Scottish Health Survey - topic report: UK comparisons

INTRODUCTION AND METHODOLOGY

BACKGROUND TO THIS REPORT

This report draws on data from the most recently published health surveys conducted in the four countries in the UK:

• The 2008 Scottish Health Survey ( SHeS) ¹
• The 2008 Health Survey for England ( HSE) (2006 for the CVD data) ²
• The 2008 Welsh Health Survey ( WHS) ³
• The 2005/6 Northern Ireland Health and Social Wellbeing Survey ( NIHWS) ⁴

Prior to the 2008 report, the Scottish Health Survey reports have all presented comparisons between Scotland and England where available. Space constraints in 2008, and the fact that the two surveys now have different reporting timetables, meant that cross-country comparisons were not included. This report has two aims. Firstly, to update the information from previous SHeS reports about how Scotland fares relative to its nearest and largest neighbour, England. Secondly, it takes advantage of the opportunity to look more widely and include information about the other countries in the UK, Wales and Northern Ireland. The report looks at three broad areas: general health and wellbeing, including long-term conditions; smoking and alcohol consumption; and diet, physical activity and obesity. As the principal aim of the report is to compare Scotland with the rest of the UK, topics were only included if data existed for Scotland. Few of the topics were measured in all four surveys in a directly comparable way. However, most were included in at least three of them. Three topics (alcohol consumption, diet and physical activity) only draw on data from two countries (Scotland and England).

Other sources of health data exist in the UK. For example, each country maintains administrative records of hospital admissions and disease registers for certain conditions. In addition, the General Lifestyle Survey ( GLF) ⁵ conducted by the Office for National Statistics collects data in Great Britain (England, Wales and Scotland) on a range of topics including health and health behaviours. Although some of the measures in the GLF are similar to those in the four dedicated health surveys, its scope is less extensive, and the sample in Scotland and Wales is much smaller, so cross-national comparisons based on the GLF are more limited.

The following sections outline why an exercise comparing the four countries is useful, and describe the methodological similarities and differences between the four countries' surveys.

WHY A GEOGRAPHIC PERSPECTIVE IS USEFUL

Responsibility for public health and NHS services is devolved to the administrations in Scotland, Wales and Northern Ireland. Although many aspects of health are truly national in that they cover the whole UK, health service delivery in each of the four countries differs in many important respects. For example, policies relating to waiting times, patient choice, prescription charges, social care costs and governance arrangements all vary. Many of these differences reflect the new arrangements in place post devolution to Scotland and Wales in 1999, and to Northern Ireland to varying degrees since 2000. However, it should also be noted that many of the structural differences in health service delivery, particularly so in Scotland and Northern Ireland, pre-date devolution by many years. ⁶ To take just one example of post-devolution policy divergence, prescription charges were abolished in Wales in 2007 and in Northern Ireland in 2010. Scotland is committed to doing the same by 2011. In contrast, England has no plans for universal free provision of this kind. Public health policy has seen similar variations in both the policies adopted and the approaches being followed. For example, Scotland was first in the UK to implement a ban on smoking in public places in March 2006, followed by Wales and Northern Ireland in April 2007, and England in July 2007. In a similar vein, all four Chief Medical Officers in the UK support minimum pricing for alcohol to help tackle excessive drinking, and Ministers in all countries support some form of alcohol pricing measures. For example, the UK Government's coalition agreement ⁷ includes commitments to ban the sale of alcohol below cost and to review alcohol taxation and pricing. However, the current Scottish Government is so far alone in its attempt to try and bring in legislation for minimum pricing of alcohol (though note that the Welsh Assembly Government does not have the power to do this).

While notable differences exist between the ways in which health services are delivered, and public health policy is implemented across the UK, there is a high degree of convergence when it comes to the advice given to each population about healthy lifestyles. For example, uniform advice exists about sensible alcohol consumption, the recommended amount of physical activity and fruit and vegetable consumption, and the dangers of smoking. Health behaviours, such as smoking, drinking, physical activity and diet form the core of the survey content in all four countries. It is therefore particularly interesting to compare the prevalence of unhealthy behaviours in all countries in relation to the advice and recommendations in place across the UK. This interesting combination of varying policy levers alongside identical advice is further complicated by a wide range of environmental, economic, social and cultural differences at the individual, regional and national levels, all of which could potentially affect health behaviours including diet, smoking, alcohol consumption and physical activity participation. In turn, these could directly or indirectly affect health outcomes, including those explored in this report (self-assessed health, long-term conditions, psychosocial health and cardiovascular disease).

It is also worth noting that the demographics of the four countries of the UK are very different. In 2008 Scotland's population was just over 5 million people, Northern Ireland's just over 1.7 million, and Wales' just under 3 million. In contrast, England's population of over 51 million exceeds the other countries by at least ten times. ⁸ The table below shows that Scotland has the lowest life expectancy for both men and women in the UK.

	Life expectancy at birth in years (2006-8) 9
	Men	Women
Scotland	75.0	79.9
England	77.7	81.9
Wales	76.9	81.2
Northern Ireland	76.3	81.2

Improvements in health have been slower in Scotland than in many other comparable countries. In 1930, Scotland's life expectancy for men was ranked 9 ^th, and women's was 11 ^th, among 17 Western European countries. By the mid-1990s its position had fallen to 16 ^th for men (only slightly ahead of Portugal) and 17 ^th for women. ¹⁰ In contrast, while the rankings for England and Wales, and Northern Ireland, were similar to Scotland's in 1930 and had also declined by 1995, the decline in these countries was less sharp. In particular, Scotland continues to have the highest death rates from heart disease and lung cancer and the second highest death rates for stroke in Western Europe. ¹⁰

This report looks at country-level variation in health outcomes and behaviours, it does not consider any geographic inequalities in the social determinants of health or other factors which might underlie these. Although, as discussed further below, there is a significant body of evidence which suggests that socio-economic circumstances alone do not explain country-level differences in health outcomes.

A significant amount of analysis has explored why many health outcomes are found to be worse in Scotland than in England. Research in the 1980s initially linked this disadvantage to Scotland's area level deprivation. However, even when matched with deprivation quintiles in England, Scotland's relative health disadvantage remained. ¹¹ Analysis of the 2001 Census found that the health divide within the same social class across Scotland, England and Wales was widest in Scotland and London. ¹² This evidence suggests that there are factors at work, other than poverty alone, affecting health in Scotland. This unexplained difference has been termed the "Scottish effect". ¹³, ¹⁴, ¹⁵,, It has also been suggested that this unexplained "Scottish effect" associated with Scotland's relative mortality disadvantage has both endured over time and worsened. ¹⁶, ¹⁷, Of relevance to the factors explored in this report, previous analysis showed that significant geographic variation remained in smoking, fruit and vegetable consumption, diabetes and obesity between areas in Scotland and England once individual economic status was taken into account. ¹⁸ However, the relationship was complex and varied depending upon the risk factor involved and by gender. In some cases, the geographic variation was not between countries; some areas within both England and Scotland had better or worse outcomes than would be predicted by deprivation alone. Obesity in women stood out as the cardiovascular risk factor that demonstrated the strongest "Scottish effect", with the difference between Scotland and England the most pronounced of the factors analysed.

More recently, work has suggested that much of the excess between Scotland's morbidity and mortality rates, and what would be expected given its socio-economic characteristics, is located in the west of Scotland. ¹⁹ One question now being explored is whether the "Scottish effect" is in fact more localised and is more accurately a "Glasgow effect". This phenomenon is not unique to Scotland, however. A study that found certain regions in the north of England and Wales also had significantly worse health outcomes than would be expected by deprivation alone, has likened this to the "Scottish effect". ²⁰

SURVEY METHODS IN THE FOUR COUNTRIES

Background and comparability

This section presents a brief overview of the methods of data collection used in each of the four surveys. Further details can be found in each of their respective published reports (see references). Appendix 1 includes details of all the questions that this report draws on and highlights any key differences in either the wording or data collection method used. Items where the question wording or collection method were very different to that used in the Scottish Health Survey were omitted from the report at an earlier scoping stage. For example, the self-reported measures of height and weight in Wales have not been compared with the direct measures taken in the other three countries. In some cases a pragmatic judgement had to be made about whether the differences between the surveys still meant that the data were functionally equivalent, even if they were not completely identical. Such examples are highlighted in the text.

Health surveys make use of many standardised scales and measures so the extent to which the surveys are alike is partly a reflection of this. In addition, the design and content of both the Scottish Health Survey and the Northern Ireland Health and Wellbeing Survey were modelled on the Health Survey for England series which began in 1991. ²¹, ²², However, each survey was also developed to suit the needs of the populations and health providers in each specific setting and variations in content and design reflect this. As will become clear, the surveys in Scotland, ²³ England ²⁴ and Northern Ireland ²⁵ are closest in terms of their content and study design. In contrast, Wales ²⁶ uses a different data collection method and its questionnaire content is more distinct.

Sample design and response

The sample in each country was designed to provide data at the national level about the adult population aged 16 and over living in private households. All surveys also provide estimates for children. Three of the surveys also provide regional estimates that relate to their specific sub-geographies and administrative arrangements for delivering health care in each country (the 14 Health Boards in Scotland, 10 Strategic Health Authorities in England, and 22 local authorities in Wales). The sample frame in each country was a list of private households (Royal Mail's Postcode Address File for Scotland, England and Wales, and the Valuation and Lands Agency list in Northern Ireland).

The Scottish, English and Welsh surveys used a multi-stage, stratified sample design (the stratifiers used were specific to each country). The addresses selected in Wales were unclustered. In contrast, Scotland and England clustered the addresses by selecting them from small geographic units to help maximise fieldwork efficiency. ²⁷ The Northern Irish Survey used a stratified unclustered simple random sample. In each country, all adults aged 16 and over in each selected household were eligible for inclusion in the survey. ²⁸

The response rates, based on the proportion of households in which at least one adult participated, differed in each country, as set out below. This variation will in part reflect differences in the burdens associated with each data collection method; as described below, the Welsh Health Survey is largely based on short self-completion questionnaires.

• Scotland 61%
• England 64%
• Wales 74%
• Northern Ireland 66%

Non-response weighting was used to reduce bias in all four surveys. Each survey used calibration weighting to match the distribution of the achieved sample to mid-year population estimates drawn from the Census.

Fieldwork and data collection

As noted above, the data presented in this report are based on the most recently conducted surveys in each of the four countries for which data are publicly available. Fieldwork in Scotland, England and Wales took place in the 2008 calendar year; fieldwork in Northern Ireland was conducted between February 2005 and February 2006. ²⁹

The survey administration was very similar in Scotland, England and Northern Ireland. Each of these surveys used face to face interviews complemented by self-completion questionnaires and direct measurements of height and weight. The main difference between them was that Northern Ireland used a computer-assisted self-completion questionnaire whereas paper questionnaires were used in Scotland and England. Data collection was very different in Wales and this has implications for the interpretation of some of the data in this report. A face to face interview was used to collect information for each participating household but the main data collection from individuals used self-completion questionnaires. There were no direct measurements of adult height or weight.

In 2006 the London Health Observatory commissioned a separate parallel study based on the Health Survey for England. This study used self-completion methods instead of face to face interviews to help reduce costs. A comparison of the HSE survey data for London (collected via interviews), and the self-completion data from the parallel study in London, drew a number of conclusions of relevance to this report. ³⁰ Firstly, measures such as general health, long-term conditions and smoking prevalence were comparable across survey types and therefore the mode of data collection did not appear to bias the results. Secondly, there were significant differences between some estimates, including the number of units of alcohol consumed on the heaviest drinking day (which is reported in Chapter 3). While it is not possible to state definitively which estimate is the more accurate in each case, it is likely that self-completion reports of sensitive behaviours such as alcohol consumption volume are more accurate than those obtained via direct interviews. The likely impact of mode effects such as these are therefore discussed further in each section of this report, where applicable.

Details of each country's survey

Scotland

The 2008 Scottish Health Survey was commissioned by the Scottish Government Health Directorates. It was the fourth of a series of surveys aimed at monitoring health in Scotland (the previous surveys were conducted in 1995, 1998 and 2003). The principal focus of the 2008 survey was cardiovascular disease ( CVD) and related risk factors. The Survey consists of two stages, all those sampled are invited to take part in the first stage interview while a sub-sample is invited to also complete a second stage visit from a nurse. The interview includes questions on general health, mental health, cardiovascular disease, respiratory symptoms, eating habits, smoking, drinking, and physical activity, as well as height and weight measurements. This report only draws on data collected in the main interview. In future years, when the sample is larger, it will be possible to compare the data collected in the nurse visit with the equivalent HSE data (the Welsh and Northern Irish surveys do not include nurse visits). The study is conducted by a consortium led by the Scottish Centre for Social Research, including the MRC/ CSO Social and Public Health Sciences Unit in Glasgow, and the Department of Epidemiology and Public Health at the UCL (University College London) Medical School.

England

The Health Survey for England is a series of annual surveys that began in 1991. The 2008 survey was the eighteenth and was commissioned by the NHS Information Centre for health and social care. The survey's main focus changes every year. The 2008 survey's primary focus was physical activity, in 2006 the focus was cardiovascular disease (data from both these years are presented in this report). Each year the survey also includes a number of measures of key health outcomes and behaviours, such as general health, smoking and drinking, fruit and vegetable consumption, and height and weight measures. As noted above, the HSE also uses a two stage design involving a first stage interview and a second stage nurse visit. Unlike SHeS, all participants at the first stage are eligible to proceed to the second. The study is conducted by the Joint Health Surveys unit of the National Centre for Social Research and the Department of Epidemiology and Public Health at the UCL (University College London) Medical School.

Wales

The Welsh Health Survey is commissioned by the Welsh Assembly Government and has been carried out annually since 2003. Prior to 2003 two health surveys were conducted in Wales, dating back to 1985, and the WHS was designed to replace these two sources. Its topics include key health behaviours such as smoking, drinking, fruit and vegetable consumption and physical activity, as well as illness burden and use of NHS services. As noted above, the study design differs to the one used in the other three countries in the UK, although the topics are broadly similar. The study is conducted by the National Centre for Social Research.

Northern Ireland

The Northern Ireland Health and Wellbeing survey has been conducted three times so far, in 1997, 2001 and 2005/6. It is commissioned by the Department of Health, Social Services and Public Safety in Northern Ireland. As noted above, the topics and data collection methods are similar to those used in Scotland and England, however the 2005/6 survey did not include a second stage nurse visit. Its principal topics include cardiovascular disease, general and mental health, physical activity, smoking, drinking and height and weight measures. The study was conducted by the Northern Ireland Statistics and Research Agency.

Data analysis

Interpreting differences between countries

The statistical significance of the differences between the figures for Scotland and the other countries presented in this report has been tested. In addition, confidence intervals for the differences have been calculated to help assess the precision of the estimates. These help to identify instances where the large samples in each country suggest that the differences in populations are statistically significant even though the size of the true population difference could be very small. Although statistical significance has long been the standard way of assessing the probability that a finding is real, precision estimates are a better way of judging the substantive importance of results of analyses such as these. ³¹ Using a hypothetical example, a difference between two countries of four percentage points (pp), with a confidence interval of 1.2-7.5, means that the difference between the two populations might be as small as just 1.2 pp, or as large as 7.5 pp. It would be unwise to place too much importance on a difference between countries as small as 1.2 pp, without additional information to support the interpretation, such as confirmation from other studies that the difference has important consequences. The conventional threshold for significance testing uses the 5% level, which means that something significant at this level has a one in twenty chance of not being real. As the samples in each country are quite large, and the number of tests being conducted was also quite high, a more conservative approach was used for this report. ³² Only differences significant at the 1% level are described as significant, while those at the 5% level are described as "marginally significant". Table A1 at the end of the tables for all the chapters presents the differences between estimates highlighted in the text and the associated confidence interval and the p value.

Data availability

Some of the data presented in this report have been taken directly from published sources. All of the figures for Scotland and Wales have been taken from the 2008 reports of those surveys, or from tables published on the web. ³³, ³⁴, Some of the data for England are taken from the 2006 ³⁵ and 2008 ³⁶HSE reports. The figures for England in the following tables are based on re-analysis by the authors and do not therefore match similar tables in the main HSE publication series: Table 1.4 (only the any CVD or diabetes row), Table 2.1 (all figures), Table 3.2 (all figures). Details of these re-analyses are provided in the Appendix. All of the figures presented for Northern Ireland are based on re-analysis by the authors using the data from the UK Data Archive.

The data for all four surveys presented in this report (and for all previous years) are available from the UK Data Archive: www.data-archive.ac.uk/.