A key public health issue in Scotland is the high and rising number of alcohol related hospitalisations and deaths. Alcohol consumption and its associated harms are strongly socially patterned, making it essential to accurately estimate changes in consumption and harms across subgroups to understand the effects of policy changes, such as the 50p minimum unit price (MUP) for alcohol introduced in Scotland in 2018. Achieving reliable estimates of alcohol consumption and harm, and of how they vary across the population, requires methods that minimise nonresponse bias to ensure representativeness and address underreporting bias. However, high nonresponse rates in health surveys, coupled with measurement errors from respondents, interviewers, and data processors, hinder our ability to obtain accurate estimates using existing approaches. We find fewer alcohol related hospitalisations and deaths among participants in the Scottish Health Survey (linked to morbidity and mortality data) than we would expect given numbers of these events in the general population. We have developed novel multiple imputation methodologies that adjust for survey non-representativeness by creating synthetic observations for non-respondents based on record-linked alcohol related hospitalisations and deaths. Yet, alcohol consumption remains under-estimated compared to sales data, likely due to residual non-response and measurement errors in the survey data.
The successful candidate will apply Bayesian and other approaches to adjust for non-representativeness and measurement errors in alcohol consumption measures within Scottish Health Survey, linked to alcohol related hospitalisations and deaths. The adjusted datasets will then be used to evaluate how the impact of Scotland’s 50p MUP policy varies across sociodemographic groups (e.g. by age, sex and deprivation). The studentship will provide opportunities for learning advanced methods for dealing with non-response and measurement biases and evaluating policies using linked survey and administrative data.
You will be part of a supportive and stimulating research environment at the School of Health and Wellbeing, University of Glasgow. The supervisory team of Professor Peter Craig, Professor Jim Lewsey, and Dr Eliud Kibuchi have wide experience of policy evaluation using population surveys and administrative data, and of methods for addressing non-response and other biases in survey data.
Supervisory Team:
- First Supervisor: Professor Peter Craig, peter.craig@glasgow.ac.uk
- Second Supervisor: Professor Jim Lewsey, jim.lewsey@glasgow.ac.uk
- Third Supervisor: Dr Eliud Kibuchi, eliud.kibuchi@glasgow.ac.uk