Spatial modelling of alcohol-related deaths and hospitalisations in Scotland

HV Social pathology. Social and public welfare;HN Social history and conditions. Social problems. Social reform;HA Statistics

Robinson, Emma ; Titterington, D.M.

University:University of Glasgow

Department:School of Mathematics and Statistics

关键词: Alcohol, whisky distillery, Scotland, Public Health, Deprivation, Bayesian Statistics, conditional autoregressive prior, Besag, York and Molli´e Model, uncorrelated and correlatedheterogeneity, random effects, hyperprior, disease mapping, Bayesian hierarchical modelling;   

Others  :  http://theses.gla.ac.uk/3580/1/2012robinsonmsc.pdf

来源: University of Glasgow

【 摘 要 】

Alcohol misuse in Scotland is a major issue which is extremely detrimentalto the health of the population and the economy (Donnelley (2008)). Thisthesis aims to explore the extent of alcohol-related health risks in Scotland ata finer geographical scale than previous research. A major objective of thisresearch is to geographically map alcohol-related health risks in Scotland formales and females separately.The Scottish data zone geographical areas are used in this study. Theseareas split Scotland into 6505 small sections each with a population of approximately500-1000 people where this is reasonable. Details of all alcoholrelateddeaths and hospitalistions in Scotland during years 2002 to 2006inclusive recorded at the data zone level are available. Information regardingarea deprivation and at-risk population structure at the data zone level hasalso been obtained. Indirect age and sex standardisation is used to work outhow many cases are expected to arise in each data zone.Firstly, the standardised incidence ratio is explored as an estimate of therelative alcohol-related health risk in each data zone in Scotland. This iscalculated separately for the combined male and female data, the male-onlydata and the female-only data. The results are mapped and discussed foreach.Further sections go on to use spatial Bayesian hierarchical modelling techniquesto estimate the relative alcohol-related health risk in each data zonein Scotland. Again these methods are considered separately for the combinedmale and female data, for the male-only data and for the female-only data.iThe basis for the models considered is the Besag, York and Molli´e model(Besag et al. (1991)). The models explore both uncorrelated and correlatedheterogeneity random effects. The correlated heterogeneity effects are fittedby means of the conditional autoregressive (CAR) prior. Fixed effects forarea deprivation are also considered.A further chapter explores a possible link between the location of singlemaltwhisky distilleries and alcohol-related health risks. This is done byincorporating the minimum Euclidian distance from the centroid of each datazone to a distillery into the Bayesian models already fitted to the combinedmale and female data.The final chapter gives a discussion of the project limitations, difficultiesand possibilities for future research.