Impact of Diagnostic Criteria on UTI Prevalence in Young Infants With Jaundice: A Meta-analysis Free

This study was conducted in accordance with the consensus statement on reporting Meta-Analysis of Observational Studies in Epidemiology¹³  and is registered with PROSPERO International Prospective Register of Systematic Reviews (CRD42018088978).¹⁴ 

A medical librarian conducted the literature search, which was completed in February 2020. The searches included Medline and Medline-in-process from 1946 to January 2020 and Ovid Embase from 1976 to January 2020. Search terms were determined by diagnoses and Medical Subject Headings and results were limited to articles in English, included infants’ birth through 23 months of age, and excluded variations of the word “pregnancy” to filter out maternal articles on UTIs. The search used a broader age range than the neonatal period as it was done in conjunction with another meta-analysis conducted on UTI prevalence in children with bronchiolitis. The full search strategy is available in the Supplemental Information. Additional studies were identified through a manual review of the bibliographies of qualifying studies and other relevant systematic reviews.

Studies were considered for inclusion if they revealed a UTI prevalence rate in infants with a primary diagnosis of neonatal jaundice. Studies were excluded if they included patients with active immunodeficiency, oncologic processes, or specifically included patients with known anatomically abnormal urinary tracts, conditions which might increase risk for UTIs. Additionally, review articles, case reports or series, editorials, meta-analyses, and guidelines were excluded. Studies identified from the search results underwent title and abstract review, which were conducted by 2 independent reviewers for inclusion. Any discrepancy regarding inclusion was discussed between reviewers until a consensus was reached. For studies that included UTI prevalence in jaundice but did not contain UA data, the corresponding authors were contacted via E-mail to determine if additional data were available. Studies were considered for subanalysis if they provided both urine culture and UA data.

For full-text articles meeting inclusion criteria, data abstraction occurred independently by 2 study investigators using a standardized data extraction form, piloted by 2 team members. Any disagreements were discussed and resolved. Final data were extracted from 32 jaundice studies (Fig 1).

We defined a positive urine culture result and positive UA result by using the individual study’s definition. If the study did not provide a formal definition, we defined a positive UA result based on the report of any positive leukocyte esterase result, positive nitrite result, or ≥5 white blood cells per high powered field (WBC/HPF). We did not apply 1 standard UA and colony count criteria because most studies did not provide enough information to do so. Because neonatal jaundice is typically expected to peak and resolve by ∼2 weeks of age, we subcategorized studies into those that included only infants ≤14 days of age and those that included infants >14 days of age. Studies that included infants with “prolonged jaundice” but did not specify exact ages were included in the >14-day-old age group.

To assess study quality, we used a critical appraisal tool for prevalence studies with modifications tailored to our study question.¹⁵  Two investigators independently evaluated the quality of all included full-text studies. The team modified the questions to fit the study population while keeping the same 7 thematic topics: a representative sample, appropriate recruitment, appropriate description of study subjects and setting, objective study definitions for UTI and jaundice, appropriate accounting for confounding factors, and, lastly, appropriate statistical analyses. Individual components for each study were rated as 0, 1, or not reported, with 0 = failing the measure and 1 = passing the measure. The maximum score for a given study was 7.

We used a κ statistic to report interobserver agreement on study inclusion and exclusion. To conduct a random-effects meta-analysis, we first stabilized the variances of the raw proportions using a Freeman-Tukey double arc-sine transformation. This approach generates a weighting estimate to reflect the contribution of individual studies and tends to have fewer problems with convergence.¹⁶  For proportions of zero, we used a continuity correction. We then calculated point estimates and 95% confidence intervals (CI) using inverse-variance weighted random effects. We calculated 2 different point estimates of UTI prevalence (1) using the study’s definition for a UTI for all 32 studies, including the 16 studies without UA data and the 16 studies with UA data, and (2) using the criteria of a positive UA result plus positive urine culture results for the 16 studies with UA data. Heterogeneity was assessed by using the I² and τ statistics. We performed all analyses using Stata version 14.2 (Stata Corp, College Station, TX).

A total of 488 studies were retrieved in our librarian search and 7 were eliminated as duplicates. We identified an additional 45 studies through reference list review. Thus, we reviewed 526 unique study abstracts and identified 53 studies for full-text review. Interobserver agreement on study inclusion/exclusion was high (κ 0.90, 95% CI 0.86–0.92). After full-text review, 21 studies were further excluded, with the most frequent reason for exclusion being studies not including UTI prevalence (8 of 21) or case reports (7 of 21) (Fig 1). Of the remaining 32 studies,^{4,8–10,17–44}  16 did not contain UA data.^9,17–31  Although corresponding authors with available contact information were emailed, only 1 author responded and was unable to provide additional UA data. Of the 16 studies that included UA data,^{4,8,10,32–44}  1 study contained aggregate UA data for patients with all infection types; the corresponding author was contacted and provided the UA data for just the UTI patients by e-mail.³⁹ 

Because we were studying prevalence, we sought to use our quality assessment tool to identify study characteristics that may increase the risk of heterogeneity in our study population. Scores within the individual categories are available in Supplemental Tables 3 and 4. Of the 16 studies without UA data, 7 included infants based on clinical jaundice alone with no reported laboratory confirmation of jaundice.^{9,18,22,24,26,30,31}  Of the 9 studies that included bilirubin levels, all included patients based on elevated unconjugated bilirubin levels. Three studies also had patients with concurrent conjugated hyperbilirubinemia.^21,25,31  One study examined infants ≤ 14 days of age,¹⁸  9 studies examined infants >14 days of age,^{9,17,20,21,23,25,28,29,31}  2 studies examined both age groups,^19,26  and 4 studies stated that they examined “neonates.”^22,24,27,30  Only 2 studies contained circumcision status.^22,29  Overall quality scores ranged from 3 to 6.

Of the 16 studies with UA data, 3 studies included infants based on clinical jaundice alone (specified within the studies as prolonged jaundice), with no reported laboratory confirmation of jaundice.^35,38,41  Of the 13 studies that contained bilirubin levels, all included patients based on elevated unconjugated bilirubin levels. Three studies also had patients with concurrent conjugated hyperbilirubinemia,^4,34,36  5 studies examined infants ≤14 days of age,^{33,40,42–44}  3 studies examined infants >14 days of age,^35,38,41  and 8 studies examined both age groups.^{4,8,10,32,34,36,37,39}  Five studies contained circumcision status.^{4,33,37,42,43}  Four studies included infants presenting with fever.^32,36,39,43  Overall quality scores ranged from 3 to 7.

The criteria for diagnosing a UTI and the definition of a positive UA result varied among studies. Definitions of a positive urine culture result ranged from any growth on a suprapubic aspiration to >100 000 colony-forming units (CFU) per mL via a urine pad (Tables 1 and 2). There were 2 studies that explicitly included mixed organisms in the urine culture results.^35,38  The most common definition of a positive UA result was >5 WBC/HPF (7 of 16 of studies^{8,33,34,39,40,42,44} ). Several of the included studies with UA data simply defined a positive UA result as “presence of pyuria”^32,43  or “white cells in the urine.”^35,38  There was no consistency in the definition of a UTI or UA criteria when stratified based on the year of article publication.

The overall reported prevalence from the 32 included studies of UTI in infants up to 8 weeks old with jaundice was 6.2% (95% CI, 3.9–8.9; I² = 93% and τ = 0.04%, Fig 2A). When limiting to the 16 studies without UA data, the reported prevalence was 4.0% (95% CI, 1.7–7.2; I² = 91% and τ = 0.03%, Fig 2B). When limiting to the 16 studies that provided UA data, the overall prevalence (by the studies’ UTI definition) was 8.7% (95% CI, 5.1–13.2; I² = 94% and τ = 0.04%, Fig 2C). With the addition of a positive UA result as a criterion, the prevalence of UTI decreased to 3.6% (95% CI, 2.0–5.8; I² = 88% and τ = 0.01%, Fig 3). Of these 16 studies that provided UA data, sensitivity analyses altered the summary statistics slightly. Removing the 4 studies that included febrile infants with jaundice (and thus symptoms beyond jaundice) produced a similar point estimate for UA-positive UTI results of 3.7% (95% CI, 2.2–5.6; I² = 77% and τ = 0.5%).^32,36,39,43  When limiting to the 5 studies that examined only infants ≤14 days of age, the point estimate of UA-positive UTI was 4.3% (95% CI, 2.7–6.4; I² = 48% and τ = 0.1%).^{33,40,42–44} 

In studies examining the prevalence of UTI in jaundice, the inclusion of a positive UA result as a diagnostic criterion substantially decreased the estimated prevalence of UTI from 6.2% (all studies) and 8.7% (studies with UA data) to 3.6%, a relative decrease of 42% and 59%, respectively. The lower point estimates held consistently across sensitivity analyses, including analyses that excluded febrile infants. Our findings reveal that diagnostic criteria for the definition of UTI can considerably impact perceptions and reports of UTI prevalence. Nonetheless, the 3.6% prevalence of UA-positive UTI is noteworthy and suggests that evaluation for UTI might be considered in some infants with jaundice.

Within the included studies, the definitions of a UTI were highly variable. Since the publication of the American Academy of Pediatrics (AAP) UTI guideline in 2011,⁴⁵  the utility of an abnormal UA for infants <2 months of age as a necessary component for the diagnosis of a UTI has been reevaluated. Two studies have revealed that the sensitivity of the UA for UTIs in young infants is substantially higher than historically reported^46,47  and call into question the presence of a true UTI in the absence of pyuria.⁴⁸  Although we are unable to conclude that all reported UTIs without pyuria were misdiagnosed, our findings suggest that previous estimates of the prevalence are inflated and may include cases of contamination or asymptomatic bacteriuria.

Most clinicians would likely obtain urine testing when the pretest probability for UTI is or exceeds 1% to 3%^45,49,50 ; the estimated prevalence of UA-positive UTI in jaundice that we found is slightly greater than this testing threshold. Routine testing for UTI in young infants presenting solely with unconjugated hyperbilirubinemia is uncommon in the United States and not recommended in the AAP Hyperbilirubinemia Guideline.²  If jaundice is indeed the primary symptom of a UTI in some infants, there will likely be some delayed or potentially undiagnosed UTI cases. Additional investigations are needed to assess the clinical course of these infants who did not initially undergo urine testing to determine the necessity of evaluating for UTI in this population. For example, it is unclear whether fever would follow shortly after the development of jaundice and whether there is any benefit to an earlier UTI diagnosis. Although some studies suggest an increased risk for renal scarring in infants with delayed treatment of UTI, these studies have only evaluated febrile UTIs.⁵¹  It is also possible many of these UTIs would self-resolve, as has been demonstrated in infants presenting to clinic with fever.⁵² 

The UTI prevalence in the 16 studies without UA data (4.0%, 95% CI, 1.7–7.2) was much lower than the culture-based prevalence in the 16 studies with UA data (8.7%, 95% CI, 5.1–13.2) and similar to the UA-adjusted prevalence (3.6%, 95% CI, 2.0–5.8). Differences in the study population and methods may account for some of these findings. Catheterization was the most common urine testing method in the studies with UA data (7 of 16 studies), whereas suprapubic aspiration was the most common method in the studies without UA data (6 of 16 studies, Tables 1 and 2). Obtaining sterile urine via catheterization may be difficult in young infants, especially uncircumcised males. Most of the studies with UA data (10 of 16 studies) contained subject inclusion based on an elevated serum bilirubin level or need for phototherapy, whereas 3 of the 16 studies without UA data contained such inclusion criteria (Tables 1 and 2). As a result, there were 9 studies with UA data and only 3 studies without UA data that reported a mean bilirubin level >15 mg/dL. Whether there might be a pathophysiologic relationship between the degree of hyperbilirubinemia and pyuria is unclear. In 3 studies, there was pyuria found in infants with unconjugated hyperbilirubinemia but negative urine cultures, and the authors reported no statistically significant difference in the proportion with pyuria between jaundiced infants with positive and negative urine culture results.^10,37,43 

This study has several limitations. One major limitation is that the studies’ age group inclusion and definitions of jaundice were heterogeneous, ranging from age <2 weeks to 8 weeks with jaundice definitions ranging from clinical jaundice to phototherapy-threshold hyperbilirubinemia. Clinically, the 0- to 1-week-old, 1- to 2-week-old, and >2-week-old infants are likely different populations in terms of their risk factors for jaundice.⁵³  Unfortunately, because of the relatively wide age range in the studies’ inclusion criteria (eg, <2 weeks, <4 weeks), it was not possible to conduct additional subanalyses based on more narrow age ranges. In addition, not every study specified clear exclusion criteria. Thus, it is unclear for exactly what population of young infants with what degree of hyperbilirubinemia urine testing is warranted.

In addition, because of heterogeneous definitions of a UTI, “positive” UA results, and “positive” urine culture results among the studies, we opted to define a UTI based on the individual study’s definition rather than 1 standard applied to all included studies because there was insufficient information provided in the studies to do so. Given the variability of UA positivity definitions and that many studies included urine culture results with bacterial growth of <50 000 CFU/mL, we are potentially still overestimating the true prevalence of concomitant UTI in young infants with jaundice.

With the incorporation of a positive UA result into the UTI definition, the estimated prevalence of UTIs in infants with jaundice decreased substantially, from 6.2% (all studies) and 8.7% (studies with UA data) to 3.6%, a relative decrease of 42% and 59%, respectively. Our findings illustrate the importance of diagnostic criteria in disease definitions and the need for more high-quality studies to elucidate exactly what subgroup of young infants with hyperbilirubinemia may warrant urine testing.