Accuracy of Urinalysis for UTI in Spina Bifida

Abstract

Download video file ^{(21.9MB, mp4)}

DOI: 10.1542/6349321851112

Video Abstract

OBJECTIVES

Urinary tract infections (UTIs) are common, but overdiagnosed, in children with spina bifida. We sought to evaluate the diagnostic test characteristics of urinalysis (UA) findings for symptomatic UTI in children with spina bifida.

METHODS

Retrospective cross-sectional study using data from 2 centers from January 1, 2016, to December 31, 2021. Children with myelomeningocele aged <19 years who had paired UA (and microscopy, when available) and urine culture were included. The primary outcome was symptomatic UTI. We used generalized estimating equations to control for multiple encounters per child and calculated area under the receiver operating characteristics curve, sensitivity, and specificity for positive nitrites, pyuria (≥10 white blood cells/high-powered field), and leukocyte esterase (more than trace) for a symptomatic UTI.

RESULTS

We included 974 encounters from 319 unique children, of which 120 (12.3%) met our criteria for UTI. Pyuria had the highest sensitivity while nitrites were the most specific. Comparatively, nitrites were the least sensitive and pyuria was the least specific. When the cohort was limited to children with symptoms of a UTI, pyuria remained the most sensitive parameter, whereas nitrites remained the least sensitive. Nitrites continued to be the most specific, whereas pyuria was the least specific. Among all encounters, the overall area under the receiver operating characteristics curve for all components of the UA was lower in children who use clean intermittent catheterizations compared with all others.

CONCLUSIONS

Individual UA findings have moderate sensitivity (leukocyte esterase or pyuria) or specificity (nitrites) but overall poor diagnostic accuracy for symptomatic UTIs in children with spina bifida.

What’s Known on This Subject:

There is limited data on the accuracy of the routine urinalysis for the diagnosis of urinary tract infections in this population. As the urinalysis is frequently used to make decisions around empiric antibiotics, understanding its accuracy is important.

What This Study Adds:

We report the accuracy of the routine urinalysis for the outcome of symptomatic urinary tract infection in a large cohort of children with spina bifida.

Urinary tract infections (UTIs) are common in children with spina bifida: Approximately half of all children with spina bifida will experience a UTI by 15 months of age, and 44% will have >5 UTIs by 15 years.¹ However, the accuracy of epidemiologic data around UTIs in children with spina bifida is questionable due of the lack of a standardized, widely-accepted definition of UTIs in spina bifida.^2,3 Two major contributors to the lack of clarity around accurately diagnosing a UTI in children with spina bifida are the frequency of asymptomatic bacteriuria⁴ and the lack of data on the diagnostic accuracy of the routine urinalysis (UA) for UTIs. Although efforts have been made to standardize the definition of UTIs in children with spina bifida, there continues to be significant variability in how clinicians diagnose UTIs in this population.^3,5

One adverse consequence to the variability in UTI diagnosis in spina bifida is overtreatment with antibiotics. Children with spina bifida have increased rates of antibiotic-resistant bacteria cultured from their urine when compared with children without spina bifida,^6,7 indicating a significant need to limit exposure to antibiotics in this population. One way to improve antibiotic stewardship around UTIs in children with spina bifida is to improve the methods of diagnosis of UTI. Because urine cultures take a minimum of 24 hours for results, the UA is the only objective, point-of-care test used to inform prescribing decisions for UTI. However, there are sparse data in the literature describing the predictive accuracy of the routine UA for children with spina bifida. Further, the extant data are limited by small patient numbers and the lack of a gold-standard definition of UTI. Thus, there is a need to better understand both the diagnostic accuracy of the routine UA for UTI in children with spina bifida and the factors that can affect interpretation of these results.

The primary objective of this study was to determine the accuracy of routine UA for the diagnosis of UTI in children with myelomeningocele, the most common and most severe survivable form of spina bifida, using an expert consensus-based definition of UTI as a gold standard. Our secondary objective was to determine how performance characteristics of the routine UA changed in children who use clean intermittent catheterization (CIC) for bladder management as compared with children who do not catheterize. We hypothesize that the diagnostic accuracy of the routine UA will be lower in children with spina bifida compared with children who do not have spina bifida, and lower in children with spina bifida who use CIC as compared with those who void/leak.

Methods

Study Design and Patients

This is a retrospective cohort study of all children with myelomeningocele who were <19 years of age and had a paired UA and urine culture sent between January 1, 2016, and December 31, 2021, at either University of Pittsburgh Medical Center Children’s Hospital of Pittsburgh or Ann & Robert H. Lurie Children’s Hospital of Chicago. We excluded children for the following reasons: Immunodeficiency, taking immunosuppressive medications, pregnancy, genitourinary surgery within the previous 4 weeks, taking systemic antibiotics to treat an infection, and a discharge diagnosis of either Kawasaki disease or multisystem inflammatory syndrome in children. We also excluded samples that were collected by bags. We did not exclude children on prophylactic antibiotics, but included this as a covariate.

Predictors

The predictors of interest in this study were presence of nitrites, leukocyte esterase (LE), and pyuria on the UA. We included nitrites as a binary variable (presence versus absence). We included LE as a binary variable in the model, using greater than trace as a cutoff. We a priori determined a cutoff of ≥10 urinary white blood cells (WBCs) per high-powered field (hpf) as the cutoff for pyuria based on the Urologic Management to Preserve Initial Renal Function (UMPIRE) study, which is a multicenter study designed to assess protocolized management of urological sequelae of myelomeningocele in young children to optimize renal outcomes.⁸ Encounters where the reported result for urinary WBCs was “too numerous to count” were considered positive for pyuria.

Outcome

The outcome of interest in this study was UTI. We modified the definition of UTI proposed by the UMPIRE team for use in this study.⁸ Children had to meet the following criteria: Growth of ≥100 000 colony-forming units/mL of 1 or 2 speciated organisms on urine culture, and at least 2 of the following symptoms: Fever ≥38°C (either documented at the point of care or reported by caregivers); gross hematuria; abdominal, suprapubic, or flank pain/tenderness; new or worsening incontinence; new or worsening urinary urgency; frequency or hesitancy; pain with catheterization or urination; and malodorous/cloudy urine. We did not include the following parameters in our definition (which are part of the initial UMPIRE definition): Requirement for pyuria or LE because these are predictors of interest, thus could not be included in the outcome definition; and specific symptoms for children <1 year of age, which include failure to thrive, dehydration, hypothermia, increasing spasticity, febrile seizures, and fussiness/irritability because the evidence behind the association of these symptoms and UTIs is not robust.⁹

Covariates

We collected data on covariates that were thought to be clinically important and relevant to UTI diagnosis in children with spina bifida. These include age, sex, circumcision status if male, bladder management (CIC versus no CIC [eg, void/leak]), history of bladder augmentation, presence of a catheterizable channel, and current use of prophylactic antibiotics. Because we could not discern between void/leak, these were combined as a comparative group against those who perform CIC.

Patient Identification and Data Collection

Eligible children were identified from patient lists maintained by the spina bifida clinic at each center. Data on all eligible children were collected through manual review of the electronic health record at each site and entered into a Research Electronic Data Capture instrument.¹⁰ No protected health information was recorded; thus, this study was deemed exempt at both sites by the local institutional review boards.

Statistical Analysis

We used χ² tests to compare categorical variables, a t test to compare normally distributed continuous variables, and Wilcoxon rank test to compare nonnormally distributed continuous variables, between groups. Because each unique patient could have multiple encounters represented in the data set, we used general estimating equations (GEE) to account for nonindependence of the within-patient repeated measures. With the binary outcome of UTI, the models used a logit link function and robust SE estimates to model the association between predictors (eg, nitrites, LE, and pyuria), assuming an exchangeable working correlation structure. We then generated receiver operating characteristics curves on the basis of the predicted probabilities of UTI from each model. For each predictor, the accuracy metrics were calculated from a confusion matrix comparing the actual versus predicted results from the model using the epi.tests R function.

Our primary analysis examined diagnostic accuracy in the entire cohort, whereas secondary analyses examined results in subgroups. In the first subgroup analysis, we restricted the cohort to children who had any symptoms related to UTI. In the second subgroup analysis, we stratified by children who use CIC versus those who void/leak. Finally, although our primary analysis accounts for the possibility of repeated measures within each patient, we also performed a sensitivity analysis that included only the first encounter from each patient. All analyses were done in R Studio (Version 2023.09.0 + 463) with packages pROC, gee, epiR, and tidyverse.

Results

We identified 319 unique patients with 974 total encounters where paired UA and urine cultures were available. Of these 974 encounters, 120 (12.3%) met our criteria for UTI. Of the 974 paired samples, 462 (47.4%) were from patients with symptoms that could be because of a UTI (Fig 1).

FIGURE 1.

Enrollment diagram of identified and excluded encounters.

The encounters in this cohort were from children with a mean ± SD age of 7.2 ± 5.5 years that were almost 50% male (of which 57.1% were circumcised). Most encounters were from children who use CIC for bladder management (67.2%). A small number of encounters were from children with a history of bladder augmentation (7.4%) or on prophylactic antibiotics (16.4%). Children with encounters that met our definition of UTI were older (9.7 ± 5.2 vs 6.8 ± 5.5 years, P < .01), were more likely to be female (51.3% vs 37.5%, P < .01), and were more likely to use CIC (95.0% vs 60.2%, P < .01) than children without UTI (Table 1). A small number of urine samples in this analysis were collected by clean catch (2.8%).

TABLE 1.

Patient Characteristics

	Entire Cohort (n = 974)	No UTI (n = 854)	UTI (n = 120)	P
Mean age (y)	7.2 (5.5)	6.8 (5.5)	9.7 (5.2)	<.01
Age, categorical, n (%)				<.01
<3 mo	33 (3.4)	32 (3.7)	1 (0.8)
3 mo–<2 y	200 (20.5)	189 (22.1)	11 (9.2)
2–<5 y	184 (18.9)	168 (19.7)	16 (13.3)
5–<11 y	265 (27.2)	232 (27.2)	33 (27.5)
11–18 y	292 (30.0)	233 (27.3)	59 (49.2)
Male, n (%)	483 (49.6%)	438 (51.3%)	45 (37.5%)	<.01
Circumcised, n (%)	276 (57.1%)	250 (29.3%)	26 (21.7%)	1.0
Void/leak, n (%)a	296 (30.5%)	291 (34.1%)	5 (4.2%)	<.01
CIC, n (%)a	655 (67.6%)	514 (60.2%)	114 (95.0%)	<.01
History of bladder augmentation	72 (7.4%)	59 (6.9%)	13 (10.8%)	.18
On prophylactic antibioticsb	153 (16.4%)	131 (16.0%)	22 (18.8%)	.57

^a Bladder management information was missing for 5 encounters. Fourteen encounters were from patients who used an indwelling catheter at the time of urinalysis and culture.

^b Only available for 934 encounters (117 in the UTI group and 817 in the no UTI group).

A microscopic urinalysis was sent in 744 (76.4%) encounters. Of these, 576 encounters reported a single value of WBCs (instead of a range). Encounters with a UTI had a median (interquartile range) urine WBC count of 43 (17–121) WBC per hpf compared with a median (interquartile range) of 12 (4–40) WBC per hpf in the no UTI group (P < .01).

In our primary analysis that included the entire cohort (n = 974 encounters), area under the receiver operating characteristics curve (AUC) for predicting UTI ranged from 0.62 to 0.67 for the UA components analyzed (Table 2). Of the 3 parameters, pyuria had the highest sensitivity (0.91, 95% confidence interval [CI] 0.84–0.95), whereas nitrites were the most specific (0.72, 95% CI 0.69–0.75). Comparatively, nitrites were the least sensitive (0.52, 95% CI 0.42–0.61) and pyuria was the least specific (0.36, 95% CI 0.33–0.40) (Table 2).

TABLE 2.

Predictive Accuracy of UA for UTI in Entire Cohort

Full Cohort (n = 974)
	AUC	Sensitivity	Specificity	PPV	NPV
Nitrites	0.62 (0.57–0.67)	0.52 (0.42–0.61)	0.72 (0.69–0.75)	0.21 (0.16–0.26)	0.91 (0.89–0.93)
Pyuria	0.64 (0.60–0.67)	0.91 (0.84–0.95)	0.36 (0.33–0.40)	0.21 (0.17–0.25)	0.95 (0.92–0.98)
LE	0.67 (0.64–0.71)	0.87 (0.79–0.92)	0.47 (0.44–0.51)	0.19 (0.16–0.22)	0.96 (0.94–0.98)

Cutoff for pyuria was ≥10 WBCs per hpf; cutoff for LE greater than trace. NPV, negative predictive value; PPV, positive predictive value.

In our first subgroup analysis of children with symptoms of a UTI (n = 462), overall diagnostic accuracy decreased across each UA component but findings remained mostly unchanged. Pyuria remained the most sensitive (0.91, 95% CI 0.84–0.95), whereas nitrites remained the least sensitive (0.52, 95% CI 0.42–0.61). Nitrites continued to be the most specific (0.66, 95% CI 0.61–0.71), whereas pyuria was the least specific (0.33, 95% CI 0.28–0.39) (Table 3).

TABLE 3.

Predictive Accuracy of UA for UTI for Symptomatic Children

All Kids With Any Symptoms (n = 462)
	AUC	Sensitivity	Specificity	PPV	NPV
Nitrites	0.59 (0.54–0.64)	0.52 (0.42–0.61)	0.66 (0.61–0.71)	0.35 (0.28–0.42)	0.80 (0.74–0.84)
Pyuria	0.62 (0.58–0.66)	0.91 (0.84–0.95)	0.33 (0.28–0.39)	0.34 (0.28–0.39)	0.90 (0.84–0.95)
LE	0.61 (0.57–0.65)	0.87 (0.79–0.92)	0.35 (0.30–0.40)	0.32 (0.27–0.37)	0.88 (0.82–0.93)

Symptomatic refers to children with any symptoms that we included within the definition of UTI: Fever; abdominal, suprapubic, or flank tenderness; new or worsening incontinence; new or worsening urinary urgency; urinary frequency or hesitancy; pain with catheterization or voiding; malodorous or cloudy urine. Cutoff for pyuria was ≥10 WBCs per hpf; cutoff for LE greater than trace. NPV, negative predictive value; PPV, positive predictive value.

In our second subgroup analysis, in which we stratified by bladder management (ie, CIC or void/leak), the overall AUCs for all components of the UA were lower in children who use CIC compared with AUCs from all others. Of the 3 components, LE had the lowest AUC (0.60, 95% CI 0.55–0.64), whereas nitrites had the highest AUC (0.62, 95% CI 0.57–0.67). Nitrites had the lowest sensitivity (0.52, 95% CI 0.42–0.61), whereas pyuria had the highest sensitivity (0.91, 95% CI 0.84–0.96). Pyuria was the least specific (0.31, 95% CI 0.27–0.35), and nitrites were the most specific (0.72, 95% CI 0.69–0.75) (Table 4). Among encounters from all children who void, any LE was the most sensitive (1.0, 95% CI 0.48–1.0), whereas nitrites were the most specific (0.86, 95% CI 0.81–0.90) (Table 4).

TABLE 4.

Predictive Accuracy of UA for UTI in Children by Bladder Management

Children Who Use CIC (n = 655 Paired Urine Results)
	AUC	Sensitivity	Specificity	PPV	NPV
Nitrites	0.62 (0.57–0.67)	0.52 (0.42–0.61)	0.72 (0.69–0.75)	0.21 (0.16–0.26)	0.91 (0.89–0.93)
Pyuria	0.61 (0.57–0.64)	0.91 (0.84–0.96)	0.31 (0.27–0.35)	0.24 (0.20–0.28)	0.93 (0.88–0.97)
LE	0.60 (0.55–0.64)	0.85 (0.77–0.91)	0.34 (0.28–0.39)	0.31 (0.26–0.36)	0.87 (0.80–0.92)
Children who void/leak (n = 296 paired urine results)
	AUC	Sensitivity	Specificity	PPV	NPV
Nitrites	0.83 (0.63–1.0)	0.80 (0.28–0.99)	0.86 (0.81–0.90)	0.09 (0.02–0.21)	1.00 (0.98–1.00)
Pyuria	0.67 (0.48–0.88)	0.80 (0.28–0.99)	0.55 (0.46–0.64)	0.07 (0.02–0.16)	0.99 (0.92–1.00)
LE	0.87 (0.79–0.96)	1.0 (0.48–1.00)	0.74 (0.68–0.79)	0.06 (0.02–0.14)	1.00 (0.98–1.00)

Cutoff for pyuria was ≥10 WBCs per hpf; cutoff for LE greater than none for children who void/leak; cutoff of LE greater than trace in children who use CIC. NPV, negative predictive value; PPV, positive predictive value.

In our sensitivity analysis, in which we included just the first encounter for each patient, results and their magnitudes were similar to the primary analysis: Both LE and pyuria were more sensitive than specific, and nitrites were more specific than sensitive (Supplemental Table 6).

Discussion

Here, we report the diagnostic accuracy of components of the UA for the diagnosis of UTI in children with spina bifida. Further, we report that the diagnostic utility of these markers differs in children who do and do not require CIC for bladder management. Together, these results suggest that the routine UA is sensitive, but not specific, for the diagnosis of UTI in children with spina bifida, and that the specificity of these markers are lower in children who require CIC compared with those who void or leak.

The diagnostic accuracy of the components of the UA differs between children with spina bifida in our study and previously published data on children without spina bifida. Table 5 shows our reported results as compared with the diagnostic accuracy of the same parameters published for the general pediatrics population,¹¹ which demonstrate differences in the performance of both pyuria and LE. Although there is variability in the reported general pediatric ranges (ranges, not the 95% CI, are presented in the table) that correspond to various cutoffs reported in the literature,¹¹ our reported specificities for LE and pyuria in children who require CIC are below the lower end of the range. This supports our conclusion that specificities for these UA parameters are worse in children with spina bifida compared with the general pediatrics population. When the population is limited to children with spina bifida who require CIC, the specificities for both pyuria and LE are even lower.

TABLE 5.

Comparison of Diagnostic Accuracies Between Children With Spina Bifida and General Pediatrics Population

Population	UA Parameter	Sensitivity (Range)	Specificity (Range)
General pediatricsa	Nitrites	0.53 (0.15–0.82)	0.78 (0.64–0.92)
Spina bifida (all comers)		0.52	0.72
Spina bifida (CIC)		0.52	0.72
General pediatricsa	Pyuria	0.73 (0.32–1.00)	0.81 (0.45–0.98)
Spina bifida		0.91	0.36
Spina bifida (CIC)		0.91	0.31
General pediatricsa	LE	0.83 (0.67–0.94)	0.78 (0.64–0.92)
Spina bifida		0.87	0.47
Spina bifida (CIC)		0.85	0.34

^a Data from Roberts.¹¹

One potential reason for the decreased specificity of both LE and pyuria in children with spina bifida is that the urothelium of children with spina bifida is inherently different than that of normal children, with evidence of chronic inflammation.¹² Thus, a higher proportion of children with spina bifida may always have LE or pyuria as a result of chronic urothelial inflammation, even in the absence of infection. A natural history study of asymptomatic children with spina bifida found that 69% of urine samples from children with spina bifida had either positive LE or >5 urinary WBCs.⁴ It is also possible that, despite the introduction of low-friction catheters, the use of CIC may cause chronic irritation to the urethra or trauma to the bladder wall, which may also affect the specificity of LE and pyuria. This theory is supported by our stratified analysis of children with spina bifida who void/leak, in which the specificity of both LE and pyuria are closer than that of the general pediatrics population. A final potential explanation is that pyuria may be associated with chronic bacteriuria seen in these children.

The nitrites in our cohort had a sensitivity and specificity closer to that of the general pediatrics population as compared with both LE and pyuria (Table 5). Nitrites indicate the presence of Gram-negative bacteria, many of which reduce urinary nitrate to nitrite,¹³ but does not suggest that the bacteria is causing an infection. Further, the urobiomes of both children with and without spina bifida can contain Gram-negative organisms.^14,15 Therefore, positive urinary nitrites only indicate the presence of Gram-negative bacteria, explaining the lack of specificity of this UA parameter. Additionally, because the presence of nitrites indicates Gram-negative organisms, sole reliance on the presence of nitrites to diagnose a UTI will also miss infections because of Gram-positive organisms, such as Enterococcus.

The consequences of relying on a poorly specific test, such as LE or pyuria, to make decisions around empirical antibiotic prescriptions for UTI are significant. Many children with spina bifida will have pyuria and/or LE in their urine, including those with a UTI, which is indicated by their sensitivity of 91%. However, a consequence of many children with spina bifida having either pyuria and/or LE in their urine is the poor specificity. Relying on these markers to prescribe empirical antibiotics at the point of care likely has led to significant overuse of antibiotics in children with spina bifida. Antibiotic overuse can lead to increased bacterial resistance, resulting in improperly treated infection and worse outcomes. Indeed, 1 study found that children with spina bifida have a 10-fold higher rate urosepsis compared with normally voiding children.⁷

The UTI definition used in the current study is stringent, but may not be widely implemented in the point-of-care medical settings where children with spina bifida present with suspected UTI. In the UMPIRE cohort, 16% of 299 newborns with myelomeningocele in their first 4 months of life were treated for UTI, but only 4% met the strict UTI definition,¹⁶ meaning the others were potentially overtreated. Because children with spina bifida are already at increased risks of antibiotic-resistant infections,⁷ antibiotic stewardship is critical in this population. There is a great need to identify novel markers of UTI in children with spina bifida with improved specificity. Although some have been reported in the literature (eg, neutrophil gelatinase-associated lipocalin),^17,18 to our knowledge, no novel biomarkers are widely used in clinical practice to accurately diagnose UTIs in children with spina bifida.

There are several limitations to this study, including the lack of a gold-standard diagnosis of UTI in children with spina bifida. Although we used a consensus-based definition generated by experts in the care of children with spina bifida, this definition is not widely implemented in clinical practice. It is possible that this definition of UTI is too strict, and thus we may have incorrectly classified some children with UTI as not having a UTI. Further, because this is a retrospective study with data pulled from the electronic health record to ascertain symptoms, misclassification bias is possible because of incomplete symptom documentation. Additionally, although we were able to account for multiple visits per patient within our analysis, GEE can only accommodate a single clustering factor, thus we were unable to account for potential clustering by center. However, because the parameter estimates derived from GEE are robust to misspecification of the correlation structure,¹⁹ this concern likely has little effect on the results. Although we limited our cohort to children with myelomeningocele, clinical heterogeneity and complexity vary widely, and we may be unable to fully account for potential differences within our patient population. However, our large sample size partially mitigates this concern. Finally, we included all paired UA and urine cultures regardless of the presence of absence of symptoms because some children may be treated for a UTI solely on the basis of presence of a positive urine culture. Because we required symptoms in our definition of UTI, inclusion of asymptomatic children may affect our results. However, to overcome this potential limitation, we included a sensitivity analysis in which we included only symptomatic children.

Conclusions

Here, we present the diagnostic accuracies of LE, pyuria, and nitrites using the largest cohort of children with spina bifida within the literature. We show that, although LE and pyuria are both sensitive for UTI, they are both nonspecific. Comparatively, the presence of nitrites is the least sensitive, but most specific, component of the UA for UTI. Relying on either LE or pyuria for the diagnosis of UTI in children with spina bifida will lead to overtreatment with antibiotics and potentially missing another cause of the child’s symptoms, whereas reliance on nitrites will lead to underdiagnosis of UTI. Until more specific markers of UTIs in children with spina bifida are identified and implemented into clinical practice, clinicians must take care to not be overly reliant on the UA when evaluating a child with spina bifida and potential UTI.

Glossary

AUC

area under the receiver operating characteristics curve

CI

confidence interval

CIC

clean intermittent catheterization

GEE

general estimating equations

hpf

high-powered field

LE

leukocyte esterase

UA

urinalysis

UMPIRE

Urologic Management to Preserve Initial Renal Function

UTI

urinary tract infection

WBC

white blood cell