| Title: | Developing a risk prediction model for chronic disease mortality using multi-level survival modelling |
| Lead Institution: | Road to Health Group Limited |
| Principal investigator: | Mr Paul Nash |
Application 24321
WARNING: the interactive features of this website use CSS3, which your browser does not support. To use the full features of this website, please update your browser.
About
Health risk assessment of future disease has a number of applications and is in widespread use among clinicians and the general public. However, most applications have been focussed on single disease areas. Predicting death from diverse disease outcomes is complex and influenced by environmental and individual characteristics. In this study, we will explore the relationship between both environmental and individual characteristics and death due to major chronic diseases (cardiovascular, diabetes, pulmonary, dementia, cancers). This will lead to developing a novel and holistic application which can predict risk of death from chronic disease, informing individuals to take preventive measures. The study proposes to use novel methodology to: (i) identify the important social and clinical determinants of chronic disease mortality, and (ii) develop and validate a new prediction model for future mortality risk. By using a current and large population based cohort, we will be able to generate findings which will inform clinicians, policy-makers, and the general public of the most pertinent determinants of health. In developing and validating a risk prediction model, we can also generate useful applications for the prediction of an individual's future risk of death. The UK Biobank study cohort will be interrogated for demographic, clinical, medical history, family history, prescribing and behavioural variables, to develop an individual patient level predictive model in the first phase. These variables will then be assessed for their relationship with chronic disease mortality. In the second phase, post-code and regional data will be added into the model to create a `multi-level model` which accounts for the clustering of patients with similar individual characteristics in relation to the area of residence. This model will then be tested on whether it predicts mortality accurately or not. Due to developing a holistic model of mortality which will require investigation and assessment of multiple determinants of health, we will require the full cohort (approx. n = 500,000). To develop and validate the novel risk model will require splitting the dataset into a 75% random sample to develop the prediction model and the remaining 25% sample to validate the model. The use of the full cohort provides adequate sample size for this procedure, and will result in the development of a robust model.1 Publication
| Pub ID | Title | Author(s) | Year | Journal |
|---|---|---|---|---|
| 7514 | Prediction of premature all-cause mortality: A prospective general population cohort study comparing machine-learning and standard epidemiological approaches | Stephen F. Weng (+3) | 2019 | PLOS ONE |
Enabling scientific discoveries that improve human health