Lung function tests relate to how well our lungs are working and measure such things as how much air our lungs can hold and how easily we can breathe in and out. All individuals have some decline in their lung function over time, particularly in later life. People with chronic obstructive pulmonary disease (COPD) lose lung function faster than the general population, particularly those who continue to smoke. Lower lung function is associated with premature death and may lead to the inability to perform simple physical tasks such as walking short distances unaided. Studies done have shown great variability in lung function decline of COPD patients, but have not yet related this to what other diseases a patient may have, which are called co-morbidities, or to how long they can expect to live. Therefore, in this project we will study whether an integrated model of co-morbidities, lung function decline and time of death can produce more accurate risk scores for COPD patients.
People with COPD have a faster decline in their lung function than people without COPD. However, little is known about how quickly lung function declines in a primary care cohort in an average COPD patient. We are beginning to address this in an already approved clinical epidemiology project (16_186R). In this methodological study we wish to take this work further to study the impact of co-morbidities, testing the ability of landmarking analysis (i.e. dynamic prediction) to model the relationship between co-morbidities and FEV1 decline and their ability to predict mortality in COPD patients. This will be performed using 10-years worth of FEV1 data. In a secondary analysis we will also study latent class analysis can identify COPD patient subtypes with different prognoses. Preliminary work will establish whether co-morbid patients have different FEV1 decline profiles using mixed effect models, using inverse probability weighting to account for selection bias in those with sufficient longitudinal data. Analyses and interpretation will be performed with informative presence in mind, for example we will focus on prediction of mortality (which is well recorded) rather than on causal analyses which would have been biased by missing data and residual confounding.
Death
Steven Kiddle - Chief Investigator - AstraZeneca Ltd - UK Headquarters
Steven Kiddle - Corresponding Applicant - AstraZeneca Ltd - UK Headquarters
Hannah Whittaker - Collaborator - Imperial College London
Jennifer Quint - Collaborator - Imperial College London
Kieran Rothnie - Collaborator - GlaxoSmithKline Services Unlimited (UK)
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation