Background Although antibiotics are recommended for the primary care management of community-acquired pneumonia, a recent UK study reported that most children admitted to hospital had not received antibiotics.
Objective To describe primary care antibiotic use for children subsequently hospitalised with community-acquired pneumonia.
Design/methods A case series of 280 children <5 years old hospitalised with pneumonia in Auckland, New Zealand. Pneumonia was defined as an acute illness with cough or respiratory distress, the presence of tachypnoea or indrawing and an abnormal chest radiograph. Receipt of antibiotics was determined by parental report and medical record review.
Results Fewer than half (108, 39%) of the children had received an antibiotic before hospital admission. For 60 children (21%) there had been no opportunity to prescribe because the illness evolved rapidly, resulting in early hospital admission. For the remaining 112 children (40%) an opportunity to receive antibiotics was missed. The parent failed to obtain the antibiotic prescribed for 23 children (21% of 112), but in 24 children (21%) pneumonia was diagnosed but no antibiotic prescribed and in a further 28 children (25%) the diagnosis was not made despite parental report of symptoms suggesting pneumonia. Missed opportunities to prescribe were not associated with increased overall severity of symptoms at hospital presentation but were associated with an increased risk of: focal chest radiological abnormalities (rate ratio (RR)=2.14; 95% CI 1.49 to 2.83), peripheral leucocytosis >15×109/l (RR=2.29; 95% CI 1.61 to 2.98) and bacteraemia (RR=6.68, 95% CI 1.08 to 58.44).
Conclusions Young children with community-acquired pneumonia may not receive an antibiotic before hospital admission because the illness evolves rapidly or the prescribed medicine is not given by parents. However, missed opportunities for appropriate antibiotic prescribing by health professionals in primary care appear to be common.
Statistics from Altmetric.com
Antibiotics are recommended as first-line treatment for community-acquired pneumonia and are almost always prescribed to preschool aged children on hospital admission with pneumonia.1 2 For example, the British Thoracic Society recommends amoxicillin as first-line treatment for children <5 years old with community-acquired pneumonia.3 However, the same guideline also says that children with ‘mild’ symptoms of lower respiratory tract infection (LRTI) do not require antibiotics. This creates a diagnostic challenge for primary care doctors. Not treating serious disease can result in death.4 Antibiotic overprescribing for lower respiratory infections increases the risk of antimicrobial resistance.5
What is already known on this topic
▶ Antibiotics are recommended for the primary care management of community-acquired pneumonia.
▶ Most children admitted to hospital with community-acquired pneumonia do not receive antibiotics prior to admission.
▶ The reasons for not prescribing antibiotics to children subsequently hospitalised with community-acquired are not known.
What this study adds
▶ Failure to prescribe in primary care is due to rapid illness evolution, prescriptions given but not filled or the opportunity to prescribe being missed.
▶ Missed opportunities to prescribe are common and are due to failure to diagnose and failure to implement recommended practice.
▶ In children hospitalised with community-acquired pneumonia missed opportunities to prescribe are associated with an increased risk of focal chest radiograph changes, leucocytosis and bacteraemia.
A case series of children hospitalised with community-acquired pneumonia in England called into doubt whether primary care doctors are meeting the diagnostic challenge effectively. It reported that only 22% of infants and 31% of children 1–15 years old had received antibiotics before admission.1 Moreover, the children not prescribed antibiotics had more severe disease on hospital admission.1 However, the report was not based on a consecutive series of cases from a defined population and no information was reported about the primary care given or the reasons for non-prescribing.
We therefore analysed a consecutive series of admissions of children from a defined catchment population in New Zealand (NZ) to confirm, in a country with a similar health system to the UK, the low rate of preadmission antibiotic prescribing for community-acquired pneumonia. We also sought to go further by consulting parents and reviewing primary care records to explore why this might happen.
Study design and setting
This was a case series of children <5 years old hospitalised with pneumonia at the Starship Children's Hospital in Auckland, NZ from 2002 to 2004. All children were resident in the city regions for which this hospital provided all inpatient care. The Ministry of Health Northern Regional Ethics Committee approved the study. Written informed consent was obtained from each child's care giver.
Children with pneumonia were identified from daily review of admissions. Children were eligible if they satisfied the WHO case definition and had an abnormal chest radiograph (consolidation or perihilar infiltrates). WHO defines pneumonia as present if a child has an illness with cough or difficulty breathing and there is age-specific tachypnoea or lower chest indrawing.6,–,8
Children with wheeze and therefore likely to have bronchiolitis, were excluded. Children were also ineligible if they had congenital heart disease, chronic lung disease, immunodeficiency, a recent history of aspiration, or had been admitted to hospital within the preceding 2 weeks.
Data collection and measurements
Consenting parents were asked to describe the symptoms they observed which indicated their child was unwell and the healthcare visits made up until the hospital presentation that resulted in admission.
For each preadmission healthcare visit made, the parent was asked if an antibiotic had been prescribed, whether the prescription was filled at the pharmacy and the number of days this antibiotic was given to the child. Children were categorised into three groups with respect to primary care antibiotic treatment: those who received antibiotics; those who had no opportunity to receive antibiotics; and those who had a missed opportunity to receive antibiotics.
From the last practice at which the child was seen before hospital admission, data were abstracted that documented acute illness assessment and management. Hospital medical records were reviewed to describe the illness at hospital presentation.
Pneumonia severity was defined from variables abstracted from the hospital medical records using the British Thoracic Society criteria (table 1).3 Socioeconomic status was measured using the NZ Index of Social Deprivation which divides households in NZ into socioeconomic deprivation deciles based upon national census data.9
Double-entered data were analysed using SAS-PC version 9.1 statistical software package (SAS Institute Inc, Cary, North Caroline, USA). Adjusted ORs and 95% CIs were estimated using logistic regression. As the ORs were determined on a cohort in which non-prescription of antibiotics was common, the rate ratio (RR) estimated from the OR was adjusted to more accurately represent the true relative risk.10
Of 415 children admitted with a diagnosis of pneumonia during the study period, 280 were included in the analysis; 65 children were ineligible and the parents of 70 children declined participation (see online supplementary figures). The child's age or household deprivation (data not shown) did not differ between the 280 children included in the analysis and the 70 whose parents declined participation.
The median age of participating children was 17 months (IQR 9–26 months). The children were from a range of ethnic groups: Maori 13%, Pacific 54% and European/other 33%. Households in socioeconomically deprived areas were over-represented (34% in the bottom quintile).
Prehospital use of antibiotics
One hundred and seventy-two (61%) of the 220 children did not receive an antibiotic before hospital admission. There had been no opportunity for 60 children (21% of 280) to receive antibiotics before admission because either they were not seen in primary care (eight) or parents were advised at the first primary care visit to go directly to hospital (52). However, for the remaining 112 children (40%, 95% CI 34% to 46%) an opportunity to receive antibiotics in primary care was missed.
Of the 220 children who had an opportunity to receive antibiotics before hospital admission, 202 (92%) had severe pneumonia and eight (4%) had mild pneumonia at hospital admission. The 10 remaining children had a respiratory rate between the criteria for mild (<50) and severe (>70 if an infant) pneumonia. Slightly more than a third of the 280 children (39%; 95% CI 33 to 44%) had received an antibiotic before hospital admission. In these 108 children, antibiotics had been given for a median of 4 days (IQR 2–6 days). Twenty-three (21%) children had received two (and in one case four) antibiotic courses. Amoxicillin, the recommended antibiotic for pneumonia in this age group in NZ,11 was prescribed to 72 (67%) and was the first antibiotic given to 67 children (62%). Other antibiotics prescribed were (in order of frequency): cefaclor, amoxicillin-clavulanate, trimethoprim-sulphamethoxazole, erythromycin and phenoxy-methyl penicillin.
Reasons for missed opportunities
Figure 1 shows the prehospital time-course of the illness. The children with no opportunity to receive antibiotics in primary care had a more rapidly evolving illness with a median time to admission of only 2 days (IQR 0–3 days). In contrast, the children with an opportunity to receive antibiotics had a median duration of preadmission illness of 4 days (IQR 3–7 days) with a median of two visits (IQR 2–3 visits) to a primary care doctor.
Table 2 shows that receipt of an antibiotic was more likely if the child was seen by their own general practitioner (GP). An antibiotic was less likely to be received if the primary care clinician failed to make the diagnosis of LRTI (21% vs 36%, p=0.03).
None of the parent-reported symptoms, including increased rate and effort of breathing, were associated with an increased likelihood of antibiotic prescription (table 3). At the last GP visit temperature was recorded for 67% and breathing rate for only 40% of the children. Based upon these two variables and the child's age, 87 of the children had severe pneumonia and 15 had mild pneumonia when seen at this GP visit. Having mild pneumonia was associated with an increased likelihood of being prescribed antibiotics (p=0.004). Table 3 also gives more detail about the clinical findings recorded in primary care, showing that even if a diagnosis of pneumonia was recorded only 45/70 (64%) children were prescribed an antibiotic.
Figure 2 explores the reasons for the missed opportunity to receive an antibiotic in primary care. Twenty-four children made one GP visit, 55 children made two GP visits and 33 children made three or more GP visits before admission. For 23 of the 112 children a prescription was written but not collected from the pharmacy. For 24 children the GP diagnosed pneumonia/bronchopneumonia but no antibiotic was prescribed.
Of the remaining 64 children, based upon parental reporting, 28 presented with increased respiratory rate or breathing difficulty, 23 with other respiratory symptoms (cough, wheeze or noisy breathing), 49 with systemic symptoms (vomiting, poor feeding, increased sleeping, irritability or lethargy). Forty-one of these children had both respiratory and systemic presenting symptoms. In 19 of these children (30%) the diagnosis of LRTI was not missed by the primary care clinicians.
Consequences of missed opportunities
Table 4 shows the potential consequences of a missed opportunity to give antibiotics before hospital admission. These analyses were adjusted for demographic and health service factors associated with the child receiving antibiotics in primary care (table 2). In a multivariate model of these variables two factors were associated with receipt of antibiotics in primary care. These were the number of primary care visits before hospital admission (p=0.007) and whether the child was seen by their own GP on any of these visits (p=0.004).
The univariate association (p=0.04) of having more severe pneumonia on hospital presentation if antibiotics had not been received was not evident in the multivariate analysis. Failure to receive prehospital antibiotics was associated with an increased risk of focal chest radiological abnormalities (RR=2.14; 95% CI 1.49 to 2.83), peripheral leucocytosis (RR=2.29; 95% CI 1.61 to 2.98) and of bacteraemia (RR=6.68; 95% CI 1.08 to 58.44).
Summary of the main findings
Most children admitted to hospital with community-acquired pneumonia (61%) did not receive antibiotics before admission. This had no impact on overall disease severity at admission but did increase the likelihood of focal chest radiograph changes, peripheral leucocytosis and bacteraemia. Failure to prescribe in primary care was explained by rapid symptom evolution requiring urgent hospital admission for about half of the children. Additionally, in about one-in-five children prescribed an antibiotic (18%), the prescription was not taken to the pharmacist by the parent. However, in about a third of all children (32%) the opportunity to prescribe in primary care was missed. This sometimes reflected diagnostic failure but also failure to implement national (and international) guidance on when to prescribe and what antibiotic to use, when a correct diagnosis of pneumonia was made.3 11 12
Comparison with existing literature
The study confirms the earlier UK report that only about a third of children admitted to hospital with community-acquired pneumonia will have received an antibiotic in the community.1 It is interesting that the impact of not receiving an antibiotic appears limited to specific disease features rather than overall severity of symptoms.
Prescribing antibiotics appears to make no difference to the time-course of prehospital illness in those eventually admitted to hospital. This might be because the pneumonia was not bacterial in origin in the majority, the first-line antibiotic was not effective at the dose prescribed, or the antibiotic was given too late in the illness.
The existing literature already makes clear the potential for diagnostic failure in primary care—half the cases of serious infection in children may be missed at first presentation.13 GPs are urged to avoid antibiotic prescribing to children with respiratory tract infection in order to minimise the real risk of causing resistance to first-line treatments.14 Most childhood respiratory tract infections for which GPs consider prescribing an antibiotic are of viral origin.15 However, secondary bacterial infection is not uncommon, particularly with influenza infection,16 and the clinical guidelines cited above argue that any LRTI in young children causing other than mild symptoms carries a sufficient risk of bacterial sepsis to warrant treating it with antibiotics.3 12
The increased risk of bacteraemia that was associated with not receiving antibiotics in primary care needs to be interpreted with caution given the small sample size and wide CIs. It also needs to be considered in an immunisation context. At the time this study was conducted conjugate pneumococcal vaccine was not part of the NZ national immunisation schedule.17
We were surprised by the lack of association between parent-reported symptoms and antibiotic prescribing. Parental observation of fast or difficult breathing identifies children with an increased risk of pneumonia.18 Parental concern that this illness is different from previous illnesses also predicts serious illness.18 19 It is impossible to know if the lack of association we observed reflects the inadequacy of our methods or a primary care failure to listen to parents. However, we did observe that appropriate antibiotic prescribing was more likely if the child was first seen by their own GP. A doctor who is familiar with a child will be more able to discern the extent to which the child's presenting behaviour differs from normal. A better relationship with the primary care provider is likely to enable a parent who has non-specific concerns that this illness is different to seek care, and for these concerns to influence treatment decisions.
One in five of the children prescribed an antibiotic in primary care did not receive it (because the prescription was not filled), which may be a particular problem in NZ. In NZ there is a co-payment for prescriptions. The government subsidises the manufacturer's price but this varies in a complex manner involving multiple patient subsidy levels and differential costs to the patient depending upon their primary care enrolment status and whether they are seen in an after-hours facility.20 The majority of the NZ public are uncertain what amount they will be charged by the pharmacy for a prescription.20 Uncertainty about cost adds another barrier to that caused by prescription co-payments.20,–,22
Strengths and limitations of the study
The strength of the study, particularly in the context of other publications, is that (1) it reports a consecutive series of admissions from a defined catchment population and (2) the data on prehospital care are based on parent interviews and primary care records. However, we were unable to conduct a face-to-face interview with each GP who saw the child and the interview with the care giver occurred a median of 6 days after hospital presentation. There was therefore a substantial risk of recall bias.
The key limitation in interpreting the study findings is that we included only children admitted to hospital. We have no data on children who avoided admission because their pneumonia was successfully treated. It is therefore impossible to determine whether primary care antibiotic treatment of pneumonia reduces the risk of hospital admission. We can be clear that existing guidance on prescribing is not being followed but we need to be cautious in drawing the conclusion that GPs are being too parsimonious with antibiotics. It is certainly possible that encouraging GPs to lower their prescribing threshold would do more harm than good (by exacerbating antibiotic resistance) unless we can help them increase their diagnostic precision.
Implications for clinical practice and research
Respiratory tract infections account for most antibiotic prescribing, and overprescribing is a key factor leading to increased rates of antibiotic resistance. Concerted efforts have been made to reduce antibiotic prescribing.23 In the UK, USA and NZ antibiotic prescribing rates for acute respiratory infections have decreased by one-third or more over the past two decades.24,–,26 Some of the largest decreases have been in preschool aged children.
As the proportion of children treated with antibiotics decreases, the risk of not treating an illness that requires antibiotics increases. Reducing antibiotic treatment for upper respiratory tract infection does not appear to have led to an increase in complications from streptococcal infection such as quinsy, mastoiditis and rheumatic fever.27 However, a recent study among adults presenting with LRTI in primary care suggested that failure to prescribe antibiotics is associated with an increased risk of hospital admission and respiratory infection-related death.28 Our findings underline the need to improve the precision of diagnosis of bacterial LRTI in primary care.
The documentation of disease severity for infants with cough and wheeze presenting to primary care in NZ has been shown to be variable but also to improve in response to implementation of a clinical guideline.12 The incomplete documentation of vital signs in primary care seen in this study implies that disease severity is still not as well assessed at primary care visits as it is on presentation to hospital emergency departments. Case management of pneumonia in children in developing countries has been shown to reduce pneumonia mortality substantially.29 Developing a successful clinical-decision rule to diagnose childhood pneumonia in primary care is not easy as it is difficult to achieve high sensitivity with adequate specificity.19 Use of near-patient C-reactive protein testing may help to target antibiotic use in primary care.30 But in order to be successful, implementation of any diagnostic decision-rule will need to recognise and deal with the socioeconomic and ethnic disparities in healthcare for pneumonia.31 As 100% sensitivity for any diagnostic strategy will never be obtained, an effective safety-netting is also likely to be crucial, particularly in economically deprived areas.32
Young children with community-acquired pneumonia may not receive an antibiotic before hospital admission simply because a rapidly evolving illness requires urgent admission. However, missed opportunities for appropriate antibiotic prescribing in primary care appear to be common. We must focus on helping GPs to diagnose pneumonia in children more effectively and better target their antibiotic use in treating LRTI. If we fail to do this, our justified efforts to avoid missed treatment of community-acquired pneumonia might easily render first-line antibiotics ineffective by encouraging overprescribing in primary care.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Funding This project was funded by research grants from the Health Research Council of New Zealand, the Auckland Medical Research Foundation and the National Child Health Research Foundation. CCG was supported by an overseas fellowship with the National Institute of Health Research National School of Primary Care.
Competing interests None.
Ethics approval The Ministry of Health Northern Regional Ethics Committee approved the study.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.