Respiratory syncytial virus (RSV) is the most important viral cause for lower respiratory infection in infants and young children throughout the world.¹ It is one of the commonest causes of respiratory tract infection leading to respiratory failure. It has been estimated that in each year 600 000 deaths occur worldwide that are directly or indirectly attributable to RSV.¹ Factors that increase susceptibility to the virus include chronological age less than 6 weeks, bronchopulmonary dysplasia, congenital heart disease, prematurity, and immunodeficiency.^2–5 Although the mortality rate for those admitted to hospital may be as low as 1–3%, it increases in those with severe bronchiolitis requiring intensive care management.^2–6 In developed countries about 2% of infants and children admitted to hospital with RSV require assisted ventilation.⁷ RSV bronchiolitis is a common cause for admission to a paediatric intensive care unit (PICU) in the winter season.^5–7

The pharmacological management of RSV bronchiolitis, other than the use of supplementary oxygen, has long been debated.^8,9 In particular, many advocate against the routine use of antibiotics in bronchiolitis because of a reported low incidence of concurrent or secondary bacterial infections in patients with RSV.^10–17 However, these studies focused on extrapulmonary bacterial co-infection and included only limited numbers of children with severe respiratory compromise/failure.

Physiologically, the lower airways are normally sterile. Nevertheless, the relationship between bacterial co-infection and viral respiratory disease has previously been recognised, having an escalating incidence with increasing severity of respiratory illness.¹⁸ Three retrospective studies investigated the occurrence of bacterial co-infection in children with severe RSV infection requiring PICU admission and found the incidence of pulmonary bacterial co-infection to vary between 17.5% and 44%.^19–21 In this study we prospectively investigated the incidence of pulmonary bacterial co-infection using established quantitative microbiology²² in patients with severe RSV bronchiolitis on admission to a tertiary PICU, and evaluated the impact of the bacterial co-infection on morbidity and mortality.

METHODS

The study group comprised children admitted to the PICU at the Royal Liverpool Children’s Hospital, a university affiliated multidisciplinary regional referral centre. The PICU is a 20-bed facility with an annual admission rate of over 1000 children. The overall mortality rate is 4.5%, with a predicted mortality of 6.25% using the paediatric index of mortality²³ and a standardised mortality rate of 0.72.

The main objectives of the study were (1) to determine the incidence of pulmonary bacterial co-infection in patients requiring admission to the PICU for severe RSV bronchiolitis; and (2) to study the impact of the co-infection on morbidity (including length of ventilation and inflammation) and mortality.

Children with RSV bronchiolitis, confirmed by RSV antigen testing and/or culture, requiring PICU admission and from whom lower airway secretions were obtained on admission were included in the study. Those with nosocomial RSV infections were excluded. Data were collected prospectively during three consecutive RSV seasons (winter) between 2002 and 2005 from RSV positive children admitted to the PICU.

The study was approved by the institutional ethics review board.

Respiratory support

Intubation was performed by our PICU retrieval team at the referring hospital, in our accident & emergency (A&E) department, or in one of the hospital wards prior to PICU admission. Alternatively, the anaesthetic team of the referring hospital intubated some of the patients before the arrival of the PICU retrieval team. It is policy that all children who require intensive care and ventilatory support are moved to the regional PICU.

The timing of extubation was judged clinically and not influenced by bronchoalveolar lavage (BAL) results.

Microbiological sampling

Diagnostic samples of nasopharyngeal aspirates (for RSV detection) and lower airway secretions (for bacterial culture) through endotracheal tube using sterile precautions²⁴ were taken on admission and processed immediately in the laboratory. Prior to routine bronchial toilet, a sterile suction catheter was passed down the endotracheal tube. Two 1 ml/kg aliquots of sterile 0.9% saline were instilled through the suction catheter, immediately followed by aspiration with constant pressure into a mucus trap. Samples were collected by specialist respiratory physiotherapists or PICU staff members. BAL was performed immediately after endotracheal intubation in children intubated in the hospital and on arrival in the PICU, and generally within 3 hours of endotracheal intubation for those admitted from other hospitals. All children within the region are only ventilated in the regional PICU, so are rapidly transferred to the PICU.

Surveillance samples of throat and rectum were obtained on admission and then twice weekly, in keeping with the routine surveillance practice in our unit.

Laboratory procedures

Viral

Nasopharyngeal aspirates were tested by the Directigen RSV test (Becton Dickinson Microbiology Systems, Maryland, USA). This is an in vitro enzyme immunoassay (ELISA) membrane test for the rapid and qualitative detection of RSV antigen directly from nasopharyngeal specimens. All samples negative for RSV using the ELISA membrane test were cultured using standard virological techniques at the Health Protection Agency.

Bacterial/yeast

Diagnostic or clinical samples were processed immediately in a qualitative and semi-quantitative way using standard microbiological methods. For all types of samples, macroscopically distinct colonies were isolated in pure culture. Standard methods for identification, typing, and sensitivity patterns were used for all micro-organisms.²⁵

Antibiotic treatment

Patients with signs of infection received intravenous cefotaxime (150 mg/kg/day four times daily for up to 7 days) as first line treatment for 48 hours while awaiting culture results. Clinical status on presentation governed whether supplementary intravenous cover with an aminoglycoside (gentamicin 7.5 mg/kg/day three times daily for up to 7 days) was added. Antibiotics were rationalised once culture and sensitivity results became available.

Definitions

Bacteria positive: the presence of micro-organisms in the lower airways which is normally sterile.

Co-infection: Infection is a microbiologically proven, clinical diagnosis of inflammation, local and/or generalised. In this study clinical signs were unreliable as all patients had bronchiolitis, so microbiological definitions were used. Bacterial co-infection required bacteria colony counts ⩾10⁵ cfu/ml of diagnostic sample for each single species obtained from lower airway secretions and, on a semi-quantitative scale of +  =  few (<50×10⁶/l), ++ = moderate (>100×10⁶/l), and +++  =  many leucocytes (>1000×10⁶/l), the presence of at least a moderate (++) number of leucocytes.^26–28

Low bacterial growth: Diagnostic samples from lower airway secretions which yielded <10⁵ cfu/ml of diagnostic sample and the presence leucocytes.

The chest radiographic appearance was not used to diagnose bacterial co-infection as changes on the chest radiograph are not pathognomonic of secondary bacterial or viral infections.^18,29

Analysis of data

Data were collected prospectively. Prediction of mortality using the paediatric index of mortality was obtained on the patient’s first contact with the PICU team.²³ Results were expressed as a percentage of the total study population; median and interquartile ranges (IQR) were used to describe the demographic distributions.

Continuous data were analysed using the Wilcoxon-Mann-Whitney (W-M-W) test. Categorical data were analysed using Fisher’s exact or McNemar’s test. Correlation was assessed using Spearman’s rank test (two tailed). Multivariate analysis was performed using linear and logistic regression analysis.

Statistical calculations were performed with the Statistical Program for Social Science release 11.0.0 (SPSS 11, Chicago, IL, USA). A p value of <0.05 was considered statistically significant.

RESULTS

A total of 181 children (103 boys and 78 girls) of median age 1.6 months (IQR 0.5–4.6) were admitted to the PICU with RSV positive bronchiolitis during the three consecutive RSV seasons (2002–5). The indication for PICU admission for these children was ventilatory/respiratory support (respiratory failure (n = 172) and/or life threatening apnoeas (n = 9)). All patients were mechanically ventilated for a median of 5.0 days (IQR 3.0–7.3). 165 children were enrolled in the study; an admission BAL sample was not available in 16 patients (8.8%).

The demographic characteristics, inflammatory markers, antibiotic history, and mortality of the RSV positive children in the subgroups RSV only, bacterial co-infection, low bacterial growth, and bacteria positive (co-infection + low bacterial growth) are shown in table 1. The white cell count, neutrophil count, and C-reactive protein (CRP) levels did not differ between the groups on admission or during days 1–5 in the PICU.

Although all patients were admitted primarily for respiratory disease, 43% (71/165) of them had other co-morbidities (congenital heart disease n = 37, chronic lung disease n = 8, immunodeficiencies n = 4, abnormality of large airways n = 5, congenital heart disease and abnormality of large airways n = 8, congenital heart disease and chronic lung disease n = 4, neuromuscular disease n = 7). Co-morbidity did not increase the risk of positive bacterial cultures (odds ratio 0.77, 95% CI 0.55 to 1.09).

Overall, 45% (74/165) received antibiotics before admission to the PICU (that is, started by the referring hospital or ward), most often cefotaxime or ceftriaxone. The breakdown between the subgroups is shown in table 1. Receipt of antibiotics before PICU admission did not affect the paediatric index of mortality (p = 0.6, W-M-W test) and length of ventilation (p = 0.2, W-M-W test). All except eight patients were continued or commenced on antibiotics in the PICU (usually cefotaxime). Antibiotics were continued for a median of 5 days (IQR 3–6). The empirical use of antibiotics was at the discretion of the attending consultant.

Sex, age, paediatric index of mortality, co-morbidity, receipt of prior antibiotics, time on antibiotics before intubation, admission oxygen and ventilation index were not predictive of positive bacterial cultures by univariate or multivariate analysis (all p values >0.16).

The organisms isolated from lower airway secretions obtained on admission are shown in table 2. All those with positive endotracheal bacteriological specimens had the same organisms isolated on admission surveillance swabs. Community organisms accounted for 83% (81/98) of the bacteria cultured.

There were 12 deaths (6.6%), five of which (2.8%) appeared to be RSV related as the patients were still RSV positive when they died. Two patients with leukaemia on chemotherapy died from RSV pneumonitis on days 1 and 16, respectively. Neither had proven bacterial co-infection and both received broad spectrum empirical antibiotic treatment. Other associated causes included singles cases of hypoplastic right heart coupled with cystic fibrosis (on day 8), B pertussis co-infection with hypoxaemic respiratory failure requiring extracorporeal membrane oxygenation (on day 26), and a child with a congenital myopathy (on day 8). The remaining seven deaths occurred 6–31 days after admission subsequent to the RSV cultures becoming negative. Causes of these RSV “unrelated” deaths included complex congenital heart disease (n = 3), multiple congenital anomalies (n = 2), congenital myopathy (n = 1), anoxic brain injury (n = 1). Positive bacterial cultures did not predict death (odds ratio 1.3, 95% CI 0.57 to 2.95), but co-morbidity did (odds ratio 0.51, 95% CI 0.37 to 0.7).

DISCUSSION

This observational study over three consecutive RSV seasons evaluating bacterial pulmonary co-infection found that 42% of children admitted with severe RSV infection harboured bacterial pathogens in their lower airways. These critically ill children run a serious risk of developing bacterial pneumonia.^18,31

BAL samples were collected very soon after intubation so significant growth densities of bacteria reflect pathogens in the normally sterile lower airways. The high number of colony forming units makes it highly unlikely that the micro-organisms isolated were “pushed down” the trachea on intubation. We acknowledge that the number of leucocytes in lower airways secretions will also be influenced by RSV infection and therefore relied on bacterial growth. The microbiological criteria were strict and avoided potentially confounding clinical factors. This microbiological approach is supported by recent literature concerning ventilator associated pneumonia (VAP).^32,33 On the other hand, it must be appreciated that our study group was very different from this VAP group as they had “virgin” iatrogenically uncontaminated lower airways. Certainly, in the group with co-infection, substantial bacterial growth densities occurred far too soon after intubation to have been oropharyngeal flora transported there by the endotracheal tube. If anything, the strict microbiological criteria probably underestimated the number co-infected by categorising many of them as having low bacterial growth. We accept that differentiating the groups into “co-infected” and “low bacterial growth” may be somewhat artificial as the lower respiratory tract should be free from bacteria.

The term “co-infection” was used as, at the time of PICU admission, these infections could either be secondary or concurrent. It would not be easy to detect the “chicken” from the “egg” as far as which was primary—the RSV or the bacteria—although a viral infection destroying cilia is in general required for a bacterial co-infection.¹⁸ The true co-infection rate is likely to be higher than the 22% rate detected, as 45% of the cases received antibiotics before admission to the PICU. These antibiotics may have converted some of the “co-infection” patients into the “low bacterial growth” group, or even prevented bacterial growth altogether.

Previous studies have examined bloodstream, otitis media, or urinary tract infections in children with bronchiolitis, very few of whom had severe RSV bronchiolitis requiring intensive care.^10–17 These studies generally found a very low incidence of secondary serious bacterial infection (1.2%)¹⁴ or bacteraemia (0.6%)¹⁶ in their hospitalised RSV patients. Because these studies did not specifically concentrate on those with severe bronchiolitis, it is difficult to extrapolate their results to this population. Duttweiler et al retrospectively studied 127 infants admitted to intensive care for RSV bronchiolitis and found that 25 (44%) of the 57 ventilated and endotracheally sampled infants had “concomitant bacteria pneumonia”.¹⁹ Similarly, the retrospective study of Kneyber et al²⁰ (82 PICU admissions with 65 (79%) ventilated) found that nine (33%) of the 24 children on whom admission endotracheal aspirates were performed had a positive bacterial culture. Randolph et al²¹ retrospectively examined 165 previously healthy infants admitted to the intensive care unit over a 12 year period with laboratory confirmed RSV infection, 63 (38%) of whom required mechanical ventilation. They found that 17.5–38% of the 63 intubated infants had “probable” or “possible” bacterial pneumonia. The incidence of bacterial pulmonary infection in these retrospective PICU reports is in keeping with that of this prospective study in which all bronchiolitic admissions were included.

Fifty one percent of the patients with bacteria in their airways and 40% of the children with RSV only had co-morbidities (congenital heart disease, chronic lung disease, large airway abnormality, immunodeficiency, neuromuscular disease). This is in keeping with well recognised risk factors associated with more severe RSV disease.^3,4,6,16,34 Co-morbidities did not account for differences in length of ventilation between the study groups, but did contribute towards mortality. The high percentage with co-morbidities is most probably also influenced by the fact that our centre is the regional paediatric cardiac referral centre, which means that children with congenital heart disease and bronchiolitis are more likely to be referred to our PICU for intensive care management.

There were fewer deaths in the bacteria positive group than in those with RSV only. However, when adjusted for those children who had recovered from their RSV infection only to die later from RSV unrelated causes, both groups had similar mortalities (2.9% v 3.2%). The paediatric index of mortality is a point of first contact score that is used to assess the risk of death while in the PICU.²³ The paediatric index of mortality scores for all the groups were similar, suggesting that all groups had matching severity of illness on admission to the PICU. Yet those with positive bacterial cultures required ventilatory support for longer than those with RSV only. Kneyber et al²⁰ reported a similar finding. Although length of ventilation was significantly different between the groups, other respiratory support and inflammation indices did not differ between them (table 1). Perhaps the general inflammatory response once triggered by RSV is not so refined as to be further enhanced by concomitant bacterial infection. Others have also found inflammatory markers unhelpful in differentiating bacterial infection in this group of patients.^20,35,36 Unfortunately, we were unable to find any early clinical measurements which would identify which RSV patients had bacterial co-infection.

Receipt of prior antibiotics and length of time on them did not predispose to bacterial co-infection. Moreover, many of the children with RSV had received antibiotics for only one day or less (often a single dose close to intubation). The fact that nearly all the RSV positive children received antibiotics in our PICU limited any interpretation on the impact of antibiotics on their outcome. All those patients with positive bacteriology in their endotracheal secretions had the same organisms isolated on admission surveillance swabs, indicating primary endogenous infection.³⁷ This reinforces the view that potential pathogens are carried first in the nasopharynx and then there is migration down the trachea into the lower airways.^37,38 The organisms isolated on admission were generally normal community organisms because most of the patients were in good health before RSV infection and PICU admission.^27,29Pseudomonas aeruginosa was the most common of the abnormal bacteria (table 2). All these patients were carriers of abnormal organisms in their throats, and in most the common denominator for their abnormal carriage was chronic illness.^37,39 Interestingly, Streptococcus pneumoniae was isolated from relatively few patients. This could be the result of prior antibiotic use.⁴⁰

Although most LRTI in children are viral in aetiology, mixed viral/bacterial infections are seen in up to a quarter of hospitalised children.^41,42 In addition, there is a risk of developing bacterial superinfection with viral LTRI.¹⁸ These issues have contributed to the recommendations by the World Health Organization that the treatment of community acquired pneumonia should include empirical antibiotics.^41,43 Concerns that using antibiotics (in our case cefotaxime) pre-emptively in this group of critically ill children would breed antibiotic resistance have been shown to be unfounded in a 4 year study.⁴⁴ Assessment of the influence of antibiotics on children with severe bronchiolitis would require a prospective randomised controlled trial.

This study has shown that up to 40% of patients admitted with severe RSV bronchiolitis were infected with bacteria in their lower airways. Co-morbidity (congenital heart disease, chronic lung disease, large airway abnormality, immunodeficiency, neuromuscular disease) predisposes to more severe RSV disease.