The development of health risk assessment models including spatio-temporal components with applications in the risks of health care associated infections (HAIs)

Risk prediction models are increasingly used within health care settings to manage patients and resources. The Scottish Healthcare Associated Infection Prevention Institute (SHAIPI) (http://www.gla.ac.uk/researchinstitutes/iii/research/researchcentres/sirn/shaipi/welcometoshaipi/)has a 3 year work stream to develop risk prediction models for health care associated infections (HAIs) in Scotland using routinely collected national, individual level data from hospitals, general practice and pharmacies.

Traditional risk prediction models use historical data collected on individuals to predict a 1 year, 5 year, or lifetime risk of a disease. Risk prediction for HAIs, however, is challenging due to temporal and fluctuating changes in risk due events such as admissions to hospital, which after a time the risk may return to a baseline level. In our research group we have recently published work the risk factors for a specific HAI, Clostridium, work which forms the cornerstone of the SHAIPI programme, whereby we will assess the ability of our model to predict HAI occurrence given previous antibiotic prescribing and other individual level risk factors. There is also potentially a spatial aspect to therisk of community acquisition of some HAIs for an individual, or groups of individuals. Aquisition may depend not only on the characteristics of the individual, for example the prescribing history, but also on the characteristics of the area (GP practice) that the individual attends. GP Practices within the same local area will share characteristics, some as a function of being within the same community health partnership and health board.

The successful PhD student will
(a) To write a comprehensive review of existing literature on statistical features of risk prediction models including the types of data sets that are used for the development of these models. This will classify the approaches used and identify the methodological gaps.
(b) develop spatio-temporal risk prediction models, test the model selection process and use simulated data with various types of spatial and temporal effects to validate the model predictions.
(c) derive a model for community acquired Clostridium Difficile and to assess the contribution of the spatial and temporal components of risk.
(d) develop novel data visualisation tools to display the risk predictions against comparator groups and to enable the user to investigate how the risk is modified by changes to the characteristics of the individual and the GP Practice.

Applicants should have a minimum of a 1st Class or 2.1 Honours degree. Preferably candidates will hold or be completing a Masters qualification in Statistics, Mathematics, or a mathematical science. To apply, email your CV, contact details of two referees and cover letter to [Email Address Removed] and [Email Address Removed].