healthscotland.com

Health Education Population Survey (HEPS) 1996-2003

Contents:

Summary 1. Introduction 2. Scotland's health 3. Attitudes towards own health 4. Perceptions of mortality, morbidity and risk 5. Physical activity 6. Diet 7. Smoking 8. Alcohol 9. Mental health 10. Oral health 11. Drugs 12. Sexual health 13. Conclusions References Appendix 1: Questions analysed in overview Appendix B: Topics covered 1996-2003 Appendix C: Base sizes Appendix D: Physical activity levels

1. Introduction

1.1 Background

The Health Education Population Survey (HEPS) monitors health-related knowledge, attitudes, behaviours and motivations to change among the adult population in Scotland. This report presents an overview of key findings and trends from HEPS data during the first eight years (1996-2003). The survey was first commissioned in 1995 by the Health Education Board for Scotland (HEBS) which was the national health promotion agency in Scotland from 1991-2002 and one of a number of agencies with a role in improving the health of people in Scotland¹. HEBS worked in three main strategic areas:

• communication with the general public to improve awareness, knowledge and motivation regarding key health-related issues
• enabling informed and effective practice among health promotion practitioners and other professionals
• developing policies, strategies and infrastructures, together with other partner agencies and sectors, to support individual and public health improvement

A core element of the work of HEBS, and health promotion generally, is to increase public awareness of health-related risk factors and how to make the lifestyle changes necessary to reduce such risks. Health education, information and communications activities seek to influence people's health-related knowledge and attitudes and to motivate and support the process of behaviour change. The main purpose of the survey is to collect the data required to monitor progress towards achieving this aim with respect to the priority topic areas identified in a series of policy documents on improving Scotland's health (Health Education in Scotland, 1991; Scotland's Health - A Challenge To Us All, 1992; Towards a Healthier Scotland, 1999; Improving Health in Scotland - The Challenge, 2003).

The indicators presented in this report concern knowledge, attitudes, motivation and behaviour/health status among adults in relation to the following topics:

• attitudes towards own health
• causes of mortality and morbidity
• physical activity
• diet
• smoking
• alcohol
• mental health
• oral health
• drug use
• sexual health

¹In April 2003, HEBS was merged with the Public Health Institute of Scotland (PHIS) to form NHS Health Scotland.

While there are some core questions on these topics that were asked at each survey wave, other questions were asked less frequently, or formed part of modules which were only included in particular waves. The questions providing the data presented in this report are listed in Appendix A, and the list of topics covered between 1996 and 2003 is shown in Appendix B.

In addition to providing performance monitoring data for the public communications and educational aspects of health promotion, the information collected by HEPS contributes towards the planning and development of future health promotion initiatives.

The data from HEPS should not be seen in isolation. It is intended to complement other health population surveys, most notably:

• the Scottish Health Survey commissioned by the Scottish Executive Health Department to provide monitoring data on the prevalence of specific health conditions and associated risk factors in the Scottish population
• the Health Education Monitoring Survey (HEMS) commissioned by the Health Education Authority and carried out between 1995 and 1998, collecting similar data for England as that provided by HEPS for Scotland
• the Health in Wales Survey commissioned by the Health Promotion Division of the National Assembly for Wales to track progress towards the achievement of health improvement targets for Wales
• local health and lifestyle surveys carried out periodically by Scottish health boards

Data collected through HEPS are presented for Scotland as a whole with breakdowns by age, sex, socio-economic grade, and deprivation category (DEPCAT). The annual sample size is too small for meaningful analysis at health board level and makes it difficult to draw reliable and statistically significant conclusions for some sub-groups of the population. Boosting the sample to allow more robust statistical analysis by sub-groups and health board regions would entail significant cost increases, and ones that have not been regarded as justifiable relative to the utility of the information. It may be useful to note that the 1995 and 1998 Scottish Health Surveys found relatively little regional variation across a range of indicators when differences associated with age and social class are taken into account (Dong and Erens, 1997; Shaw et al, 2000). These surveys are substantially more heavily resourced than the Health Education Population Survey, as they are intended to monitor progress toward national health targets. The regional differences which were observed are also difficult to interpret without taking into account contextual influences such as economic history and patterns of population migration.

However the HEPS dataset does provide a useful indication of general population trends for adults in Scotland. It can contribute to local planning activity by highlighting topic areas and population groups of particular interest. This in turn may guide more focused needs assessment to inform local service provision. By describing the broader national context within which local activity takes place, national findings can be used as a benchmark against which to assess the local situation in relation to local health promotion needs and developments.

1.2 Methodology

The survey

The survey was commissioned by HEBS in 1995 and is conducted by BMRB International. Fieldwork began in March 1996 and is carried out twice a year (March and September) in mainland Scotland. The survey was suspended for three waves during 1999-2000 while resources were diverted to data analysis activities for the 1996-1999 dataset (Walters, 2000; Phillips et al, 2001). There is thus a gap in data collection for the three survey waves covering September 1999, March 2000 and September 2000.

The survey is administered using computer assisted personal interviewing (CAPI) in respondents' homes, including a self-completion section for more sensitive issues such as mental health, sexual health and drug use. Each survey wave has an achieved sample of approximately 900 people aged 16-74 years. Respondents are selected using a multi-stage clustered random sampling design with the Postal Address File as the primary sampling frame. A 'rolling' sampling procedure allows results to be combined from consecutive waves. The data are weighted to adjust for differing probabilities of selection and response bias with respect to sex and age. Most questions are asked using prompted closed-format response categories, and those asked using unprompted open-format are identified in the text.

Sample size and response rates

This overview report presents key data from all waves of the survey (1996-2003), using combined results for each year from the two waves of the survey conducted in that year. The exception to this is 1999 when only one survey wave was carried out. The results for 1999 should therefore be treated with considerable caution since the much lower base will produce much greater levels of random variation. The number of achieved interviews and response rates with respect to the eligible sample are shown below for each year.

	1996	1997	1998	1999	2000	2001	2002	2003
Achieved interviews	1810	1795	1794	880	-	1757	1742	1720
Response rate	72%	73%	72%	72%	-	71%	72%	72%

Analysis

Annual summary reports were published for each of the first three years (HEBS, 1997a; HEBS, 1999; HEBS, 2000). These provided basic descriptive statistics noting any significant differences between socio-demographic groups for each year's data. This report focuses on changes over time for key variables, both for the whole population and for sub-groups with respect to sex, age, social grade and deprivation (base sizes for sub-groups are given in Appendix C). Differences between years are tested for statistical significance using t-tests for means or hypothesis tests for proportions, as appropriate. Unlike other significance tests, these tests also take into account the estimated design effect due to the sampling procedure (see below for more details). The report describes observed changes and explicitly points out where such a change is statistically significant (p<0.05). Differences should not be considered statistically significant unless it is specifically stated. Any use of the term significant is taken to mean statistically significant, but the use of this term does not imply substantive significance or importance. Changes over time that are significant are indicated by shaded rows in the tables. It should be noted that given the relatively small size of some of these differences, some caution is recommended in interpreting and generalising from this data in the absence of other supporting evidence.

The significance tests have been applied to look for change between two points in time (e.g. between 1996 and 2001). The period of change considered will be specified in the text, and any change will be considered in context. If the apparent change is not supported by trends of change in between the two points in time, or by change being sustained in the longer term, or by evidence of change from other sources, then caution should be used in interpreting the apparent change.

Readers should also exercise caution when considering time trend data over the relatively short period considered in this report. On the whole, changes at a population level occur gradually and underlying trends only become apparent over longer time periods. Apparent changes in the short term may in the long run turn out to be a result of random variation within the population, and predictions based on short-term patterns observed for past data may therefore prove misleading. Long-running time series, such as smoking rates from the General Household Survey, often have 'blips' in the data which can be misleading when taken out of context. Smoking rates in the UK have fluctuated between 27% and 28% since the early 1990s (27% in 2000), but the shift from 27% in 1994 to 28% in 1996 was reported as signalling a possible upward trend (ONS, 1998). The latest report on the General Household Survey notes that upward and downward movements in survey estimates are to be expected due to sampling fluctuations, even when prevalence in the population is not changing (Walker et al, 2001).

It is also not appropriate to attribute observed changes definitively and solely to health promotion activity as many other factors (e.g. macro-economic change, commercial marketing) will influence health-related attitudes and behaviours.

Sample design effects

When using survey data it is important to take the sample design into account. The aspect most likely to affect the precision of the results is clustering, where the sample is selected from a number of geographical areas in order to increase the efficiency of the fieldwork and reduce costs. While clustering does not introduce a bias, it can increase the estimation error of observed prevalences relative to a more dispersed simple random sample, and thus result in a lower level of theoretical precision. The reduction in precision can be measured by calculating the design effect, with larger values representing reduced levels of precision. This is calculated by looking at the variation in responses between sampling points. Taking into account the design effects across a range of questions, an average design effect for the survey can be defined. The variables selected in this case were those deemed most likely to reflect regional variations in health status and behaviour, and therefore the estimated design effect is likely to be larger than the true value.

For this survey, the design effect was 2.3 (the design effect for a simple random sample is 1). This is taken into account when testing for statistical significance by multiplying the standard error assumed for a simple random sample by the design effect, which means that an observed difference needs to be more marked in order to be statistically significant. The estimate of the design effect for HEPS errs on the side of caution, and is relatively large compared with bigger surveys such as the Health Survey for England and the Scottish Health Survey, which tend to have design effects of 1.2 or less. However, it can be considered small relative to school-based surveys where the design effects are larger due to the combination of more clustered sampling, institutional school effects and peer-group influences. As mentioned previously, both the Health Survey for England and the Scottish Health Survey are designed to make a substantial contribution to monitoring progress toward national health targets, and are accordingly very heavily resourced in order to provide data with the additional degree of precision.

Multiple comparisons

When using survey data it is also important to take into account the problem of multiple comparisons. If a large enough number of parameters are measured across time, then some changes will be observed, even if there are no real changes in the population. It is important to check for widespread or sustained change, or to look for evidence of change from other sources, and not to give too much weight to isolated changes. Any apparent changes in the HEPS data are considered in this context.

Self-reported behaviours

It is worth bearing in mind that the behavioural measures are self-reported, rather than observational. This is likely to mean some degree of under-reporting for behaviours such as alcohol consumption or over-reporting for behaviours such as physical activity or consumption of fruit and vegetables.

Classifications used

Age: In general, six age groups are used for analysis (16-24, 25-34, 35-44, 45-54, 55-64, 65-74). These are the standard groups used in presenting survey findings. However, in the absence of clear gradients, or in the case of small base sizes, results may be presented in terms of more aggregated age groups to clarify observed patterns of difference.

Social grade is used as a household-based proxy measure of social class. This classification is based on the normal occupation of the chief income earner in the household, which is categorised into AB (professional, managerial and technical), C1 (skilled non-manual), C2 (skilled manual), D (partly skilled and unskilled) and E (dependent on state and casual workers) (Market Research Society, 1991). The social grade of a retired person with a pension from their job is based on their previous normal occupation. The social grade of widows or widowers receiving a pension from their spouse's job is based on the previous normal occupation of the spouse. For those unemployed for two months or less, social grade is based on their previous occupation - the longer term unemployed are graded as E. The main advantage of this classification system is that it provides a relatively stable population profile over time and all respondents can be assigned a social grade, unlike occupation-based systems such as the Registrar General's Social Class based on Occupation which excludes the long-term unemployed, arguably one of the most materially and socially disadvantaged population groups.

Deprivation: DEPCAT is used as an area-based measure of deprivation and is based on the Carstairs scores derived from Census data. They are a composite measure of four variables: overcrowding, male unemployment, low social class and having no car. The Carstairs scores are used to define seven DEPCAT groups, from 1 (the most affluent) to 7 (the most deprived). Carstairs scores are updated periodically when more up to date Census data are available, or when there are changes to postcode boundaries. The division of the scores into DEPCAT groups was first done in 1981 on a pragmatic basis, using the first Carstairs scores. More recent DEPCAT groups have been achieved by dividing the population (according to the latest Carstairs scores) into seven new DEPCAT groups, each containing the same proportions of the population as those produced in 1981. The latest available DEPCAT scores were used for analysis in this report.

McLoone (2000) notes a number of weaknesses when using DEPCAT as an analysis variable. Since DEPCAT is defined at postcode sector level, an area needs to be relatively homogeneous (i.e. most residents affluent or most deprived) for it to be identified as deprived or affluent (high or low DEPCAT). Any area which is more heterogeneous will be defined as DEPCAT 3, 4 or 5 (groups covering 62% of the Scottish population). However, these postcodes contain deprived households, and it is possible that some of these 'middle' sectors contain more deprived households than those identified as DEPCAT 6 or 7. Furthermore, as postcode sectors are geographically larger in rural areas, this means that DEPCAT is probably a better descriptor of affluence in urban areas (which are likely to be more homogeneous).

Figure 1.1 shows the relationship between social grade and DEPCAT score using the HEPS data from 1996-1999. Whilst there is a clear gradient of social grade across DEPCAT scores, the chart clearly illustrates the lack of homogeneity of the different DEPCAT groups, even at the extremes. For example, when considering results for individuals classified as deprived when using DEPCAT (scores 6 and 7), a reasonable proportion will actually come from relatively affluent ABC1 households.

Fig 1.1: Relationship of social grade to DEPCAT using 1996-99 HEPS data

Base: all respondents 1996-1999

Motivation: Three mutually exclusive categories are used to classify respondents according to their motivation to change health-related behaviours. Those who:

• have tried to change in the past year
• want to change, but have not tried in the past year
• have neither tried nor want to change

Anyone who falls into either of the first two categories would be defined as 'motivated to change'.

 

Tables and figures

When using the tables and figures, the following points should be noted:

• percentages may not add up to 100 due to rounding, or the exclusion of don't know responses where they only represent a small proportion of answers
• percentages are used throughout the report, irrespective of base size - for each percentage given, the number of individuals constituting the base is given in Appendix C and should be taken into account when interpreting the findings
• the base for percentages consists of all respondents (including those for whom data are missing), unless explicitly stated, and the base size in tables is denoted by 'N'

1.3 Report structure

Section 2 gives an overview of recent health trends in Scotland, including comparison with other countries, in order to provide the context for the reported findings. Section 3 provides a discussion of respondents' self-perceived health to provide further context. Section 4 presents data on perceptions of causes of mortality and morbidity, as well as the possibility of reducing the risk of certain diseases. More detailed topic-specific findings on behaviour, knowledge and motivation to change are presented in Sections 5 to 12. Section 13 summarises the main fndings of the report and briefly discusses their implication for health promotion.

Main points

• This report presents data on time trends in health-related knowledge, attitudes, motivations and behaviours in Scotland over the period 1996-2003.
• While the aim of the analysis is to assess the degree of significant change in these indicators over time, the sample size and design mean that it is sometimes difficult to distinguish observed variations due to actual small changes from those due to random sampling error.
• The results from HEPS presented here are considered alongside other data sources, such as other long-running national surveys, in order to provide supporting evidence of observed trends.