Main findings, interpretations, and comparisons with previous studies
In this study, data from a large national survey among adults were used to examine the association between MDE and AUD and the effects of socio-economic factors on this association. In agreement with several studies [7, 23, 29-33] our study found a significant association between depressive disorder and AUD. However, our results do not provide support for the evidence of a stronger association between MDE and AUD among people with lower SES, but rather indicate that the SES impact on mental disorders is controversial which may vary between countries [19] and the types of measurements used for either depressive and alcohol use disorders or SES variables [23]. The significant relationship between MDE and AUD could be explained in terms of shared common genetic and environmental factors in the comorbidity of depressive disorder and AUD, as investigated in other studies [34-36]. It could also be explained in terms of causality where depressed persons may turn to alcohol as self-medication for their symptoms and develop AUD afterwards [37, 38]. However, confirming this explanation is beyond the scope of this study.
The prevalence of MDE was higher among harmful-dependent drinkers and non-drinkers than that in hazardous drinkers, making those with MDE seeming less likely to have hazardous drinking, compared to those with non-drinking behaviors. This should not be simply interpreted as that MDE was a protective factor against hazardous drinking because such an association was not controlled for other confounding factors. In the multivariate analyses when we controlled for confounding factors and stratified by level of SES variables, we found that most of the associations between MDE and hazardous drinking were not significant, and among the significant ones, increased likelihoods of MDE on hazardous drinking were found. Nevertheless, if we explain our findings on the relationship between MDE and AUD in terms of a causal pathway, it could be argued that if depressed individuals use alcohol for self-medication of their depressive symptoms, which are usually chronic, then they may use it in a harmful pattern and be more likely to become harmful or dependent drinkers than hazardous drinkers. This may therefore explain the higher prevalence of MDE among harmful-dependent drinkers than that among hazardous drinkers.
In agreement with other studies [23], the association between MDE and AUD, in particular harmful-dependent drinking, varied across different levels of socio-economic status. We observed significant associations between MDE with harmful-dependent drinking at high and low terciles of wealth index but not in the middle tercile. We also found a significant association between MDE and either hazardous or harmful-dependent drinking in those with a secondary school or higher level of education but not in those with a primary school level of education, and a stronger association among those living in urban areas compared to those living in rural areas. Inequalities in the prevalence of MDE and AUD between levels of wealth index, education, employment, and area of residence were also found, which supports studies investigating social inequalities in mental health conditions within a country [11, 13, 19, 39].
The reason that the strengths of association between MDE and harmful-dependent drinking among those in the highest and lowest socio-economic groups were significant and higher than that in the middle group is unclear. On the one hand, it could be explained by the fact that people in the high SES group may have more means to access alcohol or other drugs for self-medication or as a coping strategy when they get depressed. Major depression is a chronic illness and cognitive deficits are frequently observed in those suffering from major depression even at the first episode [40] which can in turn affect the individuals’ coping style. Those who are suffering from symptoms of major depression may turn to alcohol use to relieve their depressive symptoms and as a consequence, a heavy or harmful drinking pattern and alcohol dependence can occur. This phenomenon can, in fact, occur in people of any socio-economic class. However, if those who are in the high class have a better access to alcohol, the likelihood of having a harmful drinking pattern or alcohol dependence could be higher.
On the other hand, people of the lower SES group often face more social difficulties, which enhances susceptibility to negative health outcomes when exposed to risk factors [12, 15, 41]. The higher prevalence of both MDE and AUD among those in the lowest SES group could lend support to this finding. This is consistent with other studies which showed that people who have a low SES and are experiencing a greater social disadvantage generally suffer poor health outcomes, including depressive and substance use disorders, have more disabilities, and poorer access to health care [9, 10, 17, 19, 39, 42, 43]. Whether the strong association between depressive episode and harmful-dependent drinking in the high and low SES groups is driven by a high accessibility to alcohol in the high SES group or an enhanced susceptibility to negative health outcomes when exposed to chronic difficulties among the low SES group is plausible and deserves further research to elucidate mechanisms explaining this social disparity. However, chance cannot be ruled out as an explanation for the observed findings due to the low statistical power as both MDE and AUD are rare in this general population study. Furthermore, the classification of wealth index derived from the PCA might not differentiate people well, thus spurious differences could also be the case [19].
For both groups of AUD, the significant and strong associations with MDE were found only among those completing secondary school education or higher but not among those with primary school education. The reason for this may be similar to the above explanation for the high socio-economic group and the finding further supports the impact of SES on the relationship between the two disorders. Education is known to causally influence health through mechanisms such as creating a greater sense of control, better working conditions, increased social capital, and improved health behaviors [44]. If it is true that highly educated people turn to alcohol when they are depressed to alleviate their dysphoria, it is possible that a depressive condition deforms their sense of control, leading to such poor health behavior and education does not protect against this outcome. This finding highlights the importance of a comprehensive assessment of the co-occurrence of depressive disorder and AUD and providing appropriate treatment and care in all individuals regardless of their education level.
No modifying effect of employment status was found on the association between MDE and AUD. The association was only significant for that between MDE and hazardous or harmful-dependent drinking among the employed group. However, the number of unemployed individuals who were depressed and had either hazardous or harmful drinking was low or none, preventing us from determining the effect of employment status on the relationship between MDE and AUD. The inconsistency of the findings by three SES indicators warrants further studies to explicate mechanisms underlying the socio-economic inequality in the relationship between MDE and AUD. As mentioned in other studies, socio-economic inequalities in health may vary depending on the indicators used to measure SES and no single indicator can provide a full picture of SES of the population [23, 45].
Living in a rural area was found to be protective against harmful-dependent drinking in the univariate analysis and a stronger association was found between MDE and harmful-dependent drinking for people living in urban areas compared to those living in rural areas. Evidence of the relationship between urbanicity with depression and alcohol use varies in the literature [41, 46, 47]. In Thailand, people living in rural areas often have close ties and shared social activities. The poor in these rural areas also live better lives than do those in the urban areas as they have more ready access to food and green space in their surroundings. People living in urban areas can face a lot of hardships related to urban lifestyles, for example, faster pace, higher crime rates, a more crowded environment, limited green space, and higher levels of pollution [48]. Such urban environments may enhance susceptibility to negative health outcomes when exposed to risk factors and increase the risk of poor mental health [49].
Depressive disorders and AUD are common public health problems in Thailand [50] and priority conditions identified in the WHO Mental Health Gap Action Programme (mhGAP) [51]. The present findings should be considered in terms of the social context of Thailand where income inequality is in the middle range (GINI coefficient ranges between 37.5 and 39.4 since 2010) [52]. Our findings among the general Thai population suggest that mental health care and promotion would not help to improve the mental health of Thai people to its highest limit should concurrent efforts to reduce social inequalities not be implemented.
Mental health and many mental disorders are shaped by the social, economic, and physical environments where people live [15]. Our findings emphasize the need to implement and scale-up public policies and intervention programmes for depressive and alcohol use disorders among the general population. Evidence-based interventions for depressive disorder include treatment with antidepressants and psychosocial interventions such as cognitive behaviour therapy and problem-solving while those for AUD are policy and legislative interventions including regulation of the availability of alcohol, enactment of appropriate drink-driving policies, reduction of the demand for alcohol through taxation and pricing mechanisms, and interventions for hazardous drinking and treatment of AUD with pharmacological and psychosocial interventions [51]. Our results also suggest that interventions for both conditions should be provided in proportion to the needs of people of different socio-economic groups. Future research is also needed to understand the mechanisms which underlie the different relationships that exist among people of different socio-economic status.
Moreover, the present research should be helpful for clinical practice. In clinical settings where the prevalence of major depression is usually higher than that in the general population, clinicians should also seek for the co-occurrence of AUD, especially harmful-dependent drinking in depressed patients. On the assumption of a causal process of depressive disorder leading to harmful use or alcohol dependence, it can be indicated that some percentage of patients with AUD may actually improve upon treatment of their depressive disorder. There is a number of studies showing the efficacy of psychological interventions among people with co-occurring alcohol use and depressive disorders [53]. This thus suggests that treatment of MDE should include assessment and treatment of AUD or vice versa. The consistent finding across studies that the prevalence of major depression and AUD was high among those with low socio-economic status should be of concern for clinicians. Patients with a lower SES are generally less likely to access medical care whereas they tend to have more disabilities and a poorer prognosis [13, 16], thus they would need more help in a clinical setting. It should also be recognized that for patients who are in a higher SES or education level, or employed, a depressive episode increases their tendency to have alcohol use disorder, thus treatment of both disorders is recommended and should be tailored to their needs.
Study strengths and limitations
The main strength of this study is the large sample size, nationwide sampling frame, and probability sampling method, making it representative of the general population of the whole country. Furthermore, the possible role of confounding factors such as socio-demographic factors and the presence of chronic medical diseases was taken into account in the analyses. However, there are some limitations that deserve mention. A cross-sectional study is a useful design for obtaining the prevalence of MDE and AUD among the general population, but it cannot establish a causal relationship. The direction of causality is indeterminate; depression and AUD could be reciprocally related to each other by a feedback loop in which drinking increases the risk of depression and the depression leads to an increased consumption of alcohol and related problems [4]. In our study, MDE and AUD were self-reported and the questionnaire used for measuring both conditions was a screening instrument (not a diagnostic instrument), assessing symptoms that occurred in the past 12 months, which may not fit the full criteria of major depressive disorder or alcohol use disorder. Therefore, our results have limited comparability with studies that used diagnostic measures. Furthermore, there were few respondents with MDE who had hazardous or harmful-dependent drinking as the sample was taken from the general population. Although the association between MDE and harmful-dependent drinking was significant across most socio-economic levels, most of the confidence intervals were too wide to make any solid conclusions. Due to the unavailability of data, our study did not take into account other potential confounding factors, e.g. stressful life events and personality profile, which may be associated with both MDE and AUD. Finally, a collider stratification bias might exist but be undetected in our study due to our inability to measure some relevant stratification variables. Inclusion of these variables in the model could strongly mitigate the effect of MDE on AUD. However, it is not likely that the positive association between MDE and AUD found in our study would have been caused by a collider stratification bias as our finding was consistent with the true causality between MDE and AUD established in previous literature. [7, 23, 29, 33-38]. In addition, our data came from a national survey sampled from the general population so there is no possibility of collider stratification bias by design.