April 1999, VOLUME103 /ISSUE 4

Technical Report: Urinary Tract Infections in Febrile Infants and Young Children

  1. Stephen M. Downs, MD, MS


Overview. The Urinary Tract Subcommittee of the American Academy of Pediatrics Committee on Quality Improvement has analyzed alternative strategies for the diagnosis and management of urinary tract infection (UTI) in children. The target population is limited to children between 2 months and 2 years of age who are examined because of fever without an obvious cause. Diagnosis and management of UTI in this group are especially challenging for these three reasons: 1) the manifestation of UTI tends to be nonspecific, and cases may be missed easily; 2) clean voided midstream urine specimens rarely can be obtained, leaving only urine collection methods that are invasive (transurethral catheterization or bladder tap) or result in nonspecific test results (bag urine); and 3) a substantial number of infants with UTI also may have structural or functional abnormalities of the urinary tract that put them at risk for ongoing renal damage, hypertension, and end-stage renal disease (ESRD).

Methods. To examine alternative management strategies for UTI in infants, a conceptual model of the steps in diagnosis and management of UTI was developed. The model was expanded into a decision tree. Probabilities for branch points in the decision tree were obtained by review of the literature on childhood UTI. Data were extracted on standardized forms. Cost data were obtained by literature review and from hospital billing data. The data were collated into evidence tables. Analysis of the decision tree was used to produce risk tables and incremental cost-effectiveness ratios for alternative strategies.

Results. Based on the results of this analysis and, when necessary, consensus opinion, the Committee developed recommendations for the management of UTI in this population. This document provides the evidence the Subcommittee used in the development of its recommendations.

Conclusions. The Subcommittee agreed that the objective of the practice parameter would be to minimize the risk of chronic renal damage within reasonable economic constraints. Steps involved in achieving these objectives are: 1) identifying UTI; 2) short-term treatment of UTI; and 3) evaluation for urinary tract abnormalities.

  • urinary tract infection
  • infant
  • child
  • decision analysis
  • cost-effectiveness analysis
  • vesicoureteral reflux
  • Abbreviations:
    UTI =
    urinary tract infection
    VUR =
    vesicoureteral reflux
    IVP =
    intravenous pyelography
    VCUG =
    voiding cystourethrography
    RCG =
    radionuclide cystourethrography
    ESRD =
    end-stage renal disease
    IUTI =
    initial UTI
    CFU =
    colony-forming unit
    UA =
    OR =
    odds ratio
    LE =
    leukocyte esterase
    UNC =
    University of North Carolina
    WBC =
    white blood cells
    HPF =
    high-power field
    SMX =
    TMP =

    Analysis of the data on UTI consisted of several steps. The Subcommittee met to define the target population, setting, and providers for whom the practice parameter is intended. Subcommittee members identified the outcomes of interest for the analysis. A conceptual evidence model of the diagnosis and management of UTI was developed. The evidence model was used to generate a decision tree. A comprehensive review of the literature determined the probability estimates used in the tree. The tree was used to conduct risk analyses and cost-effectiveness analyses of alternative strategies for the diagnosis and management of UTI. Based on the results of these analyses and consensus when necessary, an algorithm representing the strategies with acceptable risk–benefit trade-offs was developed.


    The overall problem of managing UTI in children between 2 months and 2 years of age was conceptualized as an evidence model (Fig 1). The model depicts the relationships between the steps in the diagnosis and management of UTI. The steps are divided into the following four phases: 1) recognizing the child at risk for UTI, 2) making the diagnosis of UTI, 3) short-term treatment of UTI, and 4) evaluation of the child with UTI for possible urinary tract abnormality.

    Fig. 1.

    Evidence model for UTI guideline. The evidence model traces steps in the evaluation and management of childhood UTI from clinical suspicion to urinary tract imaging.

    Phase 1 represents the recognition of the child at risk for UTI. Age and other clinical features define a prevalence or a prioriprobability of UTI, determining whether the diagnosis should be pursued. If children at sufficiently high risk for UTI are not identified for diagnostic evaluation, the potential benefit of treatment will be lost. However, children with a sufficiently low likelihood of UTI should be saved the cost of diagnosis (and perhaps misdiagnosis) of UTI when the potential for benefit is minimal.

    Phase 2 depicts the diagnosis of UTI. Alternative diagnostic strategies may be characterized by their cost, sensitivity, and specificity. The result of testing is the division of patients into groups according to a relatively higher or lower probability of having a UTI. The probability of UTI in each of these groups depends not only on the sensitivity and specificity of the test, but also on the prior probability of the UTI among the children being tested. In this way, the usefulness of a diagnostic test depends on the prior probability of UTI established in phase 1. Overdiagnosis of UTI may result in unnecessary treatment and unnecessary imaging evaluation for urinary tract abnormalities. Underdiagnosis will result in missing the opportunity to treat the acute infection and the consequences of possible underlying urinary tract abnormalities.

    Phase 3 represents the short-term treatment of UTI. Alternatives for treatment of UTI may be compared, based on their likelihood of clearing the initial UTI (IUTI).

    Phase 4 depicts the imaging evaluation of infants with the diagnosis of UTI to identify those with urinary tract abnormalities such as vesicoureteral reflux (VUR). Children with VUR are believed to be at risk for ongoing renal damage with subsequent infections, resulting in hypertension and renal failure.1–3 Prophylactic antibiotic therapy or surgical procedures such as ureteral reimplantation may prevent progressive renal damage. Therefore, identifying urinary abnormalities may offer the benefit of preventing hypertension and renal failure. Each alternative strategy for imaging evaluation can be characterized according to its cost, invasiveness, and test characteristics (sensitivity and specificity). The potential yield of an imaging evaluation will be affected by the accuracy of the initial diagnosis of UTI. If the probability of UTI is low because a nonspecific test was used to make the diagnosis, the cost of an imaging study will yield little benefit. Therefore, the value of imaging strategies depends on the accuracy of the diagnosis of UTI in phase 2.

    Because the consequences of detection and early management of UTI are affected by subsequent evaluation and long-term management and, likewise, long-term management of patients with UTI depends on how they are detected at the outset, the Subcommittee elected to analyze the entire process from detection of UTI to the evaluation for, and consequences of, urinary tract abnormalities.

    Decision Tree

    The conceptual model was used to develop a decision tree (Fig 2). The tree quantifies the relationship between alternative strategies for the diagnosis and treatment of UTI, the evaluation of children for abnormalities of the urinary tract, and the anticipated consequences of these alternative diagnostic and treatment strategies. The tree is read from left to right. The square branch points (decision nodes) represent alternatives under the control of the caregiver. The round chance nodes represent the chance events that may be affected by the alternatives chosen. A computerized version of the decision tree was constructed and used for the analyses presented.

    Fig. 2.

    The UTI decision tree models the decision-making processes in the diagnosis and management of UTI. This model was used for cost-effective analyses.

    In the diagnosis and treatment of UTI, four alternatives are represented. As anchor points, the alternatives of treating all or treating none of the patients at risk are represented. Two alternative tests are represented in the other branches. The test characteristics and costs of these tests were adjusted to correspond with the tests the Subcommittee chose to evaluate. In this way, the Subcommittee compared alternative testing strategies.

    In the decision tree, if a testing strategy is used, the result is positive or negative. A positive result was assumed to lead to a decision to treat and a negative result to a decision to observe without treatment. If treatment is chosen, presumptively or based on the results of a diagnostic test, a treatment complication may result. Rarely, as in the case of anaphylaxis, the complication may result in death.

    For all other patients, there is a risk of urosepsis.4This risk is the probability of UTI multiplied by the prevalence of urosepsis among infants with UTI. Assuming urosepsis behaves like bacteremia from other sources in infants,5 the infection may clear spontaneously in those who have urosepsis. The probability of clearing the infection is increased among those who receive antimicrobials. If urosepsis does not clear, hospitalization will result, and the child has a risk of dying.

    Once the short-term outcome of the UTI has been resolved, the second decision is whether and how to image the urinary tract for structural and functional abnormalities. The three following options are modeled: 1) a full evaluation, including ultrasonography or an intravenous pyelogram and voiding cystourethrography (VCUG) or radionuclide cystourethrography (RCG); 2) ultrasonography alone; and 3) no evaluation.

    The results of the evaluation are determined separately for each patient type. For patients who have no UTI (false-positive diagnosis of UTI), it is assumed that abnormalities are not present and the infant is not at increased risk of renal damage. Patients who have a true UTI may or may not have VUR. Among those with VUR, the reflux may be low grade (1 or 2) or high grade (3, 4, or 5).

    The probability of having a positive result when imaging an abnormal urinary tract depends on the sensitivity of the imaging modality for the abnormality. For example, ultrasonography is much more sensitive to high-grade VUR than to low-grade VUR.6 The analysis assumes that all imaging modalities have 100% specificity (ie, no false-positive diagnoses).

    The result of the imaging evaluation would be used to select a treatment (surgical correction or antibiotic prophylaxis as appropriate) to prevent recurrent infections. An evaluation with normal results or no evaluation would lead to no therapy. Identifying the optimal therapy for a given urinary tract abnormality is beyond the scope of the present analysis.

    Patients may or may not have recurrent UTI, defined as more than three infections in a 5-year period. Recurrent infections lead to progressive renal scarring. The risk of scarring is highest among those with high-grade VUR and lowest among those with no VUR.7Therapeutic interventions reduce the risk of renal scarring by reducing VUR in the case of surgery or preventing infection in the case of antimicrobial prophylaxis.

    Patients with progressive renal scarring are at increased risk of hypertension and ESRD. Those who do not experience these outcomes may have decreased renal function but will not have clinically important outcomes.

    At the terminal nodes of each branch, the outcomes are tabulated as costs and clinical outcomes. Costs considered include the cost of diagnostic testing, treatment, complications of treatment, hospitalization for urosepsis, imaging studies, surgery or prophylaxis, management of hypertension, and ESRD. Clinical outcomes include chronic hypertension, renal failure, and death.

    The decision tree was encoded and evaluated using the Decision Maker software Version 6.0 (Sonnenberg and Pauker, New England Medical Center, Boston, MA).

    Literature Review

    Articles for review were obtained from four sources in two rounds of searching. In the first round, the MEDLINE database was searched using four separate search strategies corresponding with the four phases of the diagnosis and treatment of UTI: recognition, diagnosis, short-term treatment, and imaging evaluation (Appendix 1). The titles and abstracts resulting from these searches were distributed among the Subcommittee members who identified those that were definitely or potentially useful. These articles were reproduced in full.

    In a second round of searching, articles were identified from three additional sources: the bibliographies of two recent reviews8 ,9; a survey of the members of the Subcommittee, soliciting the articles they identified as most important and relevant to the analysis; and articles sought specifically to estimate costs for the management of chronic hypertension and ESRD. At each of the two rounds of searching, the articles were reviewed by the epidemiology consultant, and articles with no original data were removed.

    The remaining articles were reviewed and data extracted using a data extraction form designed to identify estimates necessary to evaluate the decision model (Appendix 2). In addition, the quality of each article was rated on a scale from 0 to 1 by using quality criteria adapted from Sackett and colleagues.10 Reviewers also provided a subjective rating of “good,” “fair,” or “poor” to each article. Data extracted were recorded in evidence tables, using an Excel (Microsoft Corporation, Redmond, WA) spreadsheet. A subset of 24 articles was reviewed twice by different reviewers to check interrater reliability. At the time of analysis of the decision models, the articles were reviewed again by the epidemiology consultant.


    Literature Review

    The results of the literature searches are shown in Fig 3. In the initial MEDLINE search, 1949 articles were identified. The title and abstract of each of these was reviewed by two members of the Subcommittee and identified as “useful,” “potentially useful,” or “not useful.” After this review and elimination of duplicates, 430 articles were reproduced in full for data extraction. Of these, 105 were rejected because they contained no original data relevant to the analysis. In the second round, an additional 133 articles were identified. Twenty-six of these were rejected for lack of original data. A total of 432 articles were reviewed at least once.

    Fig. 3.

    The literature review process. The diagram illustrates the number of articles examined and reviewed or rejected from the analysis.

    The quality of the articles in this area is highly variable (Fig 4), but most articles met 50% or fewer of the quality criteria. Interrater reliability for quality scores was tested using a subset of the articles (Fig 5). Correlation of scores among reviewers was only fair (r = 0.43). Correlation among the readers' subjective ratings was less good (r = 0.29).

    Fig. 4.

    Quality histogram plots the distribution of quality scores among the articles reviewed. Quality scores were assigned by reviewers based on predetermined criteria.

    Fig. 5.

    Quality score agreement. A sample of articles was reviewed twice by different reviewers. The graph plots the scores assigned by one reviewer versus the score assigned by the other reviewer. The correlation coefficient r represents only fair agreement.

    Evidence Tables

    Probability of UTI

    The prevalence of UTI among febrile infants between 2 and 24 months of age is ∼5% (Table 1). The references in Table 1 are ordered by decreasing relevance and decreasing quality scores. The references are divided additionally into two categories. The studies of infants having fever without an obvious cause are presented in the top of the table, and studies of symptomatic infants who do not strictly meet the criterion of fever without an obvious cause are presented in the bottom. The pooled prevalence represents the prevalence that is derived by pooling the data from the first 11 articles in the table. The studies are very consistent, and the pooled prevalence is insensitive to the addition or deletion of any one of them.

    Table 1.

    Prevalence of UTI

    The first three articles are cross-sectional prevalence studies of UTI in febrile infants younger than 12 months11 or children younger than 24 months.12 ,13 The data in the fourth reference14 are a subset of the study population that was younger than 6 months. North15 concluded that there was not a significant prevalence of UTI among febrile children. However, the prevalence derived from the small subset of patients younger than 2 years who had ≥10 000 colony-forming units (CFUs) is consistent with the findings from other studies. The studies by Baker and Arner,16 Kramer and associates,17 and Buys and colleagues18 have some potential for work-up selection bias because the decision to obtain a urine specimen for culture was made by the care provider. Therefore, these studies may slightly underestimate the prevalence of UTI because specimens were not obtained and cultured from all eligible patients. However, these biases would be small. The study by Pylkkanen and associates19 is the only outlier. The prevalence figure is based on a small subset of “symptomatic” infants who were reported as having fever only. This study is of lower quality than the others but cannot be excluded on any other basis. However, its small sample has little effect on the pooled prevalence.

    Included in Table 1 are results from three other studies that do not examine febrile infants per se. The first20 reports a prevalence for 187 children younger than 5 years with febrile seizures. Interestingly, the prevalence is quite similar to that of children with fever alone. Siegel and colleagues21 studied children with a variety of symptoms, including fever. The prevalence again is close to 5%. The study by Pryles and Luders22 gives the prevalence of UTI among infants with diarrhea.

    In an effort to identify subgroups of febrile infants at sufficiently low risk of UTI that they could be excluded from additional evaluation, the Subcommittee studied clinical risk factors that reduce the probability of UTI (Table 2). Factors identified were patient age, gender, circumcision, and the presence of other identifiable sources of fever.

    Table 2.

    Risk Factors for UTI

    Age and Gender

    The data indicate that the probability of finding UTI in febrile male infants is less than half that in females, and just greater than one third for males between 1 and 2 years of age (Table 2). Findings from two studies, by Roberts and associates 13 and by Hoberman and colleagues,23 report the prevalence of UTI by gender in an unbiased sample of febrile infants younger than 1 year. The studies show inconsistent results among males younger than 1 year. Roberts and co-workers13 found no males with UTI, whereas Hoberman and associates 23 found a prevalence of 2.5%. Among females, the prevalence was similar, 7.4% and 8.8%, respectively. Other studies also suggest a lower risk among males in older age groups.24 ,25

    Other studies used to estimate the effect of age and gender on the prevalence of UTI examined only children with confirmed UTI. Relative risk of UTI for a given gender was estimated from these studies using the odds ratio (OR),P(male‖UTI)/P(female‖UTI). Because this prevalence in males and females in the general population is ∼50%, the OR is essentially the same as the ratio of the prevalence of UTI among males to the prevalence among females. The prevalence itself can be derived by assuming an overall prevalence of 5% for both genders and a 50% prevalence of males in the population, using the formula,P(UTI‖male) =P(male‖UTI)·P(UTI)/P(male). The comparable formula was used for females.

    Studies also were stratified by age (Table 2). These data and the data for prevalence by gender were used to make crude estimates of the effect of age and gender on the prevalence of UTI in four subgroups: males younger than 1 year (3%), males older than 1 year (2%), females younger than 1 year (7%), and females older than 1 year (8%).


    Data from Wiswell and co-workers,26–28 Ginsberg and McCracken,4 and Craig29 show a dramatic risk reduction among circumcised males. Most (but not all) of these data are retrospective, and the studies are plagued with missing data, but the findings are quite consistent. In all these studies, the probability of circumcision given UTI [P(circ‖UTI)] was reported. Relative risk was estimated by the OR,P(circ‖UTI)/P(no circ‖UTI). Estimates of prevalence of UTI given circumcision [P(UTI‖circ)] were calculated by assuming the previous probability of UTI regardless of circumcision status [P(UTI)] is 5%, and the probability of circumcision (among males) is 70%.27 Prevalences are calculated using these estimates and Bayes' formula,P(UTI‖circ) =P(circ‖UTI)·P(UTI)/P(circ). The results suggest that the prevalence of UTI among febrile male infants who are circumcised would be ∼0.2%.

    Most data come from studies of infants younger than 1 year, but the effect seems to persist even beyond infancy,30 and Craig29 did not find that age younger or older than 1 year confounded the effect of circumcision. The bottom of Table 2 shows how the relative risk affects the probability of UTI assuming a prevalence of 5%. Making the (reasonable) assumption that circumcision status is independent of age, circumcised males older than 1 year are at lowest risk.

    Interestingly, circumcision could account for the gender differences in prevalence. If one assumes that febrile females and uncircumcised males have a UTI prevalence of 7%, that circumcised males have a prevalence of 0.2%, and that 70% of males are circumcised, then the prevalence of UTI among males would be (0.3·0.07) + (0.7·0.002) = 0.022 or 2%, a figure consistent with the data.

    Tests for UTI

    When an infant is considered to be at significant risk for UTI, the next step is to make the diagnosis. Choosing diagnostic criteria for UTI involves two competing considerations. A false-negative diagnosis will leave patients with UTI at risk for serious complications. A false-positive diagnosis may lead to unnecessary, invasive, and expensive testing. In evaluating alternative diagnostic tests for UTI, the Subcommittee defined as a “gold standard” any bacterial growth on a culture of urine obtained by suprapubic bladder aspiration (tap), ie, any bacterial growth on a culture of a urine specimen obtained by tap defines a UTI.

    For the analysis, a culture of a urine specimen obtained by a tap was considered to have 100% sensitivity and specificity. However, this was not always used for comparisons in studies of other diagnostic strategies. Alternative diagnostic strategies fall into the three following areas: 1) urine analysis (UA) for immediate diagnostic information, 2) culture of a urine specimen obtained by urine bag or transurethral catheterization, and 3) culture of a urine specimen using the dipslide culture technique.

    The sensitivities and specificities of the diagnostic tests reviewed are shown in two formats. Table 3 lists the test characteristics. In part A, data are listed by study; in B, they are listed by type of test. C presents a summary of the test characteristics of each of the tests. Figure 6 plots the sensitivities and specificities of the tests on a graph with axes corresponding to a receiver operating characteristic curve. The true-positive rate (sensitivity) is on the vertical axis, and the false-positive rate (1, specificity) is on the horizontal axis. Tests that plot nearest to the top left corner of the graph have the greatest diagnostic value. The diagonal represents test characteristics with no diagnostic value, and tests that plot below the diagonal may be misleading. The tests are identified by type (leukocyte esterase [LE], nitrite, blood, protein, microscopy for bacteria or white blood cells, combinations of tests, or special tests).

    Fig. 6.

    Reported sensitivity and specificity of diagnostic tests for UTI plotted on receiver operating characteristic curve coordinates. Also shown is the threshold line for a two-way sensitivity analysis on sensitivity and specificity. The plot shows a wide variation in reported test characteristics. Only those above and to the left of the threshold line would have sufficient sensitivity and specificity to supplant urine culture as a “gold standard” diagnostic test.


    The various components of the UA include the reagent slide tests, ie, LE, nitrite, blood, and protein, and microscopic examination for leukocytes or bacteria. The diagnostic characteristics of these tests have been evaluated individually and in combination (Table 3). Tests can be combined serially, meaning all test results must be positive for the combination to be positive or, in parallel, meaning a positive result on any one of the tests defines a positive result for the combination. The serial strategy maximizes specificity at the expense of sensitivity, whereas parallel testing maximizes sensitivity at the expense of specificity.

    Perhaps the most commonly used component of the UA used to evaluate a child who may have a UTI is the reagent strip (dipstick). Tests for UTI that are available on most reagent strips include the LE, nitrite, blood, and protein. The test characteristics of each of these are given in Table 3. The LE is the most sensitive single test. Its reported sensitivity ranges from 67% in a screening setting31 in which symptoms and, therefore, inflammation, would not be expected, to 94% in settings in which UTI is suspected.32 The specificity of LE generally is not as good. However, because the specificity describes the performance of the test on specimens from patients without a UTI, it is highly dependent on patient characteristics. The reported specificity varies from 63% to 92%.

    The test for nitrite has a much higher specificity (90% to 100%) and lower sensitivity (16% to 82%). For this reason, nitrite may be useful for “ruling in” UTI when it is positive, but it has little value in ruling out UTI. Dipstick tests for blood and protein have poor sensitivity and specificity with respect to UTI. Therefore, the use of results for blood or protein from dipstick testing has a high likelihood of being misleading.

    Carefully performed microscopic examination of the urine has high sensitivity and specificity in many studies (Table 3; Fig 6). However, the wide range of reported test characteristics of microscopy for leukocytes or bacteria (Table 3, C) presumably reflects the difficulty of performing these tests well and the hazards of performing them poorly. In studies that have shown the best test characteristics, tests were performed by on-site laboratory technicians who often used counting chambers.

    When examining the urine for bacteria, unstained33and Gram-stained34 specimens seem to be effective. However, centrifugation of the specimen reduces the specificity of the test.35 The number of bacteria also is important. Using heavy bacterial counts as a diagnostic criterion results in low sensitivity and high specificity.12 The reverse applies when observation of any bacteria is considered a positive test.34

    Microscopy for leukocytes is variably sensitive (32% to 100%) and specific (45% to 97%). Studies with accurate results generally used mirrored counting chambers, on-site technicians, or both.33 ,36 Specimens also must be examined shortly after collection. A 3-hour delay results in a 35% drop in sensitivity.37 Finally, if the number of leukocytes considered abnormal is high, the test will be insensitive,19 if the number is low, it will be highly sensitive.38 The reverse is true for specificity.

    Test characteristics resulting from combinations of UA components are shown separately (Table 3, B; Fig 6). Parallel combinations of tests maximize sensitivity. This is supported by the data presented in Table 3. Studies using careful technique in on-site laboratories found that a parallel combination of microscopy for leukocytes and bacteria had a sensitivity of 99% or greater,33 ,36 ,39 ,40 and UA, considering any component positive as a positive UA result, has ∼100% sensitivity and 60% specificity.36 ,39

    A number of special tests have been evaluated for the diagnosis of UTI. The first is a modified nitrite test.41 This involves incubating a urine specimen for several hours with added nitrite before testing for nitrate. The reported sensitivity and specificity are 93% and 88%, respectively. However, comparable results have not been reported. Immunochemical studies for early detection of bacterial growth42 have not had impressive results.

    Obtaining a Urine Specimen

    The Subcommittee defined the gold standard definition of a UTI to be growth on a culture of a urine specimen obtained by tap. Often, however, performance of a bladder tap is resisted because it is invasive. Moreover, bladder taps may not yield urine specimens. Success rates for obtaining urine specimens vary between 23% and 90%.43–45 Although 100% success has been achieved when using ultrasonographic guidance,18 practitioners often use transurethral catheterization or urine bags to obtain urine specimens for culture.

    Cultures of urine specimens obtained by catheterization have a specificity of 83% to 89% compared with cultures of urine specimens obtained by tap.8 ,11 ,45 However, if only cultures yielding >1000 CFU/mL are considered positive, catheterization cultures have a 95% sensitivity with a specificity of 99%.

    Cultures of bag urine specimens are 100% sensitive, but they have a specificity between 14% and 84%.44 46–48 Therefore, because UTI is present in a small minority (5%) of patients tested, use of culture results from urine specimens obtained from the bag to rule in UTI is likely to result in large numbers of false-positive results. Specifically, with a prevalence of 5% and specificity of 70%, the positive predictive value of a positive culture of bag urine specimens would be 15%. That is, 85% of positive cultures of urine specimens obtained from a bag would be false-positive results.

    Culturing Techniques

    The standard culturing technique involves streaking on blood agar and MacConkey media. More recently, dipslide methods have been developed. The few studies of dipslide cultures that were reviewed reported sensitivities between 87% and 100% and specificities between 92% and 98%.47 ,49 ,50

    Consequences of a Missed Diagnosis of UTI

    If the diagnosis of UTI is not made because it was not suspected or a test of insufficient sensitivity was used, three consequences may result. The first (see “Prevalence of Urinary Tract Abnormalities”) is the lost opportunity to find a urinary tract abnormality that could result in renal damage. The second is the formation of new renal scars. Although repeated UTI lead to scarring (see “Progressive Renal Damage”) and progressive scarring is associated with hypertension and ESRD, the role of scarring from a single UTI in long-term clinical consequences is unknown. Nevertheless, timely treatment of febrile UTI in children appears to be important in preventing scars.51

    The third consequence of missing the diagnosis of UTI results from the urosepsis that occurs in a small proportion of febrile infants with UTI.4 ,52 Among patients with febrile UTI, in the age group to which this analysis applies, the risk of concurrent bacteremia is between 2.2%52 and 9%.4 The natural history of bacteremia in infants with febrile UTI is not described. Common pediatric experience is that septic shock and death are rarely seen in this situation, suggesting that spontaneous resolution of bacteremia occurs in this situation as it does in others.53 ,54However, evidence exists that among infants with bacteremia attributable to Escherichia coli in the presence of a UTI, fatality rates may be as high as 10% to 12%.5

    Short-term Treatment of UTI

    In studying the short-term treatment of UTI, the two following issues were addressed: 1) What do the data suggest about the duration of outpatient antibiotic therapy? and 2) What is the best choice for presumptive oral antibiotic therapy for the infant with suspected UTI before culture results are available? From the data presented subsequently in this report, it may be reasonable to conclude the following. 1) Single-dose to 3-day therapy is not as effective as therapy of 7 days or longer (perhaps because of more rapid metabolism of antimicrobials in children), but that the minimal acceptable duration of therapy has not been demonstrated. 2) Cotrimoxazole appears to be superior to amoxicillin for presumptive antibiotic therapy of UTI, but local antibiotic susceptibility patterns should ultimately dictate the antibiotic choice.

    Duration of Therapy

    Several studies have compared treatment of pediatric UTI with varying durations of therapy (Table 4). The data have been analyzed in two ways. First, studies that directly compared different durations of therapy were examined (Table 4, A). Then, data were pooled by antibiotic and duration of therapy, and the pooled values were compared (Table 4, B).

    Table 3A.

    Characteristics of Diagnostic Tests for UTI—Listed by Study, in Descending Order of Quality Score

    Table 3B.

    Characteristics of Diagnostic Tests for UTI—Listed by Test

    Table 3C.

    Characteristics of Diagnostic Tests for UTI—Summarizing Diagnostic Characteristics of Each Test

    Table 4.

    Effectiveness of Short-term Therapies for UTI Grouped by Duration and Agent

    None of the studies of duration of therapy compared 7 days with 10 days. In seven studies with 10 comparisons between long duration (7 to 10 days) and short duration (one dose to 3 days), 8 of 10 comparisons showed better results with an attributable improvement in outcome of 5% to 21%.55–61

    When the data were pooled by agent and duration, no discernible differences were found in initial cure or relapse among one-dose, 3-day, and 10-day courses of amoxicillin. However, single-dose cotrimoxazole was 10% less effective than 1-, 3-, 7-, or 10-day therapy. Differences among the latter regimens were not discernible.


    The data pooled by agent and duration of therapy were used to compare amoxicillin and cotrimoxazole. For therapies of one dose, 3 to 4 days, or 10 days, cotrimoxazole consistently shows better cure rates (4% to 42%).

    Data comparing parenteral and oral therapy were not found. However, intramuscular ceftriaxone (one dose)62 or gentamicin (10 days)63 was 100% effective in resolving UTI in children in whom oral therapy had failed. No data are available that clarify the role of oral vs parenteral therapy for bacteremia in association with UTI. Drawing on analyses of unsuspected bacteremia in children without UTI, parenteral antibiotic therapy is 95% effective in clearing bacteremia.64

    Prevalence of Urinary Tract Abnormalities

    UTI in young children is a marker for abnormalities of the urinary tract. By far, the most common abnormality is VUR, which may be present with different degrees of severity that are graded I to V according to whether the reflux reaches the kidney and the degree of dilation of the collecting system.65 The grade of VUR is important because it determines the likelihood of detecting VUR on radiologic evaluation and the probability of renal damage, hypertension, or renal failure.


    The Subcommittee identified 77 studies that reported the prevalence of VUR among children with UTI (Table 5). The range of reported values is wide. However, examination of the graph of the prevalence of VUR plotted against the sample size (Fig 7) suggests that the variation is primarily attributable to small samples. As the sample size grows, the prevalence reported converges at 30% to 40%. The prevalence appears to decrease with age.

    Table 5.

    Prevalence of VUR Among Children With UTI

    Fig. 7.

    The prevalence of VUR as a function of study size. The variation in reported prevalence of VUR decreases among studies of larger size. A prevalence between 20% and 40% is typical of larger studies.  

    Figure 8, A and B, presents plots of reported prevalence of VUR by the average age of subjects in each of 24 studies. One outlier, a study with an average age of about 7 years and a prevalence of 73% was excluded. Figure 8, A, shows a model fit to the data in which each point is weighted by the sample size of the study it represents. The resultant graph shows a decline in the prevalence of VUR with increasing age. The intercept is 45% and the slope is −3% per year of age. Comparable graphs of prevalence of VUR by maximum or minimum age do not show a consistent relationship.

    Fig. 8.

    The prevalence of VUR as a function of patient age. The prevalence of VUR reported in 54 studies of UTI in children as a function of the average age of the patients in the study is plotted. In A, the line is a smoothing spline (λ = 100) fit to the data. In B, the line is a third-order polynomial fit to the data. The studies are weighted by sample size.

    Specific studies of the relationship between age and prevalence of VUR show similar results. For example, Buys et al18 reported a 68% prevalence among boys younger than 1 year and 25% among boys 1 to 3 years. In another study of boys, Cohen49 found some type of abnormality in 76% of boys younger than 10 years (∼54% of these were VUR) and 15% in boys older than 10 years. In a population-based study, Jodal66 showed a peak prevalence of VUR among girls 1 to 3 years of age and a rapid drop-off (perhaps by half) by age 5. In a much smaller study, Shah and associates67 found VUR in 100% of 9 boys younger than 1 year of age with a UTI and in 74% of 34 boys 1 to 5 years of age.

    This relationship was fitted to a third-degree polynomial (Fig 8, A) and a spline (λ = 1; Fig 8, B). Although the data are insufficient to define a particular functional relationship, they suggest a decline in the prevalence of VUR with age that seems to be most rapid during the first year of life. It levels off somewhat between 1 and 5 years of age, then drops off again after age 5. When studies of children younger than 3 years were pooled, the prevalence was 50% (Table 6).

    Table 6.

    Prevalence of Urinary Tract Abnormality Among Patients With UTI

    The grade of VUR is shown in Table 7. Not all studies report the grade of VUR using the international grading system. In other studies, a three-grade system is usually used. Based on the descriptions of the grading systems in these studies, grade I corresponds to grades I and II in the international system, grade II to grade III in the international system, and grade III to grades IV and V in the international system.

    Table 7.

    Distribution of Grades of Reflux Among Children With VUR

    For this analysis, grades of VUR were grouped into low grade (grades I and II) and high grade (grades III to V). This grouping reduces the precision of the analysis slightly because higher grade reflux is easier to detect with ultrasonography and poses a greater risk of subsequent renal damage than do the lower grades. However, these effects were represented in this model, making this a more detailed analysis than has been reported previously. Some studies grouped patients in a similar manner, but grade III VUR was variably grouped as high grade or low grade. For classification of results, the classification used in the articles was (by necessity) used in this analysis. Pooling the data from Table 7 gives a 51.1% prevalence of low-grade (grades I to II) VUR among patients with VUR.

    Tests for VUR

    The Subcommittee identified the gold standard test for detecting VUR as the VCUG or RCG. These studies have, by definition, 100% sensitivity and specificity. The Subcommittee also believed that renal ultrasonography or intravenous pyelography should be performed to identify obstructions or other structural renal abnormalities.

    VCUG and RCG are expensive and invasive. Data were collected on the sensitivity and specificity of renal ultrasonography alone in the detection of VUR. Overall, renal ultrasonography has poor sensitivity (30% to 62%) and good specificity (85% to 100%) for VUR.68 ,69 The sensitivity is better, however, for high-grade VUR than for low-grade VUR (82% to 100% vs 14% to 30%).68 ,70 The Subcommittee members believed that renal ultrasonography in young infants was less reliable than this. However, data supporting this point were not found.

    Progressive Renal Damage

    The evidence that links the presence of VUR to important clinical outcomes can only be assembled in a piecemeal way. The data link VUR to renal scarring in the presence of recurrent infection, and renal scarring is associated with subsequent hypertension and ESRD (Fig 9). The data support this model indirectly. However, there are no longitudinal data that link directly the presence of VUR in infants with febrile UTI and normal kidneys to the subsequent development of hypertension or ESRD. Therefore, it is difficult to quantify the strength of the relationship or the risk to such patients. The following discussion reviews evidence supporting each step in the hypothesized progression to renal damage.

    Fig. 9.

    A model of the theoretical relationship of VUR to the development of ESRD and hypertension.

    Reflux and Scarring

    The data demonstrate an association between VUR and subsequent renal scarring. More than 48 references reviewed reported renal scarring at rates between 1% and 40% in association with febrile UTI and VUR. Higher grades of VUR are associated with higher risk of scar progression. Patients with high-grade VUR are four to six times more likely to have scarring than those with low-grade VUR and eight to 10 times as likely as those without VUR.3 ,66 ,71 ,72 In one study of 84 consecutive children 2 to 70 months of age with their first recognized febrile UTI,52 the presence of VUR was 80% sensitive and 74% specific in predicting subsequent scarring in a follow-up period of 2 years or more. Of those with VUR, 30% had progressive renal scarring. Of those without VUR, only 4% did.

    Recurrent Infections and Scarring

    The association of VUR with renal scarring seems to be mediated through recurrent symptomatic infections. The association between recurrent bouts of febrile UTI and the risk of renal scarring in one study followed an exponential curve (Fig 10).66 Patients with no VUR or low-grade VUR and patients with no recurrences of UTI are at low risk for renal scarring. Thus, prophylactic antibiotic treatment of patients with VUR to prevent recurrences or surgical treatment to prevent VUR would be expected to prevent renal scarring.

    Fig. 10.

    Predicted and observed relationships between recurrent UTI and renal scarring. The plot of the risk of renal scarring as a function of the number of UTIs illustrates the close match between the data observed and what would be predicted by an exponential model.

    ESRD and Hypertension

    Retrospective examination of the causes of ESRD in registries of patients with renal failure shows that in 36% (of 9250 patients), the cause of ESRD is obstructive uropathy, renal hypoplasia or dysplasia, pyelonephritis, or a combination of these.1Unfortunately, it is not possible to determine which patients originally had normal kidneys in whom disease progressed to ESRD because of VUR and recurrent infection. In the North American Renal Transplant Cooperative Study, the fraction of transplant recipients with reflux nephropathy is 102/2033, or 5%. Patients with pyelonephritis or interstitial nephritis make up another 2% of transplant recipients.2

    Most cohort studies to quantify the risk of ESRD and hypertension among patients with renal scarring and VUR are based on highly selected patients in whom extensive scarring already has occurred. Long-term studies show that ESRD develops in 3% to 10% of these patients.3 73–75 In addition, 10% may require nephrectomy, and 0.4% renal transplantation.3 One study followed women from their first UTI in childhood and found that abnormal renal function could be documented in 21% of those with severe scarring.76 However, this study also had a selected population and did not provide data on VUR.

    The rate of hypertension in these and similar studies varies between 0% and 50%.3 ,73 75–78 It is apparently lowest in those with low-grade VUR79 and those with the least scarring.77 However, the quality of the studies relating VUR nephropathy with ESRD and hypertension was rated poor, and quantification of the relationship is weak. One study78identified retrospectively a cohort of children 1 month to 16 years of age with VUR and measured blood pressure levels 5 months to 21 years (average, 9.6 years) after VUR was identified. There was an increase in the blood pressure level associated with increasing grades of VUR. However, none of the patients had hypertension.

    Effectiveness of Prophylactic Antimicrobials and Surgical Correction

    Ultimately the value of identifying VUR depends on the potential to prevent its long-term consequences. As Fig 9 illustrates, this depends on reducing recurrent infections with antibiotic prophylaxis or eliminating VUR surgically. The efficacy of these procedures is not well documented because controlled studies have not been performed. Studies in which patients receiving antibiotic prophylaxis are compared with those not taking antimicrobials suggest a 50% effectiveness whether comparing rates of reinfection73 79–83 or progression of scarring.24 ,51

    Cost Data

    Cost data were derived from literature where available, the accounting department of the University of North Carolina (UNC) Hospital, and the physicians' fee schedule from the UNC Physicians and Associates. Estimates used in the decision model are given below.

    UA and Culture

    Based on the cost data from the UNC Hospital, the cost of a urine culture is $21.53 (charge, $26.00). The cost of a UA is $6.77 (charge, $15.00).


    The cost of short-term treatment of UTI varies substantially depending on the treatment chosen. Standard oral therapy, such as amoxicillin or trimethoprim–sulfamethoxazole, costs about $10 for a course. Newer, broad-spectrum oral therapy costs about $40 for a 10-day course. If parenteral therapy, such as intramuscular ceftriaxone, is given, the cost of the drug and the injection may be $125.

    Complications of Therapy

    For the analysis, only complications requiring a return visit to the physician were considered. This visit would cost a $50 clinic fee, plus a $50 professional fee, for a total of $100. Death attributable to anaphylaxis is rare. The additional cost of this complication was not added.

    Cost of Urosepsis

    It is assumed that sepsis that does not clear spontaneously or with antibiotic therapy will require hospital admission for intravenous antibiotic therapy at a cost of $10 000.

    Imaging Evaluation

    The two following imaging strategies were considered: 1) a full work-up, including renal ultrasonography and VCUG, and 2) renal ultrasonography alone. The cost of renal ultrasonography includes the hospital cost of $104 and the physician fee of $200, for a total of $304. The cost of the VCUG is the hospital cost of $123 plus the physician fee of $200, totaling $323. Thus, the full work-up costs $627.

    Cost of Treatment of VUR

    The estimated cost of antibiotic prophylaxis was based on low-dose nitrofurantoin. The weekly cost was $6 for an average of 3 years. The total is $1040. It was assumed that the cost of ureteral reimplantation surgery was similar.

    Complications of VUR

    Patients with VUR are at increased risk for recurrent UTI,3 ,75 ,84 ,85 at least until the VUR resolves. The period the child is at risk was assumed to be 3 years. The cost of these recurrences was based on the average cost of a clinic visit related to a diagnosis of UTI or pyelonephritis in infants at UNC, $134, plus the physician fee of $70. This was multiplied by three, the average number of infections expected over 3 years,84 for a total of $612.

    The cost of ESRD was derived from data in previous analyses.86 ,87 Based on a 10-year period, including 10 years of dialysis or a year in which transplantation occurs followed by 8 years with a functioning graft, followed by 1 year of a failed graft, the cost of ESRD is ∼$300 000. The cost of managing severe hypertension and its complications was estimated at $100 000 based on $2000 per year for 50 years.88 ,89

    Analysis of the Decision Model

    Using data extracted from the literature as described in “Evidence Tables,” probabilities and costs were inserted into the decision tree. Table 8 shows the baseline estimates for the variables in the model. Using these estimates, three types of analysis were conducted. First, a risk analysis was performed to determine the average cost and frequency of outcomes expected with each of four strategies. Next, a cost-effectiveness analysis was used to determine the incremental cost per major clinical outcome averted by the strategies. Finally, sensitivity and threshold analyses were performed to examine alternative strategies for evaluation and management of UTI.

    Table 8.

    Variables and Baseline Estimates Used in the Risk and Cost-effectiveness Analyses

    Risk Analysis

    Two risk analyses were performed. The first examined the costs and clinical outcomes expected with different strategies for making the diagnosis of UTI. The second compared alternative strategies for imaging the urinary tract of patients with UTI.

    Diagnosis of UTI

    The risk analysis of diagnostic strategies examines two anchor states. The first, “treat all,” is presumptive antibiotic therapy for occult infection without testing. The second, “observe,” is clinical observation without testing or treatment. Neither of these anchor strategies involves subsequent imaging of the urinary tract. Three testing strategies are evaluated. The first is the gold standard, culture of urine specimens obtained by suprapubic tap or transurethral catheter. The second is using a culture of the urine specimen obtained from the urine bag. Culture of bag urine specimens has 100% sensitivity and 70% specificity in the baseline analysis. The third strategy uses the cheaper but less reliable reagent strip as a test for UTI. A positive LE test result, a positive nitrite test result, or both, is considered a positive test result. The sensitivity and specificity of this strategy are 92% and 70%, respectively, based on median values given in Table 3, C. All three testing strategies include antibiotic therapy and imaging of the urinary tract of patients with positive test results.

    The results of the risk analysis are presented as the number of outcomes expected per 100 000 infants treated by using each strategy (Table 9). To observe patients without testing or treatment is the least expensive strategy. However, it results in inferior clinical outcomes, including death and hospitalization from urosepsis, hypertension, and ESRD. Presumptive treatment of all infants without testing costs slightly more than observing, but prevents hospitalization and death from urosepsis. Under the treat all strategy, it is impossible to select patients for imaging, and it is unreasonably expensive to image the urinary tracts of all patients (see “Cost-effectiveness Analysis”). Therefore, no patient in the treat all strategy undergoes imaging evaluation. As a result, there is no improvement in the rates of hypertension or ESRD.

    Table 9.

    Risk Analysis of Diagnostic Strategies—Anticipated Costs and Outcomes for a Cohort of 100 000 Patients Treated Using Each Strategy

    Using culture of urine specimens obtained by catheterization or tap offers the lowest risk of death, because all children at risk for urosepsis are identified and there are no unnecessary treatments or treatment-related deaths. Moreover, because all children with UTI are identified for imaging evaluation, this strategy minimizes the risk of hypertension and ESRD.

    Using a culture of bag urine specimens as a criterion for the diagnosis of UTI also identifies correctly all children with UTI. However, its poor specificity also results in many false-positive results. As a consequence, there are slightly more deaths (3 per million) attributable to unnecessary antibiotic treatment and many more imaging work-ups. This means an additional $47.2 million per 100 000 febrile infants with no improvement in clinical outcomes.

    Using LE and nitrite as diagnostic criteria for UTI leaves a small number of UTIs undiagnosed and results in 2½ times as many imaging work-ups compared with a culture of urine specimens obtained by catheterization or tap. The result is poorer clinical outcomes at greater expense despite the lower cost of UA.

    A fourth diagnostic strategy also was evaluated that used a full UA (reagent strip and microscopy) as an initial diagnostic test and treated any positive UA component as a positive result. A positive result would lead to presumptive treatment. However, positive results would be confirmed by a culture of urine specimens obtained by tap or catheterization before imaging was performed. This strategy has a sensitivity of virtually 100% and a specificity of 60%36 ,38 ,39 for making a treatment decision. Because positive results are confirmed with a culture of urine specimens obtained by catheterization or tap, the specificity is 100% for imaging. The cost of testing is higher for positive results because UA and culture are performed. Also, the false-positive results lead to more unnecessary treatment and the resulting costs and complications. However, there is a cost saving for patients with negative results of the UA because the UA costs less than the culture. Compared with a culture of specimens from all patients, there is a small net decrease in cost using UA followed by a culture of positive results. Moreover, this strategy has the advantage of allowing immediate treatment of patients at risk of UTI. The benefit of early treatment is difficult to quantify, but evidence suggests it reduces the risk of renal scarring.51 Although this strategy appears to be nearly as effective as (or perhaps more than) the culture of urine specimens obtained by catheterization or tap, its effectiveness depends on the accuracy of the urine microscopy. As noted above, the accuracy of urine microscopy is highly variable and apparently depends on the presence of an on-site technician. In many settings, this may be impossible or so expensive as to offset the benefits.

    Imaging Strategies

    Three imaging alternatives were evaluated by risk analysis (Table 10). The first strategy was renal ultrasonography and VCUG. The second was renal ultrasonography alone, and the third was no imaging evaluation at all. The total cost associated with each strategy includes the cost of the imaging studies, the cost of treating abnormalities identified, and the cost of any complications incurred. Together, renal ultrasonography and VCUG will identify all cases of renal abnormalities. Renal ultrasonography alone detects ∼42% of all abnormalities. However, most abnormalities (87%) missed by renal ultrasonography alone are low-grade VUR, which carries a lower risk of renal scarring. As a result, the differences in clinical outcomes (cases of hypertension and ESRD) are smaller between renal ultrasonography alone and the full evaluation.

    Table 10.

    Risk Analysis of Alternative Strategies for Imaging the Urinary Tracts of Infants With UTI—Expected Cost of Clinical Outcomes Expected for 1000 Patients Treated Under Each Strategy

    Cost-effectiveness Analysis

    Cost-effectiveness analysis was used to quantify the trade-offs between cost and clinical effect when moving from one clinical strategy to another. When one strategy offers a better clinical effect at a lower cost, it is said to be a dominant strategy. However, in most cases, the strategy with better clinical effect also has a higher cost. Cost-effectiveness analysis depicts the additional cost per unit of improvement in clinical effect. For this analysis, the units of effect were defined as cases of death, ESRD, or hypertension prevented.

    Strategies were compared using the incremental (or marginal) cost-effectiveness ratio, ie, the difference in cost among strategies divided by the difference in effect. The mathematical term is as follows:Embedded Image Table 11 gives the alternative strategies in order of increasing cost. The first column indicates the strategy, the second column shows the additional cases (per 100 000 patients) prevented compared with the preceding strategy; and the third column presents the additional cost per case prevented above the preceding strategy.

    Table 11.

    Marginal Cost-effectiveness of Strategies for Diagnosis and Treatment of UTI

    Diagnosis of UTI

    The least expensive strategy for the diagnosis of UTI is to do nothing. This also is the least effective strategy. By comparison, treating all febrile 2-month to 2-year-olds for UTI without diagnostic testing or urinary tract imaging improves clinical effect at a cost of $61 000 per death prevented. However, it does not prevent ESRD or hypertension.

    As an alternative, one could obtain a dipstick UA on all patients, culturing specimens from those for whom the results of UA were positive. Because UA costs less than culture, this strategy has the potential to save money. However, because UA does not have perfect sensitivity (Table 3), a few cases will be missed. In addition, patients for whom the results of UA are positive will incur the costs of UA and culture. To evaluate this strategy, a new branch was added to the decision tree. The new branch is a dipstick UA. If results are positive, it is linked to the “test” branch; if results are negative, it is linked to the “observe” branch. Positive results involve the cost of the UA plus culture; negative results, only the cost of the UA. This strategy prevents an additional two cases of death or serious complication per 10 000 at a cost of $261 000 per case.

    Culturing catheterization or tap urine specimens on all patients and then treating and imaging the urinary tracts of all patients with positive results is the most effective and most expensive strategy. It prevents an additional 2.5 cases of death or serious complication per 100 000 over culture-confirmed positive dipstick results at an additional cost of $434 000 per case prevented. Compared with the no testing strategy, culturing specimens from all patients prevents three cases of death or serious complication per 10 000 at a cost of $200 000 per case prevented.

    The optimal strategy depends on the decision-maker's willingness to pay for each additional case prevented. For example, if that willingness to pay is between $261 000 and $434 000, then the positive results of the dipstick confirmed by culture is the best strategy. Culturing specimens from all patients will prevent a few additional cases, but the cost to prevent each of these additional cases exceeds the decision-maker's willingness to pay. If the decision-maker is willing to pay >$434 000 to prevent a case of death or serious complication, then culturing specimens from all patients is the right strategy.

    Imaging Strategies

    The cost-effectiveness of the alternative imaging strategies also was calculated (Table 12). The least expensive alternative is to perform no evaluation. Renal ultrasonography alone will prevent almost three cases of ESRD or hypertension per 1000 studies done at a cost of $260 000 per case prevented. A VCUG prevents an additional one case of ESRD or hypertension per 1000 studies over renal ultrasonography at a cost of $353 000 per case.

    Table 12.

    Marginal Effect and Cost-effectiveness of Strategies for Imaging the Urinary Tracts of Children With UTI


    Again, the optimal strategy depends on the decision-maker's willingness to pay for each additional case prevented. For example, if that willingness to pay is between $260 000 and $353 000, then renal ultrasonography alone is the best strategy. A VCUG will prevent a few additional cases, but the cost to prevent each of these additional cases exceeds the decision-maker's willingness to pay. If the decision maker is willing to pay >$353 000 to prevent a case, then renal ultrasonography and VCUG is the right strategy.

    Sensitivity and Threshold Analysis

    Choosing one alternative over another from the risk analyses shown in Tables 9 and 10 involves striking a balance between the cost of an intervention and the improvement in clinical outcomes. When comparing alternatives, if one alternative provides better clinical outcomes at lower costs, it is the dominant alternative and is the obvious choice. However, in most circumstances, choosing an alternative requires a trade-off between costs and clinical benefit. For example, obtaining a urine specimen for culture by catheterization or tap on all febrile infants, treating those with positive cultures, and imaging their urinary tracts with VCUG and renal ultrasonography improves clinical outcome but at a substantially higher cost than doing none of these evaluations.

    To select the optimal alternative and, moreover, to identify different clinical circumstances in which different alternatives may be better, it is first necessary to make explicit the trade-off between costs and clinical outcomes. This means identifying a “willingness to pay” for each untoward clinical outcome avoided. For the following analyses, a value of $700 000 was placed on each case of ESRD or hypertension prevented. This figure is based loosely on the lifetime productivity of a healthy, young adult.90 Once this cost is assigned to each untoward clinical outcome, it is possible to use the threshold method of decision-making.91

    The threshold approach to decision-making involves changing the value of a variable in the decision analysis to determine the value at which one alternative becomes too expensive and another would be preferred. In the case of UTI, if the prior probability of UTI is sufficiently high, it costs <$700 000 to prevent a case of ESRD or hypertension by screening all febrile children. However, if the probability that a particular patient has a UTI is sufficiently low, the yield of a urine culture will be extremely low and a positive result of a UA will almost certainly be a false-positive. Under these circumstances, a strategy involving the evaluation of this child's urinary tract would cost >$700 000 per case of ESRD or hypertension prevented. Threshold analysis identifies the threshold probability of UTI above which evaluation would be cost-effective and below which it would not. A number of threshold analyses follow.

    Prevalence of UTI and Prevalence of VUR

    Because fever without an obvious cause is common in pediatric practice, culturing the urine of all febrile 2-month to 2-year-old children represents a substantial investment. The Subcommittee was interested in determining whether there were some clinical subpopulations in which the prevalence of UTI was low enough that culture was unnecessary. A threshold analysis was used to examine this question.

    Figure 11 illustrates a two-way sensitivity analysis. The x-axis shows the prevalence of UTI between 0 and 1, and the y-axis shows the prevalence of VUR among children with UTI. The decision tree was evaluated for all combinations of the prevalence of UTI and VUR. The graph in Fig 11 is divided into two parts by a threshold line. Above and to the right of the line, culturing urine specimens from all febrile infants prevents ESRD and hypertension at a cost of <$700 000 per case prevented. Below and to the left of this threshold line, the cost to prevent a case of ESRD or hypertension would be >$700 000 per case.

    Fig. 11.

    Threshold analysis, prevalence of UTI, and prevalence of VUR. This graph plots the combination of prevalence values for UTI and VUR at which evaluation for UTI (as described in the practice parameter) would be cost-effective (see text). Also plotted are the predicted prevalence values for subpopulations at risk

    Also plotted on the graph are the prevalence of UTI and VUR for febrile females younger than 1 year, females older than 1 year, males younger than 1 year, males older than 1 year, males younger than 1 year and circumcised, and males older than 1 year and circumcised (Table 2). Among circumcised males older than 1 year, the prevalence of UTI and VUR fall well below the threshold line. This implies that in circumcised males older than 1 year, obtaining a urine specimen for culture with the objective of preventing ESRD or hypertension is too expensive, and no evaluation of the urine is the preferred alternative. Circumcised males younger than 1 year fall only slightly below and to the left of the threshold line. In this group, the cost-effectiveness of obtaining a urine specimen for culture is borderline.

    Cost of Culturing Urine Obtained by Urine Bag

    Many practitioners prefer to culture urine specimens obtained by urine bag rather than by transurethral catheter or suprapubic tap because the urine bag is less invasive. However, culture of a urine specimen obtained by bag has low specificity. When the prevalence of UTI is low, as it is in febrile infants, the result is a large number of false-positive urine cultures. False-positive cultures lead to unnecessary antibiotic treatment and urinary tract evaluation. The result is additional costs with no clinical benefit. To examine these additional costs, the difference in costs between a strategy involving urine specimens obtained by catheterization or tap and a strategy involving specimens from bag urine was calculated at different levels of specificity for the urine bag. The result is plotted in Fig 12.

    Fig. 12.

    Costs of relying on a culture of urine collected by bag. This graph plots the expected cost per patient predicted from relying on urine cultures with varying degrees of specificity.

    In Fig 12, the x-axis shows that the specificity of a culture of specimens from bag urine varied between 0% and 100%. They-axis shows the difference in costs between strategies that involve the culture of urine specimens obtained by bag and strategies involving the culture of urine specimens obtained by catheterization or tap. As the specificity of the culture of bag urine varies from 0% to 100%, the additional costs varies between $1600 and $0. The shaded region shows the range of specificities reported for the culture of specimens from bag urine. At the low end, for a specificity of 14%, the cost per patient of obtaining urine by bag rather than by catheterization or tap is $1340 per person. At the high end, for a specificity of 84%, the additional cost of obtaining urine specimens by bag instead of by catheterization or tap is $293 per patient. At the baseline estimate of 70% specificity, the additional cost of the strategy of culturing urine specimens obtained by bag rather than by catheterization is $429.

    These results imply that culturing urine specimens from bag urine in this setting can only be justified if one is willing to pay between $293 and $1340 per patient to use a urine bag rather than to obtain urine specimens by catheterization or tap.

    Sensitivity and Specificity for UA

    As an alternative to obtaining urine specimens for culture from all febrile children between 2 months and 2 years of age, one could use a UA to make the diagnosis of UTI. As noted previously, a carefully performed UA has high sensitivity and specificity. Two-way sensitivity analysis was used to determine the minimal sensitivity and specificity that would be necessary to justify the use of UA rather than culture. The results are shown in Fig 6.

    Figure 6 plots the false-positive rate (1 minus specificity) on the x-axis and the true-positive rate (sensitivity) on they-axis. All the reported test characteristics for various components and combination of components of the UA are plotted on this graph (see “Tests for Urinary Tract Infection”). For all combinations of sensitivity and specificity, the decision tree was reevaluated to determine whether the UA or a culture of specimens from all patients was the most cost-effective strategy. The result is a threshold line shown in the top left corner of the Fig 6. If the sensitivity and specificity of the UA falls above and to the left of this threshold line, it would be preferred to urine culture. If the sensitivity and specificity fall below and to the right of this line, presumptive culture is preferred. Of all the sensitivities and specificities reported, only one study that used a combination of dipstick and microscopy achieved this level of sensitivity and specificity. This study had an on-site laboratory technician in the clinic who processed the urine specimens within 1 hour of collection.

    These results imply that a test must have a sensitivity >92% and a specificity >99% to be preferred over urine culture. This level of sensitivity and specificity is unlikely to be obtained in most clinical settings. Alternatively, several studies demonstrated that combinations of LE, nitrite, microscopy of fresh urine, or a combination of these, in which any abnormality constituted a positive test result, yielded a sensitivity of >92% (Fig 6). If one of these tests were used to rule out UTI, and culture of urine specimens obtained by catheterization or tap were used to confirm UTI in positive results, the requisite sensitivity and specificity would be obtained.

    Probability of UTI and Urinary Tract Imaging

    The decision to image the urinary tracts of infants with documented UTI presumes that the diagnosis of UTI is certain. If an imperfect test (eg, using a bag urine specimen) is used to make the diagnosis of UTI, a substantial number of these patients in fact may not have had a UTI and, therefore, the yield of the imaging will be lower. Threshold analysis was used to determine the minimal probability of UTI that justified a full imaging evaluation. The analysis indicates that if the probability of UTI is <49%, imaging becomes too expensive. Bayes's theorem shows that among patients with a prevalence of UTI that is 5%, those who have a positive urine culture obtained by bag have a probability (positive predictive value) of UTI that is 15%, well below 49%. This implies that if the best evidence of the UTI available in a given patient is a positive culture of urine obtained by bag, additional imaging of the urinary tract is not justified.

    Sensitivity of Renal Ultrasonography for VUR

    Because VCUG is invasive and expensive, a threshold analysis was used to explore the possible use of renal ultrasonography alone in evaluating the urinary tracts of children with UTI. Renal ultrasonography has low sensitivity for low-grade VUR, but relatively higher sensitivity for high-grade VUR. Figure 13 plots sensitivity of renal ultrasonography for low-grade VUR on the x-axis and sensitivity of renal ultrasonography for high-grade VUR on they-axis. For values that plot above and to the right of the threshold line on this graph, renal ultrasonography alone would be the preferred strategy. For values that fall below and to the left of the threshold line, a combination of renal ultrasonography and VCUG would be preferred. Also plotted on this graph are reported values of the sensitivity of renal ultrasonography for low-grade and high-grade VUR. Based on the professional opinion of the UTI Subcommittee, a sensitivity of 14% for low-grade VUR and 82% for high-grade VUR (the lower point in Fig 13) is a better reflection of the situation in children 2 months to 2 years of age. This implies that renal ultrasonography alone is an inadequate imaging evaluation for infants with UTI.

    Fig. 13.

    Threshold analysis of the sensitivity of renal ultrasonography alone for high-grade and low-grade VUR. Also shown (dots) are reported sensitivities from the literature. For values above and to the right of the threshold line, ultrasonography alone would be a cost-effective strategy (see text).

    Risk of Renal Scarring With UTI

    One area of relative uncertainty, in which quality data are lacking, is the risk of scarring in undetected or untreated VUR. This risk is higher for children with high-grade VUR than for those with low-grade VUR. Threshold analysis was used to explore how this risk of scarring affects the decision to culture urine specimens from all children 2 months to 2 years of age with fever and no obvious cause for the fever. Figure 14 plots the risk of scar progression given low-grade VUR on the x-axis and the risk of scar progression given high-grade VUR on the y-axis. For values that plot above and to the right of the threshold line, culture of urine specimens from all febrile infants is the preferred strategy. For values that plot below and to the left of the threshold line, no urine culture is the preferred strategy. The baseline estimates used in the analysis, 14% for low-grade VUR and 53% for high-grade VUR, are plotted on the graph. The risk of scarring because of untreated VUR seems to be high enough to justify culturing specimens from all febrile infants.

    Fig. 14.

    Threshold analysis of the risk of scar progression given low-grade VUR and the risk of scar progression given high-grade VUR. The dot shows the baseline estimate used in the model. Evaluation for UTI, according to the practice parameter, is cost-effective for values above and to the right of the threshold line.

    Risk of Hypertension and ESRD

    The weakest probability estimates in the present analysis relate to the risk of ESRD and hypertension among patients who have progressive renal scarring attributable to VUR. Two-way sensitivity analysis was used to determine the effect of varying these estimates on the results of the analysis. Figure 15plots the probability of hypertension on the x-axis and the probability of ESRD on the y-axis. For values below and to the left of the threshold line shown, it is not cost-effective to obtain a urine specimen for culture from all children to prevent ESRD or hypertension caused by progressive renal scarring. For values above and to the right of the threshold line, culturing urine specimens from all febrile infants is cost-effective. The baseline point estimates of these probabilities are shown, as is the range of values found in our review of the literature (shaded area).

    Fig. 15.

    Two-way sensitivity analysis of the probability of hypertension and the probability of ESRD. Evaluation for UTI, according to the practice parameter, is cost-effective for values above and to the right of the threshold line. The baseline estimate used in the model and the range of reported values also are shown.

    Our best estimate of the risk of hypertension and ESRD justifies culturing urine specimens from febrile infants as a strategy to prevent these complications. However, the quality of evidence about the future risks of ESRD and hypertension among infants with UTI later found to have VUR is tenuous at best. This is an area that needs additional investigation.


    Who Should Be Evaluated for UTI?

    Under the assumptions of the analysis, all febrile children between the ages of 2 months and 24 months with no obvious cause of infection should be evaluated for UTI, with the exception of circumcised males older than 12 months.

    Minimal Test Characteristics of Diagnosis of UTI

    To be as cost-effective as a culture of a urine specimen obtained by transurethral catheter or suprapubic tap, a test must have a sensitivity of at least 92% and a specificity of at least 99%. With the possible exception of a complete UA performed within 1 hour of urine collection by an on-site laboratory technician, no other test meets these criteria.

    Performing a dipstick UA and obtaining a urine specimen by catheterization or tap for culture from patients with a positive LE or nitrite test result is nearly as effective and slightly less costly than culturing specimens from all febrile children.

    Treatment of UTI

    The data suggest that short-term treatment of UTI should not be for <7 days. The data do not support treatment for >14 days if an appropriate clinical response is observed. There are no data comparing intravenous with oral administration of medications.

    Evaluation of the Urinary Tract

    Available data support the imaging evaluation of the urinary tracts of all 2- to 24-month-olds with their first documented UTI. Imaging should include VCUG and renal ultrasonography. The method for documenting the UTI must yield a positive predictive value of at least 49% to justify the evaluation. Culture of a urine specimen obtained by bag does not meet this criterion unless the previous probability of a UTI is >22%.


    • The recommendations in this statement do not indicate an exclusive course of treatment or serve as a standard of medical care. Variations, taking into account individual circumstances, may be appropriate.