Social Psychiatry and Psychiatric Epidemiology

, Volume 48, Issue 9, pp 1491–1501

Socioeconomic status and the risk of depression among UK higher education students

Authors

    • Department of Public Health and Community Medicine, Faculty of MedicineCommunity Health School, Assiut University
    • Division of EpidemiologyCommunity Health Sciences School, D Floor, West Block Queens Medical Centre, University of Nottingham
  • Shona J. Kelly
    • Centre for Health and Social Care Research, Sheffield Hallam University
  • Cris Glazebrook
    • Institute of Mental Health, University of Nottingham Innovation Park
Original Paper

DOI: 10.1007/s00127-013-0663-5

Cite this article as:
Ibrahim, A.K., Kelly, S.J. & Glazebrook, C. Soc Psychiatry Psychiatr Epidemiol (2013) 48: 1491. doi:10.1007/s00127-013-0663-5

Abstract

Background

Many university students experience some symptoms of depression during the course of their studies but there is evidence that students from less advantaged backgrounds may be more vulnerable.

Methods

The study was a cross-sectional online survey of 923 undergraduate students attending 6 UK Universities in the academic year 2009–2010 who completed a modified version of the Zagazig Depression Scale (ZDS).

Results

Overall, 58.1 % of female and 59.9 % of male study participants screened positive for depression (ZDS score >10). In the fully adjusted model, final year students (OR = 1.8) who lived in a more deprived area (OR = 2.3) were more likely to report higher rates of depressive symptoms. Additionally, students with high perceived control (OR = 1.6) whose mothers were highly educated (OR = 0.5) and from a family of a high affluence (OR = 0.3) were less likely to suffer from higher rates of depressive symptoms. The relationship between lower social economic status and depression was partly mediated by low sense of control.

Conclusion

Students from less advantaged backgrounds are more at risk of depression but a strong sense of control over one’s life may be protective.

Application

Since depression has strong impact on students’ learning and quality of life universities should consider confidential screening for mental health problems and provide additional support for students.

Keywords

Socio-economic statusDepressive symptomsUniversity students

Introduction

Socioeconomic status has been found to play an important role in depression [1, 2]. A recent meta-analysis of population-based surveys confirmed that the more deprived the individual the more likely he or she was to be depressed [OR = 1.81, 95 % confidence interval (CI), 1.49–1.89] and a dose–response association of increasing depression with decreasing social position was found for education and income but not occupation [3]. University students tend to come from more advantaged social backgrounds and have demonstrated a degree of resilience by their ability to meet the, often stringent, educational requirements for university entrance. Few studies have explored the impact of socioeconomic status specifically in student populations. Steptoe et al. [4] in a large multi-national study found that income inequality, family wealth, and parental education, contributed to depressive symptoms.

Reanalysis of data from a representative sample of Egyptian undergraduate students found that a range of SES indicators (parental occupation, family income, and number of persons per room) increased the risk of depression as measured by the Zagazig Depression Scale (ZDS) [5]. This raises the question of whether SES variables are particularly influential in developing countries where the economic gradient is likely to be steeper. The study also demonstrated the importance of using a multi-factorial approach to assess SES [5]. One mechanism by which higher socio-economic status may give protection against the stresses associated with the transition to college life is through the sense of control that a more affluent or educated background confers [6, 7]. It is well established that sense of control decreases with decreasing social position and the relationship between subjective personal control and health across different social classes was evident in a US study using data from three national probability samples [8]. It was found that control beliefs were significantly associated with both education and income in relation to health. Sense of control has been associated with more positive health outcomes for all social classes and lower social class individuals with strong control beliefs reported health outcomes similar to those in higher social groups for many self-reported health problems including depressive symptoms. It seems, therefore, that sense of control may mediate the relationship between SES and depression in adult populations [8]. It was further found that the strength of the depression–SES relationship was reduced when perceived control was accounted for in a university-based multi-national study in both developed and developing countries [4].

Aims and hypotheses

Study aim

This study aims to examine the impact of socioeconomic status on rates of depressive symptoms in UK undergraduate university students and to explore whether control beliefs mediate the relationship between SES and depression.

Study hypotheses

Higher socio-economic status will be associated with lower levels of depressive symptoms. The strength of the association between higher SES and lower levels of depressive symptoms will decrease as sense of control increases.

Methodology

Study design

The study was a cross-sectional online survey.

Sample size calculation

A power calculation estimated that a sample of 489 students was needed in order to detect an effect size of 0.3 in the regression of SES factors on the ZDS scores, with a p value <0.05 and 90 % power [9].

Sampling

Convenience samples of undergraduate recruited students from six UK universities were selected to reflect a range of socio-demographic profiles: University of Nottingham, Sheffield Hallam University, University of Derby, Nottingham Trent University, Newcastle University, and University of Sunderland. Inclusion criteria included being an undergraduate student, a home (UK) student and registered at one of the six universities.

Measures

Online questionnaire

  1. (A)

    Personal data questions which included: age, sex, faculty, year of study, student category (home, EU or international), and year of undergraduate study.

     
  2. (B)
    Socio-economic status of students The social class indicators were:
    • Postcode of the family house as an indicator for area-level social status via the Index of Multiple Deprivation (IMD) IMD is a measure of multiple deprivations based on data held about postcode areas, and consists of seven indices: income deprivation; employment deprivation; health deprivation and disability; education; skills and training deprivation; barriers to housing and services; living environment deprivation; and crime [10].

    • Mother’s and father’s education level Levels of education for each parent were selected from six ordinal categories according to Braun and Muller [11], which were collapsed into two categories (higher education and no-higher education).

    • Mother and father occupational prestige Participants selected a classification for their mother’s and father’s occupation using the National Statistics Socioeconomic Classification (NS-SEC) [12]. For analysis, these categories were condensed into three categories (never worked/unemployed, low/intermediate, and managerial/professional occupations).

    • Family Affluence Scale (FAS) The scale consists of four questions that measure the family material living standards. It was developed for the WHO HBCS study [13]. The range of possible scores is 0–9 and scores are recoded into low (0–2), middle (3–5), and high affluence (6–9) [13].

     
  3. (C)

    Sense of control was measured using six items adapted from the surveys conducted by the MacArthur Foundation Network [8]. Each item (e.g. “Other people determine most of what I can and cannot do”) is rated on a five-point scale (1 = strongly disagree, 5 = strongly agree). Two of the items are reversed scored (“I can do just about anything I really set my mind to” and “When I really want to do something, I usually find a way to succeed at it”). A mean score is calculated by dividing the total score by the number of items with possible scores ranging from 1 to 5. Lower scores indicate greater control [8].

     
  4. (D)

    Depressive symptoms The modified short version of the ZDS was used, which consists of 43 items representing 11 domains. The reliability of the modified ZDS was excellent (α = 0.898) [14]. We used a cut-off of 10 to indicate possible depression as proposed by the ZDS developers [15]. It was found that the ZDS as a screening tool had 76 % sensitivity and 96 % specificity when tested against the validated cut-off of 5 for the PHQ-9 for a diagnosis of possible depression in a student population [14].

     

Procedure

Administrators and Heads of Schools in targeted universities were sent details about the study and a copy of the invitation email for their students. The student’s invitation e-mail contained full information about the study, i.e., the aim of the study (to look at factors influencing psychological symptoms in university students), data about the researchers with their affiliations and contact details, why they were approached, how the study would be carried out, what are the benefits of participation, and their right to refuse to participate. The survey extended over three university semesters starting in March 2009 and ending in December 2010. The university contacts were asked to circulate the invitation email, which contained a link to the survey, to undergraduate students in their school. At the University of Nottingham medical students were also invited to participate via the Medical School’s virtual learning environment, the networked learning environment. Multiple reminders were sent to school administrators to boost the response rate. The survey took on average 20 min to complete and participants were given the option of participating in a prize draw. On completion of the survey, participants were given advice about seeking help from the relevant university counseling service (including contact details) or their general practitioner if they felt they needed to. The survey was tailored for each university as the contact details for help (e.g., counseling service) varied. The demographic questions, control scale, and revised 43-item ZDS, were administered as an online survey via a professional online survey tool (Survey Monkey®, Funchal, Madeira, Portugal). The data were downloaded directly as a flat excel file, which was opened in STATA for analysis. Data on family house postcode were sent to the East Midlands Public Health Observatory (EMPHO) to be converted into IMD scores, IMD ranks and IMD quintiles to provide area-level SES/deprivation information.

Statistical analysis

Data were analyzed using STATA version 10.1 software [16]. Frequency tables were examined to explore missing data, errors in the data (outliers and incorrect data entry), and data consistency. Missing data in the main outcome variable (ZDS) were found to be <1 % and missing at random (t = 1.45, p = 0.13). We used the series median value method for missing value replacement. It has been proposed that for purposes of analysis replacing missing values can reduce bias and often is used for this purpose if data are missing at random [17].

The outcome (dependant) variable was ZDS scores. Basic univariate analyses (χ2 and t test) were conducted to test the associations between ZDS and the exposure variables prior to multivariate analyses. Age and sex were added as a priori variables and the explanatory (independent) variables were SES measures (IMD, FAS, parental occupation, parental education), sense of control, university, faculty, and year of study.

An initial multivariate logistic regression model was built containing a priori variables (age and sex) plus the most strongly associated variables from the univariate screening analyses to report the adjusted odds ratio (AOR) for the common socio-demographic variables (e.g., age, sex, faculty, etc.). As we were examining the impact of SES on the prevalence of depression through logistic regression models, collinearity was investigated for the social variables involved in the analysis. The relationship between the ZDS categories and the domains of the SES were tested via Spearman’s rank correlations to explore any possible collinearity among the SES measures [18]. To test for the validity of these results collinearity diagnostic tests (Tolerance and Variance Inflation Factor ‘VIF’) were conducted. Multicollinearity was assumed if the tolerance below 0.5 and/or VIF above 10 [19].

Likelihood ratio test (LHR) and CIs were calculated to assess significance in the models. The final model fitting was tested in three steps. First, a priori variables and the significant variables in the initial model were included and the LHR test and CIs were used to test for the model fitting. Secondly, all non significant variables from the initial model were included in the final model one at a time to test for their effect on the model. Thirdly, to adjust for the effects of the other factors, these factors were simultaneously incorporated into the model and the likelihood ratio test was used to test for model robustness. Additionally, possible effect modification by sex was examined by testing for interactions between sex and each of the predictor variables individually.

To test for the sense of control mediation-effect on the association between SES and depression, three regression equations were estimated: first, regressing the perceived control on each of the SES variables; second, regressing ZDS scores on each of the SES variables and third, regressing the ZDS scores on the SES variables and on the sense of control [20]. The mediation effect size was explored by analyzing the change in ORs for SES measures after sense of control had been accounted for in the final regression model.

Ethical considerations

This study was approved by the Medical School Ethics Committee of Nottingham University Ref. No.N/9/2008. The online survey was anonymous. All information was identified using a study identification number and was stored separately from contact details.

Results

Characteristics of the sample

Of the 1,146 students who agreed to participate in the survey, 923 (80.5 %) were included in the study (see Figs. 1, 2 for participant flowchart). The most common reason for exclusion was being an international or EU student (14 %). The characteristics of the sample are shown in Table 1. Just over a quarter of the sample (27 %) was male and 73 % was female. The age ranged from 17 to 50 (median 20, IQR 3). The majority of fathers (62 %) and mothers (64 %) were educated to degree level or above. A total of 541 students (58.6 %) scored above the threshold for depression. Most had mild depression (N = 351) (64.9 %), 163 (30.1 %) scored in the moderate depression range and 27 (5.0 %; 95 % CI 4.5, 5.7 %) scored in the range for severe depression range on the ZDS.
https://static-content.springer.com/image/art%3A10.1007%2Fs00127-013-0663-5/MediaObjects/127_2013_663_Fig1_HTML.gif
Fig. 1

Selection and exclusion of participants

https://static-content.springer.com/image/art%3A10.1007%2Fs00127-013-0663-5/MediaObjects/127_2013_663_Fig2_HTML.gif
Fig. 2

Distribution of students according to university and sex

Table 1

Socio-demographic characteristics of the students

 

N = 923

Age group

20 year or less

502 (54.4 %)

More than 20 years

421 (45.6 %)

Sex

Male

247 (26.8 %)

Female

676 (73.2 %)

Faculty

Arts

105 (11.4 %)

Social sciences

195 (21.1 %)

Science

358 (38.8 %)

Medicine

199 (21.6 %)

Engineering

66 (7.2 %)

Year of study

1st

381 (41.3 %)

2nd

249 (27.0 %)

3rd

222 (24.1 %)

4th or more

71 (7.7 %)

IMD quintiles

1st (least deprived)

56 (6.1 %)

2nd

106 (11.6 %)

3rd

453 (48.9 %)

4th

145 (15.7 %)

5th (most deprived)

163 (17.7 %)

Father’s occupation

Never worked/unemployed

217 (23.5 %)

Intermediate occupations

329 (35.7 %)

Managerial/professional occupations

377 (40.8 %)

Mother’s occupation

Never worked/unemployed

254 (27.5 %)

Intermediate occupations

534 (57.9 %)

Managerial/professional occupations

135 (14.6 %)

Father’s education

No higher education

575 (62.3 %)

Higher education

348 (37.7 %)

Mother’s education

No higher education

589 (63.8 %)

Higher education

334 (36.2 %)

FAS

Low

34 (3.7 %)

Medium

315 (34.1 %)

High

574 (62.2 %)

ZDS

No

382 (41.4 %)

Mild

351 (38.0 %)

Moderate

163 (17.7 %)

Severe

27 (2.9 %)

Sense of control Mean ± SD

2.52 ± 0.7

Univariate predictors of depression

The rates of depression showed the expected direction of association with the levels of demographic and SES variables (Table 2). There was an inverse dose–response relationship between the percentage with depressive symptoms and the parental education domains with markedly lower prevalence in the highest SES category of the hierarchy (p < 0.001). But, no dose response gradient was seen with the FAS or IMD quintiles. The mean ZDS scores for different socio-demographic factors showed similar patterns with the mean score decreasing with increase in the rank in each SES domain (Table 2). Depressed students had poorer scores for sense of control (mean = 2.84) compared to students in the non-depressed group (mean = 2.07), where higher scores reflecting less control.
Table 2

Percentage of depression and the mean score on ZDS

Variables

Mean (SD)

Percentage of depression*

P value**

OR***

95 % CI

Sex

 Male

13.6 (7.3)

59.9

=0.16

1

 

 Female

13.4 (6.7)

58.1

0.9

0.6–1.2

Age group

 20 or less

13.5 (7.1)

57.0

=0.21

1

 

 More than 20

13.4 (6.6)

60.6

0.8

0.9–1.5

Faculty

 Medicine

12.8 (6.7)

51.3

<0.05

1

 

 Arts

13.0 (6.8)

56.2

 

1.3

1.1–2.7

 Sciences

13.9 (7.1)

60.3

 

1.7

1.2–2.3

 Social sciences

13.4 (6.7)

62.1

 

1.4

1.2–2.1

 Engineering

14.0 (6.6)

65.2

 

1.8

1.0–3.2

Year of study

 1st

13.1 (6.9)

54.6

<0.01

1

 

 2nd

13.3 (6.8)

60.6

1.2

1.1–1.6

 3rd

14.4 (6.6)

64.4

1.6

1.2–2.4

 4th or later

14.8 (6.5)

65.9

1.7

1.1–2.3

Father’s educational level

 No higher education

14.3 (6.7)

64.3

<0.001

1

1

 Higher education

12.1 (7.0)

49.1

0.5

0.4–0.7

Mother’s educational level

 No higher education

14.5 (6.7)

66.9

<0.001

1

 

 Higher education

11.5 (6.9)

44.0

0.4

0.3–0.5

Father’s job

 Never worked/unemployed

14.8 (6.6)

66.4

<0.01

1

 

 Intermediate occupations

13.6 (6.5)

60.8

0.6

0.4–0.9

 Managerial/professional

12.5 (7.2)

52.3

0.4

0.2–0.6

Mother’s job

 Never worked/unemployed

14.0 (6.6)

62.2

<0.01

1

 

 Intermediate occupations

13.5 (7.9)

59.2

0.7

0.5–1.1

 Managerial/professional

12.1 (7.6)

49.6

0.5

0.3–0.7

FAS

 Low

15.7 (5.9)

58.8

<0.001

1

 

 Medium

14.3 (6.7)

63.2

0.3

0.1–0.7

 High

12.8 (6.9)

56.1

0.2

0.1–0.6

IMD quintiles

 1st (least deprived)

12.4 (7.2)

53.6

<0.001

1

 

 2nd

13.2 (7.8)

55.2

1.7

1.1–2.6

 3rd

15.7 (5.7)

70.6

2.2

1.3–2.9

 4th

13.8 (6.8)

53.3

1.4

0.7–2.5

 5th (most deprived)

11.7 (7.5)

64.9

1.8

1.5–2.3

* We used the recommended ZDS cutoff for minor depression (ZDS > 10)

** All associations are significant at the p < 0.05 level

*** Odds ratios (ORs) as resulted from the screening regression analysis

The results of the screening logistic regression analyses are shown in Table 2. There is no sex or age difference in the proportion scoring above the cut-off for depression (p = 0.16 and 0.21, respectively). Moreover, compared to medical students, students of the faculty of arts are 1.3 times more likely to have depressive symptoms (95 % CI 1.1–2.7). Those in the faculty of science have 1.7 the risk (95 % CI 1.2–2.3), social science students have 1.4 more chance to have depressive symptoms (95 % CI 1.2–2.1) and engineering students are 1.8 times more likely to have depression (95 % CI 1.0–3.2). Regarding the year of study, a consistent trend was observed (χ2 for trend p < 0.01) with risk of depression increasing with year of study.

For the SES domains (Table 2), students with higher parental-SES levels were found to report fewer depressive symptoms. For example, students with highly educated parents (fathers and mothers) are 50 and 60 %, respectively, less likely to have depressive symptoms compared to those with less educated parents (p < 0.001) and the associations for the other SES domains (parental occupation, FAS, and IMD quintiles) are statistically significant (p < 0.001) and in the same direction. There were weak, but statistically significant, negative correlations between ZDS scores and SES domains (ρ −0.08 to −0.17) (Table 3). Analysis showed that these variables were correlated; however, the collinearity diagnostic tests provided contradictory results, so we decided to include the most statistically significant variables (p < 0.05) in the initial multivariate logistic regression model and make a decision about their inclusion in the final model depending on the results of the likelihood ratio test.
Table 3

Correlation between ZDS categories and SES domains

  

1

2

3

4

5

6

7

Co-linearity Diagnosis

ZDS categories

rs*

1

      

Tolerance

VIF

Sig.

       

Father’s education

rs*

−0.149**

      

0.13

7.5

Sig.

<0.001

      

Mother’s education

rs*

−0.160**

0.470**

     

0.2

5.6

Sig.

<0.001

<0.001

     

Father‘s job

rs*

−0.145**

0.419**

0.264**

    

0.2

5.6

Sig.

<0.001

<0.001

<0.001

    

Mother‘s job

rs*

−0.086**

0.141**

0.264**

0.252**

   

0.5

2.2

Sig.

<0.001

<0.001

<0.001

<0.001

   

FAS

rs*

−0.118**

0.232**

0.138**

0.260**

0.164**

  

0.52

1.9

Sig.

<0.001

<0.001

<0.001

<0.001

<0.001

  

IMD quintiles

rs*

−0.097**

0.014**

0.028**

−0.027**

0.006**

−0.013**

 

0.62

1.6

Sig.

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 

Mean SOC

rs*

0.176**

−0.119**

−0.154**

−0.148**

−0.114**

−0.105**

0.101**

0.82

1.2

Sig.

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

* Spearman ranked correlation

** Correlation is significant at the 0.01 level (2-tailed)

Exploratory investigation of possible risk factors for depression

In the initial model the AOR for depression symptoms was significant for the year of study, mother’s education, IMD quintiles and the mean control score using the likelihood ratio test (LRT).

The resulting final logistic regression model, after adjusting for age and sex, was composed of five predictors for depressive symptoms; year of study, mother’s educational level, FAS, IMD quintiles, and sense of control (Table 4). All interaction terms for gender that were tested failed to meet the inclusion criteria for logistic regression models (p > 0.05). The adjusted ORs changed very little from the screening regressions. Compared to first year students the odds of depression increased with increasing year (LRT test p < 0.05) the final year (OR = 1.8 and 95 % CI 1.4–2.7). Moreover, students with highly educated mothers were 50 % less likely to get depression compared to non-higher education mothers (OR = 0.5, 95 % CI 0.3–0.7; LRT < 0.001). Similarly; students from high affluent families were 70 % less likely to report depression compared to those from low FAS families (OR = 0.4 and 95 % CI 0.3–0.9) with a statistically significant LRT test result (p < 0.05). Additionally; students from the most deprived areas were 2.2 times more likely to report depressive symptoms (OR = 2.2 and 95 % CI 1.6–2.9) with a significant LRT test result (p < 0.01) compared with students from most deprived areas. The last variable in the final model is the mean sense of control; where students with low perceived control were more likely to report depression (OR = 1.6 and 95 % CI 15–1.7).
Table 4

Adjusted OR of depression for different domains of SES

Variables

Initial model

Final model*

OR

95 % CI

LRT P value

OR

95 % CI

LRT P value

Sex

 Male

1

 

=0.08

   

 Female

0.9

0.6–1.5

   

Age group

 20 year or less

1

 

=0.30

   

 More than 20 year

1.1

0.6–1.3

   

Faculty

 Medicine

1

 

=0.08

   

 Arts

1.3

1.2–4.4

   

 Sciences

1.9

0.7–2.3

   

 Social sciences

1.6

0.8–2.9

   

 Engineering

2.1

0.9–6.3

   

Year of study

 1st

1

 

<0.05

1

 

<0.05

 2nd

1.7

1.2–2.5

1.6

1.1–2.5

 3rd

1.8

1.1–3.1

1.7

1.2–3.1

 4th or later

1.9

1.4–2.9

1.8

1.4–2.7

Father’s educational level

 No high education

1

 

=0.08

   

 Higher education

0.8

0.5–1.2

   

Mother’s educational level

 No higher education

1

 

<0.001

1

 

<0.001

 Higher education

0.4

0.3–0.6

0.5

0.3–0.7

Father’s job

 Never worked/unemployed

1

 

=0.13

   

 Intermediate occupations

0.7

0.4–1.3

   

 Managerial/professional

0.6

0.3–1.7

   

Mother’s job

 Never worked/unemployed

1

 

=0.26

   

 Intermediate occupations

0.8

0.7–1.6

    

 Managerial/professional

0.9

0.6–2.0

    

FAS

 Low

1

 

0.07

1

 

<0.05

 Medium

0.4

0.1–0.8

0.4

0.3–0.9

 High

0.3

0.1–0.9

0.3

0.2–0.8

IMD quintiles

 1st

1

 

<0.01

1

 

<0.01

 2nd

2.4

1.4–2.8

2.4

1.4–2.6

 3rd

2.7

1.2–3.2

2.8

1.3–3.0

 4th

2.1

0.9–3.4

2.3

1.0–3.1

 5th

2.3

1.4–2.7

2.2

1.6–2.9

Sense of control score

1.58

1.47–1.69

<0.001

1.59

1.48–1.68

<0.001

We used the recommended ZDS cutoff for minor depression (ZDS > 10)

The effect of sense of control as a mediator for the SES–depression relationship was analyzed using logistic regression. When sense of control was added to the model, the advantage of high affluence was decreased by 52 % from 0.19 to 0.29 (95 % CI 0.07–0.65, p < 0.01), while the advantage conferred by a highly educated mother decreased by 12 % from 0.41 to 0.46 (95 % CI 0.32–0.67, p < 0.001). Furthermore, the odds ratio for students from the most deprived regions fell by 4.3 % from 2.30 to 2.20 (95 % CI 1.60–2.90, p < 0.001). Thus, the association between SES and depressive symptoms was partially mediated by sense of control.

Discussion

The key finding in this study was that risk of depression was higher in students from less advantaged backgrounds. Worryingly, over half of the students (58 %) in this sample scored above the threshold for depression on the ZDS.

Previous studies carried out on university students have shown that several socio-demographic factors increase the risk for depression; namely, age, sex, residence, family income, parental education, and parental occupation [4, 21, 22]. Although many studies in community samples find an increased risk of depression in females, the current study in a university sample could not detect any gender difference in the prevalence of depression (p > 0.05) which was in consistent with many other studies conducted with Higher Education students [5, 14, 2326]. This could reflect greater homogeneity of the student sample compared to the general population with male students experiencing many of the same stressors such as study burden, loneliness, and assessment-related anxiety and tension as females [5]. However, as we have a high proportion of female participants in the study (>70 %), it may reflect recruitment bias so that those students with depression are more likely to respond to the questionnaire, thus masking gender differences. It has been argued that males tend to avoid participation if they are not depressed [27].

Although in the univariate analysis we found, in-line with previous studies, that medical students were the least likely to report depression compared with students in other faculties [5, 28] this advantage disappeared in the adjusted model once SES factor had been controlled for. In the fully adjusted model, a dose–response relationship of increasing depression symptoms in relation to the year of study was revealed, with 2nd, 3rd, 4th, or later years were more likely to report depressive symptoms compared to first year students (OR = 1.6, 1.7, 1.8, respectively). This was consistent with other studies on college students which revealed that final year students were more at risk of depression than freshmen [29, 30].

We found that students with better educated mothers had a 50 % lower chance of having depressive symptoms compared to those with mothers with no higher education. This was in agreement with other studies conducted to test the SES correlates of depression including parental education among adolescents and university students; in which the effect size of better education as a protective factor from depression was 20 and 9 %, respectively [4, 22]. Greater affluence appeared to be strongly protective with a 70 % decrease in the odds of depression. Although we did not have a direct measure of family income which is the most commonly used measure of social class [31], it was felt that the FAS was likely to be a more valid measure in this sample as many adolescents/young adults would not be accurately aware of their family income. This strategy has been used by other studies [4, 25, 32]. Consistent with the FAS findings; students from the most deprived areas were 2.2 times more likely to report depressive symptoms (OR = 2.2 and 95 % CI 1.6–2.9) with significant LRT test result (p < 0.01) compared with students from the least deprived areas and this has been noted by others [33, 34].

As predicted, we found an inverse relationship between sense of control and depression. This was a very strong significant effect with an AOR of 1.6 for every additional point on the control scale. As expected, we also found that, a low sense of control was related with depressive symptoms. This was in agreement with a study carried out on university students from 23 countries with variable socioeconomic backgrounds where depressive symptoms were related with lower perceived control with an AOR of 2.3 [4]. Lachman and Weaver [8] theorized that sense of control would mediate SES differences in depression. Upon testing it in the current study, it was found that the SES–depression relation was partially mediated by the sense of control of students with 52 % of the variation in depressive symptoms associated with family affluence accounted for by low sense of control. Our findings are consistent with an international study which concluded that perceived control partially mediated the depression–SES relationship, i.e., 27 % of the variation in depression associated with family wealth was accounted for by low levels of perceived control [4]. Also, we found that 12 and 4.3 % of the variation associated with mother’s education and area deprivation, respectively, were also explained by lack of control. This could be explained by the observation that family affluence is partially determined by parental education and area deprivation [34].

This study found that SES factors were as important in predicting risk of depression in students in a developed country as they had been in a student population in a developing country. For example, in the UK sample high affluence, as assessed by the FAS, reduced the risk of scoring above the cut-off for depression on the ZDS by 60 % compared to 70 % in the Egyptian students. However, father’s professional occupation, rather than maternal higher education, best predicted risk of depression; reducing risk by 40 % [5].

The main strength of the present study was the large sample size (N = 923) across six universities that cover a wide spectrum of SES and the use of a well validated measures of depression. By focusing on a homogenous population it was possible to clearly demonstrate the strong contribution of SES factors to mental health. However, there were several limitations in the study. The study was a cross-sectional, which hinders the inference of causality from the identified associations between socio-demographic variables and depression. Another limitation was the self-reported nature of this study although previous validation of the ZDS with clinical interviews found it to be a valid and reliable scale [35]. It could be that students with symptoms of depression were over represented in this sample because of the topic of the survey. However, as the study aimed to explore the relationship between socioeconomic status and depression, this is not a significant limitation. As with all online surveys there is the possibility that some students may have completed the survey more than once but scrutiny of demographic data, including postcodes revealed no evidence of this. Finally, we could not gather data on some other determinants of depression prevalence such as history of psychiatric disease, family history of psychiatric disorders, history of physical disease and current treatment for psychiatric or physical diseases [36] because we elected to enhance the response rates and keep the survey as concise as possible.

In summary, in the fully adjusted model first year students, whose mothers were highly educated, from a family of a high affluence, lived in a less deprived area and with high perceived control are less likely to suffer from higher rates of depressive symptoms. There are many reasons why university degrees should become more stressful over time, including greater complexity of work and greater debt. Being closer to completion raises uncertainty about employment and career choice and there may be more risk of social isolation. This study suggests that even in a relatively advantaged social group, socioeconomic status exerts a strong impact on the risk of depression. Schemes to widen access to university education need to build in safeguards to protect student mental health.

Acknowledgments

I am very grateful for the Ministry of Higher Education, Egyptian Government specially Assiut University for sponsoring my whole studies. It is a pleasure to express my deepest gratitude and grateful appreciation to the Students at the Universities of Nottingham, Derby, Nottingham Trent, Newcastle, Sunderland and Sheffield who took part in this study.

Conflict of interest

All authors declare that they have no conflicts of interest.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013