Predictive risk assessment and risk stratification models based on geodemographic postcode-based consumer classification are widely used in the pension and life insurance industry. However, these are static socio-economic models not directly related to health information. Health information is increasingly used for annuity underwriting in the UK, using health status when the annuity is purchased. In real life, people develop new health conditions and lifestyle habits and can start and stop a certain treatment regime at any time. This requires the ability to dynamically classify clients into time-varying risk profiles based on the presence of evolving health-related conditions, treatments and outcomes. We incorporate landmark analysis of electronic health records (EHR), in combination with the baseline hazards described by Gompertz survival distributions, for dynamic prediction of survival probabilities and life expectancy. We discuss a case-study based on landmark analysis of the survival experience of a cohort of 110,243 healthy participants who reached age 60 between 1990–2000.
- School of Computing Sciences - Professor in Statistics (AVIVA)
- Business and Local Government Data Research Centre - Member
- Norwich Epidemiology Centre - Member
- Data Science and Statistics - Member
Person: Research Group Member, Research Centre Member, Academic, Teaching & Research