Antenatal depression and its potential causal mechanisms among pregnant mothers in Gondar town: Application of structural equation model

Background: Various forms of life stressors have been implicated as causes of antenatal depression. However, there is a lack of understanding of which forms of stress lead to antenatal depression and through what mechanisms. Modeling stress processes within a theoretical model framework can enhance an understanding of the mechanisms underlying relationships between stressors and stress outcomes. This study used the stress process model framework to explore the causal mechanisms underlying antenatal depression in Gondar, Ethiopia. Methods: Questionnaires, using an Online Data collection Kit (ODK) tool were administered face-to-face in 916 pregnant women in their second and third trimesters. Pregnant women were included from six randomly selected urban districts in Gondar, Ethiopia during June and August 2018. The Edinburgh Postnatal Depression Scale (EPDS) was used to screen for antenatal depression. A Structural Equation Model (SEM) was employed to explore the direct, indirect, and total effect of stressors and mediators of antenatal depression. Result: Sixty-three participants (6.9%) reported symptoms of depression. Of these, 16 (4.7%) and 47 (8.1%) were in their second and third trimesters, respectively. The SEM demonstrated several direct effects on antenatal depression scores including unplanned pregnancy (standardized β = 0.15), having a history of common mental health disorder (standardized β = 0.18) and fear of giving birth to the current pregnancy (standardized β = 0.29), all of which were associated with a higher depression score. Adequate food access for the last 3 months (standardized β =-0.11) was associated with decreased depression score. Social support (β =-0.21), marital agreement (β =-0.28), and partner support (β =-.18) appeared to partially mediate the link between the identified stressors and the risk of antenatal depression. Conclusion: Both direct and indirect effects contributed to higher antenatal depression score in Ethiopian women. The three psychosocial resources namely marital agreement, social and partner support, mediated reduced antenatal depression scores. Early screening of antenatal depression and enhancing the three psychosocial resources would help to improve maternal resilience. The current study was conducted in Gondar town, which is one of the administrative zones of Amhara Regional State, Northwest Ethiopia. Gondar town is in the Northern part of the Amhara region at 747 km away from Addis Ababa and 170 km from Bahirdar (the regional capital city). Gondar town has 12 kebeles (the smallest administrative units in the country) and in 2017/2018, the number of pregnancies in the town was expected to be 6450 [48, 49]. The town has one government-operated referral hospital, more than eight health centers, and more than 15 private medical clinics [50]. The study population included pregnant women living in the randomly selected districts and in their second and third trimester of pregnancy. A house-to-house visit was conducted to identify pregnant women who were willing to participate in this mother- child cohort study. If nobody was found at home during the initial recruitment visit and after three attempts, they were non-respondents. Identified participants were recruited and were followed until 12 weeks post-delivery. Ethical approval was obtained from the Social and Behavioral Research Ethics Committee (SBREC) of the Flinders University [51] and the Institutional Review Board of University of Gondar. A support letter was obtained from Gondar town mayoral office and the respective kebeles administration offices. Participants of the study were informed about the purpose, objectives, their right to decline participation or withdraw their participation. A written consent was then obtained. Privacy and confidentiality were maintained throughout the study. Women who were found to be seriously ill and fulfilled the following criteria were referred to University of Gondar Specialized Hospital Psychiatry unit for further diagnosis and treatment: an overall Edinburgh Postnatal Depression Scale (EPDS) score of 13 (those with ≥17 were excluded from the study for further follow up) and those who had a score 1, 2, 3 on item ten (a question about thoughts of self-harm) [52]. Structured and pre-tested electronic questionnaires were administered face-to-face in pregnant women aided by an online, Open Data collection Kit (ODK) application tool [53]. Open data collection kit is an application developed by the ODK community for collecting, managing, and using data in resource limited countries [54]. The prepared questionnaire was designed on an excel spreadsheet, converted to XLS format online, and checked for its validity using Enketo (a preview provided by ODK). The validated form was uploaded on a Lenovo 7 tablet. During collection, data were stored on the Google cloud platform. Nine qualified and registered nurses were trained as data collectors and were each provided with a Lenovo 7 tablet to administer the questionnaire to the participants. After completion of each questionnaire, the data collectors uploaded the data to Google Cloud and the principal investigator then directly downloaded the data from the system. The electronic based data collection was helpful in maintaining the quality and completeness of the data. The questionnaire collected socio-demographic information such as: age; sex; educational status (no formal education, grade 1–8, grade 9–12, diploma and above); income (low, medium, high); and marital status (single, married, separated). Information on maternal characteristics was also collected, including pregnancy intention (planned, unplanned); gestational weeks; previous history of either low weight, preterm or still birth; and previous history of a caesarian section delivery. Finally, the questionnaire collected information on psychosocial and behavioral characteristics, such as: social support (good, poor); partner support (always, most of the time, some of the time, rarely); stress coping ability (very rarely, rarely, sometimes, most of the time); coffee drinking (daily, sometimes, never); and cigarette exposure (yes, no). Antenatal depression was measured using the Edinburgh Postnatal Depression Scale (EPDS) developed by Cox and colleagues [52] and adapted for use in an Ethiopian context [55]. The EPDS, which is the most commonly used screening tool for antenatal depression [56–60], is a brief screening tool for symptoms of emotional distress during pregnancy that contains 10 specific questions with four Likert scale response options (most of the time, sometimes, not often, never) and is intended to measure the distress that pregnant women have experienced over the previous week. It is a simple and free to use tool, can be scored by simple addition and has been validated in urban settings of Ethiopia [61] with a sensitivity and specificity of 84.7 and 77.0%, respectively. The validated cut off value for possible depression in urban population in Ethiopia was 12 [43, 62, 63]. In the current study the EPDS demonstrated high reliability for the single construct of distress with an internal consistency (α) of 0.74. The Oslo Social Support Scale (OSSS-3) [64] was used to measure maternal social support during pregnancy. Although the tool has not been validated in the Ethiopian context, it has shown a good utility in various studies [62, 65]. OSSS-3 has three items measured by a few Likert scales, which are summed to 14 points and categorized as ‘poor’ if the total score is less than nine and ‘moderate’ to ‘strong’ support if the score is 9–14. In this study OSSS-3 demonstrated a high reliability for social support with an internal consistency of α = 0.76. Partner support was assessed by a question “My husband helps me a lot” with five response scales, “Always”, “Most of the time”, “Some of the time”, “Rarely”, and “Never”. Marital agreement was assessed by a question “How often do you discuss and agree with your husband in day to day life?” with a response category, “Most of the time”, “Some of the time”, “Rarely”, and “Never”. The women’s Middle-Upper Arm Circumference (MUAC) tape was used to measure nutritional status. MUAC is a validated and recommended tool for measuring nutritional status during pregnancy, with cutoff scores of 18–22 as ‘normal’ and 22.5 to 31 as ‘underweight’ [66]. Women were asked if they participated in moderate-intensity physical activity such as brisk walking, dancing, gardening, and usual housework for 2 to 3 h per week [67]. Exposure to cigarette smoking during pregnancy was assessed by the question, “Have you been smoking since your pregnancy or has there been anybody who smokes near you in your home or in your workplace?” [15]. To assess coffee exposure, we asked “How often do you drink coffee after your pregnancy?” if her answer was “daily” or “sometimes in a week”, she was labeled as exposed to coffee drinking and if not, she was labeled as non-exposed [68, 69]. Women’s health condition was assessed using the question “How do you rate your daily general health condition?” with response options of “Very good”, “Good” or “Poor”. A women’s stress coping level was assessed using the four customized internally-consistent coping subscales of the Perinatal Coping Inventory (PCI-4), which was specifically developed for pregnancy [70]. Coping styles within this tool included: (1) Preparation for motherhood, “planned how you will handle the birth” (2) Avoidance “avoided being with people in general” (3) Positive appraisal “felt that being pregnant has enriched your life” and (4) Prayer “prayed that the birth will go well”. Women were asked to report how often they used each of the above coping styles and their response was recorded using a 4-point Likert scale; 0 (Never), (1) rarely, (2) sometimes, (3) most of the time [71]. In this study, PCI-4 demonstrated a moderate reliability with an internal consistency of α = 0.50. The sample size calculation was based on the estimated effect of perinatal depression on adverse infant health outcomes. To calculate this, we used a double population proportion formula in Epi Info version 7 [72] with the following assumptions: 95% confidence level, 90% power, an exposed to non-exposed ratio of 1:2, a prevalence of underweight among those free from common mental disorder of 25%, and a difference of 1.5. A total sample size of n = 809 was estimated which was then increased by 20% to account for expected losses to follow up. The final sample size was therefore estimated as n = 970. Completed data were downloaded from the Google Cloud platform in Excel spreadsheet form, checked for completeness and imported to Stata version 14 (StataCorp, USA) for further cleaning and analysis. Descriptive statistics included mean, median, proportion/percentage, interquartile range, and standard deviations as appropriate. A chi-squared test was used to test for crude associations between the categorical stressors and evidence of depression based on a cut of score of 12. A Structural Equation Model (SEM) was constructed that reflected the stress-process model framework and which explored the direct and indirect relationships between the independent (stressors) and the dependent (antenatal depression) variables. This allowed us to assess the strength of the hypothesized direct and indirect causal paths [73, 74]. In order to better fit the measurement model for depression, the measurement items for the depression scale were parceled into three categories using a random based parceling algorithm. Parceling allows for recategorizing multiple items of a scale in order to get better model fit and convergence [75, 76]. The first parcel contained the EPDS items 1, 4, and 9. The second parcel contained the EPDS items 6, 7, and 8. The third parcel contained the EPDS items 2, 3, 5, and 10. Since the subsequent parcels displayed evidence of non-normality we used the Satorra-Bentler scaled chi-squared test when estimating model fit since this is robust to non-normality [77]. The potential stressors and hypothesized causal paths were selected based on prior subject knowledge (which informed the questionnaire). In addition, a multivariate mixed effects regression analysis was performed to help determine variables suitable for inclusion in the SEM conditioning for socio-demographic, maternal obstetrics and psychosocial factors that were significantly associated (P < 0.05) in bivariate mixed effects regression. Stressors included income, attitude to the current pregnancy, history of common mental disorders, and a fear of giving birth. Potential mediators for the association included social support, marital agreement and stress coping. Figure 2 shows the hypothesized causal pathways. Model fit was assessed for the hypothesized model and an iterative approach was used to modify the model through adding and removing paths until a theoretically supported and a statistically well-fitted model was obtained. A model that was over identified, recursive, simple, theoretically meaningful, and the best fit for the data was retained and interpreted [78]. Measurement model for elucidating the stress process for antenatal depression (N = 916), Gondar town, Ethiopia, 2018 The overall fit of the model was assessed using the following tests and recommended limits: Sartorra-Bentler Chi-squared test of fit (p > 0.05); Tucker Lewis Index (TLI) and Comparative Fit Index (CFI) value ≥0.90; and Root Mean Square Error of Approximation (RMSEA) ≤ 0.08 [79]. The direct, indirect, and total effects of the stressors on antenatal depression were reported in the form of standardized beta coefficients. Estimated effects for which p < 0.05 were considered as being statistically significant.