Scope of guidelines

These guidelines address the management of CAP in infants and children in the UK. They do not include neonates, infants with respiratory syncytial virus bronchiolitis or children with upper respiratory tract infection, mild fever and wheeze. The specific management of children with pre-existing respiratory disease or that of opportunistic pneumonias in immunosuppressed children is not addressed.

Identification of evidence

A search strategy was developed by an information specialist from the Centre for Reviews and Dissemination in York (part of the National Institute for Health Research). The Search strategy and the results are shown in appendix 1 in the online supplement.

The Cochrane Library (DARE and Cochrane Database of Systematic Reviews), MEDLINE and EMBASE were searched from 2000 onwards. There were some technical changes made to the original search strategies to reduce the chances of missing studies: a single search strategy was used rather than separate strategies for each subject. Studies were limited to English language in view of the limitations on time and resources.

Two thousand and seventy-six studies were identified by the searches, which were rerun in July 2010. The updated search identified a further 511 titles.

Assessing the literature

Initial review of the 2076 titles and abstracts was undertaken by one reviewer, screening for relevance. This was repeated after the second search by another reviewer. The relevant titles and abstracts were grouped by subject matter with many papers being relevant for more than one subject area.

Two reviewers then assessed the studies for inclusion. Studies from countries where the populations or clinical practices were very different from the UK were excluded unless they addressed questions that could be generalised to the UK (such as clinical assessment). Any differences of opinion were settled by a third party. The studies were appraised using the Cochrane data extraction template (see appendix 2 in online supplement).

Any guideline statements made were graded using the same table as that used by the group developing the adult guidelines (table 1).1 First, each paper was given an evidence level (Ia to IVb) by the authors of each chapter. Then, at the end of each chapter when evidence statements were collated, a summative evidence level was attached to each statement depending on the level of evidence underpinning that statement. Finally, each recommendation was graded (A to D) based upon a considered judgement of the body of evidence.

Provenance and peer review

The draft guideline was made available online for public consultation (January/February 2011). The draft guideline was reviewed by the BTS Standards of Care Committee (July 2010/March 2011).

2.3.1 What is the effect of seasonality?

A marked seasonal pattern with winter preponderance was seen for laboratory-reported IPD and hospital admissions due to confirmed pneumococcal infection. December and January showed a peak 3–5 times higher than August.11^[III] Senstad et al also reported a low incidence of hospital CAP in summer and a peak in January.3^[III] There is marked seasonal variation in viral infections such as respiratory syncytial virus (RSV), influenza and parainfluenza 1+2.11^[III]12^[III]13^[II] Parainfluenza 3, however, is found throughout the year.7^[II]

Mycoplasma infection occurs in clusters but has no clear seasonality.

Evidence statements

• The European incidence of CAP, defined as fever, clinical signs and chest radiograph infiltrate in a previously well child is approximately 33/10 000 in those aged 0–5 years and 14.5/10 000 in those aged 0–16 years. [Ib]

• Boys have a higher incidence at all ages. Children <5 years of age and those born between 24 and 28 weeks gestation have a higher incidence of severe disease. [III]

3.1.1 Which viruses are associated with CAP?

A number of viruses appear to be associated with CAP, the predominant one being RSV. RSV, parainfluenza and influenza are detected in similar proportions of children with pneumonia both in the community and in hospital.7^[II] Influenza virus was detected relatively infrequently in paediatric pneumonia using immunofluorescence.30^[II] However, with PCR techniques, influenza is found in 7–22% of cases.28^[Ib]32^[Ib]24^[Ib] In the UK during a 6-month winter influenza season, 16% of children with pneumonia had influenza A.31^[II] Other viruses isolated in children with pneumonia include adenovirus, rhinovirus, varicella zoster virus, cytomegalovirus, herpes simplex virus and enteroviruses.

Several new viruses have been identified and are regularly associated with pneumonia. Human metapneumovirus has been identified in 8–11.9% of cases24^[Ib]33^[Ib]34^[Ib]35^[Ib] and human bocavirus has recently been isolated from 4.5% in Thailand,36^[Ib] 14.2% in Spain24^[Ib] and 15.2% in Korea.33^[Ib] Coronavirus is identified in 1.5%33^[Ib] to 6.5% of cases.29^[Ib]24^[Ib] Overall, viruses appear to account for 30–67% of CAP cases in childhood and are more frequently identified in children aged <1 year than in those aged >2 years (77% vs 59%).28^[Ib]24^[Ib]

3.1.2 Which bacteria are associated with CAP?

Quantifying the proportion of CAP caused by bacteria is more difficult. Streptococcus pneumoniae is assumed to be the most common bacterial cause of CAP but is infrequently found in blood cultures. Overall, blood or pleural fluid culture of S pneumoniae is positive in 4–10% of cases of CAP.16^[II]17^[Ib]18^[II]19^[II]20^[II]24^[II]37^[II] It is commonly found in routine cultures of upper respiratory tract specimens, yet is known to be a commensal in this setting. A review of lung tap studies found 39% identified S pneumoniae.38^[III] A recent study of 34 children in Finland who had a lung aspirate identified S pneumoniae in 90% either by culture or PCR.39^[II] Pneumolysin-based PCR is increasingly used and validated.21^[II]22^[Ib] Studies incorporating this into diagnosis in children not immunised with the conjugate PCV have detected S pneumoniae in around 44%,28^[Ib] often as a co-pathogen with either viruses or other bacteria. The proportion of CAP due to S pneumoniae increases up to 41% in cases where serological testing is used.29^[Ib] Mixed pneumococcal and viral infections appear important and are found in 62% of pneumococcal pneumonias.29^[Ib]

Pneumococcal serotypes are important, with serotypes 14, 6B, 19F and 23F being implicated more frequently in IPD and serotype 1 in empyema. The most common isolates in IPD since the introduction of PCV7 in Europe, including the UK, were serotypes 1, 19A, 3, 6A and 7F.40^[Ib] There are no UK data on the most frequent serotypes found in pneumonia, although serotype 1 has been predominantly responsible for empyema.41^[Ib] Recent data on serotypes identified in bacteraemic pneumonia in children from Italy since the introduction of PCV7 found serotypes 1 and 19A to be the most common.22^[Ib] Both these serotypes are included in PCV13, introduced into the UK immunisation schedule in 2010.

With the introduction of conjugate pneumococcal vaccines, indirect evidence of vaccine efficacy for the prevention of pneumonia can be used to assess the contribution of S pneumoniae to CAP. In children under 2 years, all trials have consistently shown a decrease in radiologically-confirmed pneumonia from 23% in the Philippines using PCV1142^[Ib] to 37% in the Gambia with PCV943^[Ib] and 23.4% in California with PCV7.44^[Ib] The effect is most striking in the first year with a 32.2% reduction, and a 23.4% reduction in the first 2 years.44^[Ib] A recent study of PCV11 found that, although 34% of radiologically-confirmed pneumonias were prevented in children under 1 year, there was only a 2.7% decrease in those aged 12–23 months.42^[Ib] In children aged >2 years there was only a 9.1% reduction.44^[Ib] A Cochrane systematic review found a pooled vaccine efficacy for PCV11 of 27% for reduction of radiographically-confirmed pneumonia in children <2 years and 6% for clinical pneumonia.45^[Ia]

The introduction of PCV7 has dramatically decreased IPD due to vaccine serotypes in those countries where it has been universally introduced, but a steady increase in vaccine serotype replacement (ie, natural selection of pneumococcal serotypes not present in the vaccine) has been evident in the UK to 2010, so that the total IPD rate due to all serotypes is climbing back to similar rates before the introduction of PCV7 (http://www.hpa.org.uk/HPA/Topics/InfectiousDiseases/InfectionsAZ/1203008863939/). This trend is expected to reverse with the introduction of PCV13 (http://www.hpa.org.uk/web/HPAwebFile/HPAweb_C/1245581527892).

Other bacterial pathogens appear to be less frequent causes of CAP. Group A streptococcal infection is important in terms of severity as, when present, it is more likely to progress to paediatric ICU admission or empyema.30^[II]46^[III] When looked for, it may be found in 1%28^[Ib]29^[Ib] to 7% of cases.30^[II] It is increasingly associated with pneumonia complicated by empyema, as is Staphylococcus aureus.8^[Ib]

S aureus has also long been associated with increased mortality in influenza. Recent reports indicate a fivefold increase in influenza and S aureus mortality in children in the USA from 2004 to 2007.47^[Ib]

Claesson et al48^[II] assessed the antibody responses to non-capsulated Haemophilus influenzae and isolated it as the only pathogen from the nasopharynx of 43 of 336 children. A significant increase in IgG or IgM was shown in 16 (5% of all CAP). In the same study, 3% also had a significant increase in antibodies to Moraxella catarrhalis, suggesting that it too is an uncommon cause of CAP in children.49^[II] This was supported by another study by Korppi et al50^[II] in which seroconversion to M catarrhalis was documented in only 1.5% of cases of CAP.

3.1.3 What is the contribution of atypical organisms?

In aetiology studies, Mycoplasma pneumoniae previously accounted for 4–39% of isolates.51 Since 2000, those studies published where M pneumoniae is specifically sought in children admitted to hospital show remarkable consistency, with rates of detection from 27% to 36% (see table 5).52–56 Where Chlamydia pneumoniae is sought, it appears to be responsible for 5–14% of cases, but a single US study detected it in 27%.57^[II] Biases which need to be considered in these reports include whether children with mycoplasmal (or chlamydial) pneumonia are over-represented in hospital-based studies because of failure of penicillin-related antibiotic treatment in the community, or are over-represented in community studies because they are less sick and therefore less likely to be referred to hospital.

New bacteria are also being described. Simkania negevensis, a Chlamydia-like organism, is detected frequently by PCR in respiratory samples although antibody studies suggest it may be rarely implicated in pneumonia.58^[III]59^[III]

3.2 Does the aetiology differ by age?

Several generalisations are possible with respect to age. With improved diagnostic tests including serology and PCR, evidence of specific aetiology tends to be more commonly found in younger children.26^[II]28^[Ib]24^[Ib] Michelow et al28^[Ib] detected a pathogen in 92% of children aged <6 months but in only 75% of those aged >5 years. Although viral infections (especially RSV) are more commonly found in younger children,2^[II]16^[II]17^[II]19^[II]24^[II]60^[II] bacteria are also isolated in up to 50% of children aged <2 years, together with a virus in up to half of these.28^[Ib] However, bacteria are more frequently identified with increasing age,28^[Ib] hence mixed infections become less frequent with age.26^[II]61^[II] Vaccine probe studies indicate that one-third of young children with radiological changes have pneumococcal pneumonia,45^[Ia] with serological studies indicating at least 20% have a pneumococcal aetiology across all ages.26^[II] This has implications for the way in which we consider antibiotic choices.

Chlamydia and Mycoplasma species have been more commonly found in older children.16^[II]19^[II]26^[II]52^[II]54^[II]60^[II]62^[II]63^[II]64^[II] However, Block et al57^[II] found the incidence of M pneumoniae and C pneumoniae infections to be comparable in all age groups between 3 and 12 years. In particular, the finding of a 23% incidence of M pneumoniae infection and 23% of C pneumoniae infection in children aged 3–4 years is high. Recent studies have supported this, with Baer also noting a 22% incidence of M pneumoniae in children aged 1–3 years.54^[II] This raises questions about appropriate treatment in this age group, although young children may have milder M pneumoniae infection65^[IVb] and many recover without specific antibiotic treatment.66^[II]

4.4.1 Pneumococcal pneumonia

Pneumococcal pneumonia starts with fever and tachypnoea. Cough is not a feature initially as alveoli have few cough receptors. It is not until lysis occurs and debris irritates cough receptors in the airways that cough begins.

Many studies therefore emphasise the importance of the history of fever and breathlessness and the signs of tachypnoea, indrawing and ‘toxic’ or ‘unwell’ appearance.

4.4.2 Mycoplasma pneumonia

Mycoplasma pneumonia can present with cough, chest pain and be accompanied by wheezing. Classically, the symptoms are worse than the signs would suggest. Non-respiratory symptoms, such as arthralgia and headache, might also suggest mycoplasma infection.78^[IVb]

A study of 154 children by Michelow et al28^[Ib] found that, as has been proposed more recently, preschool children are just as likely as those of school age to have atypical pneumonia. There are likely to be geographical variations in these findings.

4.4.3 Staphylococcal pneumonia

This is indistinguishable from pneumococcal pneumonia at the beginning of the illness. It remains rare in developed countries where it is usually a disease of infants. It can complicate influenza in infants and older children. The incidence is increasing.

Evidence statements

• Children with CAP may present with fever, tachypnoea, breathlessness or difficulty in breathing, cough, wheeze or chest pain. These clinical features of CAP vary with the age of the child and tend not be very specific for diagnosis. [IVb]

• In children older than 3 years, a history of difficulty breathing is an additional valuable symptom. [II]

• A raised respiratory rate is associated with hypoxaemia. [II]

Recommendation

• Bacterial pneumonia should be considered in children when there is persistent or repetitive fever >38.5°C together with chest recession and a raised respiratory rate. [D]

5.1.1 Should a lateral x-ray be performed?

In a retrospective study of 1268 cases (7608 x-ray interpretations),83^[III] frontal and lateral chest x-rays of patients referred from an emergency department in the USA were reviewed by three radiologists independently. The sensitivity and specificity of the frontal x-ray alone for lobar consolidation was 100%. For non-lobar infiltrates the sensitivity was 85% and the specificity 98%, suggesting that these types of radiographic changes may be underdiagnosed in 15% of cases. The authors admit that some of the loss of sensitivity may be due to the wide variability in what is considered radiographic pneumonia. The clinical implications of these radiographically underdiagnosed pneumonias are not evident from the study.

Lateral x-rays are not routinely performed in paediatric CAP and the recommendation is that they are not necessary84^[II] and would mean exposing the child to further radiation.

5.1.2 How good is agreement on interpretation of x-rays?

There is great intra- and inter-observer variation in radiographic features used for diagnosing CAP. The WHO85 produced a method for standardising the interpretation of chest x-rays in children for epidemiological purposes but, even using this scheme, the concordance rate between two trained reviewers was only 48% (250/521).

5.1.3 Can chest radiography be used to distinguish aetiology?

It is common in clinical practice that alveolar infiltration is thought to be secondary to a bacterial cause and bilateral diffuse interstitial infiltrates to atypical bacterial or viral infections. Adequate sensitivity is lacking for either of these assignations. Chest radiography is generally unhelpful for deciding on a potential causative agent.

Toikka et al86^[II] studied 126 patients, all of whom had x-rays. Bacterial aetiology was established in 54%, viral in 32% and 14% had unknown aetiology. The x-rays were divided into two groups by three radiologists unaware of the clinical diagnoses and characteristics: group 1 (n=61) had mild or moderate changes (interstitial infiltrations not covering a whole lung, minor alveolar infiltrations, hyperaeration, perihilar pneumonia) and group 2 (n=61) had marked changes (interstitial changes covering a whole lung, major alveolar infiltrations, lobar alveolar infiltrations, pleural fluid, abscess formation, atelectasis). Of those in group 1, 39% had bacterial pneumonia and 45% viral pneumonia. Of those in group 2, 69% had bacterial pneumonia and 18% viral pneumonia. Clearly, some bacterial infections are only mild, producing less marked changes on the chest x-rays and, conversely, some viral infections are severe, producing marked changes on the x-ray. Aetiology is therefore difficult to assign on the basis of the x-ray.

Virkki et al87^[II] studied 254 children with radiographically diagnosed CAP, assigning aetiology in 215/254 patients. Radiographic findings were classified as alveolar and/or interstitial pneumonia, hyperaeration, hilar enlargement, atelectasis, pleural fluid and location in one or both lungs. Of 137 children (64%) with alveolar infiltrates, 71% had evidence of bacterial infection; 72% of 134 cases with bacterial pneumonia had alveolar infiltrates and 49% with viral pneumonia had alveolar infiltrates. Half of those with interstitial infiltrates had bacterial infection. The sensitivity for bacterial infection in those with alveolar infiltrates was 0.72 and specificity was 0.51. For viral pneumonia with interstitial infiltrates the sensitivity was 0.49 and specificity 0.72.

In a prospective study of 136 children, Drummond et al30^[II] showed that there was no significant difference in aetiology among the five radiographic groups into which their cases were divided (lobar consolidation, patchy consolidation, increased perihilar and peribronchial markings, pneumonitis and effusion).

In a study of 101 Italian children with radiographically-defined pneumonia, Korppi et al77^[II] found no association between radiographic appearances and aetiology. Alveolar infiltrates were present in 44 children (62%). In those aged >5 years alveolar infiltrates were present in 68%, although blood cultures were negative in all cases. Alveolar infiltrates were present in 46% of those with viral aetiology, 67% with pneumococcal aetiology and 70% in each of those with atypical bacterial and unknown aetiologies.

Chest x-rays are often done in research studies of CAP, but these studies do not support the routine use of chest x-rays in the investigation and management of CAP.

5.1.4 Are follow-up x-rays necessary?

Two recent studies have examined the utility of follow-up x-rays in previously healthy children with CAP.

Virkki et al88^[II] published the results of a 3-year prospective study of 196 children with CAP. They also followed the children up at 8–10 years after diagnosis. Of 196 follow-up x-rays, there were abnormalities in 30% (infiltrates 67%, atelectasis 47%, lymph nodes 28%); 20% were new abnormalities. No change in management was instituted on the basis of these radiographic findings. Follow-up at 8–10 years of 194 patients showed no new illnesses associated with the previous pneumonia. In those with an uneventful recovery, x-rays are unnecessary.

Suren et al89^[III] published the results of a retrospective study of 245 children recovering from CAP. Of these, 133 had follow-up x-rays, 106 of which were normal and 27 of which were abnormal. Of the 106 patients with normal follow-up x-rays, two went on to develop further clinical problems (both recurrent pneumonias with no established underlying cause). Of the 27 patients with abnormal x-rays, three developed further clinical problems that could be related to the previous pneumonia. Of 112 who did not have follow-up x-rays, 10 developed subsequent clinical problems. Most of these occurred within the first 4 weeks after discharge, before the regular scheduling of the follow-up x-ray. The authors established that a follow-up x-ray might have been helpful in 5/245 cases. These modest benefits should be balanced against the exposure of children to radiation.

5.3.1 Pulse oximetry

Oxygen saturation measurements provide a non-invasive estimate of arterial oxygenation. The oximeter is easy to use and requires no calibration. It does require a pulsatile signal from the patient and is susceptible to motion artefacts. The emitting and receiving diodes need to be carefully opposed. To obtain a reliable reading:

• the child should be still and quiet;

• a good pulse signal should be obtained;

• once a signal is obtained, the saturation reading should be watched over at least 30 s and a value recorded once an adequate stable trace is obtained.

In a prospective study from Zambia, the risk of death from pneumonia was significantly increased when hypoxaemia was present.68^[II]

5.3.2 Acute phase reactants

Several studies have looked at using various acute phase reactants as a means of differentiating the aetiology and/or severity of CAP.64^[II]86^[II]90^[II]91^[II]92^[II]93^[II] The utility of procalcitonin (PCT), cytokines, C reactive protein (CRP), erythrocyte sedimentation rate (ESR) and white blood cell (WBC) count individually and in combination has been assessed.

Korppi et al64^[II] examined WBC, CRP, ESR and PCT levels and chest radiographic findings in 132 cases in an effort to find combinations of markers that would differentiate a pneumococcal from a viral aetiology. For a combination of CRP >80 mg/l, WBC >17×10⁹/l, PCT >0.8 μg/l and ESR >63 mm/h, they found the likelihood ratio of the pneumonia being pneumococcal was 1.74 with a sensitivity of 61% and specificity of 65%. If alveolar infiltrates on the x-ray were included, the likelihood ratio was 1.89, specificity 82% and sensitivity 34%. None of these combinations of parameters was sufficiently sensitive or specific to differentiate bacterial (specifically pneumococcal) from viral pneumonia.

Michelow et al93^[II] investigated a panel of 15 cytokines in 55 patients with CAP. Forty-three children had an aetiological diagnosis. Twenty-one children had S pneumoniae, 17 had M pneumoniae, 11 had influenza A, three had C pneumoniae, one had S aureus and eight had viruses identified. Eleven had mixed viral and bacterial infections. Of the cytokines, interleukin 6 (IL-6) was the only one significantly associated with a rise in white cell band forms, PCT levels and unequivocal consolidation on the x-ray. However, there was no correlation with aetiology. There remains little evidence that cytokine profiles have any clinical utility.

Don et al91^[II] evaluated the usefulness of PCT for assessing both the severity and aetiology of CAP in a study of 100 patients. The cases were assigned into four aetiological groups: pneumococcal (n=18), atypical bacterial (n=25), viral (n=23) and unknown (n=34). There was no significant association between PCT levels and aetiological group. PCT levels were found to be significantly associated with severity of CAP, as defined by admission to hospital and the presence of alveolar infiltrates on the chest x-ray. Median PCT values (25th–75th centiles) for inpatients and outpatients, respectively, were 17.81 and 0.72.

Korppi et al90^[II] published a prospective population-based study of 190 children in an ambulatory primary care setting with radiologically-diagnosed pneumonia and aetiological diagnoses for five bacteria and seven viruses. They found that no association between severity of CAP (as defined by inpatient versus outpatient management) and PCT or between aetiology of CAP and PCT. The median values for each of the four aetiological groups (pneumococcal, mycoplasma/chlamydial, viral and unknown) were not significantly different (p=0.083). For inpatient versus outpatient management, PCT levels were 0.42 and 0.45 μg/l, respectively (p=0.77).

According to these two studies, there may be some alignment between PCT levels and severity, as defined by admission to hospital, but the evidence is still lacking for the ability of PCT to discriminate between viral and bacterial causes of CAP.

Toikka et al86^[II] studied 126 children with CAP, measuring PCT, CRP and IL-6 levels. Aetiology was established for six bacteria and 11 viruses; 54% had bacterial infection, 32% viral and 14% unknown. Median PCT and CRP levels were found to be significantly different, but there was marked overlapping of values. There were no significant differences for IL-6 levels. The sensitivity and specificity of CRP and PCT levels were low. If PCT, CRP and IL-6 levels are very high, then bacterial pneumonia is more likely but, generally, they have little value in differentiating viral from bacterial CAP.

Flood et al94^[Ia] performed a meta-analysis of eight studies, including several revealed in our recent search,87^[II]95^[II]96^[II] that examined the use of CRP in establishing aetiology in CAP. The pooled study population was 1230; 41% had bacterial CAP. A CRP range of 35–60 mg/l was significantly associated with bacterial pneumonia, producing an OR for bacterial versus non-bacterial CAP of 2.58 (95% CI 1.20 to 5.55). Given the prevalence of bacterial pneumonia of 41%, the positive predictive value for CRP values of 40–60 mg/l was 64%. The conclusion of the meta-analysis was that CRP was only weakly predictive for bacterial pneumonia.

5.4.1 Are there any microbiological investigations that should be performed in the community?

There is no indication for microbiological investigations to be done in the community. Some workers have investigated the feasibility of performing PCR analysis for viruses in nasopharyngeal secretions in the context of pandemic respiratory virus infections,97^[II] but this is not currently practical in the UK.

5.4.2 Which microbiological investigations should be performed on a child admitted to hospital?

It is important to attempt microbiological diagnosis in patients admitted to hospital with pneumonia severe enough to require admission to the paediatric ICU or with complications of CAP. They should not be considered routinely in those with milder disease.

Microbiological methods that may be used are several and include: blood culture, nasopharyngeal secretions and nasal swabs for viral detection (by PCR or immunofluorescence), acute and convalescent serology for respiratory viruses, M pneumoniae and C pneumoniae and, if present, pleural fluid for microscopy, culture, pneumococcal antigen detection and/or PCR.

Cevey-Macherel et al29^[Ib] identified a causative agent in 86% of 99 patients using a variety of microbiological, serological and biochemical means; 19% were of bacterial aetiology alone, 33% of viral aetiology alone and 33% of mixed viral and bacterial aetiology.

5.4.3 Which investigations are helpful in identifying a bacterial cause?

5.4.4 Which investigations are helpful for identifying atypical bacteria?

Paired serology (rising titres in antibody complement fixation tests) remains the mainstay for diagnosing M pneumoniae and C pneumoniae infections. However, two studies have investigated the use of PCR in identifying atypical bacterial infections.

Michelow et al103^[II] used PCR to diagnose M pneumoniae from nasopharyngeal and oropharyngeal swabs. They compared 21 children with serologically-proven M pneumoniae infections with 42 controls; 12 of the 21 children (57%) were PCR positive, 9 of the 12 each positive on nasopharyngeal and oropharyngeal samples, six on both. The greatest diagnostic yield was therefore when samples from both sites were combined and analysed. One of the controls was PCR positive. The OR for detecting M pneumoniae by PCR in serologically-proven cases was 54.7 (range 5.9–1279.3). When compared with ELISA, PCR had a sensitivity of 57.1%, specificity of 97.6%, positive predictive value of 97.3% and negative predictive value of 82.0%. The authors argue that PCR positivity for M pneumoniae in the upper respiratory tract is suggestive of lower respiratory tract infection. Of interest, in their study PCR-positive cases had a significantly longer duration of oxygen therapy (1.7 vs 0.78 days, p=0.045).

Maltezou et al105^[II] used PCR to diagnose Legionella and Mycoplasma lower respiratory tract infections by collecting serum and sputum or throat swabs. Of 65 children, serology (IgM EIA) was positive in 18 (27.5%) for M pneumoniae and in one (1.5%) for Legionella. Eleven of the 18 were diagnosed in the acute phase and nine (50%) of those serologically diagnosed were positive for M pneumoniae by PCR of sputum. Taken together, 15/18 were diagnosed by PCR and IgM serology; 3/18 were diagnosed by convalescent serology. The sensitivity of PCR versus IgM EIA in this study was 50%. This is consistent with recent observations that PCR can detect persistent M pneumoniae infection up to 7 months after disease onset.106^[II]

5.4.5 Which investigations are useful in identifying viral pneumonia?

Viruses are significant causes of paediatric CAP, either on their own or in mixed infections. Several studies have looked at the various techniques available for identifying viruses. These include viral culture, antigen detection, serology and PCR.

In the previously mentioned study undertaken by Cevey-Macherel and colleagues,29^[Ib] they found viral PCR of nasopharyngeal aspirates to be very sensitive. In their study, 66/99 children had evidence of acute viral infection (33/99 as co-infection with bacteria). In those with a negative PCR, viral infection could not be detected by any other method. As well as viral culture and PCR, they used viral antigen detection and serum complement fixation tests.

Shetty et al107^[Ib] subjected 1069 nasopharyngeal swabs to viral culture and direct fluorescent antibody (DFA) staining; 190 were DFA and viral culture positive (true positive) and 837 were DFA and culture negative (true negative). The sensitivity for DFA in this study was 84%, specificity 99%, positive predictive value 96% and negative predictive value 96%. One hundred and twenty of 140 hospitalised patients (86%) had viral cultures that reported positive only after the children had been discharged. The authors make the point that the viral cultures were not of any utility in making clinical management decisions.

Lambert97^[II] collected nose-throat swabs and nasopharyngeal aspirates in 295 patients (303 illnesses) and subjected them to PCR analysis for eight common respiratory viruses. Nose-throat swabs are thought to be ‘less invasive’ samples that are more easily collected by parents and therefore of possible benefit in rapid diagnosis in the context of a respiratory virus pandemic. In 186/303 (61%) paired nose-throat swabs/nasopharyngeal aspirates, at least one virus was detected. For nose-throat swabs the sensitivity was 91.9% for RSV was and 93.1% for influenza A. For adenovirus, the sensitivity of nose-throat swabs was 65.9% (95% CI 50.1% to 79.5%) compared with 93.2% (95% CI 81.3% to 98.6%) for nasopharyngeal aspirates. Concordance between nasopharyngeal aspirates and nose-throat swabs was 89.1%. The authors argue that the combination of PCR and the less invasive nose-throat swabs provides adequate sensitivity for the detection of respiratory viruses.

Evidence statements

• Children with CAP present with a range of symptoms and signs. A global assessment of clinical severity and risk factors is crucial in identifying the child likely to require hospital admission. [IVb]

Recommendations

• For a child in the community, re-consultation to the general practitioner with persistent fever or parental concern about fever should prompt consideration of CAP. [D]

• Children with CAP in the community or in hospital should be reassessed if symptoms persist and/or they are not responding to treatment. [D]

• Children who have oxygen saturations <92% should be referred to hospital for assessment and management. [B+]

• Auscultation revealing absent breath sounds with a dull percussion note should raise the possibility of a pneumonia complication by effusion and should trigger a referral to hospital. [B−]

• A child in hospital should be reassessed medically if there is persistence of fever 48 h after initiation of treatment, increased work of breathing or if the child is becoming distressed or agitated. [D]

Recommendation

• Families of children who are well enough to be cared for at home should be given information on managing fever, preventing dehydration and identifying any deterioration. [D]

7.1.1 Over-the-counter remedies

No over-the-counter cough medicines have been found to be effective in pneumonia.114^[Ia]

7.2.1 Oxygen therapy

Hypoxic infants and children may not appear cyanosed. Agitation may be an indicator of hypoxia.

Patients whose oxygen saturation is <92% while breathing air should be treated with oxygen given by nasal cannulae, head box or face mask to maintain oxygen saturation >92%.68^[II]

There is no strong evidence to indicate that any one of these methods of oxygen delivery is more effective than any other. A study comparing the different methods in children aged <5 years concluded that the head box and nasal cannulae are equally effective,115^[II] but the numbers studied were small and definitive recommendations cannot be drawn from this study. It is easier to feed with nasal cannulae. Alternative methods of delivering high-flow humidified nasal oxygen are available and increasingly used. Higher concentrations of humidified oxygen can also be delivered via face mask or head box if necessary.

Where the child's nose is blocked with secretions, gentle suctioning of the nostrils may help. No studies assessing the effectiveness of nasopharyngeal suction were identified.

No new published studies about oxygen therapy were identified in the update searches.

7.2.2 Fluid therapy

Children who are unable to maintain their fluid intake due to breathlessness or fatigue need fluid therapy. Studies on preterm infants or infants weighing <2000 g have shown that the presence of a nasogastric tube compromises respiratory status.116^[II]117^[IVb] Older children may be similarly affected, although potentially to a lesser extent because of their larger nasal passages so, although tube feeds offer nutritional benefits over intravenous fluids, they should be avoided in severely ill children. Where nasogastric tube feeds are used, the smallest tube should be passed down the smaller nostril.117^[IVb] There is no evidence that nasogastric feeds given continuously are any better tolerated than bolus feeds (no studies were identified); however, in theory, smaller more frequent feeds are less likely to cause stress to the respiratory system.

Patients who are vomiting or who are severely ill may require intravenous fluids and electrolyte monitoring. Attention is drawn to the 2007 National Patient Safety Agency alert ‘Reducing the risk of hyponatraemia when administering intravenous fluids to children’.118 Serum levels of sodium can be low in children with pneumonia and there is debate as to whether this is related to inappropriate antidiuretic hormone secretion or overall sodium depletion. Good quality evidence is lacking.

7.2.3 Physiotherapy

Two randomised controlled trials119^[Ib]120^[II] and an observational study121^[Ib] conducted on adults and children showed that physiotherapy did not have any effect on the length of hospital stay, fever or chest radiographic findings in patients with pneumonia. There is no evidence to support the use of physiotherapy, including postural drainage, percussion of the chest or deep breathing exercises.119^[Ib]120^[II]122^[IVb] There is a suggestion that physiotherapy is counterproductive, with patients who receive physiotherapy being at risk of having a longer duration of fever than the control group.119^[Ib] In addition, there is no evidence to show that physiotherapy is beneficial in the resolving stage of pneumonia.

A supported sitting position may help to expand the lungs and improve respiratory symptoms in children with respiratory distress.

There were no new studies identified.

A summary article121^[Ib] summarised the studies discussed above.

Recommendations

• All children with a clear clinical diagnosis of pneumonia should receive antibiotics as bacterial and viral pneumonia cannot be reliably distinguished from each other. [C]

• Children aged <2 years presenting with mild symptoms of lower respiratory tract infection do not usually have pneumonia and need not be treated with antibiotics but should be reviewed if symptoms persist. A history of conjugate pneumococcal vaccination gives greater confidence to this decision. [C]

8.3.1 Streptococcus pneumoniae

Despite the rapid reduction in PCV7 serotypes following the introduction of conjugate vaccine in 2000, penicillin resistance increased steadily in Cleveland, USA until 2003–4. At this time, 51% of isolates were non-susceptible to penicillin.127^[Ib]

PCVs have reduced drug-resistant S pneumoniae but, because of increased intermediate resistance among non-PCV7 serotypes, reductions in intermediate penicillin-resistant strains have not followed. Serotype 19A, which is both antibiotic resistant and a common cause of disease, is not covered by PCV7 and is now increasing worldwide, including in countries without PCV7.128^[Ia]129^[Ia]130^[Ia] However, it is included within PCV 13, the introduction of which would potentially prevent a further 50% of continuing IPD in children.

S pneumoniae macrolide resistance is also increasing, and different mechanisms of resistance drive different levels of resistance. High-level resistance also involves clindamycin resistance, whereas low-level resistance only involves macrolides. Resistance mechanisms vary geographically with mostly low-level resistance in the USA but high-level resistance in Europe.131^[Ia] US surveillance data for 2000–4 of respiratory isolates indicate a stable 30% are macrolide resistant, although an increasing proportion has high-level macrolide resistance.132^[Ib]

A study from Portugal significantly associated macrolide use with the increase of penicillin and erythromycin non-susceptible isolates from adults (p<0.01) and erythromycin non-susceptible isolates among children (p=0.006).133^[Ib]

In the UK, however, penicillin resistance is far less prevalent. Pneumococcal penicillin non-susceptibility in pneumococci causing bacteraemia rose in the 1990s to 6.7% in 2000 and has since declined to around 4% in 2007. Geographical variation ranges from 1.5% in the East Midlands to 8.0% in London. This is in contrast to much of mainland Europe where rates are 25–50% in France and Spain.134^[Ib] Erythromycin resistance in the UK is higher at 9.3% in 2007, but has decreased since 2004 and also varies across the country from 5.2% in north-east England to 14.7% in London. It is much higher in mainland Europe with 25–50% macrolide resistance in France and Italy.134^[Ib] In 2006–7, erythromycin resistance was found in 12% of invasive isolates from children, with serotype 19A still very uncommon.135^[Ib]

8.3.2 Group A streptococcus

There is also varying prevalence of macrolide resistance in Streptococcus pyogenes (group A streptococcus) worldwide, in some areas up to 40%,136^[Ib] and β-lactamase production in H influenzae is widespread. Overall, in the UK the reported resistance rates for group A streptococcus to clindamycin, erythromycin and tetracycline were 5.1%, 5.6% and 14.0% respectively in 2007, with 4.4% resistant to all three. Penicillin resistance has not been seen to date and penicillin remains the therapeutic drug of choice.134^[Ib]

8.3.3 Staphylococcus aureus

Methicillin-resistant S aureus (MRSA) is of increasing concern in the USA and has been implicated in the increase in pleural empyemas seen.137^[III] Although MRSA contributes to 31% of S aureus bacteraemia in the UK,134^[Ib] it has not yet been a significant factor in either empyema or pneumonia.30^[II]41^[II]138^[II]

8.3.4 What is the clinical impact of antibiotic resistance?

The management of pneumococcal infections has been challenged by the development of resistance and, more recently, the unexpected spread of resistant clones of serotypes such as 19A following the introduction of a conjugate PCV for use in children in 2000.

Despite the increasingly wide literature on antibiotic resistance, there is less evidence of the impact of this on clinical outcomes for children. However, series of children with pneumonia from the USA139^[III] and South Africa140^[II] found no difference in outcome between penicillin-resistant or sensitive pneumococal pneumonias, nor were differences noted in children with pleural empyema and sensitive or resistant pneumococcal disease in terms of duration of fever and tachypnoea, need for surgical treatment, bacteraemia incidence, mean duration of therapy or length of hospital stay.141^[III]

Outcomes in pneumococcal meningitis have not been shown to differ significantly between susceptible and resistant isolates.142^[III]

In the face of no widespread failure of antibiotic therapy, high-dose penicillin G (ie, in severe infection double the normal dose, as recommended in the British National Formulary for Children), other β lactams and many other agents continue to be efficacious parenterally for pneumonia and bacteraemia.130^[III]

Increased macrolide use is associated with pneumococcal and group A streptococcal resistance133^[Ib] and bacteria may acquire macrolide resistance very fast if used indiscriminately.143^[Ib] However, the clinical impact of macrolide resistance is unclear, with case reports describing clinical failure in adults with bacteraemic infection144^[III] but not in those with pneumonia.145^[II]146^[II] To date, no association with resistance and treatment failure has been demonstrated in children.

Evidence statement

• Although there appears to be no difference in response to conventional antibiotic treatment in children with penicillin-resistant S pneumoniae, the data are limited and the majority of children in these studies were not treated with oral β-lactam agents alone. [III]

Recommendations

• Amoxicillin is recommended as first choice for oral antibiotic therapy in all children because it is effective against the majority of pathogens which cause CAP in this group, is well tolerated and cheap. Alternatives are co-amoxiclav, cefaclor, erythromycin, azithromycin and clarithromycin. [B]

• Macrolide antibiotics may be added at any age if there is no response to first-line empirical therapy. [D]

• Macrolide antibiotics should be used if either mycoplasma or chlamydia pneumonia is suspected or in very severe disease. [D]

• In pneumonia associated with influenza, co-amoxiclav is recommended. [D]

Recommendations

• Antibiotics administered orally are safe and effective for children presenting with even severe CAP. [A+]

• Intravenous antibiotics should be used in the treatment of pneumonia in children when the child is unable to tolerate oral fluids or absorb oral antibiotics (eg, because of vomiting) or presents with signs of septicaemia or complicated pneumonia. [D]

• Recommended intravenous antibiotics for severe pneumonia include amoxicillin, co-amoxiclav, cefuroxime, and cefotaxime or ceftriaxone. These can be rationalised if a microbiological diagnosis is made. [D]

Recommendation

• In a patient who is receiving intravenous antibiotic therapy for the treatment of CAP, oral treatment should be considered if there is clear evidence of improvement. [D]

9.2.1 Pleural effusions and empyema

Parapneumonic effusions are thought to develop in 1% of patients with CAP171^[III] but, in those admitted to hospital, effusions may be found in as many as 40% of cases.172^[III] It has recently been reported that empyema thoracis may be increasing in incidence.173^[III]174^[III] A persisting fever despite adequate antibiotic treatment should always lead the clinician to be suspicious of the development of empyema.174^[III] Fluid in the pleural space is revealed on the chest x-ray and the amount of fluid is best estimated by ultrasound examination. A clinician should consider empyema when a child has a persistent fever beyond 7 days174^[III] or a fever not settling after 48 h of antibiotics. Where an effusion is present and the patient is persistently feverish, the pleural space should be drained, ideally in a specialist centre.

There is debate as to the best method of draining effusions. More details on the diagnosis and management of empyema are given in the BTS guidelines on pleural disease in children.113

9.2.2 Necrotising pneumonias

Lung abscess, although a rare complication of CAP in children, is believed to be an increasing and important complication.175^[III]176^[III] There are some data suggesting that some children are predisposed to this more severe form of lung infection. The predisposing factors include: congenital cysts, sequestrations, bronchiectasis, neurological disorders and immunodeficiency.177^[III] There are also emerging data that certain serotypes of pneumococcal disease are more likely to lead to necrotising pneumonia and abscess formation than others,175^[III] and that S aureus with Panton–Valentine leukocidin toxin can lead to severe lung necrosis with a high risk of mortality.178^[III] Suspicion of abscess/necrosis is often raised on the chest x-ray and diagnosis can be confirmed by CT scanning.179^[IVb] Prolonged intravenous antibiotic courses may be required until the fever settles. Lung abscess with an associated empyema may be drained at decortication if the abscess is close to the parietal pleura and is large. Ultrasound- or CT-guided percutaneous drainage can be used.180^[III]

9.2.3 Septicaemia and metastatic infection

Children can present with symptoms and signs of pneumonia but also have features of systemic infection. Children with septicaemia and pneumonia are likely to require high dependency or intensive care management. Metastatic infection can rarely occur as a result of the septicaemia associated with pneumonia. Osteomyelitis or septic arthritis should be considered, particularly with S aureus infections.

9.2.4 Haemolytic uraemic syndrome

S pneumoniae is a rare cause of haemolytic uraemic syndrome. A recent case series found that, of 43 cases of pneumococcal haemolytic uraemic syndrome, 35 presented with pneumonia and 23 presented with empyema.181^[II] Although a rare complication, in cases with pallor, profound anaemia and anuria, this should be considered.

9.2.5 Long-term sequelae

Severe pneumonia, empyema and lung abscess can lead to long-term respiratory symptoms secondary to areas of fibrosis or bronchiectasis. Children with empyema and lung abscess should be followed up after discharge until they have recovered completely and their chest x-ray has returned to near normal. There are also prospective data to suggest that children who have had an episode of CAP are more likely to suffer from prolonged cough (19% vs 8%), chest wall shape abnormality (9% vs 2%) and also doctor-diagnosed asthma (23% vs 11%).41^[Ib] The majority of children with CAP have no long-term sequelae and make a complete recovery. However, this study does suggest that some children do develop persistent expiratory symptoms, especially if they have a pre-existing diagnosis of asthma. The reasons for this are as yet unclear, but it is advised to counsel parents and carers at discharge to consult their doctor if these symptoms occur.

9.3.1 Staphylococcus aureus pneumonia

Pneumatoceles occasionally leading to pneumothorax are more commonly seen with S aureus pneumonia. The long-term outlook is good with normal lung function.182^[III]183^[III] There has been an increase in MRSA and some severe cases reported requiring extracorporeal membrane oxygenation.184^[III] Panton–Valentine leukocidin toxin-producing S aureus can lead to severe lung necrosis with a high risk of mortality.178^[III] In the UK and other developed countries, S aureus pneumonia is sufficiently unusual to warrant investigation of the child's immune system.

9.3.2 Mycoplasma pneumonia

Complications in almost every body system have been reported in association with M pneumoniae. Rashes are common, the Stevens–Johnson syndrome occurs rarely, and haemolytic anaemia, polyarthritis, pancreatitis, hepatitis, pericarditis, myocarditis and neurological complications including encephalitis, aseptic meningitis, transverse myelitis and acute psychosis have all been reported.

9.3.3 Streptococcus pneumoniae pneumonia

Pneumococcus is the most common bacterium to cause CAP and the major complication of empyema thoracis. It is increasingly being found to cause necrotic pneumonia and abscess formation that is believed to be associated with certain serotypes.175^[III] Vaccination programmes against pneumoccocus do not protect against all serotypes and surveillance studies monitoring for shift in serotype prevalence are ongoing. The rare complication of haemolytic uraemic syndrome is described with pneumococcal pneumonia.

10.2.1 Haemophilus influenzae

The impact of Hib conjugate vaccine on pneumonia in the UK is not known, but a number of clinical trials and case–control studies from the developing world have established that the introduction of this vaccine reduced radiologically-confirmed pneumonia by 20–30%.187^[Ib]188^[II] The WHO estimated that the global incidence of H influenzae pneumonia in the absence of vaccination was 1304/100 000 children aged <5 years.189^[Ib]

10.2.2 Bordetella pertussis

Whooping cough continues to be seen in the UK, with infants aged <6 months having the highest morbidity and mortality.190^[III] In the USA, from 1997 to 2000, 29 134 cases of pertussis were reported of whom 7203 were aged <6 months; 5.2% overall and 11.8% of those aged <6 months had pneumonia. There were 62 deaths, 56 (90%) of whom were aged <6 months.191^[III] Improved uptake of primary pertussis vaccination would help to prevent cases, but another important factor may be an increasing pool of susceptible older children and adults, which is why some countries have elected to have a booster vaccination programme in adolescence.190^[III]

10.2.3 Streptococcus pneumoniae

The introduction of conjugate PCVs has been the biggest recent change in pneumonia prevention. They have been hugely successful in decreasing IPD in children and there have been several studies of the effectiveness in decreasing respiratory morbidity. In the developed world, follow-up from the controlled trial of PCV7 in 37 868 children in the USA using the WHO standardisation for radiographic definition of pneumonia showed efficacy against a first episode of radiographically-confirmed pneumonia adjusting for age, gender and year of vaccination of 30.3% (95% CI 10.7% to 45.7%, p=0.0043) for per protocol vaccination.192^[Ib] Evidence that efficacy is sustained outwith a clinical trial comes from a time series analysis in the USA showing that, 4 years after the universal vaccination programme started, all-cause pneumonia admission rates in children aged <2 years had declined by 39% (95% CI 2% to 52%).193^[III] Similarly, three population-based pneumonia surveillance studies from US health maintenance organisations demonstrated fewer outpatient and emergency visits for pneumonia in children aged <2 years (a decrease of 19–33 per 1000 children per year),194^[III] a decrease of 6 (95% CI 5.4 to 6.7) per 1000 hospitalisations for all-cause pneumonia and a decrease of 40.8 (95% CI 38.8 to 42.7) per 1000 ambulatory visits in children aged <2 years,195^[III] and a significant 26% reduction in confirmed outpatient events for pneumonia in children aged <1 year.196^[III] A single-blind observational follow-up study of PCV7 in Italy also confirmed that radiologically-confirmed CAP was significantly less in the vaccinated group (RR 0.35; 95% CI 0.22 to 0.53).197^[II]

Introduction of the PCV7 conjugate vaccine in England and Wales in 2006 has almost abolished invasive disease caused by these pneumococcal serotypes in children <2 years and has substantially reduced the number in older children. However, there has been an increase in reports of invasive disease caused by non-vaccine serotypes.198^[IVb] A national time-trends study (1997–2008) recently published results on the impact of the PCV7 conjugate vaccination programme on childhood hospital admissions for bacterial pneumonia in the UK and showed a 19% decrease (RR 0.81; 95% CI 0.79 to 0.83) from 2006 to 2008.9^[III]

10.2.4 Influenza

The UK influenza vaccine programme for children is continually evolving following the H1N1 pandemic in 2009. There are no data of effectiveness in relation to childhood pneumonia in the UK. In Japan, analysis of all-age pneumonia mortality data suggested universal childhood vaccination offered population protection with prevention of one death for every 420 children vaccinated.199^[III] In Ontario, Canada the effects of introduction of a universal influenza immunisation programme were compared with targeted immunisation in other provinces.200^[II] After introduction, all-age mortality decreased more in Ontario than in other provinces, as did hospitalisations, emergency department visits and doctors' office visits in the paediatric age groups (<5 years and 5–19 years).