- Email: [email protected]
Development of a score based on urinalysis to improve the management of urinary tract infection in children
Clinica Chimica Acta 413 (2012) 478–482
Contents lists available at SciVerse ScienceDirect
Clinica Chimica Acta journal homepage: www.elsevier.com/locate/clinchim
Development of a score based on urinalysis to improve the management of urinary tract infection in children Rosa Luciano a, Simone Piga b, Leonardo Federico a, Marta Argentieri c, Francesca Fina a, Marina Cuttini b, Emilia Misirocchi a, Francesco Emma d, Maurizio Muraca a,⁎ a
Department of Laboratory Medicine, “Bambino Gesù” Children's Hospital, Rome, Italy Unit of Epidemiology, “Bambino Gesù” Children's Hospital, Rome, Italy Department of Microbiology, “Bambino Gesù” Children's Hospital, Rome, Italy d Department of Nephrology and Urology, “Bambino Gesù” Children's Hospital, Rome, Italy b c
a r t i c l e
i n f o
Article history: Received 30 March 2011 Received in revised form 4 November 2011 Accepted 6 November 2011 Available online 19 November 2011 Keywords: Urinary tract infection Urinalysis Urinary sediment
a b s t r a c t Background: The need for reducing unnecessary antibiotic treatment is being emphasized in the management of urinary tract infections (UTI), a disease frequent in childhood. An ideal test should provide early diagnosis without the waiting times of urine culture, but even a simple test of exclusion could significantly improve patient management. Methods: We evaluated the sensitivity, specificity, negative and positive predictive value of automated microscopy IRIS iQ200 combined with the dipstick analyses in children with suspected UTI. Multivariable logistic regression analysis was used to identify the set of variables that best predict positive culture results and develop a numerical risk score. Results: Of 474 consecutive urine samples retrospectively analyzed, 69 were positive at urine culture with prevalence of infection of 14.6%. Parameters significantly associated with the presence of infection in multivariable analysis were age b 1 year (p b 0.001), leukocyte esterase ≥ 15 × 10^6/L (p b 0.001), number of small particles (ASP) ≥ 5500 × 10^6/L (p b 0.001) and bacteria ≥ 3 × 10^6/L (p = 0.01). The derived score ranged from 0 to 10, with higher values indicating higher risk of UTI. The area under the score ROC curve was 79% (95% CI 0.72–0.85), and was better than those of the individual urinary chemical and microscopic analyses. Conclusions: This routine method could improve the management of UTI in children by early identifying patients with low probability of infection, for whom antibiotic treatment can be withheld until the results of urine culture become available. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Although the diagnosis and treatment of urinary tract infections (UTI) are relatively straightforward, a large proportion of the population may be unnecessarily exposed to antibiotics while waiting for the results of urine culture, with implications for costs and antibiotic resistance [1]. In infants and young children, additional issues include the severity and/or lack of specificity of symptoms, the higher risk of side effects of antibiotics, and the possibility of kidney damage in case of inappropriate treatment [2]. Urine culture is the gold standard in UTI diagnosis, but at least 24–48 h is required before results become available. Thus, antibiotics may be started in a symptomatic patient on the basis of clinical suspi-
⁎ Corresponding author at: IRCCS Ospedale Pediatrico Bambino Gesù, Piazza Sant'Onofrio, 4, 00165 Roma, Italy. Tel.: +39 06 6859 2210; fax: +39 06 6859 2014/2803. E-mail address: [email protected] (M. Muraca). 0009-8981/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.cca.2011.11.005
cion and chemical and urinary microscopic analysis, to be later withdrawn if the results of culture do not confirm the presence of UTI. The standardization of urinary sediment analysis is improved by the use of automated devices, that allow for a precise determination of the absolute number of particles per volume unit [3,4]. Reproducibility and especially efficiency of analysis are enhanced, raising attention towards the use of this method in the diagnosis of UTI and other urinary tract and renal pathology. This study explored the discriminative power and overall performance of a combination of urine chemical analysis and automatized sediment reading by IRIS IQ200 to detect UTI in the pediatric population of a tertiary children's hospital in Italy. Our aim was to identify patients for whom antibiotic treatment could be safely withheld while waiting for the results of urine culture, that in our hospital is always performed in the presence of suggestive symptoms. To this end, we aimed at identifying the combination of parameters able to maximize sensitivity while, limiting the loss of specificity and the proportion of false positive results.
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
479
2. Materials and methods
2.3. Urine culture
The study was carried out at the “Bambino Gesù” Children's Hospital, a tertiary pediatric hospital and clinical research institute in Roma, Italy. We retrospectively analyzed the data of 852 consecutive pediatric patients (≤14 years of age) assessed for UTI in three periods: March–September 2006, January–April 2007, and November– December 2007. Both hospitalized patients, and children referred to the outpatient clinic of the Laboratory Unit were included. Fresh urine samples were collected and treated according to the European Urinalysis Guidelines [5]. In children that could control voiding, urine samples were collected at mid-stream after thoroughly washing the area around the urethra. A urine collection bag was used in infants, and samples were considered adequate for urine culture only if voiding occurred within 30 min. In our hospital, suprapubic aspiration is discouraged and clean catheter samples are not routinely used. For each sample urine culture, automatized chemical (dipstick) analysis and sediment analysis were performed. The study was approved by the Ethics Committee of the hospital.
Urine culture was performed by inoculating the specimens in a combination of solid media, including Columbia CNA agar with 5% sheep blood, MacConkey agar for the detection of most Gramnegative bacilli, staphylococci, streptococci and enterococci, and Sabouraud dextrose agar for yeast. Before inoculation, the urine sample was thoroughly mixed; then using a calibrated 10 μL loop, the urine was placed at the center of the plate, from where the inoculum was spread in a line across the diameter of the plate. Once inoculated, the plates were streaked to obtain isolated colonies. In most cases, incubation for a minimum of 24 h at 35–37 °C was necessary to detect uropathogens. Urine culture was labeled as positive for UTI when a single organism was detected at a concentration of at least 10^8 CFU/L. The presence of antibiotic in the urine was assessed through the PAR (Residual Antibiotic Power) test.
2.1. Chemical urine analysis
We excluded samples with evidence of contamination (n. 50), and those with possible previous antibiotic treatment as determined by a positive (n. 270) or missing (n. 55) PAR-test. Three additional samples were excluded because urine culture was missing. Thus, data analysis was carried out on a total of 474 independent urine samples. The Chi-squared test was used to test the association of UTI with patients' age and gender. For the purpose of this study, we considered as predictors of UTI the urine analysis variables with a biologically meaningful relationship to UTI: leukocyte esterase and presence of nitrites from the chemical analysis, and concentration of bacteria, leukocytes, red blood cells and ASP from the sediment reading. Using the results of urine culture as reference standard, we computed for each variable the ReceiverOperating Characteristic (ROC) curve [6], and the full range of sensitivity, specificity, positive and negative predictive values, together with 95% confidence intervals (CI). The cut-offs to dichotomize the variables were selected a priori on the basis of reference values used in our Laboratory; a sensitivity analysis was also performed using as cut-offs the best sensitivity–specificity trade-off values obtained from ROC curves [6]. Multivariable logistic regression analysis was used to identify the combination of variables which best predicted the presence of a positive urine culture. Patient's age (below 1 year or older) and gender were also included in the model. The final model retained all the variables significantly associated with the presence of UTI at a p b 0.05 level. The Hosmer–Lemenshow test was used to measure the model performance [7]. A numerical score was developed from the multivariable model by approximately doubling the value of each β coefficient. The ROC curve was computed for the score, as well as the sensitivity, specificity, predictive values and 95% confidence intervals corresponding to different cut-offs. Statistical analyses were conducted using the Stata statistical software (StataCorp, Release 10.0; College Station, TX).
Dipstick analysis was performed with the Bayer Multistix 10SG (Siemens, Germany) on a Bayer Clinitek Atlas automated analyzer (Siemens, Germany). The following parameters were recorded: ketones, urobilinogen, creatinine, glucose, leukocyte esterase (LE), nitrite, protein, blood, pH, bilirubin, color, and protein. Calibration was performed with 1 or 2 points for all reagents, according to the manufacturer instructions. LE was calibrated to indicate trace at ≥15 leukocytes × 10^6/L, 1 + at ≥70 × 10^6/L, 2 + at ≥125 × 10^6/L, and 3+ at ≥500 × 10^6/L. 2.2. Automated urine microscopy sediment analysis We used the IRIS IQ200® analyzer developed by Iris Diagnostic Inc. (Chatsworth, California, USA). This is composed of the iQ200 Automated Urine Microscopy Analyzer, connected physically and electronically to the PCTK2000 management software which allows the complete integration of the systems responsible for the execution of urine sediment analysis, and of a computer workstation for reporting specimen composition results. After auto-classification, all abnormal result images are reviewed and edited by a dedicated trained technician. The iQ Lamina used to create the laminar flow is a particulate free isotonic buffer containing proprietary laminar flow stabilizers, bacteriostatic agents, fungicidal agents and preservatives. The day-to-day imprecision is assessed with iQ Negative and iQ Positive Control supplied by the manufacturer. The iQ200 software includes quality control charts for negative and positive control suspensions that allow to control counting accuracy over time Calibrator, Focus and Positive Control are suspensions of fixed human red cells in a particle free buffer. During data collection for this study, the coefficient of variation (CV) of positive control was 3.75% (mean 985.61 × 10^6/L). Negative Control was a particle free buffer solution. The CV of negative control was 65.71% (mean 3.15 × 10^6/L). The following solutions and materials were provided by Iris Diagnostics Inc. for the evaluation: Iris iQ Calibrator (REF 475-0059), iQ Negative Control (REF 475-0058), iQ Positive Control (REF 4750046), iQ Focus (REF 475-0060), iQ Lamina (REF 475-0047), Iris System Cleanser (REF 475-0003) and Iris Diluent (REF 475-0021). Sediment analysis by IRIS IQ200 allows for a quantitative reporting of the concentration of red and white blood cells, and of unclassified “all small particles” (ASP), that may include also bacterial elements. Bacteriuria is recorded using an ordinal scale ranging from 0 to 4. Specifically, score coding is as follows: 0 (absent), 1+ (1 or 2 bacteria × 10^6/L), 2 + (3–4 × 10^6/L), 3 + (5–9 × 10^6/L and 4 + (10 or more bacteria × 10^6/L).
2.4. Statistical analysis
3. Results Urine culture was positive in 69 patients, with an overall UTI prevalence of 14.6%. Infection was significantly more frequent in children below 1 year of age (37.0%), while no difference according to sex was found (Table 1). The discriminative and predicting power of the each individual urinalysis variable is shown in Table 2. All variables had high specificity (range from 87.9 to 100%) and negative predictive values (range from 87.9 to 91.1%). However, sensitivity values were considerably lower, ranging from 10.1% for red blood cells >18 × 10^6/L to 49.3% for LE ≥15 × 10^6/L. When LE is combined with nitrites in parallel
480
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
Table 1 Prevalence of UTI by patients'age and gender (n = 474). Urine culture
p-value
Negative (n = 405)
Age (years) b1 1–4 5–9 10–14 Gender Male Female
Positive (n = 69)
n
(%)
n
(%)
34 119 163 89
(63.0) (86.9) (87.6) (91.8)
20 18 23 8
(37.0) (13.1) (12.4) (8.3)
229 176
(86.1) (84.6)
37 32
(13.9) (15.4)
b 0.001
0.651
(i.e. LE and/or nitrite positive), sensitivity increased to 53.6%, without much loss of specificity. Table 3 shows the results of the multivariable logistic model, and the attributed values for the calculation of the total numerical score. Age younger than 12 months and urine LE ≥15 × 10^6/L increased almost five-fold the probability of UTI (adjusted ORs 4.6 and 4.7 respectively, p b 0.001). Additional variables significantly associated with the outcome were ASP ≥5500 × 10^6/L (aOR 4.0, p b 0.001) and bacteria ≥3 × 10^6/L (aOR 3.7, p b 0.01). In contrast, patient sex, presence of nitrites and sediment concentration of leukocytes and red blood cells did not significantly add to the model. The Hosmer–Lemenshow goodness-of-fit test was satisfactory (Chi squared with 3 d.f. 1.83, p = 0.61). The numerical risk score derived from the model ranged from zero (no predictor present) to 10 (all predictors present), with higher values indicating higher risk of UTI as measured by urine
culture. In the total sample, the mean risk score was 1.3 (standard deviation 2.0); however, the value was raised to 3.7 (standard deviation 2.9) among the infected children. Fig. 1 shows the ROC curve for the risk score compared to the curves for selected individual urine sediment variables and combined LE and nitrite test. The risk score had the better AUC (0.79, 95% CI 0.72–0.85), followed by the combined test (0.73). By selecting different risk score cut-offs, the discriminative and predictive values of the risk score as a test for urinary infection change (Table 4). The lower the cut-off, the higher are not only test sensitivity and negative predictive value, but also the likelihood of false positive results. With a cut-off of 2, sensitivity is 73.9% and specificity is 72.1%; 51 of the 69 patients with UTI would be identified as positives and immediately treated, while for 18 “false negatives” treatment would be delayed. However, 292 children (true negatives) would be appropriately spared antibiotics. Increasing the cut-off to 4, sensitivity drops to 49.3%, but specificity increases to 96.3%. Only 15 children would be inappropriately treated (false positives); however, as many as 35 infected patients (false negatives) would receive antibiotics only after urine culture results. When the best sensitivity–specificity trade-off values from the ROC curves were used as cut-offs for individual tests, the results of the multivariable analysis were very similar, and the ROC curve for the risk score derived from this model was almost superposable (AUC 0.79, 95% CI 0.75–0.83). 4. Discussion UTIs are the most common bacterial infections in infants and young children, and can be associated with high fever and poor general conditions [8]. The infection may involve the kidney, and in a
Table 2 Discriminative and predictive values for individual results of urinalysis. Variable
Dipstick urinalysis LE ≥15 × 10^6/L Presence of nitrites LE ≥15 × 10^6/L or presence of nitrites
Microscope IRIS IQ200® Bacteria ≥3 × 10^6/L ASP ≥5500 × 10^6/L Leukocytes >24 × 10^6/L Red blood cells >18 × 10^6/L
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
True positivesa
False positivesa
True negativesa
False negativesa
%
%
%
%
n.
n.
n.
n.
49.3 (44.8–53.8) 18.8 (15.3–22.4) 53.6 (49.1–58.1)
87.9 (85.0–90.8) 100 (100–100) 87.9 (85.0–90.8)
41.0 (36.5–45.4) 100 (100–100) 43.0 (38.6–47.5)
91.1 (88.5–93.6) 87.9 (84.9–90.8) 91.8 (89.3–94.2)
34
49
356
35
13
0
405
56
37
49
356
32
24.6 (20.8–28.5) 42.0 (37.6–46.5) 30.4 (26.3–34.6) 10.1 (7.4–12.9)
97.3 (95.8–98.8) 90.4 (87.7–93.0) 94.6 (92.5–96.6) 96.3 (94.6–98.0)
60.7 (56.3–65.1) 42.7 (38.2–47.1) 48.8 (44.3–53.3) 31.8 (27.6–36.0)
88.3 (85.5–91.2) 90.2 (87.5–92.8) 88.9 (86.0–91.7) 86.3 (83.2–89.4)
17
11
394
52
29
39
366
40
21
22
383
48
7
15
390
62
CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; LE, leukocyte esterase; ASP, all small particles. a Number of patients by classification category.
Table 3 Predictors of UTI: results of multivariable logistic regression model and attributed values for score calculation. Predictor
β coefficient
95% CI
aOR
95% CI
p-value
Attributed values for score
Leukocyte esterase ≥15 × 10^6/L Age (y) b 1 ASP ≥5500 × 10^6/L Bacteria ≥3 × 10^6/L
1.6 1.5 1.4 1.3
0.9–2.2 0.8–2.3 0.7–2.1 0.3–2.3
4.7 4.6 4.0 3.7
2.5–8.9 2.2–9.6 1.9–8.3 1.4–10.0
b0.001 b0.001 b0.001 0.010
3 3 2 2
CI, confidence interval; aOR, odds ratio adjusted for the variables included in the model; ASP, all small particles.
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
481
analysis for LE and nitrite are superior to microscopic detection of pyuria in children. A meta-analysis and a cost-effectiveness study suggested that urine dipstick negative for both LE and nitrite may be used for ruling out infections [14,15]. However, Laine et al. [16] found that in hospital care setting urinalysis by dipstick alone carries the risk of missing infections, while other authors proposed that pyuria, as measured by at least 10 × 10^6 Leuk/L on unspun urine, and any bacteriuria are best suited to identify cases where urine culture is warranted [15,17]. Recently, a new meta-analysis [11] showed that microscopy examination of urine for Gram stained bacteria is the single best rapid test, with 91 sensitivity and 96% specificity, while “either leukocyte esterase or nitrite” positivity can be rated as the best dipstick test. Most of the studies quoted (65 out of 95) used the 10^8 CFU/L cutoff for voided urine culture. However, 23 used a different value, and in 7 studies the criterion was unclear [11]. Our study confirmed that the dipstick LE or nitrite test has indeed the best performance, with a sensitivity of 56.3% and specificity 87.9%. However, the risk score developed by combining both dipstick and sediment reading results through multivariable analysis allowed us to obtain an additional improvement of the area under the ROC curve. Through the selection of different cut-off criteria, the score can be adapted to different clinical situations depending on the relative benefits of maximizing sensitivity or specificity. The selection of a very low cut-off, basically corresponding to any of the variables included in the final model being positive, leads to a 74% sensitivity and 94% negative predictive value, allowing to withhold treatment in children scored negative pending culture results. In our series of 474 children with suspected UTI, such a procedure would avoid unnecessary antibiotic administration in 292 children (72% of those without UTI); 113 however would still receive two days of treatment. A higher cut-off (e.g. ≥4) leads to 96% specificity, increasing the positive predictive value to 69% without large loss of NPV (92%). The use of this strategy would avoid unnecessary treatment in as many as 390 cases, about 96% of non infected children; however 35 patients with UTI would start antibiotics only after culture results are available. Ultimately, the selection of the best strategy will depend on the setting and on the individual characteristics and clinical conditions of the patients. Clearly, the main issue concerns balancing the harm and costs of unnecessarily starting antibiotics with the risk of twodays delayed treatment in case of UTI. Symptoms of acute UTI may be very disturbing in children, and are promptly relieved by treatment. Early and appropriate treatment is generally recommended in
Fig. 1. Receiver-Operating Characteristic (ROC) curves for the risk score and selected dipstick and urine sediment tests. (The 45° line through the origin represents the ROC curve of a test whose decision ability is no better than chance).
small proportion of patients even lead to permanent renal damage and scarring [9,10]. Culture of urine is the reference standard for the diagnosis of UTI [8], but continuing efforts are spent in searching for alternative, more rapid and still accurate tests [11]. This is especially important in children, since both a delay in treatment and the inappropriate administration of antibiotics can result in potentially serious adverse effects. Diagnostic studies in the pediatric population have however produced inconsistent results [11,12]. According to Gorelick and Shaw [13], both a Gram stain of uncentrifuged urine specimen and dipstick Table 4 Discriminative and predictive value of the clinical score by cut-off value. Score cut-off
≥2 ≥3 ≥4 ≥5 ≥6 ≥7 ≥8 10
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
%
%
73.9 (70.0-77.9) 66.7 (62.4–70.9) 49.3 (44.8–53.8) 42.0 (37.6–46.5) 26.1 (22.1–30.0) 15.9 (12.7–19.2) 8.7 (6.2–11.2) 5.8 (3.7–7.9)
72.1 (68.1–76.1) 79.5 (75.9–83.1) 96.3 (94.6–98.0) 97.0 (95.5–98.6) 98.3 (97.1–99.4) 98.8 (97.8–99.8) 99.8 (99.3–100.0) 100.0 (100.0–100.0)
PPV, positive predictive value; NPV, negative predictive value. a Number of patients by classification category.
NPV (95% CI)
True positivesa
%
%
31.1 (26.9–35.3) 35.7 (31.4–40.0) 69.4 (65.2–73.5) 70.7 (66.6–74.8) 72.0 (68.0–76.0) 68.8 (64.6–72.9) 85.7 (82.6–88.9) 100.0 (100.0–100.0)
94.2 (92.1–96.3) 93.3 (91.1–95.6) 91.8 (89.3–94.2) 90.8 (88.2–93.4) 88.6 (85.8–91.5) 87.3 (84.3–90.3) 86.5 (83.4–89.6) 86.2 (83.1–89.3)
False positivesa
True negativesa
False negativesa
n.
n.
n.
n.
51
113
292
18
46
83
322
23
34
15
390
35
29
12
393
40
18
7
398
51
11
5
400
58
6
1
404
63
4
0
405
65
482
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
infants and young children to reduce the likelihood of renal involvement during the acute phase of infection [18] and subsequent renal scarring [19]. However, recent data from two multicenter randomized controlled studies showed that the probability of developing renal scars in case of upper tract UTI is not related to the promptness of antibiotic treatment [20,21]. Although this finding derives from observational data on a subgroup of patients, and confirmation by other studies is warranted [22,23], it does suggest that, except for infants, treatment can probably be safely delayed for two or three days, until a positive culture is obtained. This study has limitations. It was carried out retrospectively, using the database of the laboratory medicine department. The full available sample was used for the development of the score, that could not be subsequently validated on a different data set. Information on the clinical conditions of the children was not available, and the final diagnosis was not compared with the medical records. Data on the indications for performing a urine culture and on the severity and duration of symptoms may have increased the predictive value of our multivariable model. We used a urine culture count ≥10^8 CFU/L to indicate UTI, as this is the criterion adopted by our hospital and in most studies [11]. Data from the 1950s have shown that this cut-off has a good sensitivity, but may lead to include false negative patients; likewise, more recent data highlight the fact that, in some patients, even lower bacterial count may be clinically significant [11,24]. The prevalence of UTI in the present study was slightly higher than the previously reported values, ranging from 7.1 to 9.1% [25], probably because of the tertiary pediatric care setting. From a methodological stand-point, moderate differences in the prevalence of UTI in the studied populations are unlikely to affect sensitivity and specificity [8]. The sensitivity, specificity, NPV and PPV obtained in our cohort were similar to those previously reported [26]. We used the automated urine microscopy analyzer IRIS200, already previously validated as a tool for the study of UTI [27–29]. The combination of this tool with automated dipstick analysis provided an excellent NPV for excluding UTI, safely supporting a clinician's decision to withhold treatment pending culture results, provided that the patient can be appropriately monitored. Further research should prospectively validate the risk score on a new set of data, and in patients with known characteristics and disease severity. References [1] Mangin D. Urinary tract infection in primary care. BMJ 2010;340:c657. [2] Bhat RG, Katy TA, Place FC. Pediatric urinary tract infections. Emerg Med Clin N Am 2011;29:637–53. [3] Hughes C, Roebuck MJ. Evaluation of the IRIS 939 UDx flow microscope as a screening system for urinary tract infection. J Clin Pathol 2003;56:844–9. [4] Van Den Broek D, Keularts IM, Wielders JP, Kraaijenhagen RJ. Benefits of the iQ200 automated urine microscopy analyser in routine urinalysis. Clin Chem Lab Med 2008;46:1635–40.
[5] European urinalysis guidelines. Scand J Clin Lab Invest Suppl 2000;231:1–86. [6] Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem 1993;39:561–77. [7] Hosmer DW, Lemeshow S. Applied logistic regression. New York NY: John Wiley & Sons; 2000. [8] Zorc JJ, Kiddoo DA, Shaw KN. Diagnosis and management of pediatric urinary tract infections. Clin Microbiol Rev 2005;18:417–22. [9] Hellerstein S. Long-term consequences of urinary tract infections. Curr Opin Pediatr 2000;12:125–8. [10] Montini G, Tullus K, Hewitt I. Febrile urinary tract infections in children. NEJM 2011;365:239–50. [11] Williams GJ, Macaskill P, Chan SF, Turner RM, Hodson E, Craig JC. Absolute and relative accuracy of rapid urine tests for urinary tract infection in children: a metaanalysis. Lancet Infect Dis 2010;10:240–50. [12] Wald E. Urinary tract infections in infants and children: a comprehensive overview. Curr Opin Pediatr 2004;16:85–8. [13] Gorelick MH, Shaw KN. Screening tests for urinary tract infection: a metaanalysis. Pediatric 1999;104:e54. [14] Devillé WL, Yzermans JC, van Duijn NP, Bezemer PD, van der Windt DA, Bouter LM. The urine dipstick test useful to rule out infections. A meta-analysis of the accuracy. BMC Urol 2004;2 4:4. [15] Whiting P, Westwood M, Bojke L, et al. Clinical effectiveness and costeffectiveness of tests for the diagnosis and investigation of urinary tract infection in children: a systematic review and economic model. Health Technol Assess 2006;10:1–154. [16] Laine P, Toivonen E, Eklund K, Hohenthal U, Siren S, Maki T. Rapid dipstick urinalysis in the internal medicine clinic: what is missed? J Intern Med 1997;242:271–3. [17] Huicho L, Campos-Sanchez M, Alamo C. Metaanalysis of urine screening tests for determining the risk of urinary tract infection in children. Pediatr Infect Dis J 2002;21:1–11. [18] Doganis D, Siafas K, Mavrikou M, et al. Does early treatment of urinary tract infection prevent renal damage? Pediatrics 2007;120:e922–8. [19] 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. [20] Hewitt IK, Zucchetta P, Rigon L, et al. Early treatment of acute pyelonephritis in children fails to reduce renal scarring: data from the Italian Renal Infection Study Trials. Pediatrics 2008;122:486–90. [21] Hewitt IK, Montini G. In reply. Pediatrics 2009;123:e174. [22] Edmonson MB, Wald ER. Treatment of pyelonephritis and risk of renal scarring. Pediatrics 2009;123:e545. [23] Fernández-Menèndez JM, Pérez-Méndez C. Early antibiotic treatment of pyelonephritis in children is still mandatory. Pediatrics 2009;123:e173. [24] Tullus K. Difficulties in diagnosing urinary tract infections in small children. Pediatr Nephrol 2011;26:1923–6. [25] Shaikh N, Morone NE, Bost JE, Farrell MH. Prevalence of urinary tract infection in childhood: a meta-analysis. Pediatr Infect Dis J 2008;27:302–8. [26] Shayanfar N, Tobler U, Eckardstein A, Bestmann L. Automated urinalysis: first experiences and a comparison between the Iris iQ200 urine microscopy system, the Sysmex UF-100 flow cytometer and manual microscopic particle counting. Clin Chem Lab Med 2007;45:1251–6. [27] Lamchiagdhase P, Preechaborisutkul K, Lomsomboon P, et al. Urine sediment examination: a comparison between the manual method and the IQ200 automated urine microscopy analyzer. Clin Chim Acta 2005;358:167–74. [28] Linko S, Kouri TT, Toivonen E, Ranta PH, Chapoulaud E, Lalla M. Analytical performance of the Iris iQ200 automated urine microscopy analyzer. Clin Chim Acta 2006;372:54–64. [29] Chien T-I, Kao J-T, Liu H-L, et al. Urine sediment examination: a comparison of automated urinalysis systems and manual microscopy. Clin Chim Acta 2007;384: 28–34.
Contents lists available at SciVerse ScienceDirect
Clinica Chimica Acta journal homepage: www.elsevier.com/locate/clinchim
Development of a score based on urinalysis to improve the management of urinary tract infection in children Rosa Luciano a, Simone Piga b, Leonardo Federico a, Marta Argentieri c, Francesca Fina a, Marina Cuttini b, Emilia Misirocchi a, Francesco Emma d, Maurizio Muraca a,⁎ a
Department of Laboratory Medicine, “Bambino Gesù” Children's Hospital, Rome, Italy Unit of Epidemiology, “Bambino Gesù” Children's Hospital, Rome, Italy Department of Microbiology, “Bambino Gesù” Children's Hospital, Rome, Italy d Department of Nephrology and Urology, “Bambino Gesù” Children's Hospital, Rome, Italy b c
a r t i c l e
i n f o
Article history: Received 30 March 2011 Received in revised form 4 November 2011 Accepted 6 November 2011 Available online 19 November 2011 Keywords: Urinary tract infection Urinalysis Urinary sediment
a b s t r a c t Background: The need for reducing unnecessary antibiotic treatment is being emphasized in the management of urinary tract infections (UTI), a disease frequent in childhood. An ideal test should provide early diagnosis without the waiting times of urine culture, but even a simple test of exclusion could significantly improve patient management. Methods: We evaluated the sensitivity, specificity, negative and positive predictive value of automated microscopy IRIS iQ200 combined with the dipstick analyses in children with suspected UTI. Multivariable logistic regression analysis was used to identify the set of variables that best predict positive culture results and develop a numerical risk score. Results: Of 474 consecutive urine samples retrospectively analyzed, 69 were positive at urine culture with prevalence of infection of 14.6%. Parameters significantly associated with the presence of infection in multivariable analysis were age b 1 year (p b 0.001), leukocyte esterase ≥ 15 × 10^6/L (p b 0.001), number of small particles (ASP) ≥ 5500 × 10^6/L (p b 0.001) and bacteria ≥ 3 × 10^6/L (p = 0.01). The derived score ranged from 0 to 10, with higher values indicating higher risk of UTI. The area under the score ROC curve was 79% (95% CI 0.72–0.85), and was better than those of the individual urinary chemical and microscopic analyses. Conclusions: This routine method could improve the management of UTI in children by early identifying patients with low probability of infection, for whom antibiotic treatment can be withheld until the results of urine culture become available. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Although the diagnosis and treatment of urinary tract infections (UTI) are relatively straightforward, a large proportion of the population may be unnecessarily exposed to antibiotics while waiting for the results of urine culture, with implications for costs and antibiotic resistance [1]. In infants and young children, additional issues include the severity and/or lack of specificity of symptoms, the higher risk of side effects of antibiotics, and the possibility of kidney damage in case of inappropriate treatment [2]. Urine culture is the gold standard in UTI diagnosis, but at least 24–48 h is required before results become available. Thus, antibiotics may be started in a symptomatic patient on the basis of clinical suspi-
⁎ Corresponding author at: IRCCS Ospedale Pediatrico Bambino Gesù, Piazza Sant'Onofrio, 4, 00165 Roma, Italy. Tel.: +39 06 6859 2210; fax: +39 06 6859 2014/2803. E-mail address: [email protected] (M. Muraca). 0009-8981/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.cca.2011.11.005
cion and chemical and urinary microscopic analysis, to be later withdrawn if the results of culture do not confirm the presence of UTI. The standardization of urinary sediment analysis is improved by the use of automated devices, that allow for a precise determination of the absolute number of particles per volume unit [3,4]. Reproducibility and especially efficiency of analysis are enhanced, raising attention towards the use of this method in the diagnosis of UTI and other urinary tract and renal pathology. This study explored the discriminative power and overall performance of a combination of urine chemical analysis and automatized sediment reading by IRIS IQ200 to detect UTI in the pediatric population of a tertiary children's hospital in Italy. Our aim was to identify patients for whom antibiotic treatment could be safely withheld while waiting for the results of urine culture, that in our hospital is always performed in the presence of suggestive symptoms. To this end, we aimed at identifying the combination of parameters able to maximize sensitivity while, limiting the loss of specificity and the proportion of false positive results.
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
479
2. Materials and methods
2.3. Urine culture
The study was carried out at the “Bambino Gesù” Children's Hospital, a tertiary pediatric hospital and clinical research institute in Roma, Italy. We retrospectively analyzed the data of 852 consecutive pediatric patients (≤14 years of age) assessed for UTI in three periods: March–September 2006, January–April 2007, and November– December 2007. Both hospitalized patients, and children referred to the outpatient clinic of the Laboratory Unit were included. Fresh urine samples were collected and treated according to the European Urinalysis Guidelines [5]. In children that could control voiding, urine samples were collected at mid-stream after thoroughly washing the area around the urethra. A urine collection bag was used in infants, and samples were considered adequate for urine culture only if voiding occurred within 30 min. In our hospital, suprapubic aspiration is discouraged and clean catheter samples are not routinely used. For each sample urine culture, automatized chemical (dipstick) analysis and sediment analysis were performed. The study was approved by the Ethics Committee of the hospital.
Urine culture was performed by inoculating the specimens in a combination of solid media, including Columbia CNA agar with 5% sheep blood, MacConkey agar for the detection of most Gramnegative bacilli, staphylococci, streptococci and enterococci, and Sabouraud dextrose agar for yeast. Before inoculation, the urine sample was thoroughly mixed; then using a calibrated 10 μL loop, the urine was placed at the center of the plate, from where the inoculum was spread in a line across the diameter of the plate. Once inoculated, the plates were streaked to obtain isolated colonies. In most cases, incubation for a minimum of 24 h at 35–37 °C was necessary to detect uropathogens. Urine culture was labeled as positive for UTI when a single organism was detected at a concentration of at least 10^8 CFU/L. The presence of antibiotic in the urine was assessed through the PAR (Residual Antibiotic Power) test.
2.1. Chemical urine analysis
We excluded samples with evidence of contamination (n. 50), and those with possible previous antibiotic treatment as determined by a positive (n. 270) or missing (n. 55) PAR-test. Three additional samples were excluded because urine culture was missing. Thus, data analysis was carried out on a total of 474 independent urine samples. The Chi-squared test was used to test the association of UTI with patients' age and gender. For the purpose of this study, we considered as predictors of UTI the urine analysis variables with a biologically meaningful relationship to UTI: leukocyte esterase and presence of nitrites from the chemical analysis, and concentration of bacteria, leukocytes, red blood cells and ASP from the sediment reading. Using the results of urine culture as reference standard, we computed for each variable the ReceiverOperating Characteristic (ROC) curve [6], and the full range of sensitivity, specificity, positive and negative predictive values, together with 95% confidence intervals (CI). The cut-offs to dichotomize the variables were selected a priori on the basis of reference values used in our Laboratory; a sensitivity analysis was also performed using as cut-offs the best sensitivity–specificity trade-off values obtained from ROC curves [6]. Multivariable logistic regression analysis was used to identify the combination of variables which best predicted the presence of a positive urine culture. Patient's age (below 1 year or older) and gender were also included in the model. The final model retained all the variables significantly associated with the presence of UTI at a p b 0.05 level. The Hosmer–Lemenshow test was used to measure the model performance [7]. A numerical score was developed from the multivariable model by approximately doubling the value of each β coefficient. The ROC curve was computed for the score, as well as the sensitivity, specificity, predictive values and 95% confidence intervals corresponding to different cut-offs. Statistical analyses were conducted using the Stata statistical software (StataCorp, Release 10.0; College Station, TX).
Dipstick analysis was performed with the Bayer Multistix 10SG (Siemens, Germany) on a Bayer Clinitek Atlas automated analyzer (Siemens, Germany). The following parameters were recorded: ketones, urobilinogen, creatinine, glucose, leukocyte esterase (LE), nitrite, protein, blood, pH, bilirubin, color, and protein. Calibration was performed with 1 or 2 points for all reagents, according to the manufacturer instructions. LE was calibrated to indicate trace at ≥15 leukocytes × 10^6/L, 1 + at ≥70 × 10^6/L, 2 + at ≥125 × 10^6/L, and 3+ at ≥500 × 10^6/L. 2.2. Automated urine microscopy sediment analysis We used the IRIS IQ200® analyzer developed by Iris Diagnostic Inc. (Chatsworth, California, USA). This is composed of the iQ200 Automated Urine Microscopy Analyzer, connected physically and electronically to the PCTK2000 management software which allows the complete integration of the systems responsible for the execution of urine sediment analysis, and of a computer workstation for reporting specimen composition results. After auto-classification, all abnormal result images are reviewed and edited by a dedicated trained technician. The iQ Lamina used to create the laminar flow is a particulate free isotonic buffer containing proprietary laminar flow stabilizers, bacteriostatic agents, fungicidal agents and preservatives. The day-to-day imprecision is assessed with iQ Negative and iQ Positive Control supplied by the manufacturer. The iQ200 software includes quality control charts for negative and positive control suspensions that allow to control counting accuracy over time Calibrator, Focus and Positive Control are suspensions of fixed human red cells in a particle free buffer. During data collection for this study, the coefficient of variation (CV) of positive control was 3.75% (mean 985.61 × 10^6/L). Negative Control was a particle free buffer solution. The CV of negative control was 65.71% (mean 3.15 × 10^6/L). The following solutions and materials were provided by Iris Diagnostics Inc. for the evaluation: Iris iQ Calibrator (REF 475-0059), iQ Negative Control (REF 475-0058), iQ Positive Control (REF 4750046), iQ Focus (REF 475-0060), iQ Lamina (REF 475-0047), Iris System Cleanser (REF 475-0003) and Iris Diluent (REF 475-0021). Sediment analysis by IRIS IQ200 allows for a quantitative reporting of the concentration of red and white blood cells, and of unclassified “all small particles” (ASP), that may include also bacterial elements. Bacteriuria is recorded using an ordinal scale ranging from 0 to 4. Specifically, score coding is as follows: 0 (absent), 1+ (1 or 2 bacteria × 10^6/L), 2 + (3–4 × 10^6/L), 3 + (5–9 × 10^6/L and 4 + (10 or more bacteria × 10^6/L).
2.4. Statistical analysis
3. Results Urine culture was positive in 69 patients, with an overall UTI prevalence of 14.6%. Infection was significantly more frequent in children below 1 year of age (37.0%), while no difference according to sex was found (Table 1). The discriminative and predicting power of the each individual urinalysis variable is shown in Table 2. All variables had high specificity (range from 87.9 to 100%) and negative predictive values (range from 87.9 to 91.1%). However, sensitivity values were considerably lower, ranging from 10.1% for red blood cells >18 × 10^6/L to 49.3% for LE ≥15 × 10^6/L. When LE is combined with nitrites in parallel
480
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
Table 1 Prevalence of UTI by patients'age and gender (n = 474). Urine culture
p-value
Negative (n = 405)
Age (years) b1 1–4 5–9 10–14 Gender Male Female
Positive (n = 69)
n
(%)
n
(%)
34 119 163 89
(63.0) (86.9) (87.6) (91.8)
20 18 23 8
(37.0) (13.1) (12.4) (8.3)
229 176
(86.1) (84.6)
37 32
(13.9) (15.4)
b 0.001
0.651
(i.e. LE and/or nitrite positive), sensitivity increased to 53.6%, without much loss of specificity. Table 3 shows the results of the multivariable logistic model, and the attributed values for the calculation of the total numerical score. Age younger than 12 months and urine LE ≥15 × 10^6/L increased almost five-fold the probability of UTI (adjusted ORs 4.6 and 4.7 respectively, p b 0.001). Additional variables significantly associated with the outcome were ASP ≥5500 × 10^6/L (aOR 4.0, p b 0.001) and bacteria ≥3 × 10^6/L (aOR 3.7, p b 0.01). In contrast, patient sex, presence of nitrites and sediment concentration of leukocytes and red blood cells did not significantly add to the model. The Hosmer–Lemenshow goodness-of-fit test was satisfactory (Chi squared with 3 d.f. 1.83, p = 0.61). The numerical risk score derived from the model ranged from zero (no predictor present) to 10 (all predictors present), with higher values indicating higher risk of UTI as measured by urine
culture. In the total sample, the mean risk score was 1.3 (standard deviation 2.0); however, the value was raised to 3.7 (standard deviation 2.9) among the infected children. Fig. 1 shows the ROC curve for the risk score compared to the curves for selected individual urine sediment variables and combined LE and nitrite test. The risk score had the better AUC (0.79, 95% CI 0.72–0.85), followed by the combined test (0.73). By selecting different risk score cut-offs, the discriminative and predictive values of the risk score as a test for urinary infection change (Table 4). The lower the cut-off, the higher are not only test sensitivity and negative predictive value, but also the likelihood of false positive results. With a cut-off of 2, sensitivity is 73.9% and specificity is 72.1%; 51 of the 69 patients with UTI would be identified as positives and immediately treated, while for 18 “false negatives” treatment would be delayed. However, 292 children (true negatives) would be appropriately spared antibiotics. Increasing the cut-off to 4, sensitivity drops to 49.3%, but specificity increases to 96.3%. Only 15 children would be inappropriately treated (false positives); however, as many as 35 infected patients (false negatives) would receive antibiotics only after urine culture results. When the best sensitivity–specificity trade-off values from the ROC curves were used as cut-offs for individual tests, the results of the multivariable analysis were very similar, and the ROC curve for the risk score derived from this model was almost superposable (AUC 0.79, 95% CI 0.75–0.83). 4. Discussion UTIs are the most common bacterial infections in infants and young children, and can be associated with high fever and poor general conditions [8]. The infection may involve the kidney, and in a
Table 2 Discriminative and predictive values for individual results of urinalysis. Variable
Dipstick urinalysis LE ≥15 × 10^6/L Presence of nitrites LE ≥15 × 10^6/L or presence of nitrites
Microscope IRIS IQ200® Bacteria ≥3 × 10^6/L ASP ≥5500 × 10^6/L Leukocytes >24 × 10^6/L Red blood cells >18 × 10^6/L
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
True positivesa
False positivesa
True negativesa
False negativesa
%
%
%
%
n.
n.
n.
n.
49.3 (44.8–53.8) 18.8 (15.3–22.4) 53.6 (49.1–58.1)
87.9 (85.0–90.8) 100 (100–100) 87.9 (85.0–90.8)
41.0 (36.5–45.4) 100 (100–100) 43.0 (38.6–47.5)
91.1 (88.5–93.6) 87.9 (84.9–90.8) 91.8 (89.3–94.2)
34
49
356
35
13
0
405
56
37
49
356
32
24.6 (20.8–28.5) 42.0 (37.6–46.5) 30.4 (26.3–34.6) 10.1 (7.4–12.9)
97.3 (95.8–98.8) 90.4 (87.7–93.0) 94.6 (92.5–96.6) 96.3 (94.6–98.0)
60.7 (56.3–65.1) 42.7 (38.2–47.1) 48.8 (44.3–53.3) 31.8 (27.6–36.0)
88.3 (85.5–91.2) 90.2 (87.5–92.8) 88.9 (86.0–91.7) 86.3 (83.2–89.4)
17
11
394
52
29
39
366
40
21
22
383
48
7
15
390
62
CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; LE, leukocyte esterase; ASP, all small particles. a Number of patients by classification category.
Table 3 Predictors of UTI: results of multivariable logistic regression model and attributed values for score calculation. Predictor
β coefficient
95% CI
aOR
95% CI
p-value
Attributed values for score
Leukocyte esterase ≥15 × 10^6/L Age (y) b 1 ASP ≥5500 × 10^6/L Bacteria ≥3 × 10^6/L
1.6 1.5 1.4 1.3
0.9–2.2 0.8–2.3 0.7–2.1 0.3–2.3
4.7 4.6 4.0 3.7
2.5–8.9 2.2–9.6 1.9–8.3 1.4–10.0
b0.001 b0.001 b0.001 0.010
3 3 2 2
CI, confidence interval; aOR, odds ratio adjusted for the variables included in the model; ASP, all small particles.
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
481
analysis for LE and nitrite are superior to microscopic detection of pyuria in children. A meta-analysis and a cost-effectiveness study suggested that urine dipstick negative for both LE and nitrite may be used for ruling out infections [14,15]. However, Laine et al. [16] found that in hospital care setting urinalysis by dipstick alone carries the risk of missing infections, while other authors proposed that pyuria, as measured by at least 10 × 10^6 Leuk/L on unspun urine, and any bacteriuria are best suited to identify cases where urine culture is warranted [15,17]. Recently, a new meta-analysis [11] showed that microscopy examination of urine for Gram stained bacteria is the single best rapid test, with 91 sensitivity and 96% specificity, while “either leukocyte esterase or nitrite” positivity can be rated as the best dipstick test. Most of the studies quoted (65 out of 95) used the 10^8 CFU/L cutoff for voided urine culture. However, 23 used a different value, and in 7 studies the criterion was unclear [11]. Our study confirmed that the dipstick LE or nitrite test has indeed the best performance, with a sensitivity of 56.3% and specificity 87.9%. However, the risk score developed by combining both dipstick and sediment reading results through multivariable analysis allowed us to obtain an additional improvement of the area under the ROC curve. Through the selection of different cut-off criteria, the score can be adapted to different clinical situations depending on the relative benefits of maximizing sensitivity or specificity. The selection of a very low cut-off, basically corresponding to any of the variables included in the final model being positive, leads to a 74% sensitivity and 94% negative predictive value, allowing to withhold treatment in children scored negative pending culture results. In our series of 474 children with suspected UTI, such a procedure would avoid unnecessary antibiotic administration in 292 children (72% of those without UTI); 113 however would still receive two days of treatment. A higher cut-off (e.g. ≥4) leads to 96% specificity, increasing the positive predictive value to 69% without large loss of NPV (92%). The use of this strategy would avoid unnecessary treatment in as many as 390 cases, about 96% of non infected children; however 35 patients with UTI would start antibiotics only after culture results are available. Ultimately, the selection of the best strategy will depend on the setting and on the individual characteristics and clinical conditions of the patients. Clearly, the main issue concerns balancing the harm and costs of unnecessarily starting antibiotics with the risk of twodays delayed treatment in case of UTI. Symptoms of acute UTI may be very disturbing in children, and are promptly relieved by treatment. Early and appropriate treatment is generally recommended in
Fig. 1. Receiver-Operating Characteristic (ROC) curves for the risk score and selected dipstick and urine sediment tests. (The 45° line through the origin represents the ROC curve of a test whose decision ability is no better than chance).
small proportion of patients even lead to permanent renal damage and scarring [9,10]. Culture of urine is the reference standard for the diagnosis of UTI [8], but continuing efforts are spent in searching for alternative, more rapid and still accurate tests [11]. This is especially important in children, since both a delay in treatment and the inappropriate administration of antibiotics can result in potentially serious adverse effects. Diagnostic studies in the pediatric population have however produced inconsistent results [11,12]. According to Gorelick and Shaw [13], both a Gram stain of uncentrifuged urine specimen and dipstick Table 4 Discriminative and predictive value of the clinical score by cut-off value. Score cut-off
≥2 ≥3 ≥4 ≥5 ≥6 ≥7 ≥8 10
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
%
%
73.9 (70.0-77.9) 66.7 (62.4–70.9) 49.3 (44.8–53.8) 42.0 (37.6–46.5) 26.1 (22.1–30.0) 15.9 (12.7–19.2) 8.7 (6.2–11.2) 5.8 (3.7–7.9)
72.1 (68.1–76.1) 79.5 (75.9–83.1) 96.3 (94.6–98.0) 97.0 (95.5–98.6) 98.3 (97.1–99.4) 98.8 (97.8–99.8) 99.8 (99.3–100.0) 100.0 (100.0–100.0)
PPV, positive predictive value; NPV, negative predictive value. a Number of patients by classification category.
NPV (95% CI)
True positivesa
%
%
31.1 (26.9–35.3) 35.7 (31.4–40.0) 69.4 (65.2–73.5) 70.7 (66.6–74.8) 72.0 (68.0–76.0) 68.8 (64.6–72.9) 85.7 (82.6–88.9) 100.0 (100.0–100.0)
94.2 (92.1–96.3) 93.3 (91.1–95.6) 91.8 (89.3–94.2) 90.8 (88.2–93.4) 88.6 (85.8–91.5) 87.3 (84.3–90.3) 86.5 (83.4–89.6) 86.2 (83.1–89.3)
False positivesa
True negativesa
False negativesa
n.
n.
n.
n.
51
113
292
18
46
83
322
23
34
15
390
35
29
12
393
40
18
7
398
51
11
5
400
58
6
1
404
63
4
0
405
65
482
R. Luciano et al. / Clinica Chimica Acta 413 (2012) 478–482
infants and young children to reduce the likelihood of renal involvement during the acute phase of infection [18] and subsequent renal scarring [19]. However, recent data from two multicenter randomized controlled studies showed that the probability of developing renal scars in case of upper tract UTI is not related to the promptness of antibiotic treatment [20,21]. Although this finding derives from observational data on a subgroup of patients, and confirmation by other studies is warranted [22,23], it does suggest that, except for infants, treatment can probably be safely delayed for two or three days, until a positive culture is obtained. This study has limitations. It was carried out retrospectively, using the database of the laboratory medicine department. The full available sample was used for the development of the score, that could not be subsequently validated on a different data set. Information on the clinical conditions of the children was not available, and the final diagnosis was not compared with the medical records. Data on the indications for performing a urine culture and on the severity and duration of symptoms may have increased the predictive value of our multivariable model. We used a urine culture count ≥10^8 CFU/L to indicate UTI, as this is the criterion adopted by our hospital and in most studies [11]. Data from the 1950s have shown that this cut-off has a good sensitivity, but may lead to include false negative patients; likewise, more recent data highlight the fact that, in some patients, even lower bacterial count may be clinically significant [11,24]. The prevalence of UTI in the present study was slightly higher than the previously reported values, ranging from 7.1 to 9.1% [25], probably because of the tertiary pediatric care setting. From a methodological stand-point, moderate differences in the prevalence of UTI in the studied populations are unlikely to affect sensitivity and specificity [8]. The sensitivity, specificity, NPV and PPV obtained in our cohort were similar to those previously reported [26]. We used the automated urine microscopy analyzer IRIS200, already previously validated as a tool for the study of UTI [27–29]. The combination of this tool with automated dipstick analysis provided an excellent NPV for excluding UTI, safely supporting a clinician's decision to withhold treatment pending culture results, provided that the patient can be appropriately monitored. Further research should prospectively validate the risk score on a new set of data, and in patients with known characteristics and disease severity. References [1] Mangin D. Urinary tract infection in primary care. BMJ 2010;340:c657. [2] Bhat RG, Katy TA, Place FC. Pediatric urinary tract infections. Emerg Med Clin N Am 2011;29:637–53. [3] Hughes C, Roebuck MJ. Evaluation of the IRIS 939 UDx flow microscope as a screening system for urinary tract infection. J Clin Pathol 2003;56:844–9. [4] Van Den Broek D, Keularts IM, Wielders JP, Kraaijenhagen RJ. Benefits of the iQ200 automated urine microscopy analyser in routine urinalysis. Clin Chem Lab Med 2008;46:1635–40.
[5] European urinalysis guidelines. Scand J Clin Lab Invest Suppl 2000;231:1–86. [6] Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem 1993;39:561–77. [7] Hosmer DW, Lemeshow S. Applied logistic regression. New York NY: John Wiley & Sons; 2000. [8] Zorc JJ, Kiddoo DA, Shaw KN. Diagnosis and management of pediatric urinary tract infections. Clin Microbiol Rev 2005;18:417–22. [9] Hellerstein S. Long-term consequences of urinary tract infections. Curr Opin Pediatr 2000;12:125–8. [10] Montini G, Tullus K, Hewitt I. Febrile urinary tract infections in children. NEJM 2011;365:239–50. [11] Williams GJ, Macaskill P, Chan SF, Turner RM, Hodson E, Craig JC. Absolute and relative accuracy of rapid urine tests for urinary tract infection in children: a metaanalysis. Lancet Infect Dis 2010;10:240–50. [12] Wald E. Urinary tract infections in infants and children: a comprehensive overview. Curr Opin Pediatr 2004;16:85–8. [13] Gorelick MH, Shaw KN. Screening tests for urinary tract infection: a metaanalysis. Pediatric 1999;104:e54. [14] Devillé WL, Yzermans JC, van Duijn NP, Bezemer PD, van der Windt DA, Bouter LM. The urine dipstick test useful to rule out infections. A meta-analysis of the accuracy. BMC Urol 2004;2 4:4. [15] Whiting P, Westwood M, Bojke L, et al. Clinical effectiveness and costeffectiveness of tests for the diagnosis and investigation of urinary tract infection in children: a systematic review and economic model. Health Technol Assess 2006;10:1–154. [16] Laine P, Toivonen E, Eklund K, Hohenthal U, Siren S, Maki T. Rapid dipstick urinalysis in the internal medicine clinic: what is missed? J Intern Med 1997;242:271–3. [17] Huicho L, Campos-Sanchez M, Alamo C. Metaanalysis of urine screening tests for determining the risk of urinary tract infection in children. Pediatr Infect Dis J 2002;21:1–11. [18] Doganis D, Siafas K, Mavrikou M, et al. Does early treatment of urinary tract infection prevent renal damage? Pediatrics 2007;120:e922–8. [19] 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. [20] Hewitt IK, Zucchetta P, Rigon L, et al. Early treatment of acute pyelonephritis in children fails to reduce renal scarring: data from the Italian Renal Infection Study Trials. Pediatrics 2008;122:486–90. [21] Hewitt IK, Montini G. In reply. Pediatrics 2009;123:e174. [22] Edmonson MB, Wald ER. Treatment of pyelonephritis and risk of renal scarring. Pediatrics 2009;123:e545. [23] Fernández-Menèndez JM, Pérez-Méndez C. Early antibiotic treatment of pyelonephritis in children is still mandatory. Pediatrics 2009;123:e173. [24] Tullus K. Difficulties in diagnosing urinary tract infections in small children. Pediatr Nephrol 2011;26:1923–6. [25] Shaikh N, Morone NE, Bost JE, Farrell MH. Prevalence of urinary tract infection in childhood: a meta-analysis. Pediatr Infect Dis J 2008;27:302–8. [26] Shayanfar N, Tobler U, Eckardstein A, Bestmann L. Automated urinalysis: first experiences and a comparison between the Iris iQ200 urine microscopy system, the Sysmex UF-100 flow cytometer and manual microscopic particle counting. Clin Chem Lab Med 2007;45:1251–6. [27] Lamchiagdhase P, Preechaborisutkul K, Lomsomboon P, et al. Urine sediment examination: a comparison between the manual method and the IQ200 automated urine microscopy analyzer. Clin Chim Acta 2005;358:167–74. [28] Linko S, Kouri TT, Toivonen E, Ranta PH, Chapoulaud E, Lalla M. Analytical performance of the Iris iQ200 automated urine microscopy analyzer. Clin Chim Acta 2006;372:54–64. [29] Chien T-I, Kao J-T, Liu H-L, et al. Urine sediment examination: a comparison of automated urinalysis systems and manual microscopy. Clin Chim Acta 2007;384: 28–34.