Estimating the changes in health inequalities across Scotland over time

Jack, Eilidh (2019) Estimating the changes in health inequalities across Scotland over time. PhD thesis, University of Glasgow.

Full text available as:
Download (40MB) | Preview
Printed Thesis Information:


Health inequalities are the unfair and avoidable differences in people’s health between different social groups. These inequalities have a huge impact on people’s lives, particularly those who live at the poorer end of the socio-economic spectrum, as they result in prolonged ill health and shorter lives. Much of the existing research into health inequalities in Scotland lacks analysis at the small area scale. The work in this thesis aims to fill that gap by estimating health inequalities in Scotland over time at the small area level used for data collection, which are known as intermediate geographies (IGs), as well as between Scotland’s 14 regional health boards, which are responsible for the protection and improvement of their populations' health. This thesis utilises conditional autoregressive (CAR) models which are the most common modelling approach for areal unit data. The first model proposed aims to estimate inequalities in risk of coronary heart disease from 2003 to 2012 across Scotland. However, focusing on a single disease gives an incomplete picture of the overall inequality in population health. Therefore, the second model proposed is a novel multivariate spatio-temporal model for quantifying health inequalities in Scotland across multiple diseases, which will enable us to better understand how these inequalities vary and correlate across diseases and how they have changed over time. This methodology is applied to hospital admissions data for cerebrovascular disease, coronary heart disease and respiratory disease, three of the leading causes of death, from 2003 to 2012 across Scotland. Finally, it was identified that a common problem in areal unit data of this type is changes to boundaries which occur during the time period for which data are available. This occured in Scotland when the IG boundaries were redrawn after the 2011 census. The final piece of work in this thesis aims to address the problem of spatial misalignment by proposing a multiple imputation approach which utilises a common latent spatial grid. This approach is applied to data containing hospital admissions for respiratory disease for the years 2006 - 2016 for the health board Greater Glasgow and Clyde, where the data from 2013-2016 are reported on the areas with redrawn boundaries.
Overall, it was found that there are still considerable health inequalities in Scotland at both the small area level and between Scotland's health boards. Although these inequalities appear to be decreasing over time for cerebrovascular and coronary heart disease, they are increasing for respiratory disease. In particular, the risk of most areas which were estimated to have a high risk of respiratory disease at the start of the time period are increasing at a higher rate than areas with low risk. It was also found that areas which experience high risk of one disease tend to experience high risk of all three diseases studied here. This highlights the issue that Scotland is facing and that more needs to be done to target the areas which are experiencing high risk of disease across multiple diseases.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: Bayesian modelling, disease mapping, health inequalities, multivariate spatio-temporal correlation, pseudo-continuous inference, spatial misalignment.
Subjects: Q Science > QA Mathematics
Colleges/Schools: College of Science and Engineering > School of Mathematics and Statistics
Supervisor's Name: Lee, Professor Duncan
Date of Award: 2019
Depositing User: Eilidh Jack
Unique ID: glathesis:2019-74312
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 12 Sep 2019 11:03
Last Modified: 05 Mar 2020 22:01
Thesis DOI: 10.5525/gla.thesis.74312
Related URLs:

Actions (login required)

View Item View Item


Downloads per month over past year