• Empyema
• pneumonia
• respiratory infection

Elemraid and colleagues raise important points about potential diagnostic misclassification and under-ascertainment using the Hospital Episodes Statistics (HES) database, as well as the absence of national pneumococcal serotype data for children.

The HES database covers all NHS hospital activity in England and has been widely used to report disease trends. It also provides the opportunity to estimate the clinical impact of major clinical policies on disease burden but, as with any large epidemiological dataset, has inherent weaknesses at the individual level.

Our study aimed to focus on common community-acquired bacterial pneumonia trends to evaluate the impact of the heptavalent pneumococcal conjugate vaccine (PCV7). We agree that pneumonia is a clinical diagnosis and that it is difficult to differentiate between bacterial and viral causes. Furthermore, a significant proportion of cases (8–40%) of community-acquired pneumonia can be of mixed aetiology.1 Hence, we applied broad pneumonia definitions. We aimed to focus on bacterial pneumonias and so excluded specific viral pneumonia ICD-10 codes (eg, ‘viral pneumonia, not elsewhere classified’—all J12 codes). The codes we searched are listed in the Appendix.

The authors highlight a useful point that some children diagnosed with pneumonia may have symptoms and/or signs recorded in the primary diagnosis field. Hence, we acknowledge under-ascertainment is possible for some pneumonia admissions. HES coding is dependent on the recording of the ‘reason for admission’ by clinicians and the subsequent coding by the trained staff, and we included this as a potential limitation in our discussion. We agree that such levels of miscoding are unlikely to have significantly changed over time. Therefore, this would suggest that the pneumonia admission trends that we observed are likely to represent real changes.

We used the Health Protection Agency cumulative weekly incidence reports of PCV7 and non-PCV7 isolates for children under 5 years,2 together with the national serotype surveillance for all ages,3 as the best available source of information on pneumococcal serotypes causing invasive pneumococcal disease. Admittedly, this covers a broader spectrum of invasive diseases. In the absence of published data specifically relating to serotype distribution of pneumococcal pneumonia for children, this is the only up-to-date national reference source available spanning our study time frame.2 3 It provides data on pneumococcal serotype distribution for cases of invasive pneumococcal disease for 2000/1 to 2005/6 and shows the most common serotypes present prior to the introduction of PCV7, which was relevant to our study.

We agree that it is not possible to determine fully the exact aetiology of pneumonia from an HES diagnosis, although we have made every attempt to do so. However, we do think the trends in diagnosed pneumonia following the introduction of PCV7 remain of interest.