Scottish Health Survey - topic report: The Glasgow Effect

2. Mental and General Health

SUMMARY

Anxiety

• The factors which were found to be significantly associated with anxiety were: age, sex, residence, economic activity, potential problem drinking, abstaining from alcohol and physical activity level.
• Even after controlling for all of these factors, residents of Greater Glasgow and Clyde were almost twice as likely to have symptoms of moderate to high severity anxiety (92% increased risk when compared to the rest of Scotland).

GHQ

• The factors which were found to be significantly associated with a high GHQ score (indicating possible psychiatric disorder) were: age, sex, residence, receiving income-related benefits, economic activity, educational qualifications, marital status, smoking status, potential problem drinking, abstaining from alcohol and physical activity level.
• Residents of Greater Glasgow and Clyde had an increased risk of having a high GHQ score even after adjusting for all of these factors (19% increased risk compared to the rest of Scotland).

WEMWBS

• Although residence in Greater Glasgow and Clyde was associated with greater odds of having a low WEMWBS score (indicating lower levels of mental wellbeing) when age and sex were adjusted for, subsequent adjustment for SIMD accounted for all of the difference.

Depression

• Although residence in Greater Glasgow and Clyde was associated with increased risk of depression when only age and sex were adjusted for, when age, sex and socio-economic variables, specifically NS-SEC, economic activity, equivalised income and marital status, were included in the model, the excess risk associated with residence in Greater Glasgow and Clyde was removed.

Self-assessed health

• Despite residents of Greater Glasgow and Clyde having higher odds of poor self-assessed health when adjusting for age and sex, the so-called 'Glasgow Effect' was fully explained when socio-economic variables were adjusted for, specifically SIMD, receiving income-related benefits, economic activity, household tenure, equivalised income, educational qualifications and NS-SEC.

2.1 Introduction

The outcomes covered in this chapter are anxiety, psychological ill health, mental wellbeing, depression and self-assessed general health.

Anxiety

Participants were classified as suffering from anxiety if they had a score of 2 or more on the anxiety scale of the Revised Clinical Interview Schedule, indicative of symptoms of moderate to high severity. Anxiety was only measured in the nurse sample.

General Health Questionnaire ( GHQ-12)

The General Health Questionnaire ( GHQ-12) is a widely used standard measure of mental distress and psychological ill-health, consisting of 12 questions on concentration abilities, sleeping patterns, self-esteem, stress, despair, depression, and confidence in the previous few weeks. As the GHQ-12 measures deviations from people's usual functioning it cannot be used to detect chronic conditions. However it allows comparisons between groups to be investigated. Responses to the GHQ-12 items were scored, resulting in an overall score between zero and twelve. A score of four or more indicates the presence of a possible psychiatric disorder.

Warwick-Edinburgh Mental Wellbeing Scale ( WEMWBS)

WEMWBS is an indicator of mental wellbeing, which comprises 14 positively worded statements with a five item scale ranging from '1 - None of the time' to '5 - All of the time'. The scores therefore range from 14 to 70. A participant was classified as having a low WEMWBS score if it was more than one standard deviation below the mean.

Depression

Participants were classified as suffering from depression if they had a score of 2 or more on the depression scale of the Revised Clinical Interview Schedule, indicative of symptoms of moderate to high severity. Depression was only measured in the nurse sample.

Self-assessed health

Self-assessed health was investigated via a question asking participants to rate their health in general as 'very good', 'good', 'fair', 'bad' or 'very bad'. Those who answered 'bad' or 'very bad' were classified as having poor self-assessed health. It is important to note that the responses will be affected by individual perceptions, as with other self-assessed measures.

2.2 Mental Health

2.2.1 Anxiety

9% of adults had a score of 2 or more on the anxiety scale of the Revised Clinical Interview Schedule. There were significant differences by gender, with women more likely to be display symptoms of anxiety than men (10% vs. 7%). There was a significant difference by age, with increasing prevalence of anxiety from age 16 - 24 to a maximum at 45 - 54, with the pattern less clear for older ages. The prevalence of anxiety significantly increased with increasing deprivation, with the prevalence more than twice as high in the most deprived SIMD quintile compared to the least deprived (13% vs. 6%). There was also a significant difference in prevalence of anxiety between residents of Greater Glasgow and Clyde and the rest of Scotland, with around twice the prevalence (14% vs. 7%).

The model development process can be found in Appendix 2, along with McFadden's pseudo R ²s for the different models. In view of these, the best fitting model was chosen, and the results for that model using the full data available are described here.

The factors which were found to be significantly associated with anxiety were: age, sex, residence, economic activity, potential problem drinking, abstaining from alcohol and physical activity level. As anxiety was only measured in participants in the nurse sample, the sample size was 1,887.

Residents of Greater Glasgow and Clyde were almost twice as likely to have moderate to severe anxiety symptoms when compared to the rest of Scotland (odds ratio of 1.92) even after adjusting for the other variables in the model. Women had significantly higher odds of having anxiety than men (odds ratio of 1.70). Age was associated with anxiety but there was no clear trend; those aged 35 - 44 had the highest odds compared to 16 - 24 year olds (odds ratio of 2.61).

Economic activity was also associated with anxiety, with those who were looking for or intending to look for work having significantly higher odds of anxiety, compared to those in paid employment, self-employed or in government training (odds ratio of 4.06). Similarly, those who were permanently unable to work had higher odds of anxiety (odds ratio of 2.68). People who were identified as potential problem drinkers according to the CAGE questionnaire had significantly increased odds of anxiety (odds ratio of 2.88). Those who abstained from alcohol also had significantly higher odds of anxiety (odds ratio of 1.67) compared to people who drank alcohol. Whilst these findings may seem contradictory, it should be noted that a proportion of those who abstain from alcohol will do so following medical advice to stop drinking.

Physical activity levels were associated with anxiety. Those whose physical activity levels were high or medium had significantly lower odds of anxiety when compared to those with low levels of physical activity, with odds ratios of 0.63 and 0.55 respectively.

Table 1

2.2.2 General Health Questionnaire ( GHQ-12)

15% of adults had a high GHQ-12 score (indicating possible psychiatric disorder), however there were significant differences within the population. Women had a significantly higher prevalence (17% compared to 12% of men). There was also significant variation by age, although the pattern was not linear, with the highest prevalence found among 45 - 54 year olds (17%), compared to 65 - 74 year olds, who had the lowest prevalence of 10%. Prevalence increased with deprivation, with prevalence in the least deprived quintile (10%) half of that found in the most deprived quintile (21%). A significant difference also existed between residents of Greater Glasgow and Clyde (18%) and the rest of Scotland (14%).

The model development process can be found in Appendix 3, along with McFadden's pseudo R ²s for the different models. In view of these, the best fitting model was chosen, and the results of that model using the full data available are presented here. It should be noted that the addition of biological variables to the model removed the effect of residence, however this model did not fit the data as well as the model described here. The results of the best fitting model were accepted over those found in other models.

The factors which were found to be significantly associated with a high GHQ score were: age, sex, residence, receiving income-related benefits, economic activity, educational qualifications, marital status, smoking status, potential problem drinking, abstaining from alcohol and physical activity level. The weighted sample size with complete data was 11,088.

Residents of Greater Glasgow and Clyde had an increased risk of having a high GHQ score when compared to the rest of Scotland after adjusting for all the other variables in the model (odds ratio of 1.19). Women had higher odds than men (odds ratio 1.59), and there was a significant association with age, with those aged 45 and above having lower odds than the youngest age group (16-24).

Moderate and heavy smokers had an increased risk of a high GHQ score when compared to those who had never smoked or who were ex-occasional smokers (odds ratios of 1.41 and 1.47 respectively). People in receipt of income-related benefits had significantly higher odds of a high GHQ score (odds ratio of 1.35). Potential problem drinkers and abstainers both had increased risks, with the odds for potential problem drinkers more than twice as high as those who were not (odds ratio of 2.09), and abstainers were two-thirds more likely to have a high GHQ score (odds ratio of 1.67) compared to those who drank alcohol.

People who were married/in a civil partnership and were living together were the least likely to have a high GHQ score. Those who were separated had the highest odds (odds ratio 1.81), followed by those who were widowed (odds ratio of 1.49). Level of educational qualifications was significantly associated with high GHQ scores but there was no clear pattern. Those who had high or medium levels of physical activity had lower odds of having a high GHQ score than those with low levels of physical activity (odds ratios of 0.67 for both categories).

Economic activity was significantly associated with high GHQ scores, with those who were looking for or intending to look for work and those who were permanently unable to work having the greatest risk of a high GHQ score when compared to those in paid employment, self-employed or government training (odds ratios of 3.36 and 2.91 respectively).

Table 2

2.2.3 WEMWBS

A similar pattern was found for having a low WEMWBS score as for having a high GHQ score. Overall 15% of adults in Scotland had a low WEMWBS score (defined as one standard deviation or more below the mean score), however there were significant differences within the population. Women had a significantly higher prevalence of low WEMWBS scores (16% compared to 14% of men). There was significant variation by age, although the pattern was not linear, with the highest prevalence found among those aged 75 and over (18%), whereas 65-74 year olds had the lowest prevalence (12%), followed by 25-34 year olds (13%). Prevalence increased with deprivation, with the prevalence in the least deprived quintile (8%) just over a third of that found in the most deprived quintile (23%). A significant difference in prevalence also existed between residents of Greater Glasgow and Clyde (17%) and the rest of Scotland (14%).

In the initial logistic regression model containing only age, sex and residence, residence in Greater Glasgow and Clyde had an odds ratio of 1.25, meaning that participants who resided in Greater Glasgow and Clyde had 25% increased odds of having a low WEMWBS score, compared to the rest of the Scotland. However once the model was also adjusted for SIMD the odds ratio dropped to 1.10, which was not significant at the 5% level. SIMD was highly significant, with increasing odds of poor mental wellbeing with increasing deprivation. This indicates that the different distribution of SIMD in Greater Glasgow and Clyde compared to the rest of Scotland explains the difference in prevalence of low WEMWBS scores.

2.2.4 Depression

8% of adults had two or more symptoms of depression on the Revised Clinical Interview Schedule, with significantly higher rates in women (10%) than men (7%). There was also significant difference by age, although there was no clear linear pattern. Prevalence of depression increased from age 16 - 24 (4%) to 45 - 54 (11%), but then decreased before increasing again for those age 75 and over. Residence was also significantly associated with depression, with a much higher prevalence among residents of Greater Glasgow and Clyde (13%) than the rest of Scotland (7%). The prevalence of depression increased with increasing deprivation, from 5% among those in the least deprived SIMD quintile to 13% in the most deprived SIMD quintile.

In the initial logistic regression model containing only age, sex and residence, residence in Greater Glasgow and Clyde had an odds ratio of 1.80 indicating that residents of Greater Glasgow and Clyde's odds of having moderate to severe depression were 80% higher than the rest of Scotland, after adjusting for age and sex. When SIMD was added to the model the odds ratio decreased to 1.66, showing that a small amount of the "Glasgow Effect" has been accounted for by deprivation. When the group of socio-economic variables were added to the model and backward selection performed, residence was no longer significant at the 5% level. The variables which remained in the model (with p<0.05) were age, sex, NS-SEC, economic activity, equivalised income and marital status. When residence dropped from the model only these variables remained in the model, showing that these variables fully explained the effect previously observed from residing in Greater Glasgow and Clyde.

2.3 General Health

2.3.1 Self-assessed health

7% of adults rated their health as 'bad' or 'very bad', with no significant difference between men and women. The prevalence increased with increasing age, from 1% of 16-24 year olds to 14% of those aged 75 and over. There was a large significant difference in prevalence by SIMD, with prevalence decreasing from 14% in the most deprived quintile to 3% in the least deprived quintile. Prevalence of poor self-assessed health was significantly higher for those living in Greater Glasgow and Clyde than for the rest of Scotland (9% vs. 4%).

In the initial logistic regression model containing age, sex and residence, residents of Greater Glasgow and Clyde had an odds ratio of 1.56 of having poor self-assessed health, meaning they were 56% more likely than those in the rest of Scotland to report poor health. When SIMD was added to the model all four variables were significant predictors of having poor self-assessed health, with the odds ratio for Greater Glasgow and Clyde reduced to 1.20, indicating that more than half of the "Glasgow Effect" in relation to poor self-assessed health was explained by SIMD. However there was still a 20% increase in odds of having poor self-assessed health for residents of Greater Glasgow and Clyde compared to the rest of Scotland. When socio-economic variables were added to the model and backward selection performed, residence was no longer significant at the 5% level. The variables which remained in the model (with p<0.05) were age, sex, SIMD, receiving income-related benefits, economic activity, household tenure, equivalised income, and educational qualifications. When residence dropped from the model NS-SEC was also in the model. This finding suggests that these socio-economic variables, alongside SIMD, explain the apparent difference in rates of poor self-assessed health between Greater Glasgow and Clyde and the rest of Scotland.

2.4 Assessing the impact of the socio-economic variables individually

When examined individually, the socio-economic variable which provided the best-fit model for the mental and general health outcomes was economic activity as it had the highest McFadden's pseudo R ², indicating that this measure was the best predictor of the general and mental health outcomes after adjusting for the rest of the variables in the final models. For anxiety and GHQ it was also possible to see which socio-economic variable best explained the difference between Greater Glasgow and Clyde and the rest of Scotland; equivalised income explained the most difference for anxiety, and economic activity for GHQ.

2.5 Conclusions and Discussion

Despite adjusting for age and sex, residents of Greater Glasgow and Clyde had increased odds of having bad or very bad self-assessed health compared to the rest of Scotland. Further adjusting for SIMD partly attenuated this increased risk and adjusting for a wider range of socio-economic variables fully attenuated the increased risk, removing the so called "Glasgow Effect". Previous analyses ¹⁰ investigating Greater Glasgow using the 1995, 1998 and 2003 Scottish Health Surveys, found that adjusting for socio-economic factors did not fully attenuate the risk. This difference in findings may represent a change over time; may be due to the different geographical areas investigated (due to the change in health board structure); may be due to the inclusion of a wider range of socio-economic variables in the current study, or the more spatially specific area-level deprivation variable used.

There was a degree of variability between mental health outcomes in relation to whether residence in Greater Glasgow and Clyde significantly affected risk. A higher prevalence of moderate to severe depression was found among residents of Greater Glasgow and Clyde, with increased odds after adjusting for age and sex, which were reduced slightly by additionally adjusting for SIMD. However the effect was completely removed by further adjusting for a variety of socio-economic variables.

A significant difference in prevalence was found for low WEMWBS scores, with a higher prevalence in Greater Glasgow and Clyde. There was an increased risk of having a low WEMWBS score after adjusting for age and sex, but once SIMD was adjusted for, the effect was removed. However, when looking at high GHQ scores, adjusting for SIMD did not remove the higher odds found among residents of Greater Glasgow and Clyde, nor did adjusting for socio-economic or behavioural variables. Adjusting for biological variables did remove the 'Glasgow Effect', but this model did not fit the data as well as a previous model, so the results were not accepted. These analyses include a much broader set of explanatory variables than Gray's analyses ¹⁰. However, the results are comparable to the extent that Gray also found that residence in Greater Glasgow was associated with significantly higher levels of possible psychiatric disorder, with the increased risk attenuated but not removed by adjusting for socio-economic factors.

GHQ-12 and WEMWBS have been compared and been found to measure different things, with participants who have the same GHQ score having a variety of scores on the WEMWBS scale ¹⁷. Confirmatory factor analysis of WEMWBS supports the hypothesis that the scale measures a single construct of wellbeing ¹⁸. Conversely, GHQ-12 has often been found to contain two or three dimensions, with the best fitting model finding three dimensions; 'Anxiety', 'Social dysfunction' and 'Loss of confidence' ¹⁹, which may explain the difference in results between the two outcomes.

The increased odds of moderate to severe anxiety in Greater Glasgow and Clyde remained significant after adjusting for socio-economic, behavioural, biological, relationship and social mobility variables. The odds ratio for residence in Greater Glasgow and Clyde when only age and sex were adjusted for was 2.92, meaning that residents were almost 3 times as likely to have anxiety than people in the rest of Scotland. The inclusion of additional variables in the model gradually attenuated this risk to the extent that once all variables were adjusted for, the odds ratio for Greater Glasgow and Clyde had reduced to 1.92. This indicates that even after taking account of all of the factors noted above, people living in Greater Glasgow and Clyde still had a 92% higher risk of anxiety than those living elsewhere.

It is important to note that the variables may not be causative, but merely predict individuals who are more likely to suffer from anxiety. People identified as potential problem drinkers by the CAGE questionnaire have 2.88 times the odds of anxiety of those who were not identified as potential problem drinkers, and abstainers had odds 1.67 times those who do drink alcohol. It may be that the anxiety causes the potentially problem drinking rather than drinking causing the anxiety, and this study is not aiming to investigate the direction of the association. Using an outcome measuring depression and anxiety, a previous study found that abstaining from alcohol was associated with common mental disorder symptoms only among previous consumers, but not among lifelong abstainers ²⁰. Unfortunately the two groups cannot be separated in this study. People who have medium levels of physical activity have close to half the odds of anxiety of those who have low levels of physical activity. Again, it is not within the scope of this report is to examine the direction of this relationship; however, many studies have found that exercise reduces levels of anxiety ²¹.