Participants
This cross-sectional study was conducted at the Hanover Park Midwife Obstetric Unit (MOU), in Cape Town, South Africa. Hanover Park is a low-income, residential suburb within the City of Cape Town, established in 1969 as part of the South African Apartheid governments Group Areas Act [22]. It has a population of approximately 45,000 people and experiences widespread gang activity and high rates of violent crimes. Coupled with this are numerous social problems such as school drop-outs, drug and alcohol abuse, prostitution, drug trafficking, and robberies. Education levels are low and unemployment is high. Only 21% of the residents have passed high school, while 36% are unemployed [23].
Every third woman, 18 years or older, attending the MOU for her first antenatal clinic visit, was invited to participate in the study between November 2011 and August 2012. Of the 559 eligible women who were invited to participate, 135 (24%) declined to participate, and 376 women were recruited. As the process was relatively time consuming, 48 women were not able to take the time to complete the questionnaires, due to work or childcare commitments.
Testing procedures
An interviewer-administered socio-demographic questionnaire was used to collect data on participants’ age, obstetric information and feelings about pregnancy, level of education, relationship status, HIV status, socio-economic status, and prior experience of depression or anxiety.
Screening tools were selected based on either local validation data or their having been used in other resource constrained settings. The US Household Food Security Survey Module (HFSSM): 6-Item Short Form [4] was used to assess household food insecurity and hunger. The scale measures the frequency of running out of food, being unable to afford balanced meals, and skipping meals because of lack of food over the prior 6 months. Perceptions of social support from three possible sources (family, friends and a significant other) were measured using the Multidimensional Scale of Perceived Social Support (MSPSS) [24]. Threatening life experiences faced by women in the preceding 6 months were measured using the List of Threatening Experiences (LTE) [25]. The Revised Conflict Tactic Scales (CTS2) [26] was used to measure Intimate partner violence.
The Expanded Mini-International Neuropsychiatric Interview (MINI Plus) Version 5.0.0 [27, 28], was used to diagnose a major depressive episode (MDE) (Module A), generalised anxiety disorder (Module P), suicidality (Module C), alcohol dependence (Module K) and substance dependence (Module L), and has been validated in several countries, including South Africa [29,30,31]. The MINI was administered by an experienced, registered counsellor, who was supervised by a clinical psychologist. The MINI Plus is available in local South African languages—Afrikaans and isiXhosa [32]. All tools were administered in English, Afrikaans or isiXhosa, the languages spoken by the women attending the MOU.
Participants were provided with refreshments mid-way through the interview process. Participants were not provided with money for participation or transport. The testing procedures are explained in more detail published elsewhere [33].
Ethical approval
Ethical approval for the study was obtained from the Human Research and Ethics Committee at the University of Cape Town (HREC REF: 131/2009). The Western Cape provincial Department of Health approved the use of the research site. All respondents who participated in the study provided written, informed consent after the procedure had been verbally explained to them. Consent forms were available in English, Afrikaans and isiXhosa. All those participants who were diagnosed with a mental disorder were offered on-site, free of charge counselling with a registered counsellor.
Data analysis
Data analysis was carried out using STATA/SE statistical software package version 14.1 (StataCorp., College Station, TX, USA). Variables were described using frequency and percentages, and associations measured using Chi-square tests. An asset index was used to stratify households based on socio-economic status [34]. Asset indices have previously been used in studies in LMICs [35, 36]. To construct the asset index, information on ownership of electronic equipment (e.g., fridge or freezer, vacuum cleaner, television, microwave, washing machine, television), transport (owning a vehicle), sources of energy (electricity) and bank accounts (including credit card) were pooled together. Principal component analysis was used to stratify households into 4 quartiles representing least poor, poor, very poor and poorest.
The nine questions comprising Module C of the MINI Plus were used to develop three categories of suicidality based on experiences in the month prior to the interview. Suicidal ideation included questions on suicidal thoughts (questions c2–c4). Suicidal behaviour included questions on plans to commit suicide as well as suicide attempts (questions c5–c8). Suicidal ideation and behaviour referred to those who endorsed items pertaining to suicidal thoughts, and those who endorsed items pertaining to planning, preparing or attempting suicide (questions c2–c8).
The primary outcome variables in the regression analyses were; (1) food insecurity and (2) MDE. Currently there is no accepted measure or standardised way of measuring food insecurity in South Africa. The cut-points suggested in the user notes of the 6-item short form of the US Household Food Security Module [37] were used: a score of 0–1 indicates high or marginal food security; a score of 2–4 indicates low food security; a score of 5–6 indicates very low food security. We assigned food security status as follows: households with a score of 0–1 are referred to as food secure, while households with a score of 2–6 [derived by combining low food security (score of 2–4) and very low food security (score of 5–6)] are referred to as food insecure. MDE was diagnosed using Module A of the MINI Plus.
Model building techniques were used to develop multivariate logistic regression models for food insecurity and MDE separately. We controlled/adjusted for the following extraneous variables so as to exclude their effect on the outcome variable: participant income, household income, employment status, number of children, having an unplanned pregnancy, and feeling happy about the pregnancy. Univariate analysis was used to identify significant associations between food insecurity and a number of correlates. The following correlates, which were significantly associated with food insecurity, were then used to build the final multivariate model: participants’ income, employment status, household income, number of children, suicidal behaviour, MDE, anxiety disorder, history of mental illness, experience of a threatening life event in the past year, experiencing intimate partner violence, perceived social support, substance dependence and alcohol dependence.
In the second model, univariate analysis was used to identify significant associations between MDE and a number of correlates. The following correlates, which were significantly associated with MDE, were then used to build the final multivariate model: household income, number of children, unplanned pregnancy, feeling happy about pregnancy, substance dependence, anxiety disorder, food insecurity, experience of intimate partner violence, perceived social support, experience of a threatening life event in the past year, suicidal ideation and behaviour, history of mental illness, and alcohol dependence.