Maternal depression in rural Pakistan: the protective associations with cultural postpartum practices

Data come from the Bachpan study, a cluster randomized perinatal depression trial nested in a longitudinal birth cohort in rural Pakistan. The study is representative of a typical low-socioeconomic rural area in the north of the Punjab Province. A more detailed description of the study is available elsewhere [21]. Briefly, pregnant women in 40 village clusters were identified and invited to be screened for depression using the Patient Health Questionnaire (PHQ-9). Women who scored 10 or higher were invited to participate in the study; roughly one in three women who scored less than 10 were asked to participate in the study [21]. Given unequal probabilities of selection into the study, sampling weights were used to represent the population of pregnant women in the area [21]. Women were then followed up at three, six, 12, 24, and 36 months postpartum through in-person interviews. We used data from pregnancy (baseline), three, and 6 months postpartum. Of the 1154 women interviewed at baseline, 331 were not interviewed at the three-month follow-up visit, resulting in 823 women in our analytic sample.

Our final analytic sample included 823 women who had complete data for baseline, three, and six-month timepoints. A comparison of baseline characteristics among women with and without available data is Table 4 in Appendix. After applying population-representative weights, women were, on average, 26.62 years old and over half of women (51.38%) had at least a secondary education (Table 1). Most had at least one other child (61.81%) and lived with their mother-in-law (68.56%). Mean social support (MSPSS) score at 3 months was 4.74 (SD: 1.15). Thirty-seven percent of women were diagnosed with MDE (SCID) at baseline and 23.06% were diagnosed at 6 months postpartum. Mean PHQ-9 score was 8.73 at baseline (SD: 6.68) and 4.70 at 6 months (SD: 5.44).

The majority of women (89.44%) participated in chilla and over 80% were satisfied with their chilla experience (Table 1). Most participated for at least 26 days (60.68%) and almost 70% participated in all three components (relief from chores, additional female support, supplemented diet). Women who participated in chilla had, on average, more education, higher SES, were more likely to be in their first pregnancy, and more baseline social support. They also had fewer MDEs at baseline and lower depression symptom severity (Table 2). We also assessed whether baseline MDE (SCID) predicted chilla participation and chilla components. More women who were non-depressed at baseline participated in chilla, had longer chilla duration, and were more satisfied with their chilla experience (Table 5 in Appendix).

In unadjusted models, chilla participation was inversely related to both MDE (Table 3 - Panel A, Model 1: OR = 0.40; 95% CI = 0.23,0.71) and symptom severity (Table 3 - Panel B, Model 1: Mean Difference (MD) = − 2.51; 95% CI: − 4.20,-0.82). After adjusting for baseline covariates, we found that these relationships hold for both MDE (Table 3 - Panel A, Model 2: OR = 0.53; 95% CI = 0.33,0.86) and symptom severity (Table 3 - Panel B, Model 2: MD = -2.03; 95% CI: − 3.10,-0.96). We then controlled for social support and, although our estimate for chilla becomes slightly less precise, chilla remains independently related to MDE (Table 3 - Panel A, Model 3: OR = 0.56; 95% CI = 0.31,1.03) and symptom severity (Table 3 - Panel B, Model 3: MD = -1.54; 95% CI: − 2.94,-0.14). We conducted sensitivity analyses to disentangle potentially different associations of chilla components and PPD, adjusting for other components, and found no significant differences among the components (Table 6 in Appendix); however, this is likely because the majority of women who participated in chilla participated in all three components.

To examine whether the association between chilla and depression differs by baseline MDE, we stratified models by baseline MDE diagnosis. In multivariable models that also included social support, we found that the association between chilla and MDE at 6 months was driven by those not clinically depressed at baseline (Table 3 - Panel A, Model D: OR = 0.44; 95% CI = 0.20,0.97) as opposed to those who were clinically depressed at baseline (Table 3 - Panel A, Model E: OR = 0.73; 95% CI = 0.39,1.34). Conversely, the association between chilla and symptom severity score was greater among those who had prenatal MDEs (Table 3 - Panel B, Model E: MD = -2.05; 95% CI: − 3.81,-0.49) compared to those who did not (Table 3 - Panel B, Model D: MD = -0.66; 95% CI: − 1.98,0.67).

Among women in rural Pakistan, we found that most participated in chilla and that most participated in all components for the majority of the traditional 40 days. Those with more education, first-time mothers, and those with higher SES were more likely to participate in chilla, as well as those without MDEs and those with lower depression symptom severity at baseline. Among chilla participants, those without baseline MDEs had more positive chilla experiences.

We found that postpartum practices hold promise for benefiting maternal mental health. Chilla was inversely associated with depression at 6 months postpartum above and beyond social support. We saw a larger effect on symptom severity (PHQ-9) among those prenatally diagnosed with depression, indicating that postpartum practices may be particularly beneficial for women with already-vulnerable mental health. We also saw a negative association between chilla participation and SCID diagnosis at 6 months among those without MDEs at baseline, meaning that chilla participation may be protective against the future development of depression. However, we did not see an association among those with prenatal MDEs. Thus, chilla participation may help prevent new onset of PPD, but may not be enough to lift women out of pre-existing depression. Our findings may point to why prior research has been inconclusive in understanding the relationship between postpartum practices and maternal depression, as such practices may have differential effects depending on both the depression measure used (i.e., MDE or symptom severity) and women’s depressive history. Additionally, our findings may be specific to using the PHQ-9 and SCID. We used the PHQ-9 so that it would be generalizable across the prenatal, postnatal, and longer-term postpartum periods given the goal of the larger study. We chose to use the SCID-IV given its cross-cultural validation and use in this study setting [24, 30, 31]. Thus, studies using other instruments to assess symptom severity and MDE may generate different results.

The negative association between participation in chilla and PPD in our study sample is consistent with prior findings in Pakistan, which found that the relative risk of depression among mothers who participated in chilla compared to those who did not participate was 0.4 (95% CI 0.3,0.6) [20]. Studies conducted in China, Vietnam, and Malaysia have also shown negative associations between participation in traditional postpartum practices and PPD. [14, 32, 33] Most often, social support is put forward as the mechanism linking traditional postpartum practices to reduced risk of PPD. For example, a study of mothers in Hong Kong found that the cultural practice of peiyue was associated with both better social support and lower risk of PPD. [32] After controlling for social support, peiyue was no longer associated with PPD, suggesting that social support is the “active ingredient” of peiyue [32]. However, in our study, we found that controlling for perceived social support did not appreciably attenuate the effect of chilla on depressive symptoms. While participation in chilla may be partially reflective of social support, we found that it was significantly associated with fewer depressive symptoms among women who were depressed during pregnancy regardless of their level of perceived social support. Thus, our results suggest that chilla captures more than perceived social support and may affect mental health through mechanisms other than social support.

It is likely that chilla affects mental health through multiple mechanisms, which are not fully captured by imagining chilla as a unidimensional proxy for social support. For example, mothers who are unable to participate in traditional postpartum practices may experience heightened role conflict, undermining their self-esteem [1]. Stern and Kruckman (1983) theorized that six primary elements influenced the presence or absence of PPD: a postpartum social support structure, recognition of vulnerability of the new mother, a mandated rest period, social seclusion, recognition of the role transition and social status of the new mother, and assistance with household tasks from female family, friends and midwives [1]. These specific cultural factors, they propose, buffer new mothers from PPD. In the Pakistani context, the postpartum practice of chilla encompasses these six primary protective elements, which suggests that we should consider the multiple dimensions of chilla and understand the full experience of chilla.

Our findings emphasize the importance of understanding informal postpartum practices for mothers that are central to the postpartum period in Pakistan and other countries [10]. Given how common these practices are in Asia and around the world, there is a specific need to better understand the role of postpartum practices on maternal mental health and well-being during this transition period for mothers and how we can incorporate these practices into mental health interventions [3‐5]. There is also a need to identify the characteristics that predict mothers’ participation in these practices so that programs and interventions can focus on those less likely to participate. Supporting women’s participation in chilla aligns with a global interest in using pre-existing structures to provide additional social support to women experiencing depression in the pre and postpartum period [34, 35]. Enhancing the practice of traditional postpartum practices can offer a sustainable and culturally-appropriate public health recommendation, potentially in both high and lower income settings.

Our study has several strengths. First, it includes a longitudinal analysis from a population-representative sample. Second, we used standardized measures of depression symptom severity and MDE that have been validated in our target population. Third, it is the first, to date, to examine the effects of chilla participation, and one of the first to examine the effects of traditional postpartum practices, independently of social support on maternal mental health. Some limitations warrant discussion. First, as chilla and other postpartum practices are rarely quantitatively measured, it is not clear if our measurement of chilla captures the full experience. There is a need to more fully understand the quality and temporality of women’s chilla experiences. Second, due to the lack of variability among chilla components, we were unable to decipher what components may be driving the association we see with maternal depression besides social support (i.e., diet or length of participation). Third, we were unable to assess if there is something unique about women who did not participate in chilla, which was only 11%. Although our models control for social support and chilla components, there may be other elements of women’s social environment that we are unable to capture. We do see that women with less education, lower SES, and higher depression severity are less likely to participate in chilla. Thus, what presumably prohibited women from participating in chilla that we were unable to measure may be driving the association between chilla participation and PPD. Fourth, it is possible that child sex can affect whether a mother participates in chilla, as mothers of female children are less likely to receive social support in the absence of chilla [36]. However, our sample size was too small to separately assess female and male births. Future research should consider the potential for effect measure modification by child sex. Lastly, we may have selection bias in our analytic sample, as women absent at three-month data collection may have been participating in chilla. However, given that chilla typically does not last more than 40 days, this is unlikely at the three-month follow-up.