Prevalence and predictors of postpartum depression by HIV status and timing of HIV diagnosis in Gauteng, South Africa

Background Postpartum depression (PPD) is a common mental health condition that can compromise the quality of life and functional capacity of mothers and cause health and developmental problems in children born to affected mothers. Objectives We set out to measure the prevalence of PPD comparing postpartum HIV-1 infected women with pre-pregnancy HIV care experience, newly diagnosed (in latest pregnancy) HIV-1 infected women and HIV negative women, and to identify predictors of major PPD among these women in a peri-urban clinic in South Africa. Methods We conducted a cross-sectional survey of 1151 adult (≥18 years) postpartum HIV-1 infected (690) and HIV negative (461) women who delivered up to 30 days before study enrolment, interviewed after their first post-natal visit (3–6 days post- delivery) at Midwife Obstetric Units in Gauteng, South Africa. PPD was categorised into no depression (CES-D 10 total score <5), low to medium depression (CES-D 10 total score ≥5 and <10) and major depressive symptoms (CES-D 10 total score≥10). We used ordered logistic regression to identify predictors of postpartum depression and report adjusted odds ratio (aOR) and 95% confidence intervals (CIs). Results Overall 288 (25.0%) women screened positive for postpartum depression, a total of 168 (14.6%) women had low to medium PPD and 120 (10.4%) had major PPD. A higher proportion of HIV negative women experienced PPD, 129/461 (28.0%) among HIV negative vs. 159/690 (23.0%) among HIV-1 infected. Among HIV positive women, there was no meaningful difference in PPD between newly HIV diagnosed and those diagnosed before the most recent pregnancy (aOR 1.3, 95% confidence interval (CI): 0.9–1.8). Predictors of PPD among HIV positive women were living with friends/in a house-share (aOR 0.5 for house-share vs. own home, 95% CI: 0.3–0.9), and attending antenatal care (ANC) for the most recent pregnancy (aOR 0.2 for ANC attendance vs. no ANC attendance, 95% CI: 0.0–0.5). Living with friends/in a house-share was also a predictor of PPD among HIV negative women (aOR 0.4 for house-share vs. own home, 95% CI: 0.2–0.8). Conclusions and recommendations Targeted symptom screening based on identified risk factors should be considered for postpartum women to increase PPD case-finding and referral to specialised social support services.


Objectives
We set out to measure the prevalence of PPD comparing postpartum HIV-1 infected women with pre-pregnancy HIV care experience, newly diagnosed (in latest pregnancy) HIV-1 infected women and HIV negative women, and to identify predictors of major PPD among these women in a peri-urban clinic in South Africa.

Methods
We conducted a cross-sectional survey of 1151 adult (�18 years) postpartum HIV-1 infected (690) and HIV negative (461) women who delivered up to 30 days before study enrolment, interviewed after their first post-natal visit (3-6 days post-delivery) at Midwife Obstetric Units in Gauteng, South Africa. PPD was categorised into no depression (CES-D 10 total score <5), low to medium depression (CES-D 10 total score �5 and <10) and major depressive symptoms (CES-D 10 total score�10). We used ordered logistic regression to identify predictors of postpartum depression and report adjusted odds ratio (aOR) and 95% confidence intervals (CIs).
Living with friends/in a house-share was also a predictor of PPD among HIV negative women (aOR 0.4 for house-share vs. own home, 95% CI: 0.2-0.8).

Conclusions and recommendations
Targeted symptom screening based on identified risk factors should be considered for postpartum women to increase PPD case-finding and referral to specialised social support services.

Background
Globally, maternal postpartum depression (PPD) is a major risk factor for non-obstetric postnatal maternal and child morbidity as well as mortality [1][2][3]. The proportion of postnatal women diagnosed with PPD varies widely from 10-15% in high-income countries to 30-60% in sub-Saharan Africa [4][5][6][7]. Untreated PPD can lead to chronic depression, disruptions of family and marital relationships, and can cause long-term health and developmental problems in children of affected women [5,8]. Early diagnosis and treatment of PPD improves the prognosis for both mother and child [9,10]. However, in high-income countries, almost 50% of PPD cases are undiagnosed, with only 30% of diagnosed cases receiving treatment [11][12][13]. In South Africa, an estimated 30-50% of women are likely to develop PPD [7,14,15], and although PPD-specific treatment estimates are unavailable, overall 75% of patients' mental health problem, including depression, do not access treatment in South Africa [16,17]. Infants are affected by their mother's mental health problems as they depend on their mothers for their developmental and nutritional needs [18,19]. Women with depressive symptoms report lower rates of breastfeeding compared to women without depressive symptoms (49% vs 61%) [19,20]. A meta-analysis of studies from developing countries found that children of mothers with PPD were 50% more likely to be underweight or stunted [18]. Furthermore, depressed mothers are often disengaged from their infants which may lead to slower cognitive development, behavioural problems and long-term psychological difficulties [19,[21][22][23][24]. The quality of infant-mother relationships appears to predict behavioural problems and disruption of cognitive abilities in children [23]. Evidence also suggests that children of depressed mothers also experience higher mortality rates. A study conducted in Taiwan, which examined mortality of children up to age five years, found that children were at a 47% greater risk of death if their mother was depressed [25]. Similarly, a cohort study in Ghana reported a nearly threefold increase in the risk of mortality by six months among infants born to women diagnosed with PPD [3].
Predictors of PPD in the South African context include relational factors such as marital status and partner's financial/ moral support, an unplanned/unwelcome baby, infant health conditions as well as the mother's educational attainment and employment situation [7,14]. Additionally, personal or family history of depression, recent stressful life events, high childcare stress, low self-esteem and neuroticism are important factors [24,26,27]. However, the impact of exposure to HIV care either through the HIV diagnostic procedures and ARV treatment before pregnancy in mitigating the risk of PPD among HIV positive women is not well described.
In general, rates of PPD are high among women living with HIV, with PPD rates of above 40% in high HIV prevalence settings [5,7,15,28,29]. Many women learn about their HIV infection during, or shortly after a pregnancy which can adversely impact on their mental health and compromise their participation in antenatal care (ANC) including the prevention of mother-to-child transmission (PMTCT) programs and antiretroviral therapy adherence [6,30,31]. Therefore, screening, referral for diagnosis and treatment of PPD among postpartum HIV-1 infected women is vital for the attainment of the HIV care and treatment goals for both the mothers and their infants [32,33].
It is unclear whether psychosocial support associated with HIV care/treatment positively impacts on the risk of PPD among HIV-1 infected with pre-pregnancy HIV care experience. In this study, we set out to measure the prevalence of PPD comparing HIV-1 infected women with pre-pregnancy HIV care experience, newly diagnosed (in latest pregnancy) HIV-1 infected women and HIV negative women, and to identify predictors of having depressive symptoms among these women.

Study design and population
This data was collected as part of the baseline survey of a randomised controlled trial (RCT) (Pan African Clinical Trials Registry: PACTR201809886446171) among adult (�18 years) women who gave birth up to 30 days before the date of enrolment at Midwife Obstetric Units (MOUs) based in Tshwane, Ekurhuleni and Johannesburg Metropolitan districts in the Gauteng Province, South Africa. New mothers were recruited consecutively via referrals from nurses at participating facilities and interviewed on the day of their first post-natal visit (scheduled three to six days after delivery). Study staff screened potential eligible women using study eligibility criteria, provided more information regarding the study, and obtained written informed consent using informed consent translated from English into Sotho and Zulu and administered in the participant's preferred language (English, Sotho or Zulu). Study enrolment was conducted from October 2016 to January 2018. The sampling for the RCT was stratified by HIV status, and participants were randomised to either Active Tracing, Active Tracing with Motivational Interviewing (MI) counselling support tracing approaches or standard of care. Six month follow-up for the RCT has been completed, and the 12 and 18 month follow-up is ongoing (Fig 1). This analysis includes cross-sectional data collected using baseline questionnaire at study enrolment, which included 690 HIV-1 infected and 461 HIV-1 uninfected women (Fig 1). Participant demographic, socio-economic and contextual data were collected using an interviewer-administered structured questionnaire translated in English, Sotho and Zulu and administered in the participant's preferred language (English, Sotho or Zulu).

Analytical variables
PPD was measured using the CES-D 10 scale, a 10-question four-point scale (scores range 0 to 3) that measures general depressive symptoms experienced up to 7 days before the interview date [34][35][36]. The total score ranged from 0 to 30 with higher total scores reflecting greater frequency of depression (Cronbach's alpha = 0.83). We created a variable for PPD categorised into no depression (CES-D 10 total score <5), low to medium depression (CES-D 10 total score �5 and <10) and major depressive symptoms (CES-D 10 total score�10) [37,38]. The CES-D 10 scale has been validated in South Africa and other low and middle-income countries (LMIC) and is used as a screening tool for PPD [35,[39][40][41].
Additional variables assessed included socio-demographic factors including age, highest education completed, marital status, employment status and work times (all-day, shift-work) and income-source. Perceived social support (PSS) was measured using a six-item scale in which participants indicated their overall level of satisfaction with available support given in each area [34]. Rating of overall satisfaction score for each item ranged from one to five. Total scores were categorised as either "high PSS" (score �26) or "medium PSS" (score �25) based on the distribution of scores in the sample. HIV knowledge was measured among HIV-positive women based on responses to 12 HIV knowledge questions, each with a possible score of 0 and 1 for incorrect and correct response respectively. Total knowledge scores were categorised as "Low" (score<11) or "Medium to high" (score �11) HIV knowledge. We also assessed factors related to ANC attendance, location and type of housing, whether the latest pregnancy was planned, new-born baby's gender, number of child dependents (own and others'), support during pregnancy, and expected childcare support postpartum. Ethics approval for this study was obtained from the Wits Human Research Ethics Committee (Medical) (HREC No. M151041). All personal identifiers were removed from the final analytic data.

Statistical analysis
Descriptive analysis was used to summarise participant characteristics at study enrolment. Categorical variables were tabulated using frequencies and percentages. Continuous variables were described using medians and interquartile ranges (IQR) where appropriate. We used Ordinal logistic regression to identify predictors of postpartum depression and the associated 95% confidence intervals (95% CI). As we could only identify prediction and not causation, factors identified with a univariate p-value <0.1 and priori variables of importance and predictors were included in the multivariate model. Significance level was set at 0.05 for the multivariate model. The Likelihood ratio test was used to test for adherence to the proportional odds assumption. We report adjusted odds ratio (aOR) and 95% confidence intervals (CIs). Data analysis was conducted using STATA version 14 (StataCorp, College Station, TX). Table 1 shows demographic and contextual characteristics of study participants which included 1151 postpartum women of a median age of 29 years (IQR: 25-33) at study enrolment, of which over 80% had at least some secondary school education. A total of 461 (40.1%) were HIV negative, and 690 (59.9%) were HIV infected. Just over half (364/690 or 52.7%) of the HIV infected women were diagnosed with HIV before their latest pregnancy, and among these 152/364 (41.8%) had a prior (in-pregnancy) HIV diagnosis and had prior experience of the PMTCT program. A total of 249 (21.6%) women were married, with a higher marriage percentage among HIV negative mothers (26.9%) compared to HIV positive mothers (125/690 or 18.1%). Overall, 78.4% of women were in non-marital relationships, with 44.6% cohabiting with a partner. Over two-thirds of the women lived with a partner or spouse (64.6%), and 46.5% lived in their own home. Nearly two thirds (63.3%) of the mothers were unemployed, with 19.5% not seeking employment. Unemployment rates were similar across HIV status. However, unemployed HIV-infected infected women (47.0%) were more likely to be searching for work compared to HIV negative women (27.5%). The majority of women reported their spouse/ partner to be their primary source of income (56.2%), with a higher proportion of HIV negative mothers relying on spousal/ partner support than their HIV-infected

Prevalence of PPD
Overall, 168/1151 (14.6%) of the sample had low to medium PPD and 120/1151 (10.4%) women screened positive for major PPD, 10.9% among HIV negative women and 10.1% among HIV positive women. Among HIV positive women, a higher proportion (12.1%) of those with a pre-pregnancy HIV diagnosis screened positive for major PPD compared to women with in-pregnancy HIV diagnosis (8.0%). Table 2 shows crude and adjusted estimates from the ordinal logistic regression model with 95% CIs for experiencing low to medium PPD and major depression respectively. HIV-infected. Among HIV-infected women, there was no difference in odds of experiencing PPD by timing of HIV diagnosis in the univariate analysis (OR 1.3, 95%CI: 0.9-1.8). Living with friends or house-mates was associated with lower risk of experiencing PPD (aOR 0.5, 95%CI: 0.3-0.9), as well as having attended antenatal care during the latest pregnancy (aOR 0.2, 95% CI: 0.0-0.5).

Predictors of PPD (multivariable analysis)
HIV negative. Among HIV negative women living with friends or house-mates was associated with lower risk of experiencing PPD (aOR 0.4, 95% CI: 0.2-0.8).
On further analysis in a multivariable analysis including all women, newly diagnosed HIVinfected women were less likely to be depressed compared to HIV negative women (aOR 0.7, 95% CI: 0.5-0.99), while there was no difference in odds of experiencing depression between HIV negative women and those with pre-pregnancy HIV diagnosis.

Discussion
This is one of the largest studies assessing PPD in the sub-Sahara Africa setting, and one of the few that looks at differences in PPD-based HIV status and timing of HIV diagnosis. Results from our study show that a quarter of women had depressive tendencies after delivery, but only 10.9% of HIV negative and 10.1% of HIV positive women showed signs of major PPD.
The risk of major PPD in both our HIV positive and negative women is lower than previously reported PPD prevalence in low and middle-income countries, including South Africa, and is much closer to rates found in high-income countries [4][5][6]. Variations in PPD rates possibly emanate from screening tool preferences as well as the varying definitions for PPD across studies, with very few elaborating on the severity of the reported risk of PPD. A large

Accommodation type
House  Risk of postpartum depression among postpartum women by HIV status and timing of HIV diagnosis prospective cohort study in the general postpartum population in Soweto, Johannesburg, reported PPD rates of 16.4% using a total score threshold of 20 on the Pitt Depression Questionnaire depressive symptoms [42]. Previous smaller studies in sub-Saharan Africa, which defined major PPD at total score thresholds ranging from 11-15 on the Edinburg Postnatal Depression Scale-10 (EPDS-10) [43], reported PPD prevalence of 33% in a mixed Zimbabwean cohort HIV-1 infected cohort [28], 45.1% was found among HIV-infected women in rural Mpumalanga (South Africa) [44]. However, our study consecutively enrolled postnatal women who delivered both at low-risk (MOU) and hospital facilities and the setting of enrolment (MOU) could have systematically excluded women with high-risk pregnancies who may also have greater risk of major PPD. There is also a possibility that symptoms were played down/ underreported due to the stigma associated with mental health disorders in many African cultures [45,46]. Women may worry that their child caring capacity may be called into question and they may also be reluctant to take up treatment interventions involving prescription drugs while breastfeeding [10,17,47]. On average 50% of depressed women often go undiagnosed, with only 20% of those diagnosed seek treatment [11][12][13]47].
We hypothesised that HIV-infected women with pre-pregnancy HIV diagnosis and hopefully some prior experience of HIV care would be more resilient and be at lower risk of major PPD than newly diagnosed HIV-infected women. However, we found no difference in major PPD by the timing of HIV diagnosis among HIV infected women, but found that women diagnosed with HIV during their latest pregnancy were less likely to experience PPD than HIV negative women. These results are contrary to results from two previous studies in Zambia and South Africa that found an increased risk of PPD among women who discovered their HIV-1 diagnosis during the last pregnancy [44,48]. Similar to previous studies, we found no difference in PPD by HIV status [28,33]. The impact of the PMTCT program and increased life expectancy of HIV-infected individuals may have lessened women's concerns about the risk of HIV transmission to their infants as well as fears of premature death [49][50][51]. Addressing depression in HIV negative women remains a crucial HIV-preventive measure as untreated depression is associated with negative coping behaviours (unprotected sex, having multiple and concurrent sexual partners, use of illicit substances) that increase the risk for HIV [52].
Predictors of PPD identified among the combined sample of HIV-infected and negative mothers include living with friends or sharing a house with others, antenatal care attendance and timing of HIV diagnosis when compared to HIV negative women. In many settings including South Africa, older female family members experienced in child care are customarily tasked with providing childcare support and guidance to new mothers. Depending on the context, these family members may come and stay with the new mother in her own home, or the new mother may move to her family home to access this support. Women who live with friends or sharing with others may, therefore, have lower expectations of this type of support. Antenatal care engagement may expose women to services that may mitigate some of the pregnancy, labour and childcare related stressors that may contribute to the risk of PPD.
Perceived social support has been previously reported as a critical factor in mitigation of PPD [5,28,53,54], but it wasn't shown as important in our cohort of women which may need to be explored further to understand what factors could be helping them to cope positively with perinatal related stressors, as well as HIV among our HIV infected women.
The cross-sectional study design limits the interpretation of the study results, and causal associations cannot be inferred. Depressive symptoms were self-reported using a validated tool, but participant recall and social desirability bias cannot be excluded. Although the CES-D 10 scale is a screening tool and not a diagnostic interview, it has been shown to have high levels of sensitivity and specificity for postpartum depression [36]. The study was conducted in an urban setting with participants hailing from a mixture of formal and informal settlements, and the study results may not be generalizable to rural settings. Child outcomes were not measured which would have strengthened the results.

Conclusions
Our results show a lower prevalence of PPD than previously reported, with no difference noted by HIV status, possibly indicating increased normalisation of HIV disease among urban populations in South Africa. However, efforts to identify depressed mothers using targeted symptom screening based on risk factors, and linked to effective treatment interventions remain essential in improving postpartum mother and child outcomes.