Secondary Logo

Journal Logo

CME Review Article

Metaanalysis of urine screening tests for determining the risk of urinary tract infection in children: CME REVIEW ARTICLE


Author Information
The Pediatric Infectious Disease Journal: January 2002 - Volume 21 - Issue 1 - p 1-11
  • Free


  • Summarize the literature on urine screening tests for urinary tract infection in children.
  • Recall the validity and applicability of the included studies.
  • Perform a quantitative synthesis (metaanalysis) of included studies.
  • Identify the test or combination of tests that best predicts the presence or absence of urinary tract infection in children.

The all prevalence of urinary tract infection (UTI) in children is ∼5%. 1, 2 It is widely accepted that UTI disposes children, particularly younger children and infants, to an increased risk of later hypertension and chronic renal failure. 3 Early diagnosis and timely and appropriate treatment are thus considered key factors to reduce these risks. 4 Unfortunately diagnosis is difficult, especially in young children and infants, because features are nonspecific and heterogeneous. 5 Confirmation of UTI diagnosis needs a urine culture, which may take 24 to 72 h. In infants <2 years old bladder suprapubic aspiration (SPA) and bladder catheterization are the only methods by which to obtain a reliable sample for urine culture. 6 These invasive procedures are performed only when the probability of a urinary tract infection is high.

Several urine screening tests have been investigated in the past; nevertheless there is a continued debate on which test or combination of tests defines best the risk of UTI in children. The search for leukocytes in a centrifuged urine sample has traditionally been the most widely used screening test. Other tests, including leukocyte esterase (LE) and nitrite through rapid dipstick tests and bacteriuria and/or pyuria in uncentrifuged urine samples, have also been studied. The results have been conflicting.

In 1999 the American Academy of Pediatrics, based on a systematic review of the literature up to 1996, recommended the use of any of the above cited tests. 7 A metaanalysis on the issue by Gorelick and Shaw was published in November, 1999. 8 The search of the literature in that report covered articles published up to 1998. The study concluded that both the presence of any bacteria on a Gram stain on an uncentrifuged urine [true positive rate (TPR) 0.93 and false positive rate (FPR) 0.05] and dipstick analysis (a TPR of 0.88 for the presence of either LE or nitrite and an FPR of 0.04 for the presence of both LE and nitrite) perform similarly in detecting UTI in children from birth through 12 years of age and are superior to microscopic analysis for pyuria. Eleven additional papers on the issue have been published since 1997. The metaanalysis by Gorelick and Shaw pointed out that there were only two studies addressing the usefulness of microscopy for presence of bacteria and leukocytes in stained unspun urine for UTI prediction. Moreover only few studies were included in the receiver operating characteristic (ROC) analysis of dipstick tests, particularly for those in combination (5 data points or substudies for LE and nitrite and 8 data points for LE or nitrite). Other limitations of the metaanalysis acknowledged by the authors comprise inclusion of only English articles and exclusion of articles published before 1966. Furthermore attempts to explain potential influence of different covariates such as age group, technique of sampling, spun or uncentrifuged urine were addressed, but not with a multivariate approach, precluding any definitive statement on such potential influences, individually or through interaction.

We performed a metaanalysis of studies on the usefulness of urinary screening tests in the diagnosis of UTI. We combined the metaanalytic method proposed for pooling through summary receiver operating characteristic (SROC) curves the reports of multiple primary studies with discordant results 9–11 with a multivariate approach, to take into account relevant variables that could potentially have an influence on TPR and FPR tradeoff and thus on the SROC curves. The main question to answer was, “Which is the urinary screening test or combination of tests that best predicts the presence or absence of UTI in children?”


Objectives and scope of the metaanalysis

The study aimed to assess the usefulness of the following index tests, alone or in combination, to predict UTI in children: leukocyturia (or pyuria) in centrifuged urine; bacteria and/or leukocytes in uncentrifuged, stained or unstained urine; and dipstick tests (LE and nitrite, alone or in combinations). The standard used was quantitative urine culture. Different criteria for defining UTI according to urine sampling technique were considered.

Search and retrieval of relevant literature

A computerized search from the National Library of Medicine Medline from 1966 through January, 2001, and from LILACS (Librería Latinoamericana de Ciencias de la Salud: Latin American Library on Health Sciences, a Spanish Latin American Database) from 1982 through 1998 was combined with a manual search of additional articles from references of primary retrieved literature, from review articles, from authors’ personal collection of papers and through contact with experts in the field of UTI. A detail of the key words used for Medline search are shown in Results.

Abstracts of identified papers were read by two independent reviewers (LH, CA) familiar with the systematic search and the critical appraisal of the medical literature. Those articles deemed to be relevant were retrieved and then read as full papers by the same independent reviewers. Any discrepancy between readers was solved by consensus.

Inclusion criteria

To be included finally for the metaanalysis the primary articles should have fulfilled the following criteria:

  • Articles addressing the usefulness of urinary screening tests in the diagnosis of UTI in children from 0 through 18 years of age.
  • Original articles. Review papers and letters to the editors were excluded.
  • Studies limited to humans.
  • Articles with enough information to judge their methodologic quality.
  • Articles with index tests and standard (urine culture) systematically performed in all patients, with specification of sampling technique used.
  • Articles with prevalence, sensitivity, specificity and predictive values explicitly stated or with data presented in such a way that calculation of this information could be feasible.
  • The index tests should have been performed with medical supervision at a hospital or at an outpatient clinic. Studies performed at home without medical supervision were not eligible.

Discrepancies on inclusion criteria between the two independent reviewers were solved by discussion and consensus.

Assessment of methodologic quality of primary studies

The methodologic quality of selected primary articles was independently assessed by the same two independent reviewers. A modified 14-point scale was used, according to a method proposed by Mulrow et al. 12 See Table 1 for details of criteria used for methodologic quality. The final score for each primary article was reached by consensus.

Table 1
Table 1:
Criteria to assess the methodologic quality of primary studies

Data extraction

Relevant data were extracted from included studies by one of the observers, whereas the other independent reviewer checked the data to assure the accuracy of the information. Discrepancies were solved by consensus. The data extracted included prevalence of UTI, sensitivity, specificity and predictive values of index tests. When not reported they were calculated from the available results, according to standard recommendations. 13 True positive rates (TPR) and false positive rates (FPR) were also calculated, to generate from them the SROC curves for each test. When different cutoffs were reported, sensitivity, specificity and predictive values were calculated separately for each cutoff.

Data extracted included age, urine sampling technique, whether urine was centrifuged, whether it was Gram-stained, the specific definitions of the standard (urine culture) and clinical data.

Age groups included young infants (0 to 6 months), infants and preschool children (up to 4 years of age), school children (5 to 10 years of age) and adolescents (11 to 18 years). When age was not reported it was assumed that all ages were included. When a single study reported data for specific age groups, each one was considered separately. If a single study reported children belonging to more than one of the above age categories, they were included in the predominant age group.

Samples obtained through SPA and bladder catheterization were grouped as “sterile” and all the others as “nonsterile,” because the former group has a lower risk of contamination.

Bacteriuria in centrifuged or uncentrifuged urine was reported as the number of bacteria per high power field (hpf) or the number of bacteria per oil immersion field (oif), respectively. Pyuria in centrifuged or uncentrifuged urine was reported as the number of leukocytes per hpf or number of leukocytes per μl, respectively.

Data analysis

SROC curves were constructed, following ad hoc methodologies to perform metaanalyses of several primary studies on diagnostic tests with discordant results. 9–11 This approach was used because sensitivity and specificity are based on a single threshold (cut point or positivity criterion) for classifying an index test as positive or negative. Changing the threshold to increase sensitivity decreases the specificity of the test and vice versa. This conflict between test sensitivity and specificity makes it imperative that they be considered jointly. 11 The summary receiver operating characteristics curve allows full capture of this underlying joint dependence, which can be profitably thought of as a tradeoff between test sensitivity (true positive rate) and test false positive rate (1 − specificity). 9

From each individual primary study, we extracted one 2 × 2 table of true positive (TP), false positive (FP), false negative (FN) and true negative (TN) results. From each of these tables we computed true positive rate (TPR) and false positive rate (FPR) as follows:MATHMATH If either the TPR or the FPR is exactly zero or 1 (as it happens when the 2 × 2 table of test results contains a zero cell), the previous equations are undefined. To avoid this we added one-half to all counts in all the tables included to calculate TPR and FPR 9, 10:MATHMATH We converted the TPR and FPR of each primary study to their logit transforms. 9, 10 The logit of TPR is MATH Similarly for the FPR MATH We then defined the difference D and the sum S of the two logits, as MATHMATH Next we modeled D as a linear function of S; i.e. we performed a weighted least squares linear regression on the 4-fold tables, using D as the dependent variable and S as the independent variable. Thus to estimate the SROC curves, we used the following linear equation:MATH where D = logit TPR − logit FPR, S = logit TPR + logit FPR, b = regression coefficient for S (slope) and i = intercept term.

Once we obtained the slope and the intercept of the transformed line, and only for graphic purposes, we used the following equation to back-transform it to yield TPR as a function of FPR, as MATH This equation describes a curve that assumes the shape of the receiver operating characteristic function. 9, 10 We extended the regression lines only over the range of the experimental data.

Multivariate analysis

A stepwise multiple regression was performed to take into account the potential influence of several variables, individually or through interaction, on the SROC curves. The final model tested included, in addition to S, the following candidate predictor variables: technique of sampling (“sterile” or not); age group; and whether the urine sample assessed was centrifuged. This analysis was performed using procedure REGRESSION of SPSS V9 with default STEPWISE parameters (alpha for entry, 0.05; alpha for deletion, 0.10). The distribution and pattern of residuals were checked visually.


Literature retrieval

Computerized search of Medline from 1996 through January, 2001. Search strategies: (1) word in title or in abstract: urinary tract infection, 27 908 references; urinary tract infection AND children, 10 302 references; urinary tract infection AND children AND urinalysis, 165 references; urinary tract infection AND children AND dipstick, 62 references; urinary tract infection AND children AND nitrite, 105 references; urinary tract infection AND children AND leukocyte esterase, 176 references. (2) Limits 0 to 18 years; human; and 1966 through 2001: urinary tract infection, 3036 references; urinary tract infection AND urinalysis, 70 references; urinary tract infection AND dipstick, 29 references; urinary tract infection AND nitrite, 38 references; urinary tract infection AND leukocyte esterase, 14 references.

Computerized search of LILACS from 1982 through 1998. Search strategies (original key words in Spanish): urinary tract infection, 680 references; urinary tract infection AND children, 160 references; urinary tract infection AND children AND dipstick, 3 references; urinary tract infection AND children AND nitrite, 8 references; urinary tract infection AND children AND urinalysis, 0 references.

From all these references and from additional articles obtained from primary and review articles and through correspondence with experts in the field of UTI, 83 deemed relevant were reviewed in detail as full papers. Finally 48 articles containing sufficient information were included for performing further analysis on diagnostic usefulness of the index tests. 14–61 Reasons for rejection of the 83 articles reviewed in full are shown in Table 2.

Table 2
Table 2:
Reasons for rejection of primarily included citations after assessment as full articles

Methodologic value of primary studies

Almost all of the studies fulfilled the main criteria of methodologic validity, namely that the index test and the standard were independently measured in all the patients and that the clinical spectrum of patients was wide. The scores of methodologic quality of primary studies are shown in Table 3.

Table 3
Table 3:
Variables concerning study group characteristics

Primary studies revealed heterogeneity regarding age groups, sampling technique, whether urine samples assessed were centrifuged, whether they were Gram-stained, definition of standard according to sampling technique used, cut points for considering a test positive or negative and clinical spectrum of studied subjects, among others. Most studies included uncentrifuged, unstained urine samples.

Diagnostic accuracy analysis

After data extraction 30 groups of individual or combined tests with different cut points were obtained. This number of tests is different from the 48 articles included because several primary articles assessed the same test or combination of tests. The following combinations were excluded, as only one isolated data-point was reported for each one of them: pyuria ≥5/hpf or bacteriuria any/hpf or LE ≥small or nitrite-positive; nitrite-positive or LE ≥trace or pyuria >5 leukocytes/hpf or bacteriuria mild; LE ≥trace and nitrite-positive or pyuria >5 leukocytes/hpf; LE ≥trace and nitrite-positive or bacteriuria any/hpf or pyuria any/oif.

Then 11 groups of urine screening tests were elaborated, grouping together tests that were similar except for small differences in cut points. Finally 9 groups of tests were included for construction of 9 corresponding bivariate SROC curves. Combinations of pyuria ≥10/hpf or bacteriuria any (P10 or B) and LE and nitrite were not plotted because there were few data points with which to construct reliable SROC curves. However, they were included in the multivariate analysis. Tables 3 to 5 show details of each of the 48 studies, including the different thresholds of tests. Overall 161 data points were used to construct the SROC curves.

Table 4A
Table 4A:
Characteristics of interest tests
Table 4
Table 4:
Table 5
Table 5:
Characteristics of patients, urine sampling and culture
Table 5B
Table 5B:

Figures 1 and 2 show the 9 SROC curves constructed with the bivariate model. The bivariate SROC curves show that pyuria ≥10/hpf and bacteriuria any/hpf (P10 and B), and bacteriuria ≥10/hpf (B10) had the best diagnostic performances (Fig. 2). The remaining tests were of intermediate or low performance.

Fig. 1
Fig. 1:
Bivariate SROC curves.N, nitrite.
Fig. 2
Fig. 2:
Bivariate SROC curves.P5, pyuria ≥5.

After the final model of multivariate analysis was run, P10 and B remained the best combination, being better when urine was collected through SPA or catheterization (Fig. 3). This performance was irrespective of age group and whether urine was centrifuged or not. Conversely P10 was consistently the test with the lowest performance for all age groups, worsening its diagnostic usefulness when urine was nonsterile and improving somewhat when it was sterile, and also improving in infants. In the final multivariate model, centrifugation did reduce performance for all tests. For only P10 as well as P10 and B, sterile tests performed better than their nonsterile counterparts. Most tests (including the rapid dipstick tests LE and nitrite alone or in combination) showed a similar intermediate performance. The phrasing of these results takes into account the interaction terms in the regression model.

Fig. 3
Fig. 3:
Multivariate SROC curves obtained from final model.S&A, school children and adolescents;INF, infants;ALL, all age groups;STE, sterile urine;CEN, centrifuged urine;NOS, nonsterile urine;NOC, uncentrifuged urine.

In addition the score of methodologic quality was tested in a separate stepwise regression along with all the other individual variables. The score did not have a significant influence on the diagnostic performance of index tests.

Table 6 shows a summary of the final regression model. Because the terms left in the final model were so because of their statistical significance, we interpreted the regression output as a test for a joint hypothesis establishing that all the terms together reflect the universe, so that no term can be deleted. Given that, we then applied our value judgment to consider, on the basis of clinical experience, that the curve modeled for the test placed highest was also high enough to be considered having a clinically important advantage.

Table 6
Table 6:
Summary of final stepwise regression model Statistically significant results are shown. Adjusted R2 for the model is 0.317.


The metaanalysis by Gorelick and Shaw 8 concluded that the presence of any bacteria revealed by Gram stain on an uncentrifuged urine specimen and dipstick analysis for nitrite and LE in different combinations perform similarly in detecting UTI in children and are superior to microscopic analysis for pyuria.

We extended the previous report and included studies in English and Spanish from 1966 to January, 2001. Furthermore we found additional studies before 1966 through review of references of primary studies and of review articles. The multivariate approach that we performed allowed several advantages. All studies assessing the usefulness of microscopy for the presence of bacteria and/or leukocytes in centrifuged or uncentrifuged urine were included, regardless of their sample size. Likewise we included the potential influence of age group and sampling technique on the performance of the index tests, thus allowing testing of confounding and interaction effects.

P10 and B and only B10 showed the best diagnostic performance in the bivariate SROC curve; but after the multivariate analysis the first combination remained the best test irrespective of the age group involved, whereas B10 for all age groups was significantly affected in its diagnostic performance. Studies for B10 were in a rather narrow range of FPR, close to 0%; therefore model extrapolation might not be warranted.

Although we did not find additional studies assessing enhanced urine examination (microscopy for presence of bacteria and leukocytes in stained unspun urine), our multivariate approach allowed us to assess the potential usefulness of combination of bacteriuria and leukocytes through microscopic analysis in the diagnosis of UTI.

Primary studies showed heterogeneity for different variables that can explain part of the differences in results. The multivariate analysis showed that the technique of urine sampling and processing for centrifugation substantially influenced the performance of tests. It is understandable that urine obtained through SPA and catheterization as well as lack of centrifugation is related to better diagnostic performance. These techniques reduce the risk of contamination and most likely lead to higher true positive and lower false positive results. This point is particularly important in infants from 2 months through 2 years of age. Thus this metaanalysis reinforces the current recommendation of using one of these procedures for obtaining a reliable urine culture for this age group. 7 The combination of P10 and B in sterile urine remained the best test, whether urine was centrifuged or not and irrespective of age group involved.

There are some potential limitations to this metaanalysis. One of them is that the methodologic quality of primary studies was variable. We grouped together sets of tests with different cut points to make manageable the large amount of data, although it is unlikely that this strategy would have influenced the shape of the different SROC curves. We also grouped together studies in which bacteriuria was counted per high power field with those in which it was counted per oil immersion field and studies in which leukocytes were counted per high power field with those in which they were counted per μl. In addition we were not able to mak e a separate analysis of Gram-stained and not stained tests for bacteriuria, given that there were few stained samples reported in all the primary studies.

In contrast the widely different characteristics of primary studies and the subgroup analyses performed through the multivariate approach allow us to extend the conclusions of this metaanalysis to a wide spectrum of subjects and settings, from infants to adolescents, from asymptomatic to ill subjects and from primary level to referral level health facilities.

The place of combinations of rapid dipstick tests such as leukocyte esterase and nitrite in the diagnosis of UTI in children could not be definitively assessed, because the number of studies taking into account such combinations was small. We think, therefore, that additional studies addressing this specific point should be performed in the future.

We conclude that pyuria ≥10/hpf (or μl) and bacteriuria-any are best suited for assessing the risk of UTI in children.

We thank Mrs. Paquita Valero, who was most helpful in the literature search and retrieval.


1. Siegel S, Siegel B, Sokoloff BZ. Urinary infection in infants and preschool children: five year follow-up. Am J Dis Child 1980; 134: 369–72.
2. Hoberman A, Chao HP, Keller DM, Hickey R, Davis HW, Ellis D. Prevalence of urinary tract infection in febrile infants. J Pediatr 1993; 123: 17–23.
3. Winberg J. Progressive renal damage from infection with or without reflux. J Urol 1992; 148: 1733–4.
4. Winberg J, Bergstrom T, Jacobson B. Morbidity, age and sex distribution, recurrences and renal scarring in symptomatic urinary tract infection in childhood. Kidney Int 1975; 8: S101–6.
5. Ginsburg CM, McCracken GH. Urinary tract infection in young infants. Pediatrics 1982; 69: 409–12.
6. Hoberman A, Wald ER. Urinary tract infections in young febrile children. Pediatr Infect Dis J 1997; 16: 11–17.
7. American Academy of Pediatrics. Practice parameter: the diagnosis, treatment and evaluation of the initial urinary tract infection in febrile infants and young children. Pediatrics 1999; 103: 843–52.
8. Gorelick MH, Shaw KN. Screening tests for urinary tract infection: a meta-analysis. Pediatrics 1999; 104: e54.
9. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med 1993; 12: 1293–316.
10. Littenberg B, Moses LE. Estimating diagnostic accuracy from multiple conflicting reports: a new meta-analytic method. Med Decis Making 1993; 13: 313–21.
11. Irwig L, Tosteson ANA, Gatsonis C, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 1994; 120: 667–76.
12. Mulrow CD, Linn WD, Gaul MK, Pugh JA. Assessing quality of a diagnostic test evaluation. J Gen Intern Med 1989; 4: 288–95.
13. Sackett DL, Haynes RB, Tugwell P. Clinical epidemiology: a basic science for clinical medicine. 2nd ed. Boston: Little Brown, 1991.
14. Lin D, Huang F, Chiu N, et al. Comparison of hemocytometer leucocyte counts and standard urinalyses for predicting urinary tract infections in febrile infants. Pediatr Infect Dis J 2000; 19: 223–7.
15. Bulloch B, Bausher J, Pomerantz W, Connors M, Mahabee-Gittens M, Dowd D. Can urine clarity exclude the diagnosis of urinary tract infection? Pediatrics 2000; 106: e60.
16. Lin D, Huang S, Lin CH, et al. Urinary tract infection in febrile infants younger than eight weeks of age. Pediatrics 2000; 105: e20.
17. Waisman Y, Zerem E, Amir L, Mimouni M. The validity of the Uriscreen test for early detection of urinary tract infection in children. Pediatrics 1999; 104: e41.
18. Shaw KN, McGowan KL, Gorelick MH, Schwartz JS. Screening for urinary tract infection in the emergency department: which test is best? Pediatrics 1998; 101: e1.
19. Craver RD, Abermanis JG. Dipstick only urinalysis screen for the pediatric emergency room. Pediatr Nephrol 1997; 11: 331–3.
20. Shaw KN, McGowan KL. Evaluation of a rapid screening filter test for urinary tract infection in children. Pediatr Infect Dis J 1997; 16: 283–7.
21. Hoberman A, Wald ER, Reynolds EA, Penchansky L, Charron M. Is urine culture necessary to rule out urinary tract infection in young febrile children? Pediatr Infect Dis J 1996; 15: 304–9.
22. Lockhart GR, Lewander WJ, Cimini DM, Josephson SL, Linakis JG. Use of urinary gram stain for detection of urinary tract infection in infants. Ann Emerg Med 1995; 25: 31–5.
23. Reed RP, Wegerhoff FO. Urinary tract infection in malnourished rural African children. Ann Trop Paediatr 1995; 15: 21–6.
24. Huicho L, Dávila M, Zerpa R. Infección Urinaria: Estudio prospectivo del valor diagnóstico de pruebas de tamizaje en niños pre-escolares hospitalizados. An Fac Med UNMSM 1995; 56: 24–9.
25. Hiraoka M, Hida Y, Hori C, Tsuchida S, Kuroda M, Sudo M. Urine microscopy on a counting chamber for diagnosis of urinary infection. Acta Paediatr Jpn 1995; 37: 27–30.
26. Hiraoka M, Hida Y, Hori C, Tuchida S, Kuroda M, Sudo M. Rapid dipstick test for diagnosis of urinary tract infection. Acta Paediatr Jpn 1994; 36: 379–82.
27. Hoberman A, Wald ER, Reynolds EA, Penchansky L, Charron M. Pyuria and bacteriuria in urine specimens obtained by catheter from young children with fever. J Pediatr 1994; 24: 513–9.
28. Villanustre Ordoñez C, Buznego R, Rodicio M, et al. Estudio comparativo de los métodos semicuantitativos (leucocituria, test de nitritos y Uricult) con el urocultivo para el diagnóstico de infección urinaria en el lactante. An Esp Pediatr 1994; 41: 325–8.
29. Hoberman A, Wald ER, Penchansky L, Reynolds EA, Young S. Enhanced urinalysis as a screening test for urinary tract infection. Pediatrics 1993; 91: 1196–9.
30. Hoberman A, Han-Pu C, Keller DM, Hickey R, Davis HW, Ellis D. Prevalence of urinary tract infection in febrile infants. J Pediatr 1993; 123: 17–23.
31. Pérula de Torres LA, De Borja F, Martínez J, et al. The validation of a rapid diagnostic method for urinary infectiom in the school-age population. Rev Clin Esp 1993; 192: 209–13.
32. Lohr JA, Portilla MG, Geuder TG, Dunn ML, Dudley SM. Making a presumptive diagnosis of urinary tract infection by using a urinalysis performed in an on-site laboratory. J Pediatr 1993; 122: 22–25.
33. Shaw KN, Hexter D, McGowan KL, Schwartz JS. Clinical evaluation of a rapid screening test for urinary tract infection in children. J Pediatr 1991; 118: 733–6.
34. Vickers D, Ahmad T, Coulthard MG. Diagnosis of urinary tract infection in children: fresh urine microscopy or culture? Lancet 1991; 338: 767–70.
35. Lejeune B, Baron R, Guillois B, Mayeux D. Evaluation of a screening test for detecting urinary tract infection in newborns and infants. J Clin Patholol 1991; 44: 1029–30.
36. Weinberg AG, Gan VN. Urine screen for bacteriuria in symptomatic pediatric outpatients. Pediatr Infect Dis J 1991; 10: 651–4.
37. Crain EF, Gershel JC. Urinary tract infections in febrile infants younger than 8 weeks of age. Pediatrics 1990; 86: 363–7.
38. Goldsmith BM, Campos JM. Comparison of urinary dipstick, microscopy, and culture for the detection of bacteriuria in children. Clin Pediatr 1990; 29: 214–8.
39. Tahirovic H, Pasic M. A modified nitrite test as a screening test for significant bacteriuria. Eur J Pediatr 1988; 147: 632–3.
40. Wiggelinkhuizen J, Maytham D, Hanslo DH. Dipstick screening for urinary tract infection. S Afr Med J 1988; 74: 224–8.
41. Powell HR, McCredie DA, Ritchie MA. Urinary nitrite in symptomatic and asymptomatic urinary infection. Arch Dis Child 1987; 62: 138–40.
42. Marsik FJ, Owens D, Lewandowski J. Use of leucocyte esterase and nitrite tests to determine the need for culturing urine specimens from a pediatric and adolescent population. Diagn Microbiol Infect Dis 1986; 4: 181–3.
43. Cannon HJ, Goetz ES, Hamoudi AC, Marcon MJ. Rapid screening and microbiologic processing of pediatric urine specimens. Diagn Microbiol Infect Dis 1986; 4: 11–17.
44. Boreland PC, Stoker M. Dipstick analysis for screening of paediatric urine. J Clin Pathol 1986; 39: 1360–2.
45. Goosens H, De Mol P, Hall M, Butzer JP. Prevalence of asymptomatic bacteriuria and comparison betwen different screening methods for its detection in infants. Eur J Epidemiol 1985; 1: 301–4.
46. Corman LI, Foshee WS, Kotchmar GS, Harbison RW. Simplified urinary microscopy to detect significant bacteriuria. Pediatrics 1982; 70: 133–5.
47. Pylkkanen J, Vilska J, Koskimies O. Diagnostic value of symptoms and clean-voided urine specimen in childhood urinary tract infection. Acta Paediatr Scand 1979; 68: 341–4.
48. Sinaniotis CA, Haratsaris MN, Papadatos CJ. Nitrite indicator strip test for bacteriuria. Lancet 1978; 1: 776–7.
49. Litlewood JM, Jacobs SI, Ramsden CH. Comparison between microscopical examination of unstained deposits of urine and quantitative culture. Arch Dis Child 1977; 52: 894–6.
50. Kunin CM, De Groot JE. Sensitivity of a nitrite indicator strip method in detecting bacteriuria in preschool girls. Pediatrics 1977; 60: 244–5.
51. Gillenwater JY, Gleason ChH, Lohr JA, Marion D. Home urine cultures by the dip-strip method. Pediatrics 1976; 58: 508–12.
52. Robins DG, White RHR, Rogers KB, Osman MS. Urine microscopy as an aid to detection of bacteriuria. Lancet 1975; 1: 476–9.
53. Randolph MF, Morris K. Instant screening for bacteriuria in children: analysis of a dipstick. J Pediatr 1974; 4: 246–8.
54. Aronson AS, Gustafson B, Svenningsen NW. Combined suprapubic aspiration and clean-voided urine examination in infants and children. Acta Paediatr Scand 1973; 62: 396–400.
55. Freyre EA, Rondón O, Bedoya J, Llerena M, Tamayo M. The incidence of bacteriuria and pyuria in Peruvian children with malnutrition. Pediatrics 1973; 83: 57–61.
56. Mond NC, Gruneberg RN, Smellie JM. Study of childhood urinary tract infection in general practice. Br Med J 1970; 1: 602–5.
57. Lam CN, Bremmer AD, Maxwell JD, Murphy AV, Low WJ. Pyuria and bacteriuria. Arch Dis Child 1967; 42: 275–80.
58. Pryles CV, Elton CR. Pyuria and bacteriuria in infants and children. Am J Dis Child 1965; 110: 628–35.
59. Houston IB. Pus cell and bacterial counts in the diagnosis of urinary tract infection in childhood. Arch Dis Child 1963; 38: 600–5.
60. Stansfeld JM. The measurement and meaning of pyuria. Arch Dis Child 1962; 37: 257–62.
61. Kunin CM. The quantitative significance of bacteria visualized in the unstained urinary sediment. N Engl J Med 1961; 265: 589–90.

Metaanalysis; urinary tract infection; children; diagnosis; urinalysis; screening tests

© 2002 Lippincott Williams & Wilkins, Inc.