Determinants of prenatal depression among women attending the antenatal clinic at a referral facility in Mombasa County,

Despite prenatal depression being a public health burden Background: and the major predictor of postnatal depression, it has not received as much attention as postnatal depression in research and policy globally. There is limited evidence on the factors associated with prenatal depression and therefore understanding these factors will inform the design of specific interventions and formulation of guidelines for the effective prevention and control of prenatal depression particularly in high-risk regions. A hospital-based case control study design was used to identify Methods: the determinants of prenatal depression among 170 women attending an antenatal clinic. Prenatal depression was assessed using the Edinburgh Postnatal Depression Scale (EPDS). A semi-structured questionnaire was administered to collect data on the socio-demographic, social network and family, lifestyle and obstetric characteristics of the participants. All eligible cases were enrolled into the study while a simple random sample of depression-free women attending the antenatal clinic were enrolled as controls. The relationship between the predictors and prenatal depression was evaluated by logistic regression. In the multivariable analysis, only marital status (adjusted odds Results: ratio (aOR)=17.1; 95% confidence interval (CI):4.0-73.0), occupation (aOR=2.4; 95% CI:1.4-4.2), domestic violence (aOR=18.3; 95% CI: 5.7-58.7) and  social support (aOR=0.2; 95% CI:0.05-0.8) were identified as significant determinants of prenatal depression. Marital status, occupation, domestic violence and lack of Conclusion: social support were the major predictors of prenatal depression in this setting. There is therefore need to implement screening for prenatal depression among pregnant women in health facilities as part of the routine antenatal care package, establish social support networks and spaces to provide an avenue for the prenatally depressed women to meet, share challenges and coping mechanisms and revise the government policy on sexual and gender based violence (SGBV) so as to strengthen efforts towards elimination of all forms of SGBV. 1,2 1


Introduction
The prevalence of depression in women is about 20% with pregnancy increasing the susceptibility to depression 1 . Depression related to child bearing can develop either during pregnancy (prenatal depression), after birth (postnatal depression) or both (perinatal depression) 2 . Prenatal depression refers to a form of clinical depression which occurs during pregnancy and is characterized by chronic anxiety, insomnia, guilt, fatigue, irritability, forgetfulness, headaches and isolation 3 . Despite prenatal depression being a significant health problem, regrettably, it has received less attention than postpartum depression 4-6 . This is partly attributable to misconceptions about existing socio-cultural structures that could shield one from mental disturbance during this period 4 . Besides, more attention is paid to the physical health of the mother and fetus than on maternal mental health during pregnancy, with a propensity to dismiss emotional episodes as the result of hormonal imbalances. Thus, depression may persist silently during this period 7 .
Besides prenatal depression being a major determinant of postnatal depression, it is also a risk factor for adverse maternal and fetal outcomes such as premature births, preterm labor, low birth weight and poor infant feeding patterns [8][9][10] . Moreover, prenatal depression is associated with impaired neurocognitive and socio-developmental disorders in the offspring such as poor motor and regulation skills, anti-social behavior and increased risk of depression and attention problems 11 Additionally, higher health expenses, poor immunization rates and frequent hospitalization have also been reported in children who are born to depressed mothers 12 .
Predictors of prenatal depression can be categorized into three domains: social, psychological and biological risk factors. Social risk factors comprise low socio-economic status, lack of social support and stressors such as economic deprivation and unplanned pregnancy 13 . Women from a low socio-economic class are likely to have fewer financial resources which may be insufficient to meet the increasing financial demands of pregnancy and this may result to prenatal depression 12 . Likewise, women who lack social support are likely to have little emotional support from their spouses, family and friends and this can lead to social instability which subsequently leads to prenatal depression 14 .
Gestational age, maternal age, genetic and hormonal susceptibility and obstetric complications are some of the biological risk factors for prenatal depression 15 . Women who are in the first trimester of pregnancy are likely to have higher depression rates than those in the third trimester that are partly ascribable to the first trimester pregnancy symptoms like fatigue, nausea, food aversions and heartburn, which most women find difficult to cope with 16 . A history of stillbirth or miscarriage can be traumatic and may result in anxiety or depression 14,17,18 . The main psychological and psychiatric predictors of prenatal depression are history of mental or anxiety disorder 19 .
Given the paucity of research on prenatal depression in Kenya there is need to understand the predictors of prenatal depression with a view to informing the design of specific interventions and formulation of guidelines for the effective prevention, control and surveillance of prenatal depression particularly in high-risk regions in Kenya. Consequently, the objective of this study was to investigate the sociodemographic, lifestyle, obstetric and social network and family determinants of prenatal depression among women attending the antenatal clinic (ANC) at a referral facility in Mombasa County, Kenya.

Study design and setting
A hospital-based case control study design was used to identify the determinants of prenatal depression. The choice of study design owes to its suitability in the investigation of rare outcomes that may be missed through random sampling. Although a population-based study would have been more optimal, a hospital-based design was selected due to the ease of recruitment of pregnant mothers (cases and controls) presenting to the antenatal clinic for care.

Study population and eligibility of participants
The study population comprised pregnant women ≥15 years attending routine ANC at CPGH during the data collection period between April and June 2019. All pregnant women who consented to participate were included in the study. Women who had already been diagnosed with depression prior to pregnancy or had concurrent chronic illnesses were excluded from the study.

Case definition and recruitment
A case was a pregnant woman aged ≥15 years residing in the hospital's catchment area and had been attending the ANC clinic at CPGH during the two-month study period and registered an Edinburgh Postnatal Depression Scale (EPDS) score of ≥13 21 . All cases meeting this definition were prospectively recruited into the study. Considering prenatal depression was not routinely screened for during ANC visits, the EPDS tool was completed in the triage room just before administration of routine ANC services.

Control definition and recruitment
Controls were pregnant women similarly defined as cases but with EPDS scores of <13 22 presenting to the same ANC clinic for care. Owing to the comparably large number of controls, these were simple randomly sampled and frequency-matched to cases by day of presentation.

Sample size determination
The appropriate sample size was estimated as per Kelsey et al. 23 for case control studies: n 1 = number of cases and n 2 = number of controls. p 1 = proportion of cases with previous history of intimate partner violence (IPV) or domestic violence (primary exposure) and p 2 = proportion of controls with previous history of IPV (set at 0.40) 24 . Of note, Z α/2 (1.96) is the value for two-tailed confidence level of 95% and Z 1-b (-0.84) is the value for a statistical power of 80%. The odds ratio for the IPV-prenatal depression association was hypothesized to be 3.41 24 . To enhance statistical power, a ratio of 1:4 (cases to controls) was employed. Based on these figures, a total sample size of 34 cases and 136 was derived.

Study variables and method of measurement
The predictor variables were collected using a semi-structured questionnaire (see Extended data) 25 and included demographic characteristics, lifestyle, social network and family risk factors and obstetrical factors. The questionnaire has not been validated. The demographic predictors included age, level of education, occupation and marital status. Social network and family related predictors consisted of social support and domestic violence. Lifestyle factors comprised smoking, use of alcohol and substance abuse. Obstetric factors were unplanned pregnancy, gestational age, history of still birth, history of miscarriage/pregnancy loss and parity. Table 1 shows the method of assessment of the study variables. Figure 1 displays the relationship between the predictor variables and outcome.

Ethical considerations
Approval to conduct the study was obtained from the Kenyatta National Hospital-University of Nairobi (KNH-UON) Ethics and Research Committee (P787/11/2018). Written Table 1. Study variables and their assessment.

Prenatal depression (nominal)
Denoted in binary form: Present or absent. This was assessed using the Edinburgh Postnatal Depression Scale which is a 10-item questionnaire which scores women's feelings and experiences of the last one week on a likert scale. The recommended cut offs for the English, version is ≥13. The sensitivity and specificity of EPDS in the African setting has been shown to be 94% and 77% respectively 22 . The EPDS scale has shown to be the most reliable instrument used for screening of antenatal depression in resource constrained settings because of the reported specificity, sensitivity and reliability 26   informed consent was secured from the participants prior to engaging in the study.

Minimization of errors and bias
Following data collection, the questionnaires were manually checked for completeness and accuracy. Data were then double entered by two data entry clerks into an Excel Spreadsheet, after which the resulting datasets were compared and revisions made accordingly. Interviewer bias was minimized by training the research assistants on the standard operating procedures (SOPs) to ensure consistency in elicitation of information from the respondents. In a bid to minimize recall bias information such as the gestational age, the obstetric history and number of antenatal clinic visits was abstracted from the mother and child health booklet.

Statistical analysis
The Excel dataset was exported to Stata version 13.0 (Stata Corporation, College Station, Texas, USA) for analysis.
Descriptive statistics (medians, means, standard deviations and inter-quartile ranges) were used to summarize continuous variables. Proportions and percentages were generated for categorical variables. In the univariable analysis, the effect of each predictor on the odds of prenatal depression was assessed using logistic regression at a liberal P-value (P≤0.20) 30 . Since inclusion of age as a continuous variable was insignificant in the univariable analysis, it was categorized into three groups: 18-25 years, 26-29 years and 30-34 years and reassessed for significance as a categorical variable.
Variables that were found to be significant in the univariable analysis were offered to a multivariable model, where a backward step-wise approach was used to eliminate variables from the model at P>0.05. Notably, the non-significant variables were eliminated from the model if their exclusion from the model did not result in a greater than 30% change in the effects of the remaining variables 30 . Two-way interactions were fitted between the remaining variables in the final model and their significance assessed. A Hosmer-Lemeshow test was used to assess the goodness of fit of the logistic model, with a P-value of > 0.05 being suggestive of a good fit.

Results
Descriptive statistics of the study data A total of 170 pregnant women (34 cases, 136 controls) were enrolled into the study.
A study flow chart illustrating the enrollment process is shown in Figure 2. Table 2 shows the descriptive statistics of the respondents.

Socio-demographic factors
The mean age of the respondents was 27.8 years (range: 18-44 years) with the mean age of cases being 27.0 years (range: 19-36 years) and that of controls being 28.0 years (range: 18-44 years). On the level of education, 44.1% (n=75) of the respondents had attained a tertiary level of education; this comprised 50.0% (n=17) of the cases and 42.7% (n=58) of the controls. Only 45.3% (n=77) of the respondents were employed, of which 26.5% (n=9) were cases and 50% (n=68) were controls.

Lifestyle factors
Respondents who reported to have consumed alcohol, tobacco or abused substances during the pregnancy period constituted 14.7% (n=25) of the population. Amongst these 29.4% (n=10) were cases and 11% (n=15) were controls.

Results of regression analyses
Of the factors assessed, only age, marital status, occupation, alcohol and substance abuse, unplanned pregnancy, gestational age, social support and domestic violence were associated with prenatal depression at P≤0.2. (Table 3). These variables were subsequently included in the multivariable model. In the multivariable analysis, only marital status, occupation, social support and domestic violence were shown to be significant predictors of prenatal depression at 5% significance level (   depression (aOR=18.3; 95% CI: 5.7-58.7) compared to those who did not experience domestic violence regardless of their marital status, occupation and social support level. Respondents who had social support had one-fifth the odds of prenatal depression (aOR=0.2; 95% CI: 0.05-0.8) in comparison to those who did not have social support controlling for their marital status, occupation and domestic violence experience.
The model had a good fit (P = 0.403).

Discussion
Marital status was shown to be a significant predictor of prenatal depression among women in the study with single women having higher odds of prenatal depression compared to those who were married. This finding is corroborated by other studies 17,31 . Being single as a result of a break up or abandonment by a partner can result in emotional problems and lack of social support from the male partners and this could lead to depression. Moreover, single parenting is stigmatized in the African culture and this may predispose one to antenatal depression 32 .
This study found that unemployed women had higher odds of prenatal depression compared to their employed counterparts. This finding is similar to that reported by a study in Italy 33 which found that participants who were unemployed had 2.17 times the odds of prenatal depression compared to those who were employed. Another study conducted among Japanese women revealed that employment is protective against prenatal depression 34 . Pregnant women who are unemployed have  fewer financial resources which may be insufficient to meet the increasing demands of pregnancy and this may predispose to prenatal depression.
Domestic violence was strongly associated with prenatal depression in this study. These findings are consistent with those from other studies, which described gender-based violence as an important predictor of prenatal depression with women who experience psychological, physical and sexual violence being prone to antenatal depression [35][36][37] . Domestic violence may cause physical injury with attendant emotional and psychological trauma that can lead to depression 38 .
Social support was protective against prenatal depression in this study. These findings support the results of other studies 14,32,39-41 . Social support from a spouse, friends or relatives provides psychosocial resources during pregnancy and these act as a cushion against difficulties that may be experienced during pregnancy hence can protect one from antenatal depression. In contrast, women who lack social support are likely to have little emotional support from their spouses, family and friends bringing about social instability which can heighten the risk of prenatal depression 42,43 .
Age did not significantly influence the likelihood of prenatal depression in this study. However, other studies have revealed that age is a significant predictor of prenatal depression owing to the fact that young pregnant women are likely to be financially unstable and may not be socially and psychologically prepared to cope with pregnancy demands and this may predispose them to depression 44 . Contrarily, some studies have demonstrated that older women are at a higher risk of developing prenatal depression as ageing increases the possibility of experiencing difficulties in conceiving and anxiety of experiencing obstetric complications. Besides, there is a high likelihood of experiencing stigma when you conceive later in life 45,46 .
Use of alcohol and abuse of drugs did not significantly predict a participant's probability of developing prenatal depression taking into account the effect of other variables. The findings of this study concur with the results of a study conducted among African American women 47 On the contrary, other studies have reported a significant relationship between alcohol and drug abuse and prenatal depression 31,48,49 . Alcohol being a depressant may inhibit neurotransmitters that regulate mood such as serotonin and norepinephrine, and this can lead to depression 50 .
After accounting for other variables, unplanned pregnancy was not found to be significantly related to developing prenatal depression in this study. This is partly ascribable to the fact that although an unplanned pregnancy might be unwanted at first, as the pregnancy progresses the shock associated with the undesired occurrence decreases and it becomes increasingly accepted, hence reducing the symptoms of depression 51 .
Other studies have reported a significant association between unplanned pregnancy and prenatal depression which is related to the fact that unplanned pregnancy is associated with lack of preparedness to deal with the financial and psychological demands of pregnancy 35,52,53 .
Gestational age was not found to be associated with prenatal depression after controlling for other variables. This finding is similar to another study that was conducted in KwaZulu-Natal 54 .
Nonetheless, other studies have demonstrated that women who are in the second or third trimester are less likely to be depressed antenatally compared to women in the first trimester owing to the fact that during the first trimester some women find it difficult to cope with pregnancy symptoms like nausea and food aversions and this can lead to depression 7,31,55 .
A couple of limitations were present in this study. There was likely to be differential recall of past exposures between the cases and controls with cases having better recall than controls. Moreover, cases were more likely to over-report their exposures and this could bias the effect estimates away from unity. The case definition of prenatal depression only relied on EPDS, which is a screening tool; this could have been supplemented by a clinical examination of the participants to improve on detection of prenatal depression. The results from this study are generalizable to similar settings in other low-and middle-income countries.

Conclusions
The present study showed that marital status, occupation, domestic violence and lack of social support were the major predictors of prenatal depression in this setting. There is therefore need to: (1) implement screening for prenatal depression among pregnant women in health facilities as part of the routine antenatal care package, (2) establish social support networks and spaces to provide an avenue for the prenatally depressed women to meet, share challenges and coping mechanisms and (3) review the government policy on sexual and gender based violence (SGBV) so as to strengthen efforts towards elimination of all forms of SGBV and improve the quality of life of the victims.

Data availability
Underlying data Harvard Dataverse: prenatal depression CPGH. https://doi. org/10.7910/DVN/QIFMOT. This project contains the following underlying data: • Prenatal depression_data.xlsx (containing responses to each question of the questionnaire from all participants).
• prenatal_depression_code.do (STATA commands file for determinants of prenatal depression evaluation).
Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).