This post is part of an innovative project to examine a series of 3300 CT scans in NHS Fife on older people aged 65 and over admitted as an emergency to hospital. The primary hypotheses are that (a) brain atrophy, and (b) white matter hyper-intensities (WMHs) are associated with higher delirium risk. The secondary hypotheses are that in patients with delirium, greater degrees of (a) brain atrophy, and (b) WMHs, predict worse outcomes such as mortality, readmission and new institutionalisation. For each of the hypotheses, appropriate variables such as age, pre-admission dementia status, physical comorbidity and activities of daily living will be controlled for.
The role of the statistician will be to analyse this cohort linked to routine datasets such as SMR01 and GRO. For outcomes that are binary over a fixed period of time such as 30 day mortality logistic regression modelling will be used to assess associations with CT factors described earlier along with binary markers for delirium, cognitive impairment, dementia, demographic factors and co-morbidity. Co-linearity will be assessed as there may be variables that are precursors or markers of each other in the model simultaneously.
For outcomes that are time to a binary event such as time to death the Cox proportional hazards model will be utilised. For outcomes such as time to readmission, it is important to account for the competing risk of mortality and this will be modelled using Fine & Gray regression.
You will have responsibility for designing and analysing the statistical modelling in Dundee. The principal investigator is Dr Vera Cvoro who is a consultant Geriatrician in NHS Fife and honorary Senior Lecturer at the University of Edinburgh, but in Dundee you will also work closely with Professor Peter Donnan (a biostatistician)