Infectious disease is the main reason for care seeking in paediatric Emergency Departments (EDs). At times, the evaluation of these patients requires microbiological studies, and blood culture (BC) is among the tests most frequently ordered.1 Fever without a source in infants at risk of potentially severe bacterial infection is the main reason for requesting BCs, and they are also frequently performed in febrile patients with immunosuppressive conditions and in children with bacterial infections requiring hospital admission for parenteral antibiotic treatment before starting its administration.2

Following the introduction of routine vaccination with the seven-valent pneumococcal conjugate vaccine (PCV), the rate of occult bacteraemia (OB) has decreased drastically to values below 0.5–1%.3–7 It is expected that the universal administration of the 13-valent PCV will lower this rate even further in reducing other forms of invasive pneumococcal disease. In this new scenario, most BCs performed in EDs that show bacterial growth turn out to have been contaminated.3,5,7 Furthermore, the continuous-monitoring BC systems that are currently available can detect bacterial growth in the first 24h in most instances, while in the past detection occurred considerably later.8 This decrease in detection time has a significant impact on the therapeutic approach to these patients, as in this reduced timeframe most of them are still febrile when they return for reevaluation, which may lead to a more aggressive management. In this emerging reality, the early identification of these patients is essential to avoid the unnecessary use of healthcare resources that often ensues, especially in the case of patients discharged to their homes (return visits, additional diagnostic testing, hospital admission and/or initiation of antibiotic treatment, etc.),3,9 thus minimising the anxiety caused to patients and their families. While different variables have been proposed as possible predictive factors for contamination in BCs with bacterial growth, such as time to positivity and Gram stain results,3,9–11 we are not aware of any studies on the subject published after the changes discussed above that have occurred in recent years.

The primary objective of this study was to identify predictive factors for contamination in BCs with detected bacterial growth performed in the ED in febrile patients 1–36 months of age, and the secondary objective was to analyse the impact of BCs found to be contaminated in patients discharged to their homes.

We conducted a prospective observational and analytical study in a tertiary referral maternity and children's hospital with 275 paediatric beds that serves a catchment population of 1800000 inhabitants and receives a mean of 95000 paediatric emergency visits a year.

Using the electronic database of the hospital's laboratory, we selected the medical records of patients 1–36 months of age with a BC ordered by the ED in which bacterial growth was detected. The period under study was 2 years (November 2011–October 2013). We included febrile patients with no risk factors for bacteraemia. The study included patients with fever without a source and also patients in which the cause of the fever was identified in the ED. Fever was defined as an axillary temperature of 38°C or higher at home or at the ED, and the risk factors for bacteraemia included the presence of an immunosuppressive condition; having an indwelling catheter, prosthetic valve or cerebrospinal fluid shunt; and/or a history of hospital admission or aggressive diagnostic or therapeutic procedures in the week preceding the ED visit.

Blood extractions for BC in the ED were ordered based on the judgement of the doctor in charge, in compliance with current protocols.12

Blood culture specimens were collected by the nursing staff of the ED by percutaneous venepuncture.13 The blood volume recommended for the type of BC bottle used is 4mL. However, minimum volumes have been established depending on the age of the patient (1mL for children younger than 3 months, 2mL for children 3 months to 3 years of age). After the blood was drawn, it was inoculated in a paediatric bottle (BacT/ALERT® PF) for aerobic culture and transported to the microbiology department for its immediate processing in the automated BacT/ALERT® incubation system (Biomérieux, Durham, USA) for a maximum period of 5 days. All BCs with bacterial growth were subjected to Gram staining and subcultured in different culture media (blood agar, chocolate agar) at 37°C. Whenever the microbiologist detected bacterial growth in a BC and, based on the patient's chart, considered it was likely to be pathogenic, the paediatrician in charge was notified, who in turn called the patient, if he or she had been discharged home, to schedule a reevaluation at the ED within a short timeframe.

We analysed the data for 169 BCs with bacterial growth, of which 30 (17.8%) were true positives and 139 (82.2%) were contaminated. Table 1 shows the clinical and epidemiological characteristics of both groups. Thirty-three (19.5%) BCs corresponded to patients 2–36 months of age with fever without a source and in good general condition that were discharged home. The bacteria isolated most frequently in positive BCs were S. pneumoniae (nine cases; 30.0%) and E. coli (nine cases; 30.0%), with a mean time to growth of 11.8 and 8.2h, respectively. Coagulase-negative staphylococci (92 cases; 66.2%) were the most frequent contaminants of BCs, and had a mean time to growth of 18.0h. Table 2 shows the distribution and time to positivity of the identified bacterial pathogens and contaminants. The Gram stain results were suggestive of a contaminant in 124 (73.4%) BCs, suggestive of a pathogen in 44 (26.0%), and negative in the rest (0.6%), with concordance between the initial Gram stain result and the final identification of the isolate in 167 (98.8%) cases.

The analysis of the predictive factors for contamination excluded seven cases due to missing data: three were missing laboratory data, three had no documented time to positivity, and one had a negative Gram stain. Table 3 shows the comparative analysis of the different laboratory and microbiological variables documented in patients with a positive BC and patients with a contaminated BC. The pathogenic bacteria isolated in the three patients with Gram stain results suggestive of contamination were S. aureus (two cases) and S. pyogenes (one case); all three had been admitted for antibiotic treatment from their initial ED visit after being diagnosed with hip arthritis, scalded skin syndrome and mastoiditis, respectively. The three predictive factors for contamination that had the highest statistical significance were CRP (area under the curve [AUC], 0.81; 95% CI, 0.75–0.87; cut-off point, 30mg/L), time to positivity (AUC, 0.80; 95% CI, 0.73–0.86; cut-off point, 16h) and the initial Gram stain result. Table 4 shows the sensitivity, specificity, PPV and NPP for these three variables, individually and combined.

The subanalysis of the 33 patients at risk of OB showed that the BC was positive in three (9.1%) and contaminated in 30 (90.9%). The bacteria isolated in patients with OB were S. pneumoniae in two cases and E. coli in one. Two of the three patients with a positive BC had C-reactive protein levels above30mg/L; all three had Gram stain results suggestive of a pathogen, and the time to positivity had been documented for two patients and was less than 16h in both.

When it came to the impact of contaminated BCs on the 48 patients that had been discharged home, four (8.3%) were scheduled for reevaluation at the ED, three had the BC redone, and two were prescribed antibiotics, with one of them admitted for parenteral administration until the definitive BC results became available.

To the best of our knowledge, this is the first study in the literature that analyses bacteraemia in febrile children 1–36 months of age in Spain following the introduction of the 13-valent PCV. The clear predominance of contaminated over true-positive BCs in cultures with bacterial growth has been described previously by other authors after the introduction of universal vaccination with the seven-valent PCV,3,4,14,15 who reported rates of contamination even higher than the one found in our study (5:1). At the same time, the remarkable proportion of bacteraemia caused by organisms other than S. pneumoniae in our setting, in which the PCV coverage during the period of the study is estimated at 60%, has also been reported by other authors.4,7,15–18 Both circumstances underscore the importance of the early differentiation between contamination and true bacteraemia in a BC with bacterial growth.

Our study found significant differences between the contaminated BCs and the positive BCs in the time to positivity and CRP values. The times to positivity of contaminants were considerably longer than those of pathogens, with a PPV of 96.9% for a time to positivity of 16h or longer. While this finding is consistent with those of other authors in previous studies,3,9–11,19 in our study the times to positivity for pathogens and especially for contaminants were remarkably lower in comparison, as the times to positivity reported in those other studies ranged from 13.8 to 19.9h for pathogens and from 31.1 to 37.6h for contaminants. These differences can mostly be attributed to the BC monitoring systems used in the previous studies, as most of them were performed in the 1990s. In recent years, the introduction of automated continuous-monitoring BC systems has brought about a significant reduction in the time elapsed until obtaining the results accompanied by a considerable increase in the overall positive rate.20 In our study, the level of CRP was the most useful laboratory parameter for the early identification of a contaminated BC, with a PPV of 95.1% for CRP levels at or above 30mg/L. As far as we know, the use of CRP levels for this purpose has not been analysed by other authors. The studies by Sard et al. and Segal and Chamberlain3,9 showed a statistically significant difference in the total leucocyte count between contaminated and positive BCs, and the former proposed a count of less than 15000cells/mm3 as a predictive factor for contamination with a PPV of 86.8%. While we also observed a statistically significant difference in the total number of leukocytes between the two groups in our sample, we found a greater difference in CRP levels. This can be attributed to the change in the distribution of pathogens that cause bacteraemia in recent years, after which leucocyte counts have become less helpful in their identification.14,16,17 In addition to these two factors, a Gram stain result suggestive of contamination stood out as the most useful individual parameter for the early identification of a contaminated BC, with a PPV of 97.5%. This finding, consistent with what has been reported by other authors,3,10 underscores the importance of the concordance between the initial Gram stain and the final identification of the isolate.21

Compared to other studies,3,22,23 there was a considerably low percentage of patients discharged home in whom the BC was contaminated that returned for reevaluation. The reason for this is that, in the centre where the study was conducted, the microbiology department generally notifies bacterial growth in BCs only if the initial Gram stain result is suggestive of a pathogen, or if it is suggestive of a contaminant but the review of the patient's chart raises the possibility of the positive result being clinically relevant. Fluid communication between paediatrician and microbiologist is required to ensure the correct selection of patients.24

There are several limitations to this study. Firstly, it focuses on febrile patients 1–3 months of age receiving care at a tertiary paediatric hospital, so the results may not be generalisable to the paediatric population outside this age group or seen in other types of healthcare facility. Second of all, the reduced number of patients with OB precluded the validation of the predictive factors for contamination in the subgroup of patients at risk of OB. Given the drastic decline in OB in recent years, it would be necessary to conduct a prospective multicentre study to validate the usefulness of the factors proposed for this population. Lastly, the retrospective design of the study did not allow us to determine the rate of BC contamination in this sample of patients, and we could not assess factors that potentially influence the contamination of a BC, especially during sample collection and transport. However, the overall contamination rate of BCs in the ED during the period under study was below 3%,2 in compliance with the generally accepted standards.

To conclude, at present most BCs with bacterial growth in febrile patients 1–36 months of age turn out to have been contaminated. Initial Gram stain results, the time to positivity of the BC and C-reactive protein values allow the early identification of contaminated cultures, so they ought to be taken into account by paediatricians when BCs show bacterial growth. The early identification of these patients will enable a reduction in unnecessary health expenditure associated with their reevaluation.

The authors have no conflicts of interest to declare.