Introduction

Major depression is a leading cause of disability in many developed countries [1], particularly among individuals with low socioeconomic position (SEP) [2–4], commonly defined as low education, occupational status, and/or income. Moreover, there is increasing evidence that these inequalities may originate in early life [5–13]. Although most studies find that low early-life SEP is associated with an increased risk of depression in adulthood [5, 7, 9, 11–13], not all aspects of early-life SEP are equally salient to depression in adulthood, including maternal vs. paternal education [9, 12]. Furthermore, depression onset occurs more commonly in early than later adulthood [2, 3], but little is known on the relation between maternal—and even less so paternal—education and depression in young adults. Differences may shed light on the mechanisms through which early-life SEP influences depression in adulthood.

Though maternal education is one of the strongest predictors of offspring physical and psychological health [14, 15], even decades later [12, 16, 17], the literature on the relationship between maternal education distinct from paternal education (or other SEP measures) and adult depression is scarce [9, 12]. Evidence of a relationship between early-life SEP and adult depression comes from a handful of cohort studies, the majority of which use the highest (or a combination) of both parents’ education or occupational status [7, 9–11, 13, 18], which makes it challenging to disentangle which aspects of early-life SEP are most relevant to depression [19]. In addition, previous findings are based on older adults [12], or combine respondents from a wide age range [9], thus potentially masking important differences related to cohort or life stage. Early adulthood is a particularly sensitive period for developing depression as depression onset occurs most commonly during this life stage [2, 3]. This is troubling given the reported associated increase in depression severity and recurrence [20], psychosocial impairment [21–23], and morbidity [20, 24, 25] with earlier age of onset.

From the life course perspective, two general models exemplify how early-life circumstances influence adult health. In the latency model, low early-life SEP directly influences adult depression through exposures at critical (or sensitive) periods of development, independently of adult circumstances; in the ‘chains of risk’ model, low early-life SEP is linked to depression indirectly through intermediate risk factors (or pathways) [26]. The majority of life course studies have examined adult SEP as an intermediate between early-life SEP and adult depression, and though most studies find evidence of a direct effect of early-life SEP [5, 9–12], some findings support an indirect relationship [7, 13]. Less is known on other potential pathways including adverse early-life experiences (such as physical and parental substance abuse) [10, 27–29], psychological stressors (or protective factors) [30, 31], and physical illness [26, 32].

To address these gaps in the literature, we sought to compare how maternal and paternal education were respectively associated with major depressive episode (MDE) in early adulthood, and to determine how adverse early-life experiences, early-adult SEP, psychosocial stressors (and resources), and physical health influence the relationship between each parent’s education and MDE.

Methods

Data (N = 17,276) were from the Canadian National Population Health Survey (NPHS) [33], a nationally representative longitudinal household survey initiated in 1994/1995 by Statistics Canada. Respondents were followed biennially and at the time of this analysis data were available up to 2006/2007 (cycle 7). The data were collected using multistage, stratified random sampling, designed to insure adequate representation across major urban centers, smaller towns, and rural areas in all provinces. People living in Native reserves, military bases, institutions, and some remote areas of Ontario and Québec were excluded. One person from each household was randomly selected to complete the health component. Parents’ education was collected only for respondents living with their parents, thus analyses were restricted to respondents residing with their parent(s) in 1994/1995 (cycle 1), who were aged 12–24 years at the time (N = 1,899). Data for the outcome, MDE, was taken from 2004/2005 (cycle 6) or 2006/2007 (cycle 7). Figure 1 shows the timing of data collection according to survey cycle and respondent’s age. Due to loss to follow-up and non-response, we had to exclude 483 (25.4 %) respondents who were missing the outcome and 149 (7.8 %) missing one or more adult covariate, leaving 1,267 cases for analysis. Loss to follow-up was associated with low or missing parental education, unmarried/common-law status, low income, being out of the labor force, and low sense of mastery.

Fig. 1
figure 1

Lexis diagram depicting the timing of data collection for the explanatory and outcome variables according to survey year and respondent’s age at the time of survey (N = 1,267). aCollected in 1994/1995, 2000/2001, 2006/2007. bCollected at each survey between 1994/1995 and 2002/2003

Full size image

Measurement of depression

Past-year MDE was evaluated by a trained interviewer using the Composite International Diagnostic Interview Short Form for Major Depression (CIDI-SFMD) [34]. The CIDI-SFMD is a structured diagnostic instrument that is scored to yield a predictive probability estimate based on the definitions and criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IIIR) [35] and the International Statistical Classification of Disease (ICD-10) [36]. Like the majority of NPHS-based depression studies, in this study MDE was defined based on a 90 % predictive probability cut-off, which required that the respondent had experienced a minimum five of the nine depressive symptoms, including depressed mood or loss of interest or pleasure most of the time and on most days, during a 2-week period in the preceding year. The overall accuracy of the CIDI-SFMD in identifying an MDE is 93 % based on the DSM-III-R criteria [34]. For the present study, respondents had to score at the 90 % predictive cut-point for MDE in either 2004/2005 (cycle 6) or 2006/2007 (cycle 7).

Measurement of parents’ education

Maternal and paternal education levels were reported by each parent or a knowledgeable adult member of the household in 1994/1995 (cycle 1), and were categorized as less than secondary school, secondary school/more, and unknown. In initial analyses, we included separate categories for secondary, some post-secondary, and post-secondary school; however, statistically significant differences were not found between these categories and therefore we collapsed them into one category (secondary school/more).

Measurement of potential intermediate variables

Early-life adversities were measured using the Childhood and Adult Stress Index (CASI) [37], which was available for 1994/1995 (cycle 1), 2000/2002 (cycle 4), and 2006/2007 (cycle 7). For respondents who provided responses to the CASI at more than one time point, we included data from their earliest completed interview. The CASI contains seven questions that ask if the respondent experienced (as a child or teenager) parental divorce, parental long-term unemployment, hospitalization for 2 or more weeks, parental substance abuse, physical abuse, a traumatic event, or being sent away from home. As evidence suggests that number of adversities experienced may be confounded by the severity of these adversities [38], we modeled the CASI items individually, rather than using a summary score.

Data on each adult risk factor were from 2002/2003 (cycle 5), or from 2000/2001 (cycle 4) if data were missing at cycle 5. Adult SEP variables included (1) education (less than secondary school, secondary school, more than secondary school), (2) employment status (employed, unemployed, not in the labor force), (3) student status (currently a full- or part-time student, or not), and (4) income adequacy (lowest, middle, highest). In the NPHS, income adequacy quartiles (lowest, middle-low, middle-high, highest) were derived from the total household income and the number of people living in the household. In the present study, middle-low and middle-high quartiles were combined due to similar results.

Adult psychosocial functioning included (1) chronic stress measured by the General Chronic Stress Index [37], a measure of activity overload, financial difficulties, and problems with relationships in day-to-day encounters (summary score from 0 to 11; higher values indicate more stress), and (2) sense of mastery measured by the Mastery Scale [39], which captures the degree to which individuals believe that their life chances are within their control (summary score from 0 to 28; higher values indicate a greater sense of mastery).

Adult physical health was based on the number of self-reported chronic conditions (range 0–22) that were diagnosed by a health professional and lasted (or were expected to last) 6 months or more, including food allergies, other allergies, asthma, fibromyalgia, arthritis or rheumatism (excluding fibromyalgia), back problems (excluding fibromyalgia and arthritis), high blood pressure, migraine headaches, chronic bronchitis or emphysema, diabetes, epilepsy, heart disease, cancer, intestinal or stomach ulcers, effects of a stroke, urinary incontinence, bowel disorder such as Crohn’s disease or colitis, thyroid condition, any other long-term condition.

Measurement of potential confounders

Covariates included age (22–36 years) in 2006/2007 [the last cycle of data collection; mean (SD) = 27.6 (4.5)], sex [male (50.8 %), female (49.2 %)], ethnicity [white (87.6 %), other (12.4 %)], marital status [married/common-law (29.6 %), or not (70.4 %)], and depression history defined as meeting the criteria for MDE at one or more interviews between 1994/1995 and 2002/2003 (cycle 1–5) or having had received a diagnosis of major depression by a health-care professional (according to the respondent’s self-report in 2004/2005 (cycle 6). Parents’ diagnosis of depression (either parent diagnosed with depression, neither parent diagnosed with depression) was available for 2004/2005 (cycle 6). Respondents were asked (1) “Have any close relatives—including your biological parents, brother and sisters—ever had one or several episodes of being sad, depressed, discouraged or uninterested most of the day, for several days, weeks and longer?”; and if they answered ‘yes’ to the first question they were then prompted to clarify (2) “was this your birth mother (or father)?”

Statistical analyses

Prevalence of MDE was calculated and logistic regression was used to estimate unadjusted odds ratios (OR) and 95 % confidence intervals (CI) of MDE for each explanatory variable. Multivariate logistic regression models were used to estimate ORs and 95 % CIs of MDE for maternal and paternal education, separately and simultaneously. These estimates were compared to evaluate possible collinearity between maternal and paternal education. The beta coefficient for maternal low education (less than secondary) increased slightly and the coefficient for paternal low education decreased; in the mutually adjusted model, however, the adjusted estimates were within the unadjusted confidence limits, with only slight increases in standard error, thus collinearity was not suspected. Four subsequent models included both maternal and paternal education, with adjustment one at a time for all early-life adversities (parental divorce, hospitalized for 2/more weeks, parental unemployment, sent away from home, parental substance abuse, experienced a traumatic event, physical abuse), adult SEP (education, student, income adequacy, employment status), psychosocial factors (stress, mastery), and physical health (number of chronic conditions), respectively. The final model was fully adjusted for all explanatory variables. Tests for interactions between sex and each parent’s level of education were not significant (p < 0.05); therefore, males and females were analyzed together, and interaction terms with sex were included for potential intermediate variables that interacted with sex. All multivariate models were adjusted for age, sex, ethnicity, marital status, and respondent’s history of depression.

In addition, final models were run adding maternal age at the respondent’s birth, family structure (single- vs. two-parent home), and household income to determine whether these factors may have influenced the relation between parents’ education and adult MDE. Finally, in sensitivity analyses, all models were adjusted for parental diagnosis of depression using the sub-sample that had complete data for this variable (N = 1,005).

Statistical analyses were performed using SAS Version 9.2 (SAS Institute Inc., Cary, North Carolina, USA). Ethics approval was obtained from the institutional review board of McGill University’s Faculty of Medicine.

Results

Nearly 11 % (N = 143) of respondents met the criteria for past-year MDE at follow-up (Table 1). MDE was much more prevalent among respondents with maternal low education (less than secondary 18.5 %) than those with maternal high education (secondary/more 9.5 %) with an unadjusted OR of 2.15 (95 % CI 1.46–3.16) for maternal low relative to high education. MDE was higher among females than among males (14.8 vs. 7.8 %), and also among respondents with a history of MDE (24.9 vs. 8.1 %). Though MDE was slightly more prevalent among individuals with low (11.4 %) than high (10.1 %) paternal education, the CI for the unadjusted OR for paternal low relative to high education included 1. Neither adult secondary nor less than secondary education were associated with MDE; however, MDE appeared to be more prevalent in these groups relative to more than secondary education (15.6 % and 13.3 vs. 10.5 %, respectively). MDE was more prevalent in the lowest quartile (19.6 %) and middle (middle-low and -high combined) quartiles (12.4 %) than in the highest quartile (7.7 %) of income adequacy, and in unemployed than employed (18.6 vs. 10.3 %). An increase of one unit on the Stress Scale was associated with 30 % higher odds (95 % CI 1.19–1.42) of MDE, and an additional chronic condition was associated with 20 % higher odds (95 % CI 1.04–1.38) of MDE. MDE was more prevalent among those who experienced an early-life adversity than those who did not.

Table 1 Characteristics of respondents with a major depressive episode, National Population Health Survey, Canada, 1994/1995–2006/2007 (N = 1,267)
Full size table

Associations between parents’ education and MDE

When the level of education for each parent was modeled separately, respondents whose mother had a low level of education were twice as likely to experience MDE in adulthood (OR 1.98, 95 % CI 1.32–2.97) compared to respondents whose mother had a higher level of education (Table 2). In contrast, paternal education was not significantly associated with MDE. After mutual adjustment, the magnitude of the estimate for maternal education increased slightly (OR 2.08, 95 % CI 1.35–3.18), while that for paternal education it did not change.

Table 2 Adjusted odds ratios (95 % confidence intervals) for the association between parents’ education and major depressive episode, with and without including potential intermediate variables (N = 1,267)
Full size table

The effect of potential intermediate variables

The associations for parents’ education did not change after simultaneously adjusting for all early-life adversities in Model 3. Among the early-life adversities, only parental substance abuse and traumatic experience remained positively associated with depression. There was an interaction with gender for both parental divorce and unemployment, wherein males (but not females) experienced greater odds of depression with parental unemployment and unexpectedly lower odds with parental divorce. In Model 4, the estimates for parents’ education did not change after inclusion of adult SEP (education, student, income adequacy, employment status), but low income adequacy was independently associated with depression. Likewise, the associations for maternal and paternal education were not substantially changed after adjustment for psychosocial factors (stress, mastery) in Model 5, or physical health (number of chronic conditions) in Model 6. Nevertheless, stress was independently associated with depression in males, but not in females. Simultaneous adjustment for all early-life and adult risk factors had, once again, little influence on the results (Model 7). Associations for potential intermediate variables did not change appreciably in the full model compared to Models 3–6, except for student status which was associated with increased odds of depression.

In sensitivity analyses, adjustment for maternal age at birth, family structure, and household income did not alter the estimates for maternal or paternal education in the final model (data not shown). Finally, adjusting for parents’ depression slightly strengthened the estimates for parents’ education, but did not change the interpretation of the results ("Appendix").

Discussion

In this prospective population-based study, we found evidence that maternal low education, but not paternal low education, was associated with an increased risk of MDE in early adulthood. Furthermore, the association of maternal education with depressive episodes among their young adult offspring was robust to adjustment for all other risk factors. Finally, the effect for maternal education was not attributable to a host of other potentially confounding factors from the family or origin, such as maternal age at the birth of the respondent, household income, family structure, or a parental history of depression.

Our finding of a stronger association for maternal than paternal education is consistent with previous studies that examined parents’ education separately [9, 12]. In contrast to these studies, however, we did not observe an effect of low paternal education. As mothers tend to play a larger role in childrearing, it may be that their parenting skills, or the social learning that takes place in terms of teaching coping skills, have a stronger impact on the risk of MDE. In turn, paternal education may have more of an indirect effect through its impact on the process of social mobility of their offspring, the direct effects of which become more evident at older ages, as attained statuses (education and income) take precedence over ascribed statuses (parents’ socioeconomic status).

Thus, the lack of paternal education effect we observed, which contrasts with previous studies distinguishing the effects of mothers’ and fathers’ education [9, 12], may be due to cohort or life-stage (or both) differences in the association of parental education with MDE. Indeed, these previous studies surveyed adults who were aged 50 years and older in 1998 [12], or adults aged 18–95 years in 1995 [9]. Thus, these respondents were both older on average when surveyed than our respondents, and predominately from earlier birth cohorts (1890–1977), whose exposure (and that of their parents) to formal education may have differed substantially from that of our younger cohort born between 1970 and 1983 [40]. Furthermore, a period effect may be operating, whereby the importance of education to health may have changed over the past century [40, 41], and hence the importance of parents’ education to offspring’s health may also be changing.

Arguably the most commonly studied pathway between early-life SEP and adult depression is adult SEP. Although some studies [7, 9, 13] find that adult SEP mediates most of the association between parents’ education and adult depression, our results are more in line with studies observing independent effects of average parents’ education [9, 12] and maternal education [12] after adjustment for adult SEP. Unlike most previous studies [9, 12, 13], though not all [7], we also did not find a statistically significant association for adult education, although we did find low income to be important, consistent with other studies [9, 12, 13]. This may have been due to the younger average age of our sample for whom the benefits of education may not have been as salient compared to older adults in previous studies [41].

Little is known on other potential intermediate variables between early-life SEP and adult major depression. Early-life adversities are plausible intermediates [10, 27–29] but have not been found to explain the influence of parental education on poor adult mental functioning [10], consistent with our observations. Rather, our observations suggest that parental unemployment (for males), parental substance use, and early-life trauma are independently associated with adult depression. Although adult education appears to influence mental health through pathways involving sense of control and social support [41, 42], our findings do not suggest that chronic stress and sense of mastery explain the effect of maternal education on adult depression. Adult physical functioning is strongly influenced by maternal education [14–16] and has been shown to predict major depression [32], but we found little evidence that chronic conditions mediated the relationship between maternal education and major depression in early adulthood. Though the potential intermediate variables included in this study were not exhaustive, our results are suggestive of a sensitive or critical period effect rather than a pathways effect of maternal education on adult depression. The less apparent role of paternal education may reflect a generally less dominant role of fathers in childrearing and managing family health matters [43], two pathways that we were unable to test for, given the lack of variables measuring parenting skills or health literacy.

This study has some limitations. Around 33 % of respondents were lost to follow-up, which was associated with low or missing parental education, unmarried/common-law status, low income, being out of the labor force, and low sense of mastery. Thus, our estimates may in fact be conservative given that we observed a strong effect in a sample with fewer risk factors. We cannot rule out the possibility of unmeasured episodes of MDE because the NPHS interviews occur every 2 years, whereas the diagnostic interview for MDE covers only the previous year, though estimates of associations are not expected to be biased [44]. Although we lacked data on potentially important intermediate variables such as quality of parenting [45], we were able to investigate a number of important risk factors that are commonly cited in the depression literature. Strengths of this study include a nationally representative sample, prospective design, a structured diagnostic measure of MDE, and parent-reported parental education, reducing bias due to self-report and recall. Finally, it is unclear how our findings apply in contexts where formal education is more or less accessible than in Canada.

Our findings suggest that maternal and paternal education are non-redundant measures of early-life SEP. Maternal education appears to be a particularly important contributor to depression in adulthood and independent of other early-life adversities, adult psychosocial factors, and physical health. Future research should therefore seek to explore the other underlying mechanisms by which maternal education impacts offspring depression. These findings could then serve to orient public health prevention initiatives that promote mental health intergenerationally, as well as across the life course.