Fever without source (FWS) in young children remains a difficult diagnostic problem, because clinical signs and symptoms are often unreliable predictors of a serious bacterial infection (SBI). Many clinical studies have addressed this problem, and the combination of a clinical evaluation associated with a total and differential leukocyte count are commonly used screening methods.1–3 The relatively poor specificity of the markers used to identify SBI, taken independently, urges physicians to give antibiotics to the majority of patients. In our study, we analyzed the predictive values of different markers in a multivariate logistic regression analysis. Our goal was to develop a simple score, which could be easily performed in the emergency room or in the office to predict SBI in a pediatric population with FWS.
MATERIALS AND METHODS
We performed a combined analysis of data collected from 2 prospectively and consecutively enrolled cohorts of children with FWS in a single university center.4,5 Both cohort studies had the same inclusion and exclusion criteria and had followed similar methodology. The study protocol was approved by the Ethical Committee of the Child and Adolescent Department, University Hospitals of Geneva. The study included all children aged from 7 days to 36 months who were consecutively admitted to the Emergency Department of the University Children's Hospital of Geneva with a rectal temperature above 38°C and without localizing signs of infection in their history or at physical examination. Criteria of exclusion are notified in the previous studies.4,5 All children had a clinical score based on the Infant Observation Scale (IOS),6 a urine analysis with culture and blood drawn for white cell count, determination of C-reactive protein (CRP), procalcitonin (PCT), and culture. Lumbar puncture was performed when meningitis was suspected. The pediatric resident in charge of the patient decided which child should receive antibiotics. All children had a clinical follow-up with physical examination by a pediatrician in the following 48 hours or by a telephone contact. The diagnosis was registered at the end of the clinical follow-up. Technical laboratory determinations and definition of SBIs: bacteremia, pyelonephritis, lobar pneumonia, bacterial meningitis, and criteria of benign infection are described elsewhere.4
Statistics.
The study population was divided by stratified randomization in a derivation set (2/3) and a validation set (1/3). The sensitivity, specificity, negative, and positive predictive values for the detection of a SBI were determined in the derivation set for the different laboratory parameters using the cutoff points derived from our previous studies.4,5 Univariate logistic regression was performed considering the dichotomized predictive parameters as independent values and SBI as the outcome value. Then, parameters significantly associated with SBI were entered forward stepwise into a multiple regression model and only those remaining independently significantly (P < 0.05) associated with SBI were retained. For ease of use in the clinical setting, we then created a Laboratory-score using only the predictive variables independently associated with SBI. The sensitivity, specificity, and predictive values of the Laboratory-score were determined in the derivation set and in the validation set.
RESULTS
Two hundred twenty-two children were consecutively included from March 1998 to February 2002. Twenty children were excluded. The data of 202 children were analyzed. The final diagnosis was: SBI in 54 children (27%) (7 bacteremia, 40 pyelonephritis, 5 lobar pulmonary condensation, 1 retropharyngeal abscess, and 1 mastoiditis), benign focal infection in 26 children (13%) (cystitis, acute otitis media, adenitis, Campylobacter gastroenteritis), and probable viral infection in 122 children (60%) (negative culture and no signs for focal infection at clinical follow-up). One hundred thirty-four of 202 (66%) of the children received antibiotics. The study population was divided in a derivation set (n = 135) and a validation set (n = 67). The 2 sets were comparable in terms of age, fever, incidence of SBI, clinical observational scores, and laboratory parameters. The sensitivity, specificity, and predictive values for the different parameters associated with SBI are listed in Table 1.
;)
TABLE 1: Predictive Value (%) of Different Variables Between Children With and Without Severe Bacterial Infections
Logistic Regression.
We first performed univariate logistic regression with variables potentially associated with SBI. PCT [odds ratio (OR): 35.6] showed the strongest association followed by CRP (OR: 12.9), urine dipstick (OR: 9), and leucocytosis (OR: 3). Left shift and IOS score were not statistically associated with SBI.
Then PCT, CRP, urine dipstick, and leucocytosis were entered into a forward stepwise multiple logistic regression model to identify independent predictor of SBI. The PCT value remained the most significant predictor of SBI (OR: 37.6; 95% CI: 5.8–243). The other variables independently associated with SBI in this analysis were CRP (OR: 7.8; 95% CI: 2–30.4) and urine dipstick (OR: 23.2; 95% CI: 5.1–104.8). Leucocytosis was not independently associated with the occurrence of SBI (P = 0.49).
Laboratory-Score.
Based on the results of the logistic regression analysis, we developed a risk index score, named Laboratory-score. The relative weighting of each component variable of the Laboratory-score was based on its odds ratio in the univariate analysis. Two points were attributed to PCT or CRP above the cutoff values (0.5 ng/mL and 40 mg/L, respectively) and 4 points for values of PCT above 2 ng/mL, and for CRP above 100 mg/L. One point was attributed for a positive urine dipstick (Table 2).
TABLE 2: Laboratory Score
The performance of the Laboratory-score was then tested both on the derivation population and the validation set (Table 1). In the derivation set, the Laboratory-score (≥3) had a sensitivity of 94% and a specificity of 81%. When compared with the other parameters commonly used to predict SBI, the Laboratory-score had the best accuracy associating good sensitivity and specificity. In the validation set the Laboratory-score had similar performances with a sensitivity of 94% (95% CI: 74–99) and a specificity of 78% (95% CI: 64–87) (Table 1).
DISCUSSION
Our data showed that PCT, CRP, and urine dipstick are independent predictors of SBI in this population of children with FWS. In our study, the IOS score and left shift were not statistically different between children with and without SBI. Moreover, leucocytosis was not an independent predictor of SBI when PCT, CRP, and urine dipstick have been taken into account.
We have developed a scoring system (Laboratory-score) based on the 3 predictive variables independently associated with SBI: PCT, CRP, and urinary dipstick. The principal advantage of the Laboratory-score is its good specificity (81%) for the prediction of SBI associated with the security of a high sensitivity (94%). The good specificity of the Laboratory-score should enable the reliable selection of children who need antibiotic treatment, without over treating children with viral infection. Based on this study, if antibiotics had solely been administrated for children with a positive score, only 40% of the population would have received antibiotics. In comparison, based on the clinician's decisions, more than 65% of the studied population received antibiotics. The use of the Laboratory-score could, thus, substantially reduce antibiotic use.
Potential limitations of our study should be considered. Our study population is relatively small explaining the wideness of the confidence intervals around the estimates of sensitivity and specificity. The incidence of SBI (27%) in our study seems higher than reported in other studies,7–9 but similar to the incidence of a recent study from Italy (23%) that analyzed comparable populations of children in a tertiary hospital.10 This likely reflects referral bias, as pediatricians refer ill-appearing children to our hospital for initial work-up. Because this bias affects the prevalence of SBI in our patient population, the predictive values of the Laboratory-score must be interpreted with caution, and the performance of the Laboratory-score might vary if applied to other cohorts of children. In contrast, the sensitivity and specificity of our scoring system are not affected by this potential bias. An internal validation of the score was performed on a subset of the population. However, this sample is small and the potential bias associated with our entire population remains.
In conclusion, PCT, CRP, and urine dipstick are independent predictors of SBI in this study. White blood cell count is not an independent predictor, when these 3 variables are taken into account. A Laboratory-score including PCT, CRP, and urine dipstick provides a security equivalent to the standard work-up, is easier to use, and could considerably diminish antibiotic use in children with benign infection. However, children should be carefully followed up, to identify the small proportion with SBI not initially detected by a positive score. Finally, the Laboratory-score should be prospectively validated and evaluated in different clinical settings before its use in clinical guidelines of children with FWS.
REFERENCES
1. Baraff LJ, Schriger DL, Bass JW, et al. Practice guideline for the management of infants and children 0 to 36 months of age with fever without source.
Pediatrics. 1993;92:1–12.
2. Baraff LJ. Management of fever without source in infants and children.
Ann Emerg Med. 2000;36:602–614.
3. American College of Emergency Physicians Clinical Policies Committee; American College of Emergency Physicians Clinical Policies Subcommittee on Pediatric Fever. Clinical policy for children younger than three years presenting to the emergency department with fever.
Ann Emerg Med. 2003;42:530–545.
4. Galetto-Lacour A, Zamora SA, Gervaix A. Bedside procalcitonin and C-reactive protein tests in children with fever without localizing signs of infection seen in a referral center.
Pediatrics. 2003;112:1054–1060.
5. Galetto-Lacour A, Gervaix A, Zamora SA, et al. Procalcitonin, IL-6, IL-8, IL-1 receptor antagonist and C-reactive protein as identificators of serious bacterial infections in children with fever without localising signs.
Eur J Pediatr. 2001;160:95–100.
6. McCarthy P, Sharpe M, Spiesel S, et al. Observation scales to identify serious illness in febrile children.
Pediatrics. 1982;70:802–809.
7. Thayyil S, Shenoy M, Hamaluba M, Gupta A, Frater J, Verber IG. Is procalcitonin useful in early diagnosis of serious bacterial infections in children?
Acta Paediatr. 2005;94:155–158.
8. Pulliam PN, Attia MW, Cronan KM. C-reactive protein in febrile children 1 to 36 months of age with clinically undetectable serious bacterial infection.
Pediatrics. 2001;108:1275–1279.
9. Bachur RG, Harper MB. Predictive model for serious bacterial infections among infants younger than 3 months of age.
Pediatrics. 2001;108:311–316.
10. Andreola B, Bressan S, Callegaro S, Liverani A, Plebani M, Da Dalt L. Procalcitonin and C-reactive protein as diagnostic markers of severe bacterial infections in febrile infants and children in the emergency department.
Pediatr Infect Dis J. 2007;26:672–677.
