Evaluating the impact of minimum unit pricing on alcohol attributable hospital admissions and deaths in Scotland

Public Health Scotland (previously NHS Health Scotland) has been commissioned by the Scottish Government to lead the evaluation of minimum unit pricing (MUP) for alcohol, which was implemented in Scotland on 1st May 2018. A portfolio of studies has been developed by Public Health Scotland to evaluate MUP, which are summarised on the Public Health Scotland website: http://www.healthscotland.scot/health-topics/alcohol/evaluation-of-minimum-unit-pricing/mup-evaluation-overview.

Understanding the impact of introducing MUP in Scotland on the harm to health that alcohol can cause is a key outcome of the evaluation. To that end the evaluation includes a package of studies concerned with the impact of MUP on alcohol-attributable health harms, including hospital admissions and deaths. While the evaluation programme will assess a wide range of outcomes, the importance of a robust and credible assessment of the impact of MUP on these health outcomes has been emphasised by senior government officials and members of the MUP Governance Board. Indeed, much of the evidence that informed the legislation was focused on the potential impacts on hospital admissions and deaths caused by alcohol.

An important aspect in achieving a robust evaluation of the impact of MUP on alcohol-attributable health harms, is the inclusion of a concurrent geographical control. The purpose of this application is therefore to request alcohol-attributable hospital admissions data for England. England makes an ideal candidate as a control for Scotland in this study due to the geographical proximity, similarity in demography, culture and behaviours, and the fact that MUP was not implemented there.

The study will examine trends and patterns in alcohol-attributable hospital admissions in Scotland, making comparisons with England and two large, sub-national regions of England. The study will use a natural experimental design to assess the impact of MUP on hospital admissions and deaths caused wholly or partially by alcohol in Scotland, using routine administrative data.

Research Questions include:

• What is the impact of the introduction of MUP on alcohol-attributable hospital admissions in Scotland?

• What is the impact of the introduction of MUP on alcohol-attributable deaths in Scotland?

• To what extent does any impact of the introduction of MUP on alcohol-attributable hospital admissions and deaths vary by sex, age group and socioeconomic deprivation?

It is planned that data for England will be used as the geographical control group (head count of population, not a physical geographical area); data for sub-national English regions (North East and North West) will be used in supplementary analyses. The impact of MUP on overall admissions and separately for a range of demographic groups (sex, age group, area deprivation decile) will be assessed. Statistical models will be adjusted for sociodemographic characteristics (age group, area deprivation decile), seasonality and underlying trend.

The main outcome measures will be:

• All wholly alcohol-attributable deaths/admissions

• Acute wholly alcohol-attributable deaths/admissions

• Chronic wholly alcohol-attributable deaths/admissions

• All alcohol-attributable deaths/admissions (those wholly and partially caused by alcohol)

• All acute alcohol-attributable deaths/admissions

• All chronic alcohol-attributable deaths/admissions

• A selection of condition-specific outcomes.

To minimise the request, ICD10 codes pertaining to alcohol related conditions only are being used.

The ICD10 codes included in the request are limited to those defined as being either wholly or partially attributable to alcohol based on the latest available evidence (Sherk et al, 2017, Tod et al, 2018, Webster et al, 2018). In wholly attributable conditions, such as alcoholic liver disease, alcohol is the sole cause. In partially attributable conditions alcohol may be one of several factors that cause the disease. For these conditions alcohol-attributable fractions are used to quantify the proportion of those conditions that are caused by alcohol. Acute conditions are those that come on suddenly, have immediate symptoms and are limited in their duration. Chronic conditions are long lasting and can worsen over time.

To assess the impact of MUP on alcohol-attributable hospital admissions in Scotland, the Scottish Morbidity Record, a national data scheme that records comprehensive information relating to all inpatients and day cases admitted to either general acute or psychiatric hospitals in Scotland, will be used. Monthly data on the number of person-specific admissions by sex, age group, socioeconomic deprivation and condition group will be obtained by a request to the relevant department within Public Health Scotland.

Obtaining equivalent data for England (including sub-national regions) is the subject of this request. The inclusion of a concurrent geographical control, one which has not been exposed to the intervention in question, is considered good practice in natural experimental studies and is key in attributing any observed effect to the intervention (Craig et al, 2012, 2017; Leatherdale, 2018). The use of England or England & Wales (combined) as a concurrent geographical control is consistent with the approach taken as part of the wider evaluation of Scotland’s alcohol strategy that has been undertaken since 2009 through the well-established and highly regarded Monitoring and Evaluating Scotland’s Alcohol Strategy (MESAS) work programme. Work examining the impact of MUP on alcohol sales in Scotland published earlier this year used England & Wales as a geographical control in primary analyses (Robinson et al, 2020). Similarly, the proposed approach of using data for the whole of England is consistent with previous natural experimental studies to evaluate impact of the Alcohol Act legislation (including the multi-buy discount ban) on alcohol-related hospital admissions and deaths (Robinson et al, 2017) and alcohol sales (Robinson et al, 2014) in Scotland, as well as to evaluate the impact of lowering the drink driving alcohol limit (Haghpanahan et al, 2018).

While England represents a large geographical control area it is appropriate for the purposes of this study due to external factors that could potentially have an impact on both the Scottish and comparator population. Using data for the whole of England will reduce the impact of local policies and strategies that may impact on alcohol-related hospital admissions in specific areas of England. Data at national and government office regional level will be less susceptible to short-term change and random variation, thus strengthening the case for using a larger geographical control area.

It has been suggested that Northern England is a more appropriate control group for Scotland than the whole of England due to a more similar socio-demographic make-up and industrial and cultural history. Data for these regions are therefore being requested for supplementary analyses to test the robustness of our main results, where the whole of England will be used as the control. This is consistent with previous studies both within the MESAS programme (Robinson et al, 2014, 2017, 2020) and of other researchers in the area (O’Donnell et al, 2019).

The use of England or England & Wales (combined) as a geographical control is consistent with the approach taken as part of the wider evaluation of Scotland’s alcohol strategy that has been undertaken since 2009 through the MESAS work programme. The availability of data for England as the primary geographical control group (head count of population, not a physical geographical area) is therefore a key part of the proposed natural experimental study. There is no single perfect control area in this instance; the application therefore includes a request for data for the north of England. Analysis of both data for the whole of England and the northern regions are of equal importance in this study; the former allow us to address the issue of local strategic variation while the latter provide us with a closer and more demographically similar control area. The analysis of both areas provide the opportunity for triangulation of results and strengthening any potential causal inference. This is the approach taken in our recent assessment of the impact of MUP on sales-based consumption (Robinson et al, 2020).

Obtaining unsuppressed aggregated data for the whole of England is therefore necessary to ensure comparability with Scotland and to ensure that the impact of MUP is evaluated based on complete data. It is essential that unsuppressed data is obtained as a key part of the analytical process is the application of appropriate alcohol-attributable fractions. Using unsuppressed data will allow accurate calculation of the true burden of alcohol in relation to hospital admissions, both pre and post MUP-implementation. It will also allow examination of the impact of MUP on certain condition-specific causes as detailed in the research questions.

Three data extracts will be required based on the following definitions:

1. Admissions to hospital where the primary diagnosis is a code wholly attributable to alcohol.

2. Admissions to hospital where the primary diagnosis is any alcohol-attributable code or a secondary diagnosis is an alcohol-attributable external cause code

3. Admissions to hospital where the primary diagnosis or any of the secondary diagnoses are an alcohol-attributable code. Where there is more than one alcohol-attributable ICD10 code among the 20 possible diagnostic codes the code(s), the one from the lowest diagnostic position is selected.

Altogether, data from January 2012 to April 2021 are required and the data will be requested in two waves:

Wave 1 – up to and including the most recent available data

Wave 2 – an update to Wave including everything up to and including the end of April 2021.

In total this will provide data for over six years before, and three full years after the implementation of MUP.

The rate of each outcome measure will be calculated during each calendar month in each population subgroup. Monthly population counts for each subgroup will be estimated using interpolation of mid-year estimates. Data for the outcome measures will first be analysed descriptively to enable trends and other key information to be presented in tables and charts. In addition, the time series for each outcome will be separated to show trend and seasonal components separately which will aid interpretation of the trends.

Multiple regression (statistical technique used to analyse the relationship between variables) will be used to evaluate the impact of MUP on alcohol-attributable hospital admissions and deaths in Scotland. Both immediate and lagged effects will be assessed and a range of additional sensitivity analyses will be performed to test the robustness of the results.

Data will be processed by Public Health Scotland and the University of Glasgow on behalf of Public Health Scotland. In addition to the existing Service Level Agreement, a Data Sharing Agreement and Data Processing Agreement will be established between Public Health Scotland and the University of Glasgow for the purposes of this specific study. To that end Public Health Scotland will be the data controller and processor, while the University of Glasgow will be a data processor.

While the Scottish Government has tasked Public Health Scotland with leading the MUP evaluation and is partially funding the evaluation, Public Health Scotland remains the decision making authority in relation to the design and delivery of this study including the data being requested and how it is processed, therefore the Scottish Government are not considered to be a data controller for the purposes of this Agreement.

As the data will be aggregated and calculated as rates, there is no risk of anonymity being jeopardised or the potential for stigmatisation of individuals or groups. In addition, the study is overseen by an Evaluation Advisory Group which serves to provide advice and expertise and to provide assurance that will maximise the quality and relevance of the study.

The proposed research is therefore in accordance with Article 6(1)(e) of the GDPR which gives a lawful basis for processing personal data where:

“processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the controller”

Data on health is considered special category data by the GDPR. The purpose of processing NHS Digital Hospital Episode Statistics meets the condition set out in Article 9(2)(j) of the GDPR:

(j) processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes in accordance with Article 89(1) based on Union or Member State law which shall be proportionate to the aim pursued, respect the essence of the right to data protection and provide for suitable and specific measures to safeguard the fundamental rights and the interests of the data subject.

All organisations party to this agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract i.e.: employees, agents and contractors of the Data Recipient who may have access to that data).

No data will flow from Public Health Scotland to NHS Digital. The following data from HES APC will flow from NHS Digital to PHS:

Year

Month

Condition (ICD-10)

Sex

Age group

Government Office Region

Index of Multiple Deprivation (IMD) decile

Admission count

Aggregate data (which will include small numbers) received by Public Health Scotland will be used to calculate the count of alcohol attributable hospital admissions, by condition, in each population subgroup (age group, sex, IMD decile) by applying alcohol attributable fractions (AAFs). These data will then be shared with the University of Glasgow using secure encrypted email. Alcohol attributable admissions will be compared with those in Scotland, but there will be no data linkage performed as part of this project.

The researchers at the University of Glasgow will apply statistical analysis methods (controlled interrupted time series analysis) to the data. The team providing the processing service for Public Health Scotland have extensive experience in analysing hospital and mortality data (including highly sensitive linked data). The team regularly update their knowledge regarding data security and confidentiality via regular training sessions available at the University of Glasgow and have completed the MRC Research Data and Confidentiality course.

Both Public Health Scotland and the University of Glasgow will access data via a secure organisational network, requiring passwords to access. Any transfer of data will be via secure encrypted email.

The data will be stored by Public Health Scotland (data controller and processor) on a Storage Area Network (SAN) located at ATOS Data centre. ATOS provides a warehousing facility for Public Health Scotland’s IT hardware. This physical location provides multi-layered security access procedures to ensure the integrity of access to the physical systems. No staff member of ATOS has direct access to Public Health Scotland’s hardware or IT systems/networks.

ATOS provide storage for Public Health Scotland but do not access data held under this Agreement and are not listed as a data processor as they only supply the building. Therefore, any access to the data held under this Agreement would be considered a breach of the Agreement. This includes granting access to the database[s] containing the data.

There are no subsequent flows of data.

MUP as implemented in Scotland is unique and so this high-profile project will provide the first assessment of its effectiveness on alcohol-attributable health harms. The findings will be of interest to politicians, researchers and advocacy organisations globally. The results will be made available in a stand-alone Public Health Scotland research report. They will also be included in the final MUP Evaluation report that will inform the review report being laid before the Scottish Parliament. Both of these will be published on the Public Health Scotland website. All outputs will only contain results in highly aggregated format and as statistical summaries and measures of association. Statistical disclosure control will be applied where necessary; specifically, where numbers for a particular group are between 1 and 5, or other statistics derived from numbers between 1 and 5, these will be suppressed so as to avoid any potential confidentiality breach. Record level information will not be released to any third party.

Public Health Scotland is registered as a producer of official statistics with the UK Statistics Authority. As required under the code for producing official statistics, Public Health Scotland adheres to strict information governance protocols including that around statistical disclosure control.

Publication for the individual report is expected to be by spring 2023 and for the full evaluation report, November 2023. Following publication of the Public Health Scotland research report, the findings will be submitted for publication in a high-impact peer-reviewed journal. Opportunities to present at local, national and international fora will also be sought, so as to make the results available to a wide-reaching and global audience.

The findings will be used by local, national and international policymakers, public health practitioners & advocacy groups with whom the team already have strong, long-standing links. Consistent with the MUP Evaluation Communication and Engagement Strategy (available upon request) Public Health Scotland will: (i) involve key stakeholders throughout the project including via a study advisory group (ii) disseminate findings via social media, prepare a lay summary, and issue a press release (iii) send findings directly to MSPs & MPs on health committees/cross-party groups; (iv) present findings to alcohol policy and academic communities (e.g. Faculty of Public Health; Global Alcohol Policy Alliance; Scottish Parliament Health and Sport Committee; Kettil-Bruun Society).

The data will not be used for commercial purposes, will not be provided to any third parties, and will not be used for direct marketing.

Within Scotland, understanding the impact of minimum unit pricing for alcohol on alcohol-attributable health harms is a vital part of the evaluation, and arguably the most important in the chain of expected outcomes. It is hypothesised that MUP may reduce population consumption, particularly among harmful drinkers; it would be expected that reduced alcohol consumption would result in a reduction of alcohol-related health and social harms and, in turn, the demand on health and social services. Data on alcohol-related hospital admissions is therefore a key indicator of alcohol-related harms and, if MUP achieves its aim, would also indicate the potential to release capacity to meet unmet need and/or enhance quality, in health and social care services.

This evaluation is a legislative requirement and the findings from this project could inform the decision of the Scottish Parliament on whether the legislation should continue after six years of implementation. The importance of a robust and credible evaluation of its impact on health outcomes has been emphasised by senior government officials and members of the MUP Governance Board overseeing the overall evaluation. Indeed, much of the evidence that informed the legislation was focused on the potential impacts on hospital admissions and deaths caused by alcohol (Holmes et al, 2018; United Kingdom Supreme Court, 2017). The public benefit of this work to Scotland is therefore that it may be key in informing the decision as to whether MUP continues in Scotland beyond the 6th year of implementation.

The evaluation as a whole, and specifically understanding the impact of MUP on alcohol-attributable health harms, offers significant benefit to a much wider health and social care audience, including those in England. Broadly, the work serves to benefit the public interest by:

• providing an evidence base for public policy decision-making;

• providing an evidence base, nationally and internationally, for decisions which may impact significantly on population health

• significantly extending researchers' understanding of the impact of social policy on health.

This is of specific importance to the population of England as the potential introduction of minimum unit pricing in England remains under consideration: in July 2019, the UK Government stated that "There are currently no plans to implement minimum unit pricing in England. However, this will be kept under review as evidence emerges from Scotland" (House of Commons, 2019). The results of the proposed study are therefore directly relevant to policy development and decisions in England. In particular, if the evaluation of minimum unit pricing in Scotland is more robust due to the availability of sufficient and stable comparison population data (as is being requested here), this is hoped to provide better quality evidence to inform policy decision-making in England. Furthermore, the evidence is more likely to be generalisable to England if the comparison population consists of the entire English population because it would take into account the entirety of current alcohol policy there and examine the ‘added value’ of minimum unit pricing. It is evident through our regular MESAS Monitoring Reports that, since 2013, alcohol sales through the off-trade (supermarkets and off-licences) have risen steadily in England & Wales (Giles et al, 2020). Coupled with the increasing burden of alcohol attributable disease since the early 1990’s throughout the whole of the UK (Global Burden of Disease, 2016), understanding the impact of alcohol policy such as MUP on alcohol-attributable harm to health remains as important to the English policy audience, as it does globally.

The MUP legislation includes a review and sunset clause meaning that the findings from this project, alongside findings from a range of other projects, hope to inform the decision of the Scottish Parliament on whether the legislation should continue after six years of its implementation. Indeed, the UK Supreme Court considered the experimental nature of the legislation to be a key factor in their decision: “The system will be experimental, but that is a factor catered for by its provisions for review and sunset clause. It is a significant factor in favour of upholding the proposed minimum pricing regime” (UK Supreme Court, 2017). It is therefore crucial that the most robust study design possible is adopted; we believe that to be through the use of a concurrent geographical control.

It is hoped the findings from this study may be published in an individual report on the Public Health Scotland’s website no later than spring 2023. The findings may be synthesised alongside all other studies and evidence generated throughout the evaluation, into a final report to be presented to the Scottish Parliament no later than November 2023; it is hoped, this may form the basis of the decision-making process as to whether MUP will be continued beyond 6 years of implementation in Scotland. It is hoped the findings are submitted for publication in an international peer-reviewed journal and opportunities to present at international conferences will be sought, so as to make the results available to a global audience.

The data from NHS Digital will not be used for any other purpose other than that outlined in this Agreement.

Yielded Benefits is not a requirement for new applications.