Predicting future life expectancy is an important task in modern societies. Governments need predictions to estimate health and social care needs of the ill and the elderly, and to set equitable retirement ages between different generations.
Current models of life expectancy developed by actuaries (risk modellers) use aggregated mortality data from the Office for National Statistics, insurers and pension schemes. However, these models have not considered the more detailed data on diagnosed medical conditions and their treatment that can be derived from de-identified patient electronic health records (eHR). As such, current models of life expectancy could be overly simplistic and more likely to make poor predictions.
Around 100,000 people in the UK are admitted to hospital with a stroke each year, and there are around 1.2 million stroke survivors across the UK. Stroke is the 4th leading cause of death in the UK, and different studies have found hypertension to be the main controllable risk factor. This study will develop and test a new mortality model methodology using de-identified data from patient eHR to produce future predictions of UK stroke mortality based on data analysis of the impact of risk factors, such as hypertension, and the impact of anti-hypertensive treatments. This methodology will be used to estimate what further improvements could be achieved in UK stroke mortality by 2030 through better management of hypertension.
The Framingham Heart Study first established that there was a dose-relationship without a lower limit between high blood pressure, or hypertension, and both the likelihood and severity of a stroke. The Quality Outcome Framework (QOF) evaluates annual performance in primary care across the UK, and QOF estimates that 17% of hypertensive patients have a blood pressure of more than 150/90 mmHg. This study aims to develop a novel methodology for predicting future UK stroke mortality, and hence quantify the potential for further improvements in period life expectancy by 2030 through better management of hypertension.
This study will consider a 10% sample of those patients in CPRD GOLD meeting the following criteria:
1) Registered for 1+ year between 1 January 2010 and 30 June 2021
2) Registered at GP practices contributing "Up To Standard" data
3) Age 40+ at last birthday
4) 1+ readings of systolic blood pressure >=130 mmHg
5) No history of stroke recorded
Primary exposure data for calendar years 2010-2019 will be collected on blood pressure measurements and anti-hypertensive medications, as well as other covariate data. The outcomes of interest are first-ever stroke incidence and all-cause mortality for those with a prior stroke. A multi-state mortality model will be initially constructed using historical data on stroke incidence and deaths following a prior history of stroke. Various multivariate risk factor models will be constructed from the targeted exposures and outcomes to estimate the risk of BOTH first-ever incidence of stroke and subsequent death. Results from the optimal multivariate risk factor model will be used to enable the multi-state mortality model to estimate future trends in UK stroke mortality from now until 2030 according to different scenarios on the future effectiveness of hypertension management in primary care. All of the models will segment exposures/outcomes by sex and 10-year age group.
Incidence of first-ever stroke; All-cause mortality after prior stroke
Anita McGrogan - Chief Investigator - University of Bath
Daniel Ryan - Corresponding Applicant - University of Bath
Julia Snowball - Collaborator - University of Bath
Rachel Charlton - Collaborator - University of Bath