This prospective study of pregnant women in Kenya demonstrated a relatively high prevalence of any perinatal depression and antenatal depression. This is one of few studies to evaluate depressive symptoms longitudinally in both the antepartum and postpartum periods. In this cohort, more women had any perinatal depression (32.9%), which is mainly driven by the large proportion of women who had antenatal depression (28.5%) while only 9.1% of women had postpartum depression, which is lower than anticipated given previous studies. One study of pregnant Sudanese women established a similar postpartum depression prevalence of 9.2% at 12 weeks postpartum21, but most African studies have reported a higher prevalence of postpartum depression22. For instance, the estimated postpartum depression rate in Kenya has been cited as ranging from 13-18.7% in previous studies2,8,22. Neighboring countries of Tanzania and Ethiopia demonstrated similar postpartum depression rates, 12% and 20.9%, respectively, while Uganda was most notable for its 43% postpartum depression rate when using EPDS as the assessment tool22. Of note, 51.9% of women with postpartum depression in our study had antenatal depression. This is consistent with Gelaye et al.’s finding that a little more than half of women who have postpartum depression subsequently were found to have depressive symptoms present in the antenatal period4. Other studies have similarly demonstrated that antenatal depression is a significant risk factor for the development of depression in the postpartum period2,20. Taken together, all of these results highlight the importance of screening for depression in the entire perinatal period and emphasize the need to include the antepartum period in perinatal depression research, particularly given its impact on both maternal, pregnancy and infant outcomes.
It is evident that rates of perinatal depression are affected by psychosocial stressors and the broader social context of a woman’s life. Experiencing abuse was the strongest risk factor for perinatal depression. Other studies corroborate our findings and have demonstrated strong associations between abuse, particularly in the form of intimate partner violence, and increased risk of incident and persistent perinatal depression2,4,20. Consistent with other studies4,23,24, women with less than a primary education had a significantly increased risk of postpartum and persistent perinatal depression. Participants living with HIV had an increased risk for persistent perinatal depression. The association between HIV and perinatal depressive symptoms has been well-documented in both the antenatal and postnatal periods25,26. Finally, maternal unemployment was a significant risk factor for antenatal depression, with an even stronger association between maternal unemployment and persistent perinatal depression. Previous literature in high-income countries has provided evidence that maternal unemployment is a risk factor for perinatal depression, while a recent study in South Africa found a borderline association of antenatal depression associated with maternal unemployment22,27. More studies are needed to elucidate this relationship.
Accessing prenatal care is often challenging for adolescent women for a variety of systemic reasons, including social stigma and fear of judgment and stigma by health care providers28. Surprisingly, adolescent age was not significantly associated with any increased risk of developing perinatal depression. One reason for this finding may be that we defined adolescents as women aged 14-19 in this study, consistent with the WHO definition of adolscence29s. Thirty-eight (65.5%) of the 58 adolescents enrolled in this study were 19 years old. Previous studies examining the relationship between adolescence and perinatal depression often define adolescence as between ages 15-18, which is more consistent with legal and societal definitions of adolescence30–32. If we had utilized this aforementioned age range to define adolescence, our results may have been more consistent with established literature.
Becoming pregnant without the intent of having a/another baby at some time was associated with significantly higher risk of any perinatal depression and antenatal depression, but not postpartum depression. Surprisingly, we found that having an undesired pregnancy was in fact associated with a decreased risk of persistent perinatal depression after adjusting for a priori confounders. There have been a plethora of studies suggesting that undesired pregnancies are associated with an increased risk of perinatal depression23,33, which is consistent with most of our findings. There is some evidence that undesired pregnancy is associated with depressive symptoms in the first trimester, with waning significance over time 23,32. One U.S. study demonstrated that regardless of pregnancy intention, severity of depressive symptoms diminished in the postpartum period33. Additionally, our metric did not measure unintended pregnancies, which include both undesired and mistimed pregnancies, i.e., pregnancies occurring earlier than desired20,33. This decision to separate undesired pregnancies from mistimed pregnancies may have confounded our results since desired and mistimed pregnancies were lumped together as one measure while undesired pregnancy was a separate measure. Thus, our results may reflect either a true diminishment of depressive symptoms after childbirth despite initial lack of maternal desire for a/another child; a distortion by our inability to detect subclinical depressive symptoms in the postpartum period; and/or evidence of the complex relationship between depressive symptoms and maternal attitudes toward pregnancy.
In this cohort, a history of miscarriage was significantly associated with an increased risk of persistent perinatal depression but a decreased risk of any perinatal depression and antenatal depression on multivariate analyses. Having to deal with a previous pregnancy loss may fortify a woman’s capacity to handle the stress of, and even provide hope/enthusiasm for, a current pregnancy in the antenatal period. In the long term, the memory of a pregnancy loss in the face of a new maintained pregnancy with a subsequent live birth may predispose women to having a higher risk persistent perinatal depression if antenatal depression was present. These findings emphasize how crucial it is to longitudinally track maternal mood symptoms as maternal feelings and attitudes about pregnancy and motherhood change over time.
There is ample research that has documented the association between antenatal depression and increased likelihood of preterm birth4,5,34. Interestingly, we did not find any significant associations between antenatal depression and preterm birth. Additionally, antenatal depression was associated with a significant reduction in the risk of a woman’s infant dying in the immediate postpartum period in our univariate analysis. However, infant mortality was defined only up to 14 weeks of age, as compared to most research which extends mortality in children up to one year of age35. Due to this small sample size, an adequate multivariate analysis for an association between antenatal depression and infant mortality could not be conducted; therefore, this remains inconclusive in our study. Previous studies in rural Ghana and Ethiopia did not find any association with antenatal depression and infant mortality5.
Previous literature has focused how postpartum behaviors, e.g. supplementing infants’ diets instead of exclusive breastfeeding, attribute to the risk of having an ill infant with diarrheal or other infectious illnesses, which are the leading causes of under-five mortality in Kenya 4,8,34,36,37. However, our study focused on the relationship between antenatal depression and infant morbidity, and we found antenatal depression correlated with a higher risk of having an ill infant necessitating a clinic or hospital visit. One study conducted in Pakistan found an increased risk of ≥5 diarrheal episodes per year for infants whose mothers had antenatal depression38. Yet, this study emphasized that a unidirectional relationship between maternal mental health and infant well-being cannot be assumed since poor infant health may act as a maintaining factor for a mother’s depressive symptoms38. Our study suggests that in utero exposure to maternal antenatal depression may have lasting effects on infant health. Physiologic changes related to antenatal depression, i.e. higher levels of maternal cortisol, alterations in the maternal hypopituitary axis and catecholamine function, could predispose infants to developing less robust immune systems and may lead to higher rates of illness5,35. Further work investigating the relationship between depressive symptoms preceding delivery and infant outcomes is necessary to further understand how infants and their health trajectories are impacted by mothers with antenatal depression.
Typically, depression is characterized by social withdrawal and isolation, leading to reduced communication and weakened interpersonal relationships. As a group, mothers with antenatal depression and those with persistent perinatal depression sent significantly less total and less long SMS messages, respectively, compared to their counterparts without depression. Moving forward, tracking messaging behaviors could allow for the development of a mechanism to screen for depressive symptoms amongst women with a lower SMS response rate. Another feature of an mHealth system could provide more tailored messages to women with depressive symptoms to provide better individual support and psychoeducational messages. We currently have integrated this approach into our large, randomized trial of this SMS program. Women with antenatal and postpartum depression now receive specific tracks of messaging that address their mental health condition longitudinally and attempt to engage women around these symptoms. Further qualitive analysis of the message content would enhance our understanding of how mothers with depressive symptoms communicate via this platform as well as allow for the incorporation of diagnostic tools and therapeutic modalities. Additionally, eliciting more feedback about women’s experiences with the messaging platform to identify barriers to use, likes and dislikes of this communication modality, or other general reflections about interactive SMS messaging would provide valuable information about utility and customization of such mHealth interventions in the future.
Women with persistent perinatal depression had a non-significant trend for fewer nurse consultations than women with antenatal or no depression. All women who initiated a nurse consult for any reason found them helpful. There was little difference in the topics for which women with or without depression consulted the study nurse, except women with antenatal depression were less likely to consult the nurse about infant health questions. A similar trend was found amongst mothers whose infants were seen at a clinic/hospital for an illness or whose infants were admitted after delivery. In this group of women with ill infants, mothers with antenatal depression were significantly less likely to have consulted the nurse about their infant’s health. This raises questions about the ways in which antenatal depression affects prenatal and infant care engagement and overall treatment-seeking behavior. Further investigation into health-seeking behaviors of women with perinatal depression, especially antenatal depression, is warranted, and there should be a continued effort to explore how mHealth interventions can benefit prenatal and postnatal engagement of women regardless of their depression status.
Our findings should be interpreted in the context of our study’s limitations. One limitation is selection bias. There were 1039 eligible women and 800 women enrolled in the study. Those who were eligible but did not enroll mainly noted they did not feel they had sufficient time to participate in the study or were otherwise busy. One-hundred forty-three women (18%) were lost to follow-up before their exit visit, and 83 women (10%) attended their exit visit by phone and did not complete an EPDS survey because phone-based collection of EPDS data was not approved. A total of 226 women (28%) were therefore missing postpartum EPDS data. It is important to note that 64 of the 143 women lost to follow-up had antenatal depression and another 29 (out of 83) women who had some follow-up data but did not complete a second EPDS also had antenatal depression (see Supplementary Table A in Appendix). Thus, 93 (41.2%) of the 226 women who do not have follow-up information had antenatal depression as compared to the 163 (28.5%) of women in the retained cohort who had antenatal depression (see Supplementary Table B in Appendix). This level of missingness and difference between those with vs. without follow-up data means our findings on postpartum depression are limited by selection. Additionally, sample sizes for some outcomes were small, i.e., persistent perinatal depression, WLWH and infant mortality, making it difficult to draw robust conclusions about these groups/outcomes.
It is important to note that we did not conduct diagnostic evaluation to identify depression; we used the Edinburgh Postnatal Depression Scale (EPDS) to screen for depressive symptoms. Only two EPDS scores were collected for each participant, at baseline and at the end of the study at 14 weeks. Follow up time after delivery was limited to one timepoint of data collection, measuring only immediate maternal and infant outcomes. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) only allows for the diagnosis of a major depression with peripartum onset when onset of symptoms occurs during pregnancy or in the 4 weeks following delivery39. Our second EPDS score was not collected until 14 weeks postpartum, outside of the DSM-5 timeframe for diagnosis. Thus, our assertion of perinatal depression in this study relies on a screening tool, which is suggestive of an underlying pathology of depression but not a definitive diagnosis as outlined by the DSM-5 criteria. However, the EPDS has been validated for use in Kenya40 and literature suggests that the presence of depressive symptoms negatively influences maternal and child health outcomes in a manner similar to clinical depression4.
Finally, the majority of this analysis relies on self-reported answers to questions. Self-reported data is subject to bias (i.e., reporting bias and social desirability bias), which may result in exaggerated and/or understated answers. Depression itself may contribute to memory bias41. Subsequently, SMS message data and self-reported data are not comparable but should reflect similar trends.