Based on our results, we can conclude that psychological distress partly accounted for employment status and household income level differences in unnatural mortality (suicide, accidents and violence, and alcohol-related mortality) in both genders, and for educational level differences in unnatural mortality among men; among women no significant educational differences were found in unnatural mortality in the first place. The contribution of psychological distress variables to socio-economic differences in CHD mortality, on the other hand, was negligible.
The strength of our study is the nationally representative data from repeated population surveys, which was supplemented with extensive socio-economic register data and national causes of death register data, providing for a prospective study design with a 28-year follow-up. However, the cross-sectional measure of socio-economic factors and psychological distress variables allows for no conclusions about the direction of the association, that is, health selection versus causation, which may both contribute to the associations between socio-economic position and psychological factors .
The response rate of the survey is similar to that of other population surveys . However, in our non-respondent analysis of this data  we found lower response rates for the lower educated. Total and cause specific (for example, alcohol, external causes, suicide) excess mortality rates were higher among survey non-respondents and this is partly explained by educational and income differences between respondents and non-respondents . These results indicate that non-respondents have more severe illnesses, mental health problems and depression as well as unhealthy lifestyles, such as smoking and alcohol use. They also indicate that the comparability of the results of the different socio-economic groups may be biased and, therefore, the socio-economic differences may actually be stronger than those observed in this data. Additional analyses for respondents with missing data on psychological distress variables (N = 1129, 1,6%), although containing relatively small number, showed that those with missing data on psychological distress measures were also more likely to be in the lower SES groups.
One principal limitation of the study is that the measures of psychological distress are very simple self-reported single-item questions. These measures may cover a variety of transient or chronic psychological symptoms, a wide range of meanings from the temporary decrease of psychological well-being to deeply impaired, even life-threatening disorders. Therefore, the main focus of these indicators is not to detect clinical disorders but to reflect the subjective experience of mental health, and to study mental well-being at an extensive population level . Nevertheless, single-item psychological distress variables demonstrated significant associations with cause-specific mortality, indicating that self-reported psychological distress have an implication for health. Another limitation concerning measures used in this study is the unemployed versus employed classification, which is a crude measure of employment status.
In the previous studies psychological factors only weakly or moderately mediate the relationship between SES and all-cause mortality [40, 41]. In this study, we analysed three different measures of psychological distress and found some mediation for unnatural mortality and SES, and weak mediation for CHD mortality by employment status. It has been proposed that the excess CHD mortality among those in a lower socio-economic position is dependent on socio-economic differences in behavioural and biological risk factors, such as smoking, blood pressure and serum cholesterol levels . A previous study based on the same data examined health behaviours as explanations for educational level differences in CHD mortality . Health behaviours, most importantly smoking, physical activity and vegetable intake, explained about 50% of the educational differences in CHD mortality among men, but did not explain much of the differences among women. Compared to these results, psychological factors examined in the present study did not add to the contribution made by behavioural factors in explaining socio-economic differences in CHD mortality. However, psychological distress explaining some of the inequalities in suicide, accidents and violence, and alcohol-related mortality indicates that in these specific causes of death, poor mental health is related to more severe consequences in the lower socio-economic status groups than in the higher SES groups. It is possibly due to poor coping strategies of psychological distress in the lower SES. Obviously, that includes risky behaviour and, above all, heavy alcohol consumption which may be aimed at relieving psychological symptoms.
Theories and models which propose psychosocial factors as mediators in the SES-health relationship also emphasize that health status is the result of complex causes. Health behaviour, socio-demographic factors and early environmental, genetic, biomedical and medical factors are all seen as related to this phenomenon. Our results suggest that psychological distress may explain some of the cause-specific mortality disparities between socio-economic groups.