Infectious acute respiratory failure in patients under 5 years of age: a retrospective cohort study

Introduction

According to the WHO, pneumonia is responsible for 20% of deaths of children under 5 years old, ergo 900 000 deaths worldwide in that age range every year; it is the first cause of death in this population.1–3 This infectious pathology disease raises interrogations on numerous topics: making a clinical diagnosis, assessing its severity, determining its aetiological origin, deciding on the appropriate therapy and the optimum type of respiratory support.

The presence of moderate hypoxaemia and signs of respiratory struggle are the symptoms that are most frequently associated with the diagnosis of pneumonia in children under 5 years old, while fever, tachycardia or auscultatory signs are not always specifically associated with this illness.1 4

Signs of severe pneumonia include paradoxical breathing, altered consciousness, hypoxia, decreased food intake and cyanosis.5 6 Overall, 12%–20% of patients under the age of 18 hospitalised with community-acquired pneumonia require paediatric intensive care management making this the first cause of hospitalisation in paediatric intensive care units (PICUs).7 8

Community-acquired pneumonia, as well as bronchiolitis, is essentially caused by viruses.4 There are no radiological or biological criteria that can absolutely distinguish between a bacterial or viral pneumonia. The chest X-ray, for example, may be normal even though the patient has a bacterial lung infection or a bacterial superinfection of a viral infection.5

Similarly, no single biological parameter or any combination of them, can formally differentiate between the bacterial and viral origin of a pulmonary infection.9 Some studies suggest procalcitonin (PCT) as a biological parameter: a literature review notes that a PCT of less than 0.25 ng/dL has a negative predictive value of 93% for the diagnosis of bacterial pneumonia.4

Respiratory infectious diseases account for almost 75% of antibiotic prescriptions in children under 18 years old in the USA and these prescriptions are often unnecessary.1 Limiting antibiotic prescriptions is, therefore, a real public health challenge given the majority of viral respiratory infections.

Optimal management of paediatric patients under 5 years old admitted to PICU for respiratory infections complicated by acute respiratory failure is a major diagnostic, therapeutic and public health challenge. In France, there is no precise description of this population and of the severity of factors likely to lead to the death of these children.

The aim of the study was to describe the characteristics of infectious acute respiratory failure in this population, PICU management and outcomes in order to better identify the needs of these patients.

Our secondary aim was to compare the characteristics and PICU management of this population (1) depending on their age (less or more than 6 months old) and (2) depending on the pulmonary imaging (absence or presence of an alveolar condensation on the chest X-ray or lung ultrasound).

Materials and methods

We conducted a retrospective study in two PICUs in the Ile-de-France region: one at the Centre Hospitalier Universitaire (CHU) Raymond-Poincaré (RPC) in Garches and one at the CHU Bicêtre (KB) in Kremlin-Bicêtre, within the Département Médico-Universitaire (DMU) of the Groupe Hospitalo-Universitaire (GHU) Paris Saclay at the Assistance Publique des Hôpitaux de Paris (APHP).

We included children under 5 years old hospitalised between 1 January 2017 and 31 December 2021 in one of those two departments for a respiratory infection complicated by acute respiratory failure.

Patients hospitalised for reasons other than pulmonary infection were excluded. A total of 707 hospitalisation records were included in this study (figure 1).

Request permission

The data of interest concern patients’ anamnestic characteristics, parameters on admission to the PICU, biology, microbiology, imaging, antibiotics and adjuvant treatments during hospitalisation (ventilation, amines), as well as complications during their stay.

It was necessary to group certain patients according to a common diagnosis. We, therefore, decided to code each patient according to the final diagnosis established in the record by the clinician in charge of said patient: 0=bronchiolitis without antibiotic therapy initiated; 1=bronchiolitis with antibiotic therapy for 5 days or less; 2=bronchiolitis with antibiotic therapy longer than 5 days; 3=pulmonary infection without bacterial documentation; 4=pulmonary infection with bacterial documentation; 5=severe acute asthma with antibiotic therapy for 5 days or less; 6=severe acute asthma with antibiotic therapy over 5 days. The 5-day limit corresponds to the validated treatment duration for acute bacterial community-acquired pneumonia in children according to the Groupe de Pathologie Infectieuse Pédiatrique.10

The first part of the statistical analysis was meant to provide a precise description of the population. Gradients were used for qualitative variables and means with SDs as well as medians with the values of the first and third quartiles were employed for quantitative variables.

The following two statistical analyses were carried out in subgroups in order to distinguish potentially different infectious aetiologies and management. The first one is a comparison between patients aged more or less than 6 months old. We used a χ² test, a Fisher’s exact test and a Wilcoxon test, and set the significance level at 0.05. The second one was a comparison between patients with normal imaging (chest X-ray or lung ultrasound) and those with a pulmonary condensation or pleuro-pneumopathy. We also used a χ² test, a Fisher’s exact test and a Wilcoxon test, and set the significance level at 0.05.

Results

There were initially 1669 hospitalisation records. A total of 707 hospitalisation records were eligible and included in this study. The median age of the cohort was 3 months (figure 2), and 444 patients (63%) were female. The majority of children were at term, with 583 (82%) born after 37 weeks’ gestation. We note that 229 patients (32%) had at least one comorbidity. The median length of stay in intensive care was 4 days (0–88) (table 1).

Request permission

On arrival in the PICU, 447 patients (63%) required the initiation of non-invasive ventilation (NIV) ventilatory support and 14 patients (2%) were in need of invasive ventilation. Regarding the patients on NIV, 398 (73%) were in continuous positive airway pressure (CPAP) mode. Patients required oxygen support, with a mean fraction of inspired oxygen (FiO2) of 34% (25–40), mean pH 7.35 (6.88–7.53), mean partial pressure of carbon dioxide (pCO2) 48 mm Hg (20–122) and mean blood bicarbonates 26 mmol/L (11.5–77.0). Patients had a moderate to severe inflammatory syndrome, with a mean C reactive protein (CRP) of 42 mg/L (0–421) and 253 patients (48%) had a PCT higher than 0.5 µg/L. Most patients were admitted to the intensive care unit and treated with a betalactamine—amoxicillin was the first choice for 165 patients. We note that 320 patients (45%) were not receiving antibiotics on arrival in the PICU (table 2). Finally, lung imaging (chest X-ray or lung ultrasound) showed an alveolar condensation in 300 patients (43%), an interstitial syndrome in 69 patients (10%) and was normal in 203 patients (29%). Only 23 patients (3%) were diagnosed with pleuropneumopathy on imaging. The interpretation of lung imaging was done by the resuscitation team during the patient’s hospitalisation.

During hospitalisation, several parameters were collected, including the most invasive or maximal mode of ventilation required by each patient, with, in ascending order: nasal cannula or inhalation mask then high-flow oxygen therapy, then NIV, then invasive ventilation and finally tracheoventilation. 500 patients (71%) required NIV ventilatory support and 69 (10%) required tracheal intubation. A total of 125 patients (18%) required volume expansion, and 30 (4%) required vasopressors. The two most frequently observed complications were neurological (as loss of consciousness or seizure, eg) and cardiac failure, affecting 56 (8%) and 47 (7%) patients, respectively. It can be seen that 86 patients (12%) were treated with antibiotics for less than 5 days, 131 patients (19%) for between 5 and 7 days, and 51 patients (7%) for more than 7 days. It should be noted that data on the duration of antibiotic therapy was missing for 437 patients (table 3).

Microbiological results during the PICU stay are shown in online supplemental table. 80% (593/671) of PCRs for respiratory viruses were positive and respiratory syncytial virus (RSV) was found in 64% of cases. A clear majority of viral infections (79%) were due to a single virus.

When looking at blood cultures’ analyses, 15% were tested positive, and coagulase-negative staphylococci (Staphylococcus hominis, Staphylococcus epidermidis, Staphylococcus haemolyticus, Staphylococcus capitis and Staphylococcus warneri) were found in 61% of cases. In addition, it was noted that pulmonary infections appear to be very rarely bacteraemic, and positive blood cultures often suggest contamination by cutaneous flora. Haemophilus influenzae, Moraxella catarrhalis, Pseudomonas aeruginosa, Staphylococcus aureus, other streptococci (Streptococcus parasanguinis, Streptococcus oralis, Streptococcus agalactiae) and other germs (Enterococcus faecalis, Acinetobacter junii, Acinetobacter odontolyticus, Moraxella osloensis, Klebsiella pneumoniae, Candida) were also found.

Respiratory samples were 56% (10/18 samples) positive for sputum, 66% (119/181) for bronchial or tracheal aspirates, 67% (16/24) for protected distal samples or bronchoalveolar lavages, and 84% (21/25) for pleural punctures. The three germs most frequently found in respiratory samples were Streptococcus pneumoniae, M. catarrhalis and H. influenzae: respectively in 40%, 10% and 33% of positive sputum samples, in 36%, 39% and 39% of positive bronchial or tracheal aspirates, and in 13%, 25% and 44% of positive protected distal samples or bronchoalveolar lavages. The above-mentioned respiratory samples also include M. catarrhalis, P. aeruginosa, S. aureus, Mycoplasma pneumoniae, Streptococcus pyogenes, other streptococci (Streptococcus dysgalactiae, S. oralis, Streptococcus mitis), Bordetella pertussis and Bordetella parapertussis and other germs (Serratia marcescens, Acinetobacter spp, Acinetobacter baumannii, Elizabethkingia miricola, Delftia acidovorans, Stenotrophomonas maltophilia, K. pneumoniae, Escherichia coli, Enterococcus cloacae, Klebsiella oxytoca, Candida albicans, Candida krusei, Candida striatum, S. epidermidis, A. odontolyticus, Corynebacterium pseudodiphtheriticum, E. faecalis). For pleural punctures, the three most frequently found germs were S. aureus, followed by S. pneumoniae and S. pyogenes, in 52%, 24% and 19% of positive samples, respectively.

Respiratory samples showed 105 cases of viral and bacterial coinfection.

Two subgroup analyses were performed. The first compared two groups: patients with an alveolar condensation on lung imaging and those without (table 4). There were significantly more positive PCRs for respiratory viruses in the group without an alveolar condensation than in the other group (89% vs 79%, p=0.002). More respiratory samples (sputum, bronchial or tracheal aspirate, protected distal swab or bronchoalveolar lavage) were positive for bacteria in the group with an alveolar condensation (25% vs 19%, not significant with p=0.06). There was also a significant difference in each mode of ventilation used during the stay: patients with an alveolar condensation required more nasal cannulas, inhalation masks, high-flow oxygen therapy or invasive ventilation (68% vs 44% with p<0.001, 30% vs 16% with p<0.001 and 13% vs 6% with p=0.002 respectively), whereas patients without an alveolar condensation required more ventilatory support with NIV (87% vs 73% with p<0.001). There was no significant difference in mean pH between the two groups. There was a significant difference in mean FiO2, with higher oxygen requirements in patients with an alveolar condensation (37% vs 30% with p<0.001). Finally, the duration of antibiotic therapy was longer in patients with an alveolar condensation (p<0.001).

The second subgroup analysis compared patients aged under 6 months old and over 6 months old (figure 3). It showed that the type of lower respiratory infection differs according to age. Thus, bronchiolitis was more common before the age of 6 months old, whereas this diagnosis was less frequent after 6 months old. Severe acute asthma was more common after 6 months old. The diagnosis of pulmonary infection with or without bacterial documentation was more frequent after 6 months old and accounts for the majority of lower respiratory infections in this age group.

Request permission

In addition, a subgroup analysis comparing these two populations was performed (table 5). Significantly, patients aged less than 6 months required more NIV (92% vs 56%, p<0.001), and less nasal cannula or inhalation mask, high-flow oxygen therapy or invasive ventilation (52% vs 64% with p=0.002, 13% vs 43% with p<0.001 and 8% vs 14% with p=0.004, respectively). It was also noted that patients under 6 months old had less frequent alveolar condensations (39% vs 64%, p<0.001), slightly lower oxygen requirements (mean 31% vs 39%, p<0.001), a less significant inflammatory syndrome (mean CRP at 31 mg/L vs 65 mg/L, p<0.001; mean PCT 4 µg/L vs 11 µg/L, p<0.001) and a more frequently positive PCR for respiratory viruses (91% vs 72%, p<0.001).

Discussion

This study included 707 patients hospitalised for pulmonary infection in 2 PICUs over a 5-year period. To the best of our knowledge, this is the first French study carried out in PICU describing this population. This population represents the world’s highest mortality rate in children under 5 years old. With this study, we sought to better describe this population and better characterise its management needs, to help implement new strategies in PICU.

In this study, it is noted that almost 90% of PCRs performed in search of a respiratory virus are positive, which is similar to what is observed in an American study conducted on 2219 patients hospitalised between 2010 and 2012 in 3 paediatric hospitals in Memphis (Tennessee), Nashville (Tennessee) and Salt Lake City (Utah) for management of pneumonia.11

A Brazilian review of the literature4 cites two studies similar to our own. First, a multicentre study, the Perch study, carried out in several developing countries (Bangladesh, Gambia, Kenya, Mali, South Africa, Thailand and Zambia) looked at the causes of pneumonia in children aged between 1 month and 5 years old: it estimated that viruses were responsible for pneumonia with an alveolar condensation in 61.4% of cases, compared with 79% in our study, and bacteria in 27.3% of cases, compared with 25% according to our findings. Moreover, RSV was the germ most frequently associated with radiological pneumonia in this study, as in ours, where RSV was found in the vast majority of positive viral PCRs. A possible explanation for the higher proportion of viral infections with an alveolar condensation in our study is the probably higher vaccination coverage in France against S. pneumoniae, leading de facto to a decrease in the proportion of bacterial infections. However, it is important to note that even if we found that there were significantly more positive PCRs for respiratory viruses in the group without an alveolar condensation than in the group with an alveolar condensation in our study, this result may not be clinically significant as the percentage of patients with a positive PCR for respiratory viruses and without an alveolar condensation is very high too. A Brazilian study also looked at the causes of community-acquired pneumonia with an alveolar condensation. It showed that Rhinovirus and S. pneumoniae were, respectively, the viral and bacterial agents most frequently associated with the diagnosis of pneumonia with an alveolar condensation. Our study found that RSV was the viral agent most frequently found, while S. pneumoniae was also the most frequent bacterial agent. The high frequency of RSV in lower respiratory infections has been confirmed in several paediatric studies.2 12 Nevertheless, the results observed are very similar, which seems to us to be an argument in favour of the relevance of our approach in characterising the ventilatory support needs of these patients.

Tracheal or bronchial aspirations were the main method of sampling for bacteria. In our study, the three bacterial germs most frequently found in the various respiratory samples were S. pneumoniae, M. catarrhalis and H. influenzae. These results differ from those observed in the literature; in fact, in an Australian study conducted over eleven years in the PICU in Melbourne,13 the most frequent Gram-negative bacillus isolated from bronchoalveolar lavage or blood cultures was H. influenzae, while the most frequent Gram-positive bacillus was S. aureus, well ahead of S. pneumoniae. There are three possible explanations for these results: a different bacterial epidemiology, the inclusion of older patients in the Australian study and the inclusion of patients who acquired pneumonia during hospitalisation. It is interesting to note that the proportion of positive blood cultures and bronchoalveolar lavages is more or less the same in both studies: 15% and 67%, respectively, in our study, vs 17% and 78%, respectively, in the Australian study.

Aetiological guidance in the face of respiratory infection with respiratory failure is crucial for several reasons: considering the severity of patients, antibiotic treatment must be instituted quickly if necessary. Antibiotic resistance is increasing worldwide, it is, therefore, harmful to contribute further to this trend by prescribing antibiotics when they are not needed. A Canadian study looked at antibiotic prescribing in paediatric patients hospitalised in PICU.14 It showed that detection of RSV by PCR in a respiratory specimen was the only variable significantly associated with a reduction in antibiotic prescribing among resuscitators and infectiologists. The vast majority of patients responded favourably to amoxicillin therapy, the antibiotic most frequently prescribed in this study, in line with recommendations in the literature3 4 whereas cephalosporins are prescribed much less frequently. Amoxicillin is an easily accessible antibiotic, which could also contribute to a reduction in morbidity and mortality worldwide.

In addition, it is interesting to note two findings concerning the management of these lower respiratory infections. First, we saw that over 70% of patients in our cohort required ventilatory support with NIV, compared with only 10% of patients who required invasive ventilation. The proportion of patients on NIV is higher in patients under 6 months of age, as well as in patients without an alveolar condensation on lung imaging, as shown by the two subgroup analyses: this result is consistent with the distribution of diagnosis observed according to age, and the greater frequency of bronchiolitis and viral infections before the age of 6 months. Pneumonia, whether viral or bacterial in origin, is therefore mainly managed by NIV, and this result shows the major and ever-growing interest in access to NIV rather than invasive ventilation and should lead to optimising its use and offering medical and paramedical training. This could drastically reduce mortality from respiratory infections, as NIV requires a lower technical platform, cost and training than invasive ventilation and is therefore accessible to developing countries. For instance, two randomised controlled trials in Ethiopia and Bangladesh, which are two developing countries, have tried to assess the effect of bubble CPAP in children aged 1–59 months with severe pneumonia and hypoxaemia by comparing it to low-flow and high-flow oxygen therapies. Bubble CPAP is a mean to deliver CPAP without a mechanical ventilator, using simple and low-cost material: it generates positive-end-expiratory pressure by connecting the expiratory limb of a breathing circuit to a tube, which is submerged in water. Both studies showed that bubble CPAP reduced the risk of treatment failure and in-hospital mortality in children with severe pneumonia and hypoxaemia compared with the use of standard low-flow oxygen therapy.15 16

A weakness of this study is its retrospective nature, which may have led to errors in the information recorded in the files and transcribed in this study. In addition, the descriptive nature of this study does not allow us the drawing of any formal conclusions on the parameters studied.

Nevertheless, it has several strengths: it is multicentric, including two PICUs with different patient recruitment (more patients with neuromuscular comorbidities at the CHU Raymond-Poincaré), covers a substantial number of patients and accurately describes this population of patients under 5 years of age hospitalised in PICU, with few missing data.

Conclusion

In conclusion, our work shows a certain similarity between patients hospitalised for severe lower respiratory infections in PICUs in the Ile-de-France region and those in other intensive care units in Australia, the USA and Brazil. It is clear that optimal management relies mainly on NIV, oxygen therapy with FiO2 under 40% and widely available antibiotics. These results are very encouraging and lead us to believe that the implementation of NIV training and equipment programmes could help reduce mortality from lower respiratory infections affecting children worldwide.