• Bacterial infection
• pneumonia
• respiratory infection

We read with interest the comments of Dr Singh and colleagues.1 We note that in the data submitted from their local population they found a lower mortality for patients with high severity community acquired pneumonia (CAP) (CURB65 score 3–5) than was found in the National cohort.2 It is not possible to precisely explain this difference from the available information, but there are a number of likely explanations. The first is sampling error, which will be inevitable when small numbers of cases are submitted from a large number although their local data collected over a longer time period than the British Thoracic Society (BTS) audit suggest that their lower mortality score is real. A differential response to treatment is a consideration, but since only selected aspects of patient care were documented in the audit, this aspect cannot be properly examined. The most likely explanation of all is probably case selection, which may be slightly different in each contributing centre as we alluded to in the paper. Contributors may have used a number of different methodologies to identify and corroborate CAP cases, but data on how this was done were not collected. The 2010/2011 audit may go some way in answering this difference since additional data, such as nursing home residence, were obtained.

High severity (CURB65 score 3–5) CAP has a mortality of between 15% and 40%, so treatment should primarily be used that is most likely to limit this mortality. Ideally, antibiotic recommendations would be based on robust randomised controlled comparative treatment trials, but unfortunately such trials are lacking in this area. What data are available are mainly from weak, non-robust trials or retrospective or observational data such as the study referred to by the authors.3 It is therefore impossible to be certain that one antibiotic regime is categorically preferable to another in this setting. The ecological importance of good antibiotic stewardship is acknowledged in the BTS guidelines and where possible narrow spectrum agents have been recommended in preference to broad spectrum agents.4 Since only one-third of admissions for CAP are of high severity, we consider the choice of penicillin plus clarithromycin by Singh et al compared with the BTS recommendation of co-amoxiclav plus clarithromycin to be one that is unlikely to make a significant difference to antibiotic resistance or Clostridium difficile rates. Of much more importance as a driver of resistance and C difficile rates are the failures to accurately diagnose CAP and to accurately assess CAP severity. These result in overuse of antibiotics, including combination regimens primarily indicated for high severity CAP, in patients who should not require such broad therapy as identified in the audit. Penicillin plus clarithromycin may be as effective as co-amoxiclv plus clarithromycin in high severity CAP, but there are no robust trial data to demonstrate this. Co-amoxiclav has the advantage of a broader antibacterial coverage, which is important for this severely ill group and easy conversion to an oral formulation for the continuation phase of treatment. We acknowledge that local data on resistance and C difficile rates may mean that an antibiotic choice other than that recommended in the BTS guidelines may be appropriate for that local setting.

We agree with the authors that National audit data should be interpreted carefully as the data collection processes at each contributing centre may not be as robust as in carefully conducted cohort studies by interested experts, but one of the strengths of the National cohort is its size. For the first time, we have an up-to-date picture of how CAP is currently managed in the UK, where this differs from guideline recommendations and where attention should be paid to lead to improvements in practice. We believe that the BTS guidelines are a reasonable translation of the available scientific evidence with regard to this topic, but we also acknowledge that they are not perfect and may not be appropriate for all settings. Inevitably, they are weakest where there is least evidence and choice of antibiotics is one such area. We would like to see the guidelines improve, but this can only occur with better evidence. This requires future funding for clinical research in this important, but research-neglected area.