The use of linked electronic health data to investigate the burden and outcomes of community-acquired pneumonia among older individuals in the United Kingdom.

The aim of this thesis was to use large linked electronic health datasets from primary and secondary care to better estimate the burden of community-acquired pneumonia (CAP) in older adults in the UK, and to identify the determinants of severe outcomes of these common infections. Hospitalisation for CAP is increasingly common in this older age group, and these patients remain at an elevated mortality risk for over a year after hospital discharge, making this an important area of study. CAP incidence was estimated at 7.99 episodes/1000 person-years (IQR:7.92-8.07/1000); rates were higher in men than women and rose strikingly with age. CAP incidence generally increased between 1997 and 2011, but this growth was attenuated when the rates were age-standardised. To separate trends in incidence from trends in treatment location, CAP episodes admitted to hospital within 28 days of diagnosis were compared to CAP episodes that were not hospitalised. A wide range of factors potentially associated with hospital admission were investigated, and 14 co-morbidities, five frailty factors, and four medications/vaccinations were identified. Despite adjusting for these factors, the average predicted probability of hospitalisation after CAP rose from 57% (1998-2000) to 86% (2009-2010), while duration of hospitalisation and 28-day mortality decreased. Finally, prognostic models were developed with the aim of assisting GPs in identifying CAP patients with an unexpectedly high mortality risk in the year after hospital discharge. Among 17 factors identified, increasing age, dementia, congestive heart failure, low weight, residential care and leukaemia/lymphoma were the strongest positive predictors of mortality, while being female, an ex-smoker and pneumococcal vaccination received more than a year ago had the strongest negative effects. The model showed a reasonable ability to distinguish between patients who died and survived. The linked data used in this study allowed greater capture of incident CAP episodes and thus better estimates of the burden of disease. They also provided enriched patient medical histories, enabling detailed examination of the determinants of hospitalisation or death after CAP. The results presented will be of use to both clinicians and health planners as the UK’s population ages, and burden of CAP increases. Further work is needed to fully understand the increasing hospitalisation trend seen, and to externally validate the prognostic models developed.