Our results suggest that perinatal depression represents a significant disease burden among refugee and labour migrant women on the Thai-Myanmar border. Overall, 18.5% of women experienced moderate-severe depression and 39.8% experienced depression of any severity during the perinatal period. The prevalence of moderate-severe depression found in our study is in line with estimates from non-migrant women in other LMIC. For example, Woody et al. (2017) found a pooled antenatal prevalence of 19.2% and postnatal prevalence of 18.7% in low-income settings [3]. Our estimates are also in line with findings from a meta-analysis of migrant women from LMIC who resettled in HIC, which reported pooled estimates of 17% for major depressive disorder and 31% for any depressive disorder [13]. The finding that over one in three women in our setting exhibited depressive symptoms during the course of her pregnancy and post-partum period is worrying, especially in light of the serious and potentially long-lasting consequences of perinatal depression to women, their children and society at large. The relatively high incidence of depression during the course of pregnancy and the early post-partum period – 15.4% for moderate-severe depression and 28.5% for any depression – highlights the need for healthcare workers to remain vigilant for new-onset symptoms of depression during this period. The slight increase in incidence between the third trimester and the first month post-partum suggests that the time immediately preceding and following delivery may be a time at which women are particularly prone to developing the condition.
The slight decline of both moderate-severe depression and any depression prevalence from the first trimester of pregnancy through to the end of the pregnancy is consistent with findings from other studies [23]. However, this trend must be interpreted with caution given that in our study, women identified as having depression were offered treatment which may have affected prevalence estimates at subsequent assessment times. Overall, the percentage of participants who received treatment was small, at less than 10% of the total sample. But even amongst women who were not treated, the repeated SCID interviews in and of themselves may have had a therapeutic effect by providing women with an opportunity to talk about symptoms in an open manner to a healthcare professional – an opportunity which in this setting is rare [24]. Depression of any severity was more likely to persist over the course of the perinatal period than depression of moderate-severe severity. Again, these results must be interpreted cautiously given the treatment of women with depression. However, the trend suggests that there may be an aspect of milder depressive symptoms – which are included in the ‘any’ depression category but not in the moderate-severe depression category – that are less amenable to treatment. This might be explained by milder symptoms of depression being reflective of on-going difficult life circumstances – which are unchanging over time in this setting – rather than of more severe underlying psychopathology.
The decision to include the less severe DSM-IV category of Depression ‘Not Otherwise Specified’ (NOS) was based on a number of factors. First, we felt it was important to capture the full epidemiological burden of distress within this population, and including depressive symptoms across the entire spectrum of severity was deemed important to ensuring all cases were identified. Second, despite it being regarded as less clinically meaningful than Minor and Major Depressive Disorders, Depression NOS can nevertheless cause significant emotional distress and impact negatively upon individuals’ social and occupational functioning. The many women who fell into this category in our setting, along with the persistence of their symptoms over time and the impact of these symptoms on their everyday lives suggest that this category represents a clinically important state of distress which cannot be ignored. Excluding this category from data collection or data analysis risks falsely labelling these women as psychologically unaffected by their circumstances, fails to recognise the true extent of the burden of depression in this setting and risks leaving a considerable number of women less able to access support.
Risk factors differed between moderate-severe depression and any depression. For moderate-severe perinatal depression, the main risk factors identified were psychosocial, including interpersonal violence (OR 4.5), a history of trauma (OR 2.4), a self-reported history of depression (OR 2.3), perceived insufficiency of social support (OR 2.1) and labour migrant (as opposed to refugee) status (OR 2.1). Low levels of social support have consistently been associated with perinatal depression across multiple and diverse settings, and it is unsurprising that this was one of the key factors associated with depression in our population [8, 10, 14]. In this context of geo-political tension, women often live in separation from family members who have remained in home towns or villages in Myanmar or resettled abroad. The support networks normally available to these women from extended family members is therefore often lacking. Associations between interpersonal violence and perinatal depression are also well-documented: an important finding given the increased occurrence of interpersonal violence during pregnancy [25]. Similarly, experiences of trauma are common among many migrant populations, and mechanisms to sensitively assess and manage trauma histories are called for as part of a wider assessment of mental well-being [26]. The relevance of a self-reported history of depression is of particular interest. On the Thai-Myanmar border, mental disorders are rarely formally diagnosed by medical professionals. The strong association we found between women’s own subjective accounts of prior experiences of depression suggests that asking women directly about depressive histories may be useful as part of a wider programme to detect perinatal depression in this setting.
In the analysis of factors associated with any depressive disorder – i.e. including the milder category of Depression NOS – interpersonal violence (OR 4.4), a history of depression (OR 3.4) and experience of trauma (OR 2.2) remained important risk factors. However, telephone ownership (OR 1.6) was also significantly associated, and an additional member of the household employed (OR 0.57) was a significant protective factor. These variables were included as proxies for household income (telephone ownership) and income stability (additional household member working). Their significance in the multivariable model suggests that when milder experiences of depression are included in the outcome, socio-economic factors come into play, as these impact heavily upon the daily lives of migrant and refugee women in this setting. This adds weight to the idea suggested above that the lower-severity depressive episodes are reflective of general life situations rather than an endogenous depression. If this is indeed the case, moderate-severe perinatal depression and perinatal depression of any severity represent distinct diagnostic categories, each important in its own right. Finally, the magnitude of the association between antenatal and postnatal depression (OR 5.1) was striking and emphasises the importance of identifying and supporting pregnant women with depression as early as possible to avoid persistence of the condition postnatally.
We found that labour migrants had a higher likelihood of depression than refugees. Although both populations endure hardships, labour migrants may arguably be more vulnerable to day-to-day stressors such as securing an income and living under the constant threats conferred by their undocumented status. In a study of labour migrants in the urban centre of Mae Sot, forced overtime labour, verbal abuse and withholding of documents and salaries were common and linked to depression [12]. Refugees – though they face multiple other psychological stressors – have at least some of their basic needs such as a minimal ration of food per family and housing requirements met within the camp setting, and thus may be slightly less at risk of depression relative to their labour migrant counterparts.
Strengths and limitations
To our knowledge, this represents the first cohort study of perinatal depression among labour migrant and refugee women in a LMIC setting. Our findings offer a direct comparison of labour migrants and refugees within the same setting. Few other studies have included different categories of migrants, and those that did have not always conducted comparative analyses [10]. Hard-to-reach migrants including those with insecure legal status and those who do not speak the host country language – such as the women included in this study – have rarely been included in research to date, and the importance of this group being represented has been emphasized [14]. The use of a diagnostic rather than screening instrument to identify women with depression constitutes a further strength, enabling greater accuracy and addressing a significant gap in the existing evidence, which relies heavily on screening tools [13]. Our study adopted a comprehensive approach to perinatal depression: outcomes encompassed the rarely studied category of Depression NOS, while inclusion of a wide range of socio-demographic, psychosocial and clinical exposures enabled the effects of multiple potential risk factors to be explored simultaneously. The longitudinal design allowed associations between antenatal and postnatal depression to be examined. Finally, given that over 90% of migrant women in this region access antenatal care, we believe that the results are representative of the local migrant population [27].
There were also a number of limitations. The SCID may have missed culture-specific manifestations such as somatic symptoms which are common in Asian cultures [4, 28]. This could have led to an under-estimation of the true prevalence of perinatal depression. However, previous work on the Thai-Myanmar border found that the SCID elicited more information than a number of screening tools [29, 30]. Working across different languages introduces a risk of misinterpretation and loss of meaning. Social desirability bias and stigma may have influenced women’s disclosure of symptoms [24]. Due to resource constraints, some interviews were conducted by trained counsellors rather than a physician. However, the ability of local counsellors to establish trust, understand, recognise and respond to participating women can be regarded as a considerable strength. Furthermore, in resource-constrained settings it is neither desirable nor sustainable to rely on specialists to identify or manage mental disorders and indeed, training and engaging local staff in this study built up experience and expertise within the community and ensured its sustainability.
The treatment of women who were diagnosed with depression introduces uncertainty into the point prevalence estimates of depression from the second trimester of pregnancy onwards, as the symptoms of those who were followed-up and treated may have resolved as a result and led to an under-estimation of natural rates. It also meant that the duration of depressive episodes could not be ascertained – a potentially significant omission when one considers, for instance, that depression which lasts for the whole pregnancy differs considerably from depression that starts at the end of pregnancy and ends at birth. However, we felt strongly that not treating women experiencing depression would not have been ethical practice and it was important for all those participating in the study to feel supported. The point prevalence at baseline (first trimester of pregnancy) remains unaffected, and the period prevalence also resolves this issue to a degree, as any woman who experienced depression at any timepoint (even if it subsequently resolved following treatment) is accounted for in this measure. Nevertheless, we cannot rule out the possibility that treatment of participants may have led to the under-estimation of true prevalence.
The lack of a non-migrant control group makes it difficult to establish to what extent depression was associated with migrant status rather than general socio-economic adversity. The political tensions in this region meant it was not possible to recruit non-migrant Burmese or Karen women living in comparable regions of Myanmar, Although the overall sample size was sufficient for our study aims, the smaller number of women who received treatment meant that the effectiveness of counselling and medication offered could not be reliably assessed, nor could a stratified analysis according to treatment status be conducted. Finally, the inclusion of a large number of co-variates in logistic regression may have increased the probability of type II errors. However, the fact that risk factors identified in this study are highly plausible and well-aligned with findings from other settings lends credence to the validity of the methods and final model.
The Thai-Myanmar border region is a resource-constrained area with diverse health priorities, and recommendations for practice must remain feasible and realistic within the context of this environment. Evidence from HIC suggests some benefit of primary prevention efforts such as strengthening social networks, women’s empowerment and addressing interpersonal violence [31]. The effectiveness of such efforts in LMIC needs evaluating. Secondary prevention efforts for the early detection of perinatal depression may also be indicated [28]. The perinatal period offers a valuable window of opportunity of frequent contact between women and health providers, and routine screening of all pregnant women is recommended across many HIC settings. Optimal timing of screening initiatives needs to be established, especially in light of continued onset of new cases throughout the perinatal period. Ongoing staff training and regular debriefing opportunities are essential for long-term sustainability. Finally, effective and integrated treatment and referral mechanisms are required. The evidence base around effective interventions in LMIC settings is growing. Initiatives delivered by trained non-specialists have been well-received, and embedding programmes within the community can reduce the stigma of attending [31]. Ultimately, addressing perinatal depression in this setting requires long-term, multi-sectoral collaborations at community, regional and national levels to tackle the wider socio-economic determinants.