Background
Critically ill children are at high risk for severe healthcare associated infections (HAI) due to invasive devices and procedures, secondary immunosuppression and underlying diseases [
1‐
3]. Multidrug-resistant (MDR) Gram-negative infections are an increasing threat to children admitted to the Paediatric Intensive Care Unit (PICU). However, predisposing factors for colonization and infection with MDR Gram-negative organisms are unclear in this vulnerable patient population.
MDR Enterobacteriaceae have become a particular concern for mechanically ventilated patients [
4]. Placement of an endotracheal tube is followed rapidly by tracheal colonization with potentially pathogenic microorganisms from the oropharyngeal flora, including MDR organisms [
5,
6]. Colonization of the lower respiratory tract by endogenous flora or opportunistic pathogens provides the major route to acquiring ventilator-associated pneumonia (VAP) [
7]. Moreover, increased nasopharyngeal bacterial density is associated with a higher risk of invasive respiratory disease [
8].
Data characterizing tracheal colonization in mechanically ventilated children independent of respiratory infection are scarce and vary substantially among hospitals and across countries as well as throughout the course of mechanical ventilation [
9]. A clear predominance for Gram-negative organisms in ventilated PICU patients has been described in a paediatric colonization study in India [
10]. Furthermore, endotracheal colonization was dominated by Enterobacteriaceae, in particular
E.coli and
Enterobacter, in two recent studies of VAP in adults and children [
4,
11].
However, bacterial colonization does not necessarily imply infection and tracheal aspirates lack specificity for VAP [
5]. Nevertheless, tracheal aspirates are part of the Center for Disease Control’s (CDC) criteria for the diagnosis of VAP and are frequently used to guide antibiotic therapy in PICU [
12].
Empiric antibiotic treatment of Gram-negative infections is becoming increasingly difficult, because antibiotics that were previously considered the treatment of choice are no longer useful in MDR Gram-negative organisms [
13‐
15]. If initial antimicrobial therapy is ineffective and only few treatment options remain in critically ill children, recurrence of infection, morbidity, mortality, length of PICU and hospital stay as well as healthcare costs will rise [
13,
14]. Empiric antibiotic therapy might be improved by the identification of risk factors for colonization and infection with Enterobacteriaceae.
Risk factors for the acquisition of Enterobacteriaceae, especially extended-spectrum ß-lactamase (ESBL) producing organisms were investigated in adults during a stay in the intensive care unit. Major risk factors for infection due to ESBL-producing bacteria were: travel to high-prevalence countries, prior antibiotic use and mechanical ventilation [
16]. Similar potential risk factors for infection due to MDR Enterobacteriaceae have been identified in neonates and children including the presence of chronic disease, previous hospitalization, invasive ventilation, pre-term low birth weight and antibiotic intake [
17]. In particular, prior use of cephalosporins has been defined as an independent risk factor for the acquisition of MDR Enterobacteriaceae [
18,
19]. Preceding antibiotic therapy may lead to a disruption of the intestinal flora and facilitate colonization and overgrowth of nosocomial MDR Gram-negative organisms [
20]. These MDR strains increase the risk of infection by progressive colonization of the gastrointestinal and subsequently the respiratory tract during a hospital stay [
21].
To our knowledge, clinical risk factors for infection of mechanically ventilated PICU patients with MDR Enterobacteriaceae have not been described to date. Knowledge of potential clinical risk factors for infection with MDR Enterobacteriaceae might help to initiate appropriate infection control precautions to prevent transmission of MDR bacteria in PICU and improve empiric antibiotic treatment.
Therefore, we aimed to determine the incidence and spectrum of MDR Enterobacteriaceae in children with suspected respiratory tract infection admitted to a large academic PICU from 2005 to 2014. Additionally, we analysed risk factors for the finding of MDR Enterobacteriaceae in these mechanically ventilated children.
Discussion
In this study we analysed the spectrum of Enterobacteriaceae in tracheal aspirates of intubated PICU patients from 2005 to 2014. The spectrum of Enterobacteriaceae in lower respiratory tract material revealed
Enterobacter spp., E.coli and
Klebsiella spp. as the most common isolates (86%). Comparable data of matching study settings are scarce in the current literature. Wilson et al. [
5] collected daily tracheal aspirates from intubated children. Consistent with our findings, the most common Gram-negative organisms isolated were
Klebsiella spp. and
E.coli, followed by
Citrobacter freundii and
Enterobacter cloacae. Lee et al. [
34] described the microbiological spectrum and susceptibility pattern of clinical isolates from a PICU and found a rate of 20% ESBL-positive
Klebsiella in 2005. Our study displayed a lower rate with only 6.5% of
Klebsiella spp. isolates being ESBL-positive. However, more than half of
E.coli isolates (55%), about a quarter of all
Klebsiella spp. (28%) isolates and 4 out of 6
Morganella (67%) isolates were MDR in the present study (Additional file
3). This finding is consistent with two large, nationwide studies of antibiotic drug use and bacterial resistance in the United States and in 53 German ICUs [
35,
36]. In the latter, the most striking result was the ten-fold increase of 3
rd generation cephalosporin-resistant
E.coli from 2001 to 2008. Highly resistant
E.coli are known to have a 30% increase in infection rate and a higher mortality compared to susceptible isolates [
37]. Keeping this in mind, MDR
E.coli should alert PICUs to the prevalence of MDR
E.coli.
In total, 167 Enterobacteriaceae isolates from lower respiratory tract samples were identified in 123 ventilated PICU patients. 43 (35%) patients were infected with MDR Enterobacteriaceae. Patient characteristics, including the proportion who were infants, days of antibiotic pre-exposure, presence of a CVC and underlying diseases did not differ between patients infected with susceptible versus MDR organisms. The only exception were patients with immunodeficiencies, who were significantly more likely to be infected with an MDR organism.
Clinical outcome of infection with MDR and susceptible Enterobacteriaceae was investigated in this study. Infection with MDR Gram-negative organisms has been associated with longer length of hospital stay or length of PICU stay [
38]. We investigated whether patients infected with MDR organisms had more ventilated days in total, a longer length of PICU stay, a higher incidence of VAP or a higher all-cause mortality. Clinical outcome was similar in both groups in our study. This finding is to be expected, since MDR organisms commonly do not feature higher pathogenicity than their more susceptible counterparts of the same genera [
39]. However, mortality from infection with MDR organisms is known to be higher due to the delay and a lower rate of appropriate empiric treatment [
37,
40]. In our institution, susceptibility testing is rapidly available and empiric antibiotic therapy is rather aggressive. Differences in outcome may hardly be detectable in our study based on a low number of cases and the fact that PICU patients have several other underlying factors which may influence ventilated days, length of PICU stay, incidence of VAP and mortality.
Knowledge of potential clinical risk factors for infection with MDR organisms might help to improve infection control precautions, diagnostics and empiric antibiotic therapy. We investigated several factors described in the literature that could potentially increase the risk of infection with MDR Enterobacteriaceae [
41,
42]. Of these potential risk factors, two were identified by the decision and regression tree analysis as the most important: (i) the duration of antibiotic exposure and (ii) gastrointestinal comorbidity were most relevant for infection with MDR Enterobacteriaceae.
Following the tree-based structure of the CRT analysis, antibiotic pre-treatment for ≥ 7 days increased the risk of infection with an MDR isolate to 60%. Furthermore, gastrointestinal comorbidity increased the chance of MDR infection from 25 to 40% in patients with a short (<7 days) duration of antibiotic therapy. It is well known that the stomach represents a reservoir for Gram-negative bacilli, especially in critically ill children who are fed by nasogastric tubes or treated with H (2) antagonists [
43]. Transmission of pathogens to the lower respiratory tract may be facilitated in ventilated children with gastrointestinal comorbidities due to a higher rate of contamination of the hands or apparel of healthcare workers, contaminated respiratory equipment and micro-aspiration of stomach contents and flora into the lower respiratory tract. Long-term antibiotic use in these children (especially with cephalosporins) increases selective pressure on Gram-negative bacilli of the gastrointestinal and oropharyngeal flora, resulting in a higher rate of MDR Enterobacteriaceae [
28].
Multivariable logistic regression supported the findings obtained in the CRT model with slight differences. Compared to the tree-based model, the adjusted odds ratio for gastrointestinal comorbidity did not reach significance. However, antibiotic exposure for ≥ 7 days remained the most relevant prognostic factors for infection with MDR Enterobacteriaceae. The OR of 4.56 (95% CI 1.69–12.30) for antibiotic exposure for ≥ 7 days indicates that after antibiotic therapy of 7 days or more, the risk of infection with MDR Enterobacteriaceae increases 4.56 times for every additional day of treatment. Our finding is consistent with numerous papers that have demonstrated antibiotic exposure as a strong risk factor for infection with MDR organisms [
6,
18,
44]. Furthermore, this study investigated another important issue: namely, at which particular time an antibiotic might select MDR organisms in ventilated PICU patients. Our CRT analysis revealed a critical cut-off at 7 days of prior antibiotic treatment to increase the risk of infection with MDR Enterobacteriaceae. This cut-off at 7 days of antibiotic pre-treatment was validated by multiple logistic regression. In a previous study, children received antibiotics for clinician-suspected Ventilator-Associated Tracheitis (VAT) and the cut-off at 7 days of antibiotic therapy discriminated between short- and prolonged-course therapy, similar to our study. The hazard of colonization or infection with an MDR organism was more than 4 times greater if children had received a prolonged-course of antibiotic treatment, whereas short-course therapy had a significantly lower incidence of MDR infection and did not affect clinical outcome [
45].
The present study has several limitations. First, this study was conducted in a retrospective observational manner and comprised a single PICU in a University Children’s Hospital. Therefore, data may be of limited applicability for PICUs in different settings. Second, assessment of risk factors for infection with MDR Enterobacteriaceae may have methodological deficits and may be biased with regards to estimates and associations. Selection bias towards infants and sicker patients may have affected our results. It is possible that tracheal aspirates were more likely to be taken in infants who were ventilated for a longer time after surgery for congenital heart or gastrointestinal defects. Consequently, infants might be overrepresented in our study. Additionally, sicker patients might have been more likely to be infected with an MDR pathogen. Furthermore, we could not exclude that the association of potential risk factors and infection with MDR Enterobacteriaceae was influenced by confounding factors. The patient population investigated in this study is a highly heterogeneous group with many potential confounders that are not uniform. Multiple hospitalisations, chronic diseases, congenital malformations, medication like H (2) blockers and antibiotic prophylaxis (e.g. with cephalosporins) may influence the rate of infection with MDR organisms. In particular, it is likely that severely ill children were treated with longer courses and a broader spectrum of antibiotics. These children may have a higher risk for infection with MDR organisms. Third, since there was no active screening policy for Enterobacteriaceae in tracheal aspirate on our ward, only patients with suspected infection were included into the study. Lastly, it should be noted, that the baseline classification odd (MDR versus susceptible organism) was 50%. Consequently, an overall predictive value of 71% of the CRT model implies a 21% increase in predictive accuracy.