Income and education predict elevated depressive symptoms in the general population: results from the Gutenberg health study

Socioeconomic status (SES) is a combined measure of a person’s social position regarding education, occupation and income. There is strong evidence for socioeconomic inequalities in physical and mental health [1]: Persons with lower SES are more likely to develop medical diseases (e.g. diabetes or coronary heart disease) and mental disorders (e.g. depression), are less likely to have access to specialized care and more prone to unfavourable outcome [2‐5]. These associations of SES and health have been demonstrated in different subgroups regarding age or culture. Different mediating factors such as psychosocial stress, negative emotions or adverse life events are supposed to contribute to socioeconomic health inequalities [6‐11]. Despite the vast amount of studies, the explanation of these associations is limited due to the heterogeneity of SES measures. Further, combined measures obscure the role of the single SES dimensions (e.g. education versus income versus occupation).

Major depression and subthreshold depressive symptoms impair quality of life, increase use of health services and expenditures, and reduce life expectancy [12, 13]. Therefore, studies are needed, which identify subgroups with an increased risk for depressive symptoms. Regarding the importance of SES for depression, Lorant and colleagues [4] summarized the results of 56 studies comprising n = 215,490 participants investigating single dimensions of SES and found a higher incidence and persistence of depressive symptoms in persons with lower scores of SES, especially for the dimensions education and income [4]. However, the studies included only investigated one dimension at a time, thus no conclusions about the importance of specific SES factors could be drawn. A Canadian study analyzed the influence of education and household income (amongst others) on major depressive episodes. They found that low education level was associated with a higher risk of major depressive episodes, but only in employed participants. In those who had not worked during the previous 12 months, low education was a “protective” factor which shows the importance of searching for interactions with explanatory worth. In both groups income was negatively associated with the risk of major depressive episodes [14]. However, this study included neither occupational position nor baseline depression as covariates. Similar results with assimilable limitations have been found in the Longitudinal Aging Study Amsterdam [15]. One major limitation of previous studies on the relationship of SES with depression is that they – to the best of our knowledge – did not include baseline depression as covariates or included all three recommended SES dimensions [16].

Hence, we want to determine a) the associations between the main SES factors and elevated depressive symptoms at baseline and after 2.5 years in a population-based sample in Mid-West-Germany and b) whether the contribution of SES is independent from common predictors of depression (e.g. baseline depression, medical history of depression or anxiety, sociodemographic variables, major medical diseases (MMD)) and c) whether there are any interactions of the SES factors with sex, partnership or MMD in persons from this area. The results will extend the knowledge about socioeconomic inequalities in depression.

The main findings are: 1) Persons without elevated depressive symptoms at baseline had a higher socioeconomic status (on all three indicators compared to persons with elevated depressive symptoms). 2) 50.1% of the group with elevated depressive symptoms at baseline had elevated depressive symptoms 2.5 years later versus 15.1% of the group without baseline depressive symptoms. 3) In the group without elevated depressive symptoms at baseline, SES and its dimensions income and education were strong and independent predictors of elevated depressive symptoms at the 2.5-year follow-up. In the group of persons with depressive symptoms at baseline, however, SES was not related with the occurrence of elevated depressive symptoms at the 2.5-year follow-up. 5) In the explorative analyses we found interactions of household net-income with major medical diseases as well as occupational position with living with a partner. The detrimental effect of MMD on the development of elevated depressive symptoms was buffered by high income and living in a partnership buffered the detrimental effect of a low occupational position.

Regarding the prediction of elevated depressive symptoms by SES, we found that particularly the dimensions educational qualification and household net-income were important predictors, which is in line with the results of the meta-analysis of Lorant and colleagues [4]. Important to note, the effects of these SES dimensions were as strong as or even stronger than the effects of major medical disease, partnership or sex in our sample (Additional file 3). In contrast to Lorant and colleagues [4], in our sample the influence of SES on the development of elevated depressive symptoms was negligible for persons who were already bothered by depressive symptoms at baseline. This differential effect of SES in the two groups might mean that once a person is depressed, the effect of SES for the further 2.5 year course is negligible in comparison with direct measures of mental health or disease burden. In persons without current elevated depressive symptoms, however, SES represents an important factor. Further reasons for our divergent findings, might be that we applied different baseline measures and included a wide range of relevant covariates (e.g., baseline depression and anxiety, MH of depression or anxiety, MMD, sociodemographic variables vs. only sociodemographic variables, prevalence of depression), that we analyzed the sample stratified by baseline depression, that we used slightly different predictor variables (SES indicators as dimensional variables vs. as binary variables) and that Lorant and colleagues [4] included studies from different countries with different health care systems.

Limitations of our study concern the lack of clinician-administered interviews of depression. However, it has been repeatedly demonstrated that even subthreshold depressive symptoms or elevated depressive symptoms according to the criterion PHQ-2 ≥ 2 impair quality of life, increase use of health services and expenditures, and reduce life expectancy [12, 13, 25], emphasizing the public health relevance of our findings.

Further, the study impresses through its sample size and its representativeness regarding the general population as well as through the standardized and therefore high comparable measurement of the SES indicators as Lampert and Kroll would recommend them [16]. Moreover, we were able to investigate the sample at two times of measurement and include baseline depressive symptoms as group variable and covariate as well as severe medical diseases. In the interpretation of the results, it should be considered that, due to the design, collider stratification may occur with regard to persons with already depressive symptoms at the first measurement point; PMID: 19689488. In this context it can also happen regression to the mean which means that those who start a cohort high are not going to stay high. These issues are due to stratification based on baseline symptoms [26, 27]. Due to the age range of our sample starting with 35 years, conclusions about the direction of the association between depression and SES indicators are limited. We assume that education and occupational position were largely determined at study entrance. Nevertheless, it can be assumed that the SES still varies over time in this age range, which could not be assessed in the present study. However, Gilman and colleagues found that childhood SES is associated with depression in adulthood [28]. Therefore, the results support the argument that causation as a mechanism is more important than selection. It can also be assumed that the effects cannot be applied equally to both sexes. This possible gender effect cannot be given sufficient attention in this paper; it is a research question in itself. In the present study, however, sex has been included as a control variable. It can also be assumed that not all ethnic groups are equally influenced by the SES indicators. This, too, is a research question in its own right, which is investigated in Beutel and colleagues [29], among others. Further, conclusions can be drawn with regard to the population group studied, since it can be assumed that countries differ in terms of the social determinants of health and well-being [30‐33].

In the present study, education and household net-income were independent predictors for the incidence of elevated depressive symptoms in persons from Mid-West-Germany who were free from elevated depressive symptoms at baseline. Occupational position seemed to have no independent longitudinal influence on depressive symptoms in this sample. Therefore, we conclude that education and household net-income might play a causative role for the elevated risk of persons with lower SES to develop depressive symptoms in persons from this area.