predictors of contamination in clean-catch urines
In clean-catch urines, a positive dipstick test for nitrites, collection of urine at home, sex of child (female) and a higher number of nappies used in the previous 24 hours were predictors of contamination, the last exhibiting a consistent gradient.
In nappy pad urines, a positive dipstick test for leucocytes, higher pH, protein and blood, fewer hours taken in transport and sex (male) were predictors of contamination (Table 74).
TABLE 74
Predictors of contamination in nappy pad urines
Discussion
Summary of main results
Using a definition selected to reflect ‘frank contamination’, this study showed a 10% greater risk of contamination when urine was collected using nappy pads than with a clean-catch sample. Depending on the definition selected, contamination was present in as many as 26% of nappy pad samples, compared with 6% for clean catch. This corroborates the findings in Chapters 4 and 5 in addition to other reports, stating that a clean-catch sample is preferred for diagnosing UTI in young children. The two most common sources of contamination in urine from young children are faeces and skin. E. coli and enterococci are common faecal organisms and as such would be represented highly in contaminated urines; however, E. coli is also the most common cause of UTI. In this study, as expected, E. coli was not exclusively associated with contamination of urine, but E. coli was dominant in urines from patients diagnosed with a UTI. Enterococci were more prevalent in contaminated urines than UTI positive urines. Contamination of the urine with both E. coli and enterococci is almost certainly via faeces. In contrast, coagulase-negative staphylococci, part of normal skin flora, were found to be a dominant contaminant. The data also show no association between presence of these faecal organisms and nappy pad use, which is corroborated in the lack of association by univariate analysis of presence of diarrhoea.
Contamination of urine can happen with all collection methods in the young, but nappy pads have been implicated with increased contamination rates.171 Only coagulase-negative staphylococci showed a significant association with the nappy pad collection method. Coagulase-negative staphylococci are considered skin contaminants and the close association with nappy pad reflects the close contact between nappy pads and skin. The clean-catch samples showed an increasing risk of contamination with use of an increasing number of nappies per day, but there was no relationship with nappy rash. Relaxing the criteria for contamination would increase the number of potentially contaminated samples, but we chose to base our analysis on a conservative definition.
Urinalysis by dipstick test is commonly used in primary care to inform diagnosis of UTI. This study found that in both clean-catch and nappy pad urines the variables showing association with contaminated urines are NOT exclusive to contaminated urines. Thus, they are not reliable indicators of when urines are contaminated. A positive leucocyte test is generally considered to be an indicator of pyuria. Reports have shown that pyuria can be present in feverish children without UTI172 and so this is probably not a good predictor of contamination. The pH of urine is dependent upon the patient’s acid–base status, but bacteria present in the urine can raise pH, as can a recent meal.173 A positive blood dipstick result was associated with contaminated urines, a finding reported previously.174 Similarly, positive protein dipstick results have been shown to be caused by bacteriuria and fever in children.173 A positive nitrite dipstick test is used to detect bacteruria by common pathogens known to cause UTI. In this study, high numbers of squamous epithelial cells were an indicator of contamination only in clean-catch samples. Squamous epithelial cells should be ignored if samples are derived from nappy pads.
The study considered if the time taken for the urine sample to reach the laboratory was a factor in contamination. Probability of contamination was not increased by increased time taken to arrive at the laboratory, but home sampling was a risk for contamination with clean-catch samples, perhaps reflecting less controlled collection technique.
Comparison with existing literature
Contamination rates in collection methods vary, as indeed do the definitions of contamination. Contamination rates have been shown to vary from 0% in clean-catch urine of 23 samples to 48% of bag urines.98 In a retrospective observational cohort study, contamination in clean catch, catheter specimen of urine and bag was 1%, 12% and 26%.98 Definitions of contamination vary from single organism growth < 105 OR ≥ 2 organisms to ≥ 2 organisms present at > 105 CFU/ml of urine.89,91
In a US study, no institutional factors, such as access to refrigeration, were found to associate with either low or high contamination rates.99 Similarly, sex and diarrhoeal symptoms have been shown to have no association with higher contamination rates.100 Perineal cleansing in female adults had no association with contamination rates, while urine contamination rates were higher in midstream urine collected from toilet-trained children when obtained without perineal/genital cleaning.89,94 One of the only factors shown to reduce contamination rates published was changing nappy pads every 30 minutes.91
Clinical implications
To reduce the risk of contamination, urine samples should ideally be clean catch, taken in the GP surgery and not taken from children where there has been more than usual nappy use within 24 hours of presenting at surgery. Further, our data suggest that contamination is more likely in clean-catch urines if the child is female, the sample was taken at home as opposed to the GP surgery and if the child had increased nappy use 24 hours prior to presenting at the surgery. These factors may be useful in addition to microbiological criteria in clinical decisions.
