PEDIATRICS Vol. 108 No. 2 August 2001, pp. 311-316

Predictive Model for Serious Bacterial Infections Among Infants Younger Than 3 Months of Age

Richard G. Bachur, MD^* and Marvin B. Harper, MD^*,

From the Divisions of ^* Emergency Medicine and  Infectious Disease, Children's Hospital, Boston, Massachusetts.

	ABSTRACT

Top Abstract Methods Results Discussion Conclusion References

Objective.  To develop a data-derived model for predicting serious bacterial infection (SBI) among febrile infants <3 months old.

Methods.  All infants 90 days old with a temperature 38.0°C seen in an urban emergency department (ED) were retrospectively identified. SBI was defined as a positive culture of urine, blood, or cerebrospinal fluid. Tree-structured analysis via recursive partitioning was used to develop the model. SBI or No-SBI was the dichotomous outcome variable, and age, temperature, urinalysis (UA), white blood cell (WBC) count, absolute neutrophil count, and cerebrospinal fluid WBC were entered as potential predictors. The model was tested by V-fold cross-validation.

Results.  Of 5279 febrile infants studied, SBI was diagnosed in 373 patients (7%): 316 urinary tract infections (UTIs), 17 meningitis, and 59 bacteremia (8 with meningitis, 11 with UTIs). The model sequentially used 4 clinical parameters to define high-risk patients: positive UA, WBC count 20 000/mm³ or 4100/mm³, temperature 39.6°C, and age <13 days. The sensitivity of the model for SBI is 82% (95% confidence interval [CI]: 78%-86%) and the negative predictive value is 98.3% (95% CI: 97.8%-98.7%). The negative predictive value for bacteremia or meningitis is 99.6% (95% CI: 99.4%-99.8%). The relative risk between high- and low-risk groups is 12.1 (95% CI: 9.3-15.6). Sixty-six SBI patients (18%) were misclassified into the lower risk group: 51 UTIs, 14 with bacteremia, and 1 with meningitis.

Conclusions.  Decision-tree analysis using common clinical variables can reasonably predict febrile infants at high-risk for SBI. Sequential use of UA, WBC count, temperature, and age can identify infants who are at high risk of SBI with a relative risk of 12.1 compared with lower-risk infants.  Key words:  fever, bacteremia, meningitis, urinary tract infection, infant, predictive, management, urinalysis, white blood cell, temperature.

The management of febrile infants <3 months old is challenging because of the relatively high prevalence of serious bacterial infection (SBI) and the inability to easily discriminate those with serious bacterial disease from those with simple viral illness. Individual studies have shown the prevalence of SBI to be from 6% to 10%, and in the largest meta-analysis of previous data, the rate of SBI was 9%.¹ Despite having the problem well-defined, there is no consistent management approach to infants <3 months old.^2-6 Several strategies or protocols have been reported, compared, and retested in the pediatric literature, but no protocol has been universally adopted. The debate over appropriate management has also been complicated by studies involving varying populations with respect to age, different temperature cutoffs to define fever, and varying clinical and laboratory criteria for selecting out febrile infants at high risk or low risk for SBI.

Definitive identification of SBI requires a positive culture of the cerebrospinal fluid (CSF), blood, or urine or an identifiable bacterial focus by physical examination or radiograph. However, because most of these infants have unrevealing physical examinations and the results of the cultures are not immediately available, clinicians must decide on appropriate patient management based on the history, physical examination, and laboratory testing. The most commonly promoted strategies require a complete sepsis evaluation with the primary goal of identifying patients at low risk for SBI, and these low-risk infants are selected for outpatient therapy with or without antibiotics.^7-9 Our goal was to develop a data-derived model for febrile infants <3 months old to predict infants at high risk for SBI using common, objective, clinical variables.

	METHODS

Top Abstract Methods Results Discussion Conclusion References

We retrospectively identified all patients seen in the emergency department (ED) from 1993 to 1999 who were 90 days old with a rectal temperature 38.0°C. For the purposes of this study, SBI was defined as a pathogen isolated from the CSF or blood or a urinary tract infection (UTI). A UTI was defined as a urine culture with a single pathogen growth 1000 colony-forming units (cfu)/mL of urine from a suprapubic aspiration or 10 000 cfu/mL from a bladder catheterization. Charts of all patients with positive CSF, blood, or urine cultures were reviewed for confirmation that the culture isolate was considered a true pathogen and not a contaminant. For the study period 1996-1999, charts of patients with SBI were reviewed by one of the authors (R.G.B.) for the description of the patient's appearance in the ED: patients were coded as ill-appearing or well-appearing according to the medical record.

A positive urinalysis (UA) was defined by a positive test for leukocyte esterase or nitrite by dipstick (Multistix, Bayer Corporation, Elkhart, IN) or by pyuria as defined by 5 white blood cells (WBC) per high-power field (hpf) on a spun urine sample. Gram stain results of the urine and CSF were not studied.

Tree-structured analysis by Classification and Regression Tree (CART) Version 3.6.1 (Salford Systems, San Diego, CA) was used to develop the predictive model. CART's methodology is technically known as binary recursive partitioning. Binary refers to the fact that all decisions involve a parent node that is broken into 2 child nodes. Recursive refers to the fact that each child node can then becomes a parent node, and in so doing, a decision tree with expanding branches is created. To begin the process, a single outcome variable and all potential predictor variables are assigned. Splitting rules are developed in a stepwise fashion by analyzing each potential predictor and all possible cut points. Splits are made to minimize false-negative or false-positive assignments for the outcome variable at each step. Partitioning is repeated until either of the subgroups contains a homogeneous group or the subgroups are too small for further subdivision. CART then proceeds to prune back some branches by recombining subgroups if classification errors are not significantly increased in the process of simplifying the tree. A parameter representing the significance of misclassifications can be modified such that the model maximizes sensitivity or specificity. If a case has missing data for a particular split, a surrogate variable that behaves like the missing variable (as determined by the software) is used for the decision. Finally, the model was tested by V-fold cross-validation: the data set is divided into 10 equal parts with similar distribution of dependent variable, and then the model is derived with 9 parts (the learning set) and tested with 1 part (the validation set). This cross-validation is repeated 10 times, and the results are combined to develop the predictive accuracy and error rates for the tree.

For our model, SBI or No-SBI was used as the dichotomous outcome variable, and age, temperature, WBC count, UA, absolute neutrophil count, and CSF WBC count were entered as potential predictors. CART then determined the order and use of the variables as well as any cut points for any continuous variables.

Statistical analyses were conducted using the Statistical Program for the Social Sciences Version 6.1.1 (SPSS, Chicago, IL). Medians and interquartile ranges ([IQR]: 25th-75th percentile) were provided for nonnormal data. Mean values of interval data were compared between groups by using a 2-tailed Student's t test. EpiInfo Version 6.04b (Centers for Disease Control and Prevention, Atlanta, GA) was employed for ² and Fisher's exact tests of nominal data. Confidence intervals for proportions were calculated using Stata Version 6 (Stata, College Station, TX). The institutional review board of the hospital approved the study.

	RESULTS

Top Abstract Methods Results Discussion Conclusion References

Study Population

Over the study period, 5279 febrile infants 90 days old were evaluated in the ED. Median age and temperature were 1.6 months (IQR: 1.0-2.2 months) and 38.4°C (IQR: 38.1-38.9°C). The number of patients who had tests and cultures performed are shown in Table 1.

SBIs

SBIs were diagnosed in 373 patients (7%): 316 UTIs, 17 meningitis, and 59 patients with bacteremia including 8 (47%) of those with meningitis and 11 (3.5%) of those with UTIs. The following pathogens were isolated from the blood: group B streptococci (21), Escherichia coli (20), Staphylococcus aureus (5), Streptococcus pneumoniae (4), Enterococcus faecalis (2), Enterobacter cloacae (1), group A streptococci (2), Salmonella spp. (2), Neisseria meningitidis (1), and Citrobacter koseri (1). Pathogens isolated from the CSF were: group B streptococci (5), Escherichia coli (5), Citrobacter spp. (3), Enterococcus faecium (2), S aureus (1), and S pneumoniae (1). Median age, temperature, and WBC count of those with SBI were: 1.5 months (IQR: 0.9-2.2 months), 38.8°C (IQR: 38.3-39.3°C), and 14 860/mm³ (IQR: 10 230-19 160/mm³). By age groups, SBIs were diagnosed in 8.8% (105/1193) of those 1 month old, 7.3% (143/1961) of those 1 to 2 months old, and 7.1% (125/1753) of those 2 to 3 months old. Appearance of the patient was reviewed in all SBI cases from 1996 to 1999: 9% of patients with SBI (14/204) were described as ill-appearing including 26% (9/35) of those with bacteremia or meningitis.

Model for Predicting Patients at High Risk for SBI

The model, as created through recursive partitioning, is shown in Fig 1. Absolute neutrophil count and CSF cell count were not chosen (by the software) as a predictor. The distribution of patients through the model is shown in Fig 2. The sensitivity of the model for SBI is 82% (306/372, 95% confidence interval [CI]: 78%-86%), specificity 76% (3748/4907, 95% CI: 75%-77%), negative predictive value (NPV) 98.3% (3748/3814, 95% CI: 97.8%-98.7%), and positive predictive value 21% (306/1465, 95% CI: 19%-23%). The performance of the model can also be represented by the dichotomous distribution of SBI cases: 5279 patients enter the model with a 7% rate of SBI, and then the model defined a high-risk group of 1465 patients of whom 307 (21%) had SBI and a lower risk group from which 66 of 3814 (1.7%) infants had SBI. The associated relative risk (between the high and lower risk groups) is 12.1 (95% CI: 9.3-15.6).

View larger version (38K): [in this window] [in a new window]   Fig. 1.   Decision tree for predicting serious bacterial infection among febrile infants <3 months old.

View larger version (31K): [in this window] [in a new window]   Fig. 2.   The distribution of study patients by the decision tree. The number and percentage of patients with SBI in a node are shown in parentheses.

In this model, UA was found to be the single best discriminator for SBI. The overall sensitivity of the UA for any SBI was 71% (95% CI: 66%-76%), and for UTI the sensitivity of the UA was 81% (95% CI: 76%-85%). Fifty-one (77%) of 66 misclassifications were UTIs with a negative UA. The other 15 misclassifications included 14 patients with bacteremia and 1 patient with meningitis (Table 2). The NPV of the model for bacteremia and meningitis is 99.6% (95% CI: 99.4%-99.8%).

	DISCUSSION

Top Abstract Methods Results Discussion Conclusion References

Febrile infants are commonly evaluated by pediatricians and emergency physicians.¹⁰ For those infants who appear ill or who have an evident focal infection, management is straightforward. The vast majority of infants, however, appear relatively well with nonspecific symptoms and no localized source of infection. Because of the relatively high prevalence of occult infections, many of these infants undergo invasive testing to include cultures of the CSF, urine, and blood. Many of the infants are hospitalized and given empiric antibiotic therapy pending culture results. Outpatient management of a select minority of infants, who are considered to be low risk for SBI by published criteria, has become more popular over the last decade, but the selection process remains controversial.^7-9,11 Although management guidelines (based on meta-analyses of available data) have been offered in the medical literature,¹² the management of young febrile infants continues to be debated.^2,313-17

In review of the literature regarding febrile infants, the available data are further complicated by differing and arbitrarily determined subpopulations being studied (0-3 months, 1-2 months, 1-3 months), varying temperature limits to define fever, different definitions of fever without source and SBI, and variability of testing and the criteria used to define high- versus low-risk infants. Despite the lack of universal acceptance of any one strategy, previous publications have consistently noted the inability to discriminate those with significant bacterial disease from those with simple viral illness based on observation scales or simple parameters such as height of fever or WBC count.^1,818-28 Without such simple, reliable, discriminating tests, current strategies involve complete sepsis evaluations with outpatient treatment of the lower risk patients with⁸ or without antibiotics.^7,9 Three common strategies for managing febrile infants without a fever source are: 1) the Philadelphia protocol for infants 29-60 days old,⁷ 2) the Rochester criteria (60 days old),⁹ and 3) the Boston criteria (28-89 days old).⁸ Each has its own definitions of the high-risk and low-risk infants based on a combination of factors including history, physical examination, and laboratory parameters (Table 3). In the first 2 strategies, the lower risk patients are selected for outpatient therapy without antibiotics, whereas the Boston strategy treats all patients with empiric antibiotics but selects a smaller high-risk population for hospitalization. The Philadelphia protocol and Rochester criteria have been promoted by their high NPV (ability to remove patients with SBI from the low-risk group)99.7% and 98.9%, respectively. However, each suffers from a relatively low positive predictive value14% and 12%, respectivelyattributable to the large numbers of patients considered higher risk and therefore hospitalized for antibiotics. The Boston criteria has been argued to be the more cost-effective strategy (despite treating all with antibiotics) because fewer patients require admission.²⁹

For the current study, we wanted to develop a model to predict infants at high risk for SBI. Unlike the previously discussed criteria that had been preselected, we allowed the data to determine which clinical variables (as well as the corresponding cutoffs for continuous variables) were used. The order and use of variables are derived by the software without any bias other than the choice of potential predictors. Additionally, decision-tree analysis is easy to understand and apply as compared with other statistical methods such as logistic regression. The predictive model sequentially used four common clinical parameters: UA, WBC count, temperature, and age. With such a simple model, the NPV was 98.3% for SBI and 99.6% for bacteremia or meningitis. In a comparison to the Rochester criteria, the recursive partitioning model had a comparable enrichment of the higher versus lower risk group (Fig 3). The value of selectively enriching the high-risk group versus the low-risk group (Rochester Criteria) can be seen in the proportion of high- versus low-risk patients in each model; if the intervention was to admit only higher risk patients, the recursive partitioning model would admit only 28% versus 53% (Rochester criteria).

View larger version (48K): [in this window] [in a new window]   Fig. 3.   Comparison of recursive partitioning model to Rochester criteria.9

Despite the excellent NPV of our current model (similar to the Philadelphia and Rochester criteria), the limitations of any management strategy for this population are evident when looking at the SBI patients who were misclassified into the lower risk group: 1.7% (66/3814) of these patients had SBI (51 UA-negative UTI, 14 bacteremia, 1 meningitis). The relatively low sensitivity of the standard UA (dipstick plus microscopy) in this study (80%) is comparable to previous reports.³⁰ Despite the predominance of UTIs among the misclassifications, the relative frequency of misclassified patients with bacteremia and meningitis must be more seriously considered when evaluating any management strategy. Therefore, despite a comparable performance of our simple model to more complex management schemes, the significant misclassifications of SBI patients into a lower risk group suggest that an outpatient strategy using empiric therapy may be justified.

A true, direct comparison to the Philadelphia and Rochester criteria cannot be performed because each applies to specific age groups (and not all febrile infants <3 months old) and certain key data elements that are used by the Philadelphia criteria (appearance of patient, band-neutrophil ratio, urine Gram stain, CSF Gram stain, results of chest radiograph and stool smears if obtained) and the Rochester criteria (patient appearance, absolute band count, results of stool smear if obtained) are not available in our dataset. Although some patients that were misclassified by the recursive partitioning model would be properly classified by the other strategies, other patients that were correctly classified by the recursive partitioning model may be misclassified by these other strategies. In our dataset of infants <3 months with temperature 38.0°C, the Philadelphia criteria (using WBC count <15 000/mm³, UA 10 WBCs/hpf, CSF <8 WBCs/mm³) would misclassify 69 patients (16 bacteremia, 2 meningitis, 51 UTI) and the Rochester criteria (using WBC count >5000 and <15 000/mm³, UA 10 WBCs/hpf) would misclassify 77 patients (14 bacteremia, 6 meningitis, and 57 UTI)the value of this analysis is only to compare the recursive partitioning model to other criterion (using our data) and not to test the other strategies. Interestingly, the Philadelphia protocol is intended for use in infants 29 to 60 days old with temperature 38.2°C; in this age subgroup alone, we had 1175 infants with temperatures of 38.0 or 38.1°C who would not require any diagnostic testing under the Philadelphia protocol47 (4%) of these infants had SBI (6 bacteremia, 1 meningitis, 40 UTI), therefore the 38.0°C threshold for diagnostic testing seems justified in this age group. Additionally, both the Rochester criteria and Philadelphia protocol are directed at infants <2 months old; in our study, 7.1% (125/1753) of infants 2 to 3 months old had SBI (including 3 patients with meningitis and 14 with bacteremia).

This study and its methodology have certain limitations. By using a limited definition of SBI to include only meningitis, bacteremia, and UTI, we underestimated the prevalence of SBI in our population. Not all patients had all cultures performedagain underestimating the prevalence of SBI and potentially affecting the development of the model. We were not able to assess the value of some other predictors used by other strategies, such as Gram stains of the urine, which could easily diminish the number of misclassifications.³¹ And although a positive urine culture is the gold standard for diagnosing a UTI, it is worth mentioning that some patients with a positive urine culture and a negative UA could simply have asymptomatic bacteriuria or, theoretically, contamination. Furthermore, the appearance of the patient was not prospectively determinedideally, ill-appearing children and those with obvious sources of focal bacterial infection would be removed before development of any management strategy (aimed at the well-appearing febrile child without an identifiable source of infection). Finally, the past medical history of the patientincluding perinatal history, underlying conditions, and infectious exposures were not considered in our model but would be important factors in practice.

	CONCLUSION

Top Abstract Methods Results Discussion Conclusion References

Decision-tree analysis can be used to reasonably predict febrile infants at high risk for serious bacterial infection. A simple model of sequential use of UA, WBC count, temperature, and age can identify infants who are at high risk for SBI with a relative risk of 12 compared with the not high-risk patients. Despite the relatively low sensitivity of the UA, it remains the single most important diagnostic test for SBI, and improvements in sensitivity of the UA has the greatest potential for improving detection of SBI in young febrile infants at the time of the initial evaluation. Based on the inability to reliably discriminate infants with SBI from those with simple viral illness, conservative management of this population continues to be warranted. This decision tree model may initially be most useful in settings where physicians do not routinely perform complete sepsis evaluations (including lumbar puncture) on all febrile infants or in settings where not all infants receive empiric antibiotics pending results of the cultures; based on the proposed strategy, those infants determined to be higher risk based on age, temperature, WBC count, or UA would be considered for additional testing and empiric treatment.

	FOOTNOTES

Received for publication Aug 28, 2000; accepted Dec 1, 2000.

This work was presented at the Pediatric Academic Societies Meeting; May 12-16, 2000; Boston, MA.

Reprint requests to (R.G.B.) Division of Emergency Medicine, Children's Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail: bachur{at}tch.harvard.edu

	ABBREVIATIONS

SBI, serious bacterial infection; CSF, cerebrospinal fluid; ED, emergency department; UTI, urinary tract infection; cfu, colony-forming unit; UA, urinalysis; WBC, white blood cell count; CART, Classification and Regression Tree; hpf, high-power field; IQR, interquartile range; CI, confidence interval; NPV, negative predictive value.

	REFERENCES

Top Abstract Methods Results Discussion Conclusion References

1. Slater M, Krug SE. Evaluation of the infant with fever without source: an evidence based approach. Emerg Med Clin North Am. 1999;17:97-126, viii-ix
2. Baraff LJ Management of the febrile child: a survey of pediatric and emergency medicine residency directors. Pediatr Infect Dis J 1991; 10:795-800 [Medline]
3. Jones RG, Bass JW Febrile children with no focus of infection: a survey of their management by primary care physicians. Pediatr Infect Dis J 1993; 12:179-183 [Medline]
4. Greene JW, Hara C, O'Connor S, Altemeier WD Management of febrile outpatient neonates. Clin Pediatr 1981; 20:375-380
5. DeAngelis C, Joffe A, Willis E, Wilson M Hospitalization v. outpatient treatment of young, febrile infants. Am J Dis Child 1983; 137:1150-1152 [Abstract/Free Full Text]
6. Pantell RH, Naber M, Lamar R, Dias JK Fever in the first six months of life: risks of underlying serious infection. Clin Pediatr 1980; 19:77-82
7. Baker MD, Bell LM, Avner JR Outpatient management without antibiotics of fever in selected infants [see comments]. N Engl J Med 1993; 329:1437-1441 [Abstract/Free Full Text]
8. Baskin MN, O'Rourke EJ, Fleisher GR Outpatient treatment of febrile infants 28 to 89 days of age with intramuscular administration of ceftriaxone [see comments]. J Pediatr 1992; 120:22-27 [CrossRef][Medline]
9. Jaskiewicz JA, McCarthy CA, Richardson AC, Febrile infants at low risk for serious bacterial infectionan appraisal of the Rochester criteria and implications for management. Febrile Infant Collaborative Study Group [see comments]. Pediatrics 1994; 94:390-396 [Abstract/Free Full Text]
10. Baskin MN, Fleisher GR, O'Rourke EJ Identifying febrile infants at risk for a serious bacterial infection [letter; comment]. J Pediatr 1993; 123:489-490 [CrossRef][Medline]
11. Dagan R, Powell KR, Hall CB, Menegus MA Identification of infants unlikely to have serious bacterial infection although hospitalized for suspected sepsis. J Pediatr 1985; 107:855-860 [CrossRef][Medline]
12. Baraff LJ Management of infants and children 3 to 36 months of age with fever without source [review]. Pediatr Ann 1993; 22:497-498 [Medline]
13. Baraff LJ Outpatient management of fever in selected infants [letter; comment]. N Engl J Med 1994; 330:938-939 [Free Full Text]
14. Cox RD, Wagner M, Woolard DJ Infants and children with fever without source [letter; comment]. Ann Emerg Med 1994; 23:598-600 [CrossRef][Medline]
15. Kramer MS, Shapiro ED Management of the young febrile child: a commentary on recent practice guidelines [see comments]. Pediatrics 1997; 100:128-134 [Free Full Text]
16. Long SS Antibiotic therapy in febrile children: "best-laid schemes" [editorial; comment]. J Pediatr 1994; 124:585-588 [CrossRef][Medline]
17. Pantell R, Bergman D, Takayama J, Detecting serious bacterial illness in febrile infants: do guidelines help? Pediatr Res 2000; 47:216A
18. Berkowitz CD, Orr DP, Uchiyama N, Variability in the management of the febrile infant under 2 months of age. J Emerg Med 1985; 3:345-351 [CrossRef][Medline]
19. Baker MD, Avner JR, Bell LM Failure of infant observation scales in detecting serious illness in febrile, 4- to 8-week-old infants. Pediatrics 1990; 85:1040-1043 [Abstract/Free Full Text]
20. Bonadio WA, Hagen E, Rucka J, Shallow K, Stommel P, Smith D Efficacy of a protocol to distinguish risk of serious bacterial infection in the outpatient evaluation of febrile young infants. Clin Pediatr 1993; 32:401-404
21. Bonadio WA, Webster H, Wolfe A, Gorecki D Correlating infectious outcome with clinical parameters of 1130 consecutive febrile infants aged zero to eight weeks. Pediatr Emerg Care 1993; 9:84-86 [CrossRef][Medline]
22. Bonadio WA, Hennes H, Smith D, Reliability of observation variables in distinguishing infectious outcome of febrile young infants. Pediatr Infect Dis J 1993; 12:111-114 [Medline]
23. Bonadio WA, McElroy K, Jacoby PL, Smith D Relationship of fever magnitude to rate of serious bacterial infections in infants aged 4-8 weeks. Clin Pediatr 1991; 30:478-480
24. Bonadio WA, Romine K, Gyuro J Relationship of fever magnitude to rate of serious bacterial infections in neonates. J Pediatr 1990; 116:733-735 [CrossRef][Medline]
25. Bonadio WA, Hegenbarth M, Zachariason M Correlating reported fever in young infants with subsequent temperature patterns and rate of serious bacterial infections. Pediatr Infect Dis 1990; 9:158-160
26. King JC Jr, Berman ED, Wright PF Evaluation of fever in infants less than 8 weeks old. South Med J 1987; 80:948-952 [CrossRef][Medline]
27. Klassen TP, Rowe PC Selecting diagnostic tests to identify febrile infants less than 3 months of age as being at low risk for serious bacterial infection: a scientific overview [see comments]. J Pediatr 1992; 121:671-676 [CrossRef][Medline]
28. Roberts KB, Borzy MS Fever in the first eight weeks of life. Johns Hopkins Med J 1977; 141:9-13 [Medline]
29. Lieu TA, Baskin MN, Schwartz JS, Fleisher GR Clinical and cost-effectiveness of outpatient strategies for management of febrile infants [see comments]. Pediatrics 1992; 89:1135-1144 [Abstract/Free Full Text]
30. Shaw KN, McGowan KL, Gorelick MH, Schwartz JS. Screening for urinary tract infection in infants in the emergency department: which test is best? Pediatrics. 1998;101(6). URL:http://www.pediatrics.org/cgi/content/full/101/6/e1
31. Lockhardt G, Lewander W, Cimini D, Josephson S, Linakis J Use of urinary gram stain for detection of urinary tract infection in infants. Ann Emerg Med 1995; 25:31-35 [CrossRef][Medline]