Infections in infants and children are very common. Differential diagnosis between viral and bacterial agents is important in order to avoid unnecessary antibiotic treatment. Classic infectious indices such as C - reactive protein (CRP), full blood count with differential and procalcitonin concentration are used in the evaluation of febrile episodes in children. The current bibliography demonstrates the superior performance of procalcitonin and CRP instead of the total white blood cell count and absolute neutrophil count [1-4].

The objective of our study was to evaluate whether the haematological leukocyte subpopulations in combination with the CRP and the Erythrocyte Sedimentation Rate (ESR) could reliably demonstrate the presence from the absence of acute infection. Additionally we questioned whether the application of the above composite formula could be helpful for the differential diagnosis between viral and bacterial infection in infants and young children who visit the emergency department due to fever without source.

Patients and study design

This case-control study was conducted during a three-year period between 2008 and 2011. Children of Greek origin and nationality participated voluntarily in the study. Informed consent from parents was obtained in advance. The Ethics Committee of our Hospital and the Athens University Medical School approved the research protocol. A total number of 138 children, admitted to the Athens’ University 2nd Department of Pediatrics, were enrolled in the study. The study group consisted of children with fever and good clinical appearance, who visited the emergency department and admitted to the hospital for further evaluation. The subjects’ parents filled in a detailed questionnaire. Children under antibiotic treatment and those with medical history of a chronic disease or other co-morbidities were excluded from the study.

According to the protocol, the study population was divided into two groups. Group A consisted of 69 infants and children (42 boys, mean age 28.28 ± 29.58 months, range 2 to 118 months) with acute febrile episode lasting 6 hours or less. The discrimination between bacterial and viral infection was based on the combination of positive blood/urine culture along with CRP >40 mg/dl and leukocytosis with neutrophilia (white blood cell count >15.000/mm³, neutrophils >60%). The viral infections were documented either by the presence of IgM antibodies or positive polymerase chain reaction tests (PCR). The most common infections were infections of the upper respiratory system due to inluenzae A or B, parainfluenzae, adenovirus or syncytial respiratory virus. The group B (control group) consisted of 69 children age and sex matched to study group. Those were otherwise healthy children without fever or infection which were admitted to the hospital for a routine surgical procedure (tonsilectomy, correction of strabismus). All children of the control group were afebrile with CRP <10mg/dl, leucocytes <10.000/mm³, normal renal and liver function.

Methods

Full blood count with differential, ESR and CRP measurements as well as cultures of blood and urine and either IgM antibodies or PCR for viral agents were performed in all subjects using standard methods by the same laboratory within the first six hours from the onset of fever.

Statistical analysis

All statistical analyses and data management were performed using STATA for Windows v 8.5, (StataCorp, Texas, USA, 2006). Data are expressed as mean ± SD, median. Group data were compared using Mann-Whitney U test (independent samples) and p<0.05 was considered significant. Observing the data that were derived from study population we noted statistically significant marked differences between children with infection and control group and between children with viral and bacterial infection regarding neutrophils, lymphocytes, monocytes, ESR and CRP. We therefore used these parameters as discrimination markers between the mentioned above groups. Neutrophils were higher in children with bacterial infection compared to viral infected children. Respective changes were noted in lymphocyte and monocyte counts between groups. In order to maximize the effect of those differences we calculated the ratio of neutrophils (N) to the sum of lymphocytes (L) plus monocytes (M) and we further added the parameters of C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR). The new proposed formula (formula 1) was as follows:

We examined the optimal cutoff value of the Composite Bacterial Infection Index (CBII) in two distinct cases, using Receiver Operating Characteristic (ROC) curve analysis. First to discriminate between the absence and the presence of infection and secondly to discriminate between bacterial and viral infection. The area under the curve (AUC) is an index reflecting the overall accuracy of the diagnostic test. The AUC is actually the probability that the diagnostic test, which in our case is the composite index (CBII), correctly ranks a random pair of infected and noninfected children or a random pair of bacterial infected and viral infected children, based on their calculated Composite Bacterial Infection Index values.

The first analysis was performed in the whole study population, consisting of 69 children with infection (group A) and 69 healthy children (group B). Neutrophils, Monocytes, ESR and CRP were significantly higher; whereas Lymphocytes were significantly lower in group A. Composite Bacterial Infection Index was also significantly higher in group A (Table 1). Most of the children were correctly classified between two groups without much loss (cost) in sensitivity and specificity. One typical Composite Bacterial Infection Index cutoff value of ≥ 32.45 can correctly classify 88.41% of study children with 85.51% sensitivity and 91.3% specificity. AUC value (± S.E.) of 0.9389 ± 0.02 can be considered highly accurate (Table 3).

Table 1: Characteristics of the study population. Figures are expressed as mean (±SD), median

The ROC scatterplot of Composite Bacterial Infection Index in diagnosis of infection in the whole study sample confirms a highly accurate test (Figure 1).

Figure 1: ROC curve of composite index in the diagnosis of infection in whole study sample

The second analysis was performed in the group of children with acute infection, consisting of 52 children with viral (group A1) and 17 children with bacterial infection (group A2). Neutrophils, ESR and CRP were significantly higher whereas Lymphocytes and Monocytes were significantly lower in children with bacterial infection (group A2). Composite Bacterial Infection Index was also significantly higher in children with bacterial infection (Table 2).

Table 2: Characteristics of the subgroups in children with infection. Figures are expressed as mean (± SD), median

Table 3: Cut off value of Composite Bacterial Infection Index in the diagnosis of infection

The second ROC analysis of Composite Bacterial Infection Index capability to discriminate between children with bacterial from those with viral infection also resulted in remarkable findings. To discriminate between viral or bacterial agent one typical Composite Bacterial Infection Index cutoff values of ≥ 2044 can correctly classify 86.96% of infected children with 82.35% sensitivity and 88.46% specificity.

The AUC value (± S.E.) of 0.9174 ± 0.03 can similarly be considered highly accurate (Table 4). The ROC scatterplot of Composite Bacterial Infection Index in diagnosis of bacterial infection in the subgroup of children with infection indicates also a highly accurate test (Figure 2).

Table 4: Cut off value of Composite Bacterial Infection Index in the diagnosis of Bacterial Infection

Figure 2: ROC curve of composite index in the diagnosis of bacterial infection

Acute febrile infections are very common among infants and young children and a frequent reason for visits in primary care settings. Febrile children usually have viral self-limiting infections. Even though, many times they receive unnecessarily antibiotic treatment. Nowadays, the cost of antibiotics along with their over-the-count use is a common problem for every medical unit. In the emergency department the differentiation of viral from bacterial illness is based on the clinical appearance of the patient along with basic biochemical and hematological parameters. Many inflammatory indices have been studied in order to discriminate viral from bacterial infection in febrile infants and young children. This age group is considered highly vulnerable to serious bacterial infections even though the vaccination against H. influenzae and S. pneumoniae had decreased the percentage of severe invasive disease.

The aim of our study was to evaluate the parameters from complete blood count in combination with CRP and ESR in order to find a low-cost, reliable and easily applicable formula to distinguish the presence from the absence of infection and to investigate whether the proposed formula might be used in differentiating bacterial from viral infection in the emergency department setting avoiding thus unnecessary antibiotic treatment.

Well established infection markers such as total white blood count, absolute neutrophil count, CRP and procalcitonin have been evaluated in different underlying pathologies [5-9]. Several studies had evaluated the total white blood count as a screening tool in children with serious bacterial infection with disappointing results. White blood cell count by itself was not an accurate predictor for bacterial infection as it has low sensitivity and specificity [10-16].

The most recent study by Hornik et al. evaluated the total white blood count, absolute neutrophil count, high immature – to – total neutrophil ratio and low platelet count in infants with late-onset sepsis and found that no complete blood count index possessed high sensitivity in anticipating late-onset sepsis reliably [10]. Gendrel et al. compared the concentration of CRP, procalcitonin, interleukin 6 (IL-6) and interferon – a in paediatric patients aiming at differentiation of viral from bacterial infection They concluded that the best parameter was procalcitonin (83% sensitivity and 93% specificity with a cutoff of 1.0 ng/ml) as the concentration of the CRP was above 10.0 in 47% and above 20.0 in 26.9% of children presenting with viral infection [8].

Pulliam et al. demonstrated that CRP performs better in predicting severe bacterial infection in febrile children less than 36 months of age compared to leukocyte and neutrophil count [14]. In a recent study performed by Andreola et al. on a group of patients under 36 months of age demonstrated that CRP and procalcitonin were the only factors that can predict severe bacterial infection [1]. The same results concerning the C-reactive protein and procalcitonin value in evaluating young children with bacterial or viral infection were demonstrated by the study of Olaciregui et al. [17]

In our study we evaluated the possible role of the combination of separate haematological parameters along with CRP and ESR (formula 1). Our goal was to evaluate the aforementioned formula as an indicator of the presence of acute infection and to examine whether the application of the Composite Bacterial Infection Index (CBII) in infants and young children could reliably discriminate the bacterial from viral agent. As showed, the ROC scatterplot of CBII in diagnosis of infection in the whole study sample yielded in a high accuracy of the proposed formula as far as the differentiation between the presence and the absence of infection is concerned. In the second analysis, which was performed in the group of children with acute infection, CBII was also significantly higher in children with bacterial infection. The second ROC analysis of CBII capability to discriminate bacterial from viral infection similarly yielded a highly accurate screening test. The application of CBII in the emergency department provides a cost-effective and easily performed solution in the evaluation of febrile children with good clinical appearance by without a source. Differential diagnosis between bacterial and viral infection may benefit further from the inclusion of CBII, Additionally the future use of the CBII in children with co-morbities (such as febrile neutropenic cancer children) might be useful in discriminating viral from bacterial infections in such a vulnerable group of patients.

Our study has several limitations due to the small number of patients of the study group. Additionally the CBII could not be applied in neutropenic cancer patients receiving chemotherapy or in neutropenic septic patients. In those cases the underlying pathology along with the deteriorated clinical condition are the basic criteria of the evaluation. Our goal was the detection of bacterial infection in a previously healthy and well-appearing febrile child.

In conclusion, we have shown, that CBII could be a reliable diagnostic tool to discriminate the presence from the absence of infection as well as differentiate viral from bacterial infection in infants and young children. The application of the CBII is superior to total white blood count or CRP alone for anticipating bacterial infection, especially in the emergency department where prompt decision making is of significant importance in order to avoid unnecessary antibiotic treatment. Before we can recommend BCII for more widespread use we recommend the formula to be used in larger studies that may aid in finding cutoffs with clinically useful predictive values.