Is there a "Scottish effect" for self reports of health? Individual level analysis of the 2001 UK census

The sample

The 2001 Samples of Anonymised Records (SARs) are a collection of random samples – created from individual and household level census data – made available to the research community to allow micro level analysis of the UK census. The samples are all anonymised and procedures are in place to ensure that individuals cannot be identified. Here the 2001 individual licensed SAR is used. This is a 3% sample of the 2001 UK census incorporating 1,843,525 individuals. Details of the individual SAR's methodology have been published elsewhere [17]. While the statistical office of each UK country is responsible for conducting their own census there is much co-ordination, making the individual SAR an attractive dataset to compare the health of individuals in Scotland and England.

Variables

Analysis

All analysis was conducted in Stata version 9 [24]. Initial descriptive analysis compared the distribution of the variables by country of residence and birth. Chi-square tests were used to assess associations.

Next, logistic regression models were used to further assess the association of country of residence and birth with limiting illness and general health. Models were fitted in two stages. The initial model was adjusted for age and sex only. At the next stage I additionally adjusted for SEP (using the three separate measures of SEP) and employment status.

To explore whether the Scottish excess in not good health and limiting illness was seen across all employment groups, models were then stratified by employment group. These models were initially age and sex adjusted and then additionally adjusted for the three measures of SEP. As men and women have very different rates and patterns of economic inactivity (14.5% of men, 30.4% of women) and employment (81.3% of men and 67.2% of women) the analysis for these employment groups was further stratified by sex to explore whether any Scottish excess in these employment groups was seen for both men and women.

Main findings

Working age people born and living in Scotland were, on average, more likely to report limiting illness and not good general health in comparison to people born and living in England. Adjustment for differences in SEP and employment status largely explained this association. This suggests that there is unlikely to be a universal and unidentified "Scottish effect" for self reported general health and limiting illness. Controlling for the impact of SEP or employment using either a single measure or measures from one time point in the lifecourse is unlikely to fully capture differences between populations [9, 25]. In further analysis (available from the author) adjusting separately for only one of the measures of SEP, or just employment status, attenuated the relationship but not to the degree shown in table 2 when controlling for all these measures. It is likely that including further measures of SEP and employment status from across the lifecourse would have attenuated the relationship even further.

The stratified analysis supports the notion that SEP and employment status differences account for Scotland's excess in poor general health and limiting illness. Scotland's excess was only consistently seen amongst the economically inactive born and living in Scotland and this relationship was partly attenuated after controlling for the measures of SEP. People born and living in Scotland were most likely to be economically inactive and, amongst this employment group, most likely to classify themselves as permanently sick or not to identify with one of the census economic inactivity categories and (for men) least likely to be retired.

Why might people born and living in Scotland be more likely to be economically inactive? One possible reason is the lasting impact of industrial decline and labour market restructuring which has been particularly dramatic in Scotland since the 1980s [13]. In this period in Britain, not only did rates of employment for the manual social classes decline the most, the impact was felt greatest among those with a limiting illness [20]. So "a man has to be 'healthier' to remain employed in a manual rather than in a managerial, professional or clerical occupation" [20]. Moreover, there is strong evidence that unemployment figures, whether based on claimant counts or more widely accepted international definitions, seriously underestimate real levels of unemployment in old industrial areas because many (ill) long term unemployed with little prospect of finding work (and so regarding themselves as economically inactive) in these labour markets have been diverted from unemployment to sickness benefit [26]. Rates of sickness benefit have continued to rise for both men and women while unemployment has been on a overall downward trend in Britain [26]. In age and sex adjusted models (not shown) it was the economically inactive, rather than the unemployed, who had the highest odds of a limiting illness or not good general health in comparison to those in employment. This was true even after excluding people who described themselves as permanently sick or disabled.

Of course, the diversion from unemployment to sickness benefit has occurred in former industrial regions across Britain. However, a recent review of sickness benefit in Scotland shows that at the country level at least, Scotland has a higher percentage of the working age population claiming sickness benefit than England [27]. This review concludes that "Scotland's enormous number of incapacity claimants should really be interpreted as the legacy of twenty years of de-industrialisation and job destruction" [27].

In contrast to those who were economically inactive, people born and living in Scotland who were in employment seemed to be at slightly lower risk of poor health than people born and living in England. When accounting for SEP differences between the employed populations this association was actually strengthened slightly.

As discussed in the next section, there is the possibility that these results reflect national perceptions of poor health rather than actual differences in morbidity.

Comparison with other studies

Ecological analysis of census results amongst those of working age showed that people living in Scotland overall reported worse general health than residents of England but compared to some regions of England and to the whole of Wales, Scotland's general health was actually better [7]. In the present study the analysis was extended by using individual level census data and additionally accounting for country of birth to compare the health of residents of Scotland and England. However, comparing self reports of health across countries is not unproblematic given possible variations in cultures and health expectations [28]. While self reported general health is a valid measure of health status, [29] and is predictive of future mortality risk independent of other risk factors, [30] it seems that the relationship between mortality and self reports of health varies in the countries of the UK, with people living in Scotland, in comparison to residents of England, self reporting better general health and less limiting illness than their mortality profile suggests they should [19, 28].

Why this is the case is not clear. Given that Scotland's premature mortality rate is higher perhaps people living in Scotland have lower health expectations or perhaps cultural differences mean that they are more stoical. Recent work did not find any evidence that people in lower socioeconomic groups were more stoical when self reporting their health than people in higher groups, however the research was limited to Scotland [31]. Similar work comparing countries is required to help resolve the ongoing debate about the comparability of self reports of health across countries. In the meantime, the potential remains for self reports of health from Scotland to underplay the extent of any "Scottish effect" on morbidity.

Of course, there are major SEP differences in morbidity within countries and regions of the UK. In Scotland, social class differences in self reported health between the highest and lowest were the most acute in Britain for both men and women [7]. As Leon et al. argue there is a tendency, however, when discussing the possibility of a "Scottish effect" to conflate what is a between country issue with discussions of within country variations in inequalities in health [2].

The "Scottish effect" identified in relation to mortality remains unexplained. It would be fruitful to replicate the analysis here using nationally representative individual level data on mortality. The forthcoming (as of May 2006) census based Scottish Longitudinal Study, which links individual census data with mortality records, will allow direct comparison with the established ONS longitudinal study of English and Welsh censuses.

In ecological analysis of mortality data, an excess in Scotland existed in all deprivation categories from the most affluent to the most deprived, with differences in deprivation prevalence explaining less of the Scottish excess over time [1, 4]. However, the possibility remains that established risk factors may explain the "Scottish effect". Scotland's mortality excess compared to England and Wales was greatest in the most deprived areas suggesting that if there is a "Scottish effect" influencing Scotland's mortality excess, it does not apply evenly across Scotland's population [2]. Moreover, the Carstairs score fails to explain all the current variation in mortality between English regions [2]. There is also evidence that another deprivation index largely explains regional mortality variation in England [32].

Future analysis should also explore cause specific mortality as well as all cause mortality as Scotland's comparative position in Western Europe varies by disease type [2]. For example, Mitchell et al. explored ischemic heart disease prevalence as Scotland's record is particularly poor for this disease [5].

Strengths and limitations

To the best of the author's knowledge, this is the first study to compare rates of self reported general health and limiting illness in Scotland and England using individual, rather than ecological, census data. The individual SAR provided a large random sample of a whole population survey. Importantly, it allowed not just the exploration of between country of residence differences but also allowed account to be taken of country of birth.

Of course, using country of birth was a crude way to identify inter-country migration (for example, no information was available on when a person moved or why). The census does record the address of individuals who moved in the last year, however the level of inter country migration in one year was, relatively, small. As this study was cross-sectional it was not possible to explore issues of causality in this sample.