Article Text
PDF
Abstract
Background The National Institute for Health and Care Excellence (NICE) neonatal jaundice guidance recommends a urine culture for investigation of babies with prolonged jaundice. However, the evidence cited for this guidance is limited. We aimed to review local data and the existing literature to identify evidence to either support or refute this guidance.
Method We retrospectively reviewed 3 years of urine cultures from our outpatient prolonged jaundice clinic. We then conducted literature review with meta-analysis of studies presenting original data on urine tract infection (UTI) rates in jaundiced and prolonged jaundiced babies.
Results From our local data, none of the 279 patients met our unit clinical criteria for UTI. Literature review revealed considerable differences worldwide in UTI rates in both jaundiced and prolonged jaundiced cases. Using pooled data from our literature review and our local population, the incidence of UTI in prolonged jaundiced babies is 0.21% (95% CI 0.0% to 0.73%) in the UK. This is significantly lower than the figure indicated from the data from elsewhere in the world, 8.21% (95% CI 4.36% to 13.0%).
Conclusions The findings both from our local data and the current literature do not support the practice of routine screening for urine infection in well babies with prolonged jaundice. In view of the above, we no longer include urine culture in screening of well infants with prolonged jaundice. We hope that NICE will re-examine the evidence and recommend changes to their guidance on the role of routine screening for urine infection in babies with prolonged jaundice.
- Evidence Based Medicine
- General Paediatrics
- Neonatology
- Screening
- Jaundice
Statistics from Altmetric.com
What is already known on this topic
Urine infection is quoted as a cause of prolonged jaundice in infants.
National Institute for Health and Care Excellence guidelines currently recommend a urine culture be taken when screening a baby with prolonged jaundice.
What this study adds
The reported urine tract infection rates in prolonged jaundice vary worldwide.
The rate of positive urine culture in infants with prolonged jaundice is consistently low in the UK (0.21% (0.0% to 0.73%)) and may represent coincidentally identified asymptomatic bactiuria.
The need for screening for urine infection should be reconsidered when assessing well infants with prolonged jaundice in the UK.
Introduction
Although urinary tract infection (UTI) is often quoted in textbooks1–7 as a cause of prolonged jaundice (PJ) in infants, there is little physiological explanation for either a cause or effect relationship. The current National Institute for Health and Care Excellence (NICE) guideline8 recommends that the investigation of PJ (jaundice persisting beyond 2 weeks old) should include a urine culture to assess for UTI. This recommendation is based on three published papers. Hannam et al9 reported two cases of UTI (out of 137 screened infants) of which one, the authors admit, did not meet local UTI criteria. Unal et al10 is a case series of 26 infants, none of whom had UTIs. Finally, Tiker et al11 is a case series of early onset conjugated hyperbilirubinaemia and, although 15/42 infants had culture proven sepsis, none were described as having UTI.
Because these studies were small and varied in quality, we decided to review our own local data and evaluate the current evidence in the literature to identify evidence to support this guidance for urine cultures as part of the PJ screen.
Local review of practice
Methodology
We retrospectively reviewed infants who had urine samples sent for microscopy and culture over the 3-year period from 1 June 2010. All urine samples sent from the neonatal outpatient clinic during the study period were reviewed and filtered to identify the phrases ‘jaundice’ and ‘prolonged jaundice’ in the clinical details for infants over 2 weeks of age at time of screening.
Urine samples in our outpatient clinic are collected by sterile bag sample in the first instance with the bag applied before venipuncture. In our experience, the infant has usually passed urine by the end of the venipuncture process and, if not, then the family are recalled after the next patient. Samples are repeated at the discretion of the clinician leading each clinic by clean catch method.
Our unit guideline for defining a UTI is a single pure growth of an organism with >100 000 colony-forming units (cfu)/mL (as per the NICE guideline for UTI12). Dipstick tests and microscopy for white cells are not used for decision-making as neither have been shown to perform particularly well at excluding UTI in children under 3 years old.12
Mixed growths are excluded and lesser pure growths are discounted or repeated at the discretion of the clinician in charge of the clinic. Treatment should only be started if a second sample (clean catch) confirms pure growth above the >100 000 cfu/mL threshold with the same organism. All infants treated with antibiotics undergo renal imaging to exclude renal tract anomaly and assess the need for prophylaxis.
Results from local data
During the study period, our unit delivered 23 243 infants (11 941 males, 11 302 females). In total, 423 infants had urine samples sent for culture during the study period, of which 279 (179 males, 100 females) were from outpatients sent from the PJ clinic. This equates to 1.2% of all deliveries (1.5% of males, 0.9% of females). The average age at PJ screening was 27.3 days (14–61 days).
We identified 145 negative screens, 114 patients with mixed growth and 20 samples (11 males, 9 females) were reported as a pure growth (at varying cfu/mL numbers; see figure 1). The organisms seen were coliforms (12), Escherichia coli (4), Klebsiella (2), Citrobacter (1) and coagulase-negative staphylococci (1). There were only five significant (>100 000 cfu/mL) pure growths on first sample, which on repeat were all lesser or mixed growths.
Patient breakdown from local data. UTI, urine tract infection.
However, as the prescribing of antibiotics is at the discretion of the clinician leading the clinic, two infants did receive oral antibiotics. The first was a 38/40 gestation male with an uneventful neonatal course. He grew E. coli (10 000–100 000/mL) from a bag urine and was treated with trimethoprim before the negative culture was available from his clean catch (taken prior to antibiotics). Unfortunately, there is insufficient documentation in the case notes for why this decision was made. The second was a 37+1/40 gestation female with an initial growth of >100 000/mL E. coli. She had been admitted twice to the neonatal unit for jaundice before discharge during which time she had an insignificant growth of E. coli (10 000–100 000/mL) in urine. Her repeat sample grew E. coli at 100 000 cfu/mL, but given her neonatal history she received a 7-day course of oral trimethoprim. Both infants had normal imaging findings.
Although 2 infants received antibiotic courses, of the 279 infants who attended outpatients screening, none of our cohort had a confirmed UTI based on the NICE or our local unit guidelines.
Literature review
Methodology
Given the low UTI rate identified in our population, we wished to establish rates in existing data to assess for any notable difference. We aimed to identify published papers and abstracts presenting data on urine infections in patients with PJ to allow comparison data.
On local advice, the literature search was conducted of the Medline and Embase databases using the terms for articles relating to jaundice (jaundice, icter*, hyperbili*) in infants combined with AND function to look for causes or investigations of this (UTI, Urin*, microbio*, culture, cause*, reason, aeitiol*, investig*). The full search strategy (see online supplementary appendix A1) yielded 59 articles, the analysis of which is in figure 2. Papers were only included in our analysis if they included original statistics.
Identification of papers.
Supplementary appendix
Data analysis
The identified papers were reviewed in full text form by SS and IA. The key points were summarised into a spread sheet to facilitate easy comparison in table 1. CIs were generated (at 95% significance) using the proportional test from the R statistics package, which tests the null that the proportions (probabilities of success) in several groups are the same, or that they equal certain given values and are displayed graphically in figure 3. The meta-analysis was conducted using the R metafor package. The CIs generated by the proportional test were pooled using the random effects model, which used the maximum-likelihood estimator to pool the subgroups into non-UK prolonged; non-UK not prolonged; UK prolonged and UK not prolonged. The random effects model is more suited to the levels of study heterogeneity experienced. The typical measure of heterogeneity is Cochran's Q, which in this analysis has a value of Q(df=18)=185.3052, p <0.0001, the I2 value (total heterogeneity/total variability) is 89.43%, given these values we can assume there is substantial heterogeneity.
Papers for literature review
Forest plot of meta-analysis data.
Results
We found 15 original articles, 1 letter (including statistics) and 2 conference abstracts that provided data on UTIs in populations of jaundiced infants (table 1). We have included our data on this table for comparison. Seven studies were from hospitals in the UK, a further four from Europe, and the remainder from other parts of the world including the USA, Taiwan and the Middle East. One conference abstract discussing infants with exclusively PJ did not specify a country of origin.14
Six papers exclusively reviewed infants with PJ, a further three stated patients were drawn from their PJ clinic, but did not give age ranges (presumed PJ for our analysis), and the remainder covered mixed populations, infants under 2 weeks, or did not clearly specify the age of their study population.
Eight papers9 ,13 ,16 ,19 ,23–26 selected patients from a general outpatient population attending for outpatient PJ investigation. A further four18 ,20–22 described that their patients were ‘admitted for assessment’, but as there were no further exclusion/inclusion criteria, this may simply reflect local assessment practice. Other study populations were taken from patients with ‘marked jaundice’, conjugated jaundice, cohorts previously admitted to the neonatal unit for jaundice or emergency department attendances. All of the UK studies are drawn from an appropriate outpatient general population.
Nine papers clearly defined their urine collection methods and how they defined a baby as having a UTI (table 1). The diagnostic threshold for UTI varies between papers based on urine collection method and local criteria. All of the UK studies that specify collection methods are using either clean catch method, a sterile pad or a urine bag. We have considered patients as having UTI if the authors of the paper include them in their statistics. However, it should be noted that in some papers, such as Hannam et al,9 patients have been stated as having UTI because they received treatment, but discussion in the paper acknowledges that patients did not meet their local UTI criteria. Only one paper25 described in any detail the use of dipstick test, and although a few others say cell counts were done, none present or discuss the results of this.
Not all patients in every study had urine samples taken to assess for UTI. As we are trying to ascertain incidence values within the PJ population, we have used the number of patients tested for UTI as the denominator when assessing rates.
The reported UTI rates in jaundiced infants varied greatly from paper to paper (0–21%). Because of this reported variation and that our interest lies in the UK PJ population, we have categorised our studies into four subgroups: UK prolonged; UK non-prolonged; non-UK prolonged and non-UK non-prolonged.
Overall UTI rates—UK vs non-UK
Looking by individual papers, the UTI rates in the non-UK studies are on the whole higher than the UK groups. One large study from Lebanon21 had a rate of 21.1% (32/152 infants), although notably their threshold for diagnosing UTI was >10 000 cfu/mL. As some studies do not quote their diagnostic threshold, it is difficult to comment on if their rates may be higher due to lower thresholds or other factors within the local population. All of the studies that state a lower diagnostic threshold than the one used by NICE and ourselves are papers that fall into the non-prolonged group and so should not affect the subgroup analysis for PJ rates.
PJ rates
For studies exclusively reporting infants with PJ, the range of UTI was 0–14.9%. From our forest plot (figure 3), the UK rate for PJ is 0.21% (0.0–0.73%) which is significantly lower than the non-UK PJ rate of 8.21% (4.36–13.0%).
Discussion
The literature review
Although the data presented from the gathered studies are mostly from small, heterogeneous single-centre studies, the UTI rates reported in UK studies looking at infants with exclusively PJ are all consistently low and similar to that found in our local data. For UK studies only including infants over 2 weeks old, there were only 5 reported cases of the 1093 infants screened with a rate of 0.21% (0.0–0.73%) from the meta-analysis. When considering the appropriateness of a screening investigation, the background rate of the condition must be taken into account. Some degree of the variance in reported worldwide UTI rates may be due to differences in sampling methods or diagnostic thresholds.
Urine collection methods
As identified by this study, there are a large number of methods of urine collection available for patients at this age range but none are entirely satisfactory for both parents and clinicians. A suprapubic aspirate is the gold standard for avoiding false positives but is an invasive and painful test, making it inappropriate for an otherwise well outpatient population. Sampling via catheter has similar problems and carries the risk of introducing infection. Urine bag and urine pad samples are the least invasive and most convenient but carry a high rate of contamination from skin flora and because of this are only useful in isolation in the event of a negative sample. A clean catch sample, as recommended by NICE,12 remains the best non-invasive collection method but is time consuming and samples are often missed, ending up on the parent or the floor.
Definition of UTI
There is much discussion about how UTI should be defined and this is detailed in the NICE guidelines for UTI.12 Their conclusion is that “in terms of ruling out a UTI, no test performs particularly well, with only marginal differences between dipstick and microscopy”. For this reason, culture remains the gold standard in diagnosis for UTI in infants and this is what our unit uses.
Asymptomatic bacteriuria or UTI
There is still no agreement in the previously published literature, given the lack of a physiological pathway or mechanism, if indeed UTI is a cause of PJ or simply a coincidental finding. The presence of bacteria in the urine of the screened infants could be one of several possibilities; the organism causing a UTI, which in turn causes the infant to become or remain jaundice; the jaundice in some way makes the infant more susceptible to UTI; or that this is a coexisting asymptomatic bacteriuria (the presence of bacteria without manifestation of clinical symptoms that is recognised to resolve spontaneously and only need treatment if the infant is unwell). We were unable to find any good quality recent studies looking at incidence of UTI or asymptomatic bacteriuria in non-jaundiced infants but the historical reported rates for the latter are between 0.5% and 1.0%.28–31 We feel that this supports the theory that any cases identified through PJ screening are most likely coincidentally identified asymptomatic bacteriuria.
Consequences of not screening
If these patients were not screened by culture, and the bactiuria not identified, then the possible outcomes would be either chronic asymptomatic bactiuria with eventual resolution (remaining unidentified unless the infant is screened for another reason) or the infection would manifest with other recognised symptoms (eg, poor feeding, vomiting, lethargy). As the first several months of life are a time of high parental anxiety and frequent interaction with healthcare professionals, there would be multiple opportunities for parents to discuss and raise concerns regarding other UTI symptoms.
By not screening well infants with PJ, there would also be resource and time savings. For example, the laboratory costs and staff time associated with follow-up appointments for repeating samples. In addition, there would be a reduction in parental anxiety.
Finally, the accepted criteria for a good screening test describe that the natural history of a screened condition should be adequately understood32 and that the cost of screening should be balanced. Given the lack of physiological evidence for a mechanism linking PJ and UTI, coupled with the low rate of infants with positive cultures in the UK, we feel that screening for urine infections in infants with PJ is not a valid or cost-effective screening test and this practice should be discontinued. As such our unit has now changed its practice and do not perform urine culture in otherwise well infants attending our PJ clinic. This has not only saved time in collecting the samples but also reduced lab costs and associated paperwork in checking and recording of negative results.
Conclusion
Our findings do not support the practice of routine screening for urine infection in well babies with PJ. In view of this, we no longer include urine culture in screening of well infants with PJ. We hope that NICE will re-examine the evidence and recommend changes to their guidance on the role of routine screening for urine infection in babies with PJ.
Acknowledgments
Thanks to Ioannis Gallos, clinical lecturer at University of Birmingham, for additional support with statistical analysis.
References
Footnotes
Twitter Follow Sarah Steadman at @DrSLJ
Contributors The project was the idea of VR and IA. Data collected and papers reviewed by SS and IA. Data analysis and statistics by KM.
Competing interests None declared.
Ethics approval Hospital audit board.
Provenance and peer review Not commissioned; internally peer reviewed.