Pyuria and Urine Concentration for Identifying Urinary Tract Infection in Young Children

In this retrospective cross-sectional study, we examined the urinalysis results of children <24 months of age with suspected UTI who presented to the emergency department (ED) of a quaternary children’s hospital between January 2012 and December 2017 and had paired urinalysis and urine culture. Eligible patients were identified from the electronic medical record (Epic, Verona, WI). For patients with >1 ED visit, only the first visit was included. Patient demographics, method for urine collection, urinalysis, and urine culture results were extracted directly from the hospital’s Epic database. Patients with unknown urine collection source, indwelling catheter, bag urine, missing urinalysis results, urine culture growing mixed or multiple organisms or normal genital flora, or missing colony count were excluded from the analysis.

At our institution, dipstick urinalyses are performed on the Clinitek Atlas automated urine chemistry analyzer (Siemens Medical Solutions USA, Inc, Malvern, PA). Microscopic urinalysis is performed with Iris iQ200 Series automated urine microscopy analyzer (Beckman Coulter, Inc, Brea, CA) on all specimens with positive dipstick results for LE, nitrite, or blood. A combined dipstick and microscopic urinalysis may also be performed, but this is not the typical practice. Specimens with negative dipstick results for LE, nitrite, and blood are not reflexed to microscopic urinalysis. In patients with no microscopic urinalysis, a negative dipstick result was considered equivalent to a negative microscopic urinalysis result.¹⁵  Urine specific gravity (SG) was used as a surrogate marker for urine concentration. Patients were assigned into 3 urine SG groups (low <1.011, moderate 1.011–1.020, high >1.020) for analysis. We calculated the likelihood ratios for 1 to 10 WBCs per HPF cut points to determine the optimal WBCs per HPF for microscopic pyuria in these urine SG groups. The positive likelihood ratio (LR+) of 10 was considered as the ideal cutoff for ruling in disease.^14,16  We also calculated test characteristics for pyuria cut points of ≥5 and ≥10 WBCs per HPF, trace and small LE, and positive nitrite for positive urine culture results across 3 urine SG groups. The area under the curve (AUC) values were obtained from the receiver operating characteristic (ROC) curves to determine the diagnostic accuracy of microscopic pyuria for positive urine culture results across 3 urine SG groups.

The primary outcome measure in our study was positive urine culture result, which was defined in accordance with the AAP clinical practice guideline (CPG) for UTI, as follows.^3,17,18  Transurethral in-and-out catheterization specimens with growth of ≥50 000 colony-forming units (CFU)/mL of a single uropathogen were defined as positive. Standard midstream specimens were positive if ≥100 000 CFU/mL of a single uropathogen grew in culture. For this study, pathogenic urogenital organisms included Escherichia coli, Proteus species, Klebsiella species, Serratia marcescens, Citrobacter species, Enterobacter species, Pseudomonas species, Enterococcus species, Streptococcus agalactiae, and Staphylococcus saprophyticus. Urine cultures growing a single uropathogen at colony counts lower than the culture threshold for positive urine culture result by the AAP CPG or a single organism in significant quantity but not defined as a urogenital pathogen, as above, were defined as negative in this study. Urine cultures with growth of multiple organisms or urogenital flora were interpreted as contaminated specimens and were excluded from the study analysis.

The electronic medical record was used to gather demographic and clinical data that included age, sex, race, ethnicity, peak ED temperature, urine specimen source (transurethral bladder catheterization, suprapubic aspiration, clean catch, or bag collection), urine culture results, SG, WBCs per HPF, LE amount, and nitrite.

Statistical analyses were performed by using SAS Enterprise Guide 7.1 software (SAS Institute, Inc, Cary, NC). Standard descriptive statistics were used to describe patient characteristics. Median and interquartile range (IQR) were used to summarize sample characteristics for continuous data, whereas frequencies and percentages were used for categorical data. Likelihood ratios and their 95% confidence intervals (CIs) were calculated for microscopic pyuria and trace and small LE because we determined this statistical value would be the most useful measure to inform clinical decision-making. Sensitivity, specificity, and odds ratios and their 95% CIs were also calculated for ≥5 and ≥10 WBCs per HPF and trace and small LE cut points.

We identified 30 462 children <24 months of age with paired urinalysis and urine culture. A total of 6291 patients were excluded from the analysis for the following reasons (Fig 1): bagged specimen (n = 388), indwelling catheter (n = 46), unknown or other urine collection methods (n = 20), missing urine SG (n = 27), urine culture with mixed or multiple urogenital organisms (n = 5177), organism without a specified colony count (n = 599), and nonpathogenic organism (n = 34). The remaining 24 171 patients constituted the study group, of whom 2003 (8.3%) had a positive urine culture result. Urine was obtained by transurethral catheterization in 97.9% (n = 23 663). No urine specimen was obtained by suprapubic catheterization. The median age of the study group was 7.3 months (IQR 2.5–12.9 months), and female patients were 58.8% (n = 14 216). Hispanic individuals constituted 54.5% of the overall population, whereas white individuals were 21.1%, African American 17.5%, Asian American 2%, others 3.1%, and unknown 1.8%. Seventy-eight percent of the patients had a urine SG ≤1.020 (Table 1). The most common organism was E coli (88.7%) followed by Klebsiella species (5.6%), Enterobacter species (1.6%), Proteus species (1.5%), Enterococcus species (1.5%), and others (1.1%).

The LR+ decreased from 16.1 (95% CI 14.7–17.5) at low urine SG to 6.1 (95% CI 5.4–6.8) at high urine SG for ≥5 WBCs per HPF cutoff. The cutoff that resulted in LR+ >10 was 3 WBCs per HPF at low (LR+ 10.5; negative likelihood ratio [LR−] 0.12), 6 WBCs per HPF (LR+ 12; LR− 0.14) at moderate, and 8 WBCs per HPF (LR+ 11.1; LR− 0.35) at high urine concentrations (Table 2).

The sensitivity, specificity, and likelihood ratios of various urinalysis results in identifying a positive urine culture result was also compared between SG groups, including pyuria, trace or small LE positivity, nitrite positivity, and combined LE and nitrite positivity (Table 3). Notably, a significant decrease in test sensitivity for detecting positive urine culture results was seen at high urine SG (>1.020) for both ≥5 and ≥10 WBCs per HPF pyuria cut points and trace and small LE compared to ≤1.020 SG. Specificity decreased with increasing urine SG for both ≥5 and ≥10 WBCs cut points but increased for both trace and small LE cutoffs with increasing urine SG. The LR+ for small LE increased from 25.2 (95% CI 22.7–28.0) in the low SG group to 37.6 (95% CI 29.4–48.2) in the high SG group as shown in Table 3.

The odds ratio of positive urine culture results for ≥5 WBCs per HPF cut point was 92 (95% CI 75.0–111.2; P < .001) in the low urine SG group, which decreased to 19.9 (95% CI 14.2–27.9; P < .001) in the high urine SG group. Higher odds ratios for positive urine culture results were also seen for pyuria cutoff of ≥10 WBCs per HPF and trace and small LE at low and moderate urine SG, which were decreased at high urine SG (Table 4). The ROC curves revealed statistically significant excellent diagnostic accuracy for microscopic pyuria at low and moderate urine SG (AUC of 0.95 and 0.94, respectively). At high urine SG, a statistically significant AUC value of 0.88 signified a good diagnostic accuracy (Fig 2). The cutoffs identified for microscopic pyuria are highlighted on the ROC curve for the SG groups and reflect good sensitivity and false-negative values for positive urine culture results.

In this data set of 24 171 children with paired urinalysis and urine culture, we identified a correlation between urine concentration and the accuracy of different components of urinalysis in predicting positive urine culture results. We found that a cutoff of 3 WBCs per HPF at low (SG <1.011), 6 WBCs per HPF at moderate (SG 1.011–1.20), and 8 WBCs per HPF at high (SG >1.020) urine concentrations yields optimal likelihood ratios for presumptive diagnosis of UTI in young children. At low urine concentration, there are higher odds for having positive urine culture results with a microscopic pyuria cut point of 5 WBCs per HPF and any positive LE. This study expands on the findings by Chaudhari et al⁷  that the diagnostic test characteristics for dipstick and microscopic urinalysis for UTI vary by urine concentration and agrees with some of the similar findings reported in a recent study by Shaikh et al.¹⁹  Interestingly, with increasing urine concentration, LR+ for positive urine culture results decreased for microscopic pyuria, whereas it increased for positive LE. LR− increased similarly for both microscopic pyuria and positive LE with increasing urine concentration.

Our study is unique because we have studied the test characteristics of urinalysis elements for presumptive UTI at 3 different urine concentrations in a much larger population of young children than previous studies.^7,19  We noted a higher LR+ for positive urine culture results for positive trace and small LE than previously reported.^7,19  Unlike the Shaikh et al¹⁹  article, we noted increasing LR+ for positive LE with increasing urine concentration as previously reported by Chaudhari et al.⁷  Trace LE had higher or comparable sensitivity, specificity, and values for LR+ than ≥5 WBCs per HPF cutoff across all urine SG groups. Values for LR− were similar for positive trace LE and ≥5 WBCs per HPF cutoff.

The prevalence of a positive urine culture result (8.3%) was slightly higher in our population compared to previously reported UTI prevalence of 7% to 8% in young children. However, a higher prevalence of 14% has been reported by authors of an urban study²⁰  using a lower colony count of 10 000 for catheterized specimens, 1000 for suprapubic specimens, and 100 000 for bag specimens. We used strict criteria for positive urine culture results as described in the 2016 reaffirmation of AAP CPG culture threshold. Additionally, we excluded patients with mixed or normal urogenital flora even with a colony count <10 000, but some of the previous studies may have included them and classified them as negative, thus probably increasing prevalence in our study sample. Moreover, higher prevalence of positive urine culture results in our population could be due to Hispanic patients who constituted 55% of the overall population and had a higher prevalence of positive urine culture results (10.1%). Higher UTI prevalence in Hispanic patients has been described before²⁰  and could be due to certain physiologic differences, such as blood group antigens, which confer varying degrees of risk for UTI,²¹  and presumed lower circumcision status in Hispanic male individuals.²²  Male uncircumcision has been associated with increased risk of UTI.^1,23,24 

Interestingly, this study, like in the studies by Chaudhari et al⁷  and Shaikh et al,¹⁹  we noted decreased (3.5%) prevalence of positive urine culture results at higher urine concentration (SG >1.020). There were 2.9-fold higher odds (95% CI 2.5–3.4; P < 1.0001) of having a positive urine culture result in the less concentrated urine (SG ≤1.020) compared to concentrated urine (SG >1.020). More effect on the test characteristics was seen at lower urine concentration (≤1.020), which could be because of the high prevalence of positive urine culture results in the low (8.8%) and moderate (10.7%) urine SG groups.

Trace LE on automated dipstick outperformed the microscopic pyuria cut point of 5 WBCs per HPF in our population of young children <24 months of age. These findings have been previously noted in infants aged 1 to 90 days.^10,14 

The reasons for decreased test sensitivity with concentrated urine across microscopic pyuria and LE cutoffs are unclear. A possible explanation could be low prevalence of positive urine culture results (3.5%) in this group. We presume that urinalysis has suboptimal sensitivity at high urine concentration. We do see that test sensitivity decreases as expected with increasing WBC cutoff points from ≥5 to ≥10 WBCs per HPF. Specificity decreases with increasing urine concentration and increases with an increase in WBC cutoffs from ≥5 to ≥10 WBCs per HPF. Overall, our study suggests that both microscopic pyuria and positive LE are not sensitive tests for predicting positive urine culture results at high urine concentration.

We had several limitations to this study. First, we did not review patients’ charts to correlate with clinical history, including urogenital abnormalities, intermittent urethral and/or bladder catheterization, renal disease, history of recurrent UTI, previous antibiotic exposure, duration of fever, or circumcision status. However, we investigated the performance of urinalysis results for detecting likely UTI strictly by culture results alone; therefore, additional clinical data would not impact final diagnosis according to these criteria.

Second, there is a possibility of selection bias because we included all patients who had a paired urinalysis and urine culture. It is possible that some patients might have received urinalysis for reasons other than clinical symptoms of UTI and were unintentionally paired with urine culture by the provider. On the other hand, it is also possible that we might have missed patients with UTI who were diagnosed on the basis of urinalysis, or UTI was overlooked for febrile viral illness or fever of unknown source.

Third, we defined positive urine culture result thresholds as described in the 2016 reaffirmation of the AAP UTI CPG,¹⁶  but we did not include pyuria or bacteriuria, which is also required to make a diagnosis of UTI. We used strict culture criteria to determine the predictive value of positive LE and microscopic pyuria in alignment with previous studies.^7,24  With this methodology, patients with asymptomatic bacteriuria or culture contamination may have been included. However, the prevalence of asymptomatic bacteriuria is reported to be low (∼1%).^8,25  Moreover, in our study, we had in-and-out transurethral catheterization in 97.9% of the patients and excluded patients with mixed or multiple organisms, urogenital flora, or nonpathogenic organisms to decrease the chances of culture contamination. Because colony counts <50 000 CFU/mL are currently being considered for the diagnosis of UTI,²⁶  there is also a possibility of missing early UTI by using the strict urine culture threshold per the AAP CPG, in particular at lower urine concentration.

Fourth, urine SG is not an ideal surrogate for urine concentration. Urine osmolality is a better test to assess the degree of urine concentration, which was not possible because of the retrospective nature of the study. Urine SG is affected by urine glucose, protein, ketones, bilirubin, urobilinogen, and hemoglobin.²⁷  In general, there is a good correlation between urine osmolality and urine SG unless the patient has an underlying condition such as uncontrolled diabetes mellitus or nephrotic syndrome.²⁸  Patients with fever sometimes have transient proteinuria²⁹  and that can increase the urine SG; however, we see higher odds of having positive urine culture results in the lower urine SG groups.

We identified that the accuracy of the microscopic pyuria for presumptive diagnosis of UTI varies with urine concentration. The likelihood of having positive urine culture results for microscopic pyuria decreased with increasing urine concentration. Our data suggest using a cutoff of 3 WBCs per HPF at low (SG <1.011), 6 WBCs per HPF at moderate (SG 1.011–1.20), and 8 WBCs per HPF at high (SG >1.020) urine concentrations for presumptive diagnosis of UTI in young children. We noted that positive leukocytes esterase is a strong indicator of a positive urine culture result regardless of urine concentration.

AAP

American Academy of Pediatrics

AUC

area under the curve

CFU

colony-forming unit

CI

confidence interval

CPG

clinical practice guideline

ED

emergency department

HPF

high-power field

IQR

interquartile range

LE

leukocyte esterase

LR−

negative likelihood ratio

LR+

positive likelihood ratio

ROC

receiver operating characteristic

SG

specific gravity

UTI

urinary tract infection

WBC

white blood cell