Every year about 1% of the UK population will develop pneumonia. Between 20-40% of these people will have to stay in hospital to be treated. Some of these hospital stays could be avoided through better out of hospital healthcare and public health programmes. Any unplanned hospital stay is hard on patients, can affect planned health services and costs the local health service extra money. Being able to identify which groups of people are at risk of a hospital stay for pneumonia will help to target prevention programmes to their needs.
This study will use data from the health service, as well as death certificates, to build tools which can identify groups of people at risk of a hospital stay for pneumonia. The study will also describe these groups - for example, how long they stayed in hospital, how many died, and how many had to come back to hospital after they went home. We will work with NHS staff who plan health services to develop these tools. In this way, any tools we develop should help planners to make care for patients at risk of pneumonia better.
Study Objective: To identify and characterise discrete adult sub-populations at risk of unplanned hospitalisation following community acquired lower respiratory tract infection (CA-LRTI) in UK population using linked electronic health records to inform the development of group-oriented admission avoidance schemes.
Methods: We will use linked primary care data (CPRD GOLD), A&E and hospital admission data (HES) and mortality data (ONS) to identify sub-groups of patients with shared characteristics at risk of hospitalisation for CA-LRTI. We will predict this risk over three time intervals (6 months, 1 or 2 years) using as explanatory variables those describing patients' demographic, social and personal situation, frailty markers, health behaviours, biomarkers, co-morbidities, medication and immunisation status, healthcare utilisation and healthcare and area features. Once identified, we will describe the outcome of hospitalisations (length of stay, unplanned re-hospitalisation, mortality), prevalence, population attributable fraction, and costs of healthcare utilisation for each sub-group.
Data analysis: We will utilise decision-tree based supervised statistical learning methods to identify sub-groups at risk of hospitalisation and evaluate them in terms of accuracy and model interpretability. We will estimate the proportion of admitted patients with subsequent adverse outcomes for each subgroup. We will calculate the mean and median cost of hospitalisation for each sub-group, using the relevant NHS tariffs.
Primary Outcome:
Unplanned hospitalisation (all cause)
Secondary Outcomes:
Unplanned hospitalisation for community acquired lower respiratory tract infection / pneumonia; unplanned hospitalisation >7 days; unplanned hospitalisation > 21 days; unplanned hospitalisation for less than median length of stay; unplanned re-hospitalisation within 30 days; all-cause mortality within 30 days of unplanned hospitalisation
Julie George - Chief Investigator - University College London ( UCL )
Julie George - Corresponding Applicant - University College London ( UCL )
Ana Torralbo - Collaborator - University College London ( UCL )
Andrew Hayward - Collaborator - University College London ( UCL )
Carlos Andres Valencia-Hernandez - Collaborator - University College London ( UCL )
Dionisio Acosta Mena - Collaborator - University College London ( UCL )
Jennifer Quint - Collaborator - Imperial College London
Liam Smeeth - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Sarah Sharer - Collaborator - University College London ( UCL )
Sonya Crowe - Collaborator - University College London ( UCL )
Spiros Denaxas - Collaborator - University College London ( UCL )
HES Accident and Emergency;HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation