Predictors of alcohol and other drug use among pregnant women in a peri-urban South African setting

Ethics

The Western Cape provincial Department of Health and the University of Cape Town Human Research and Ethics Committee provided ethical approval for the study (HREC REF: 131/2009). Written informed consent was obtained from all respondents after the study was verbally explained. The consent forms were available in English and the local languages.

Setting

This study was a cross-sectional study and was conducted at the Hanover Park Midwife Obstetric Unit (MOU). This facility provides maternity services in the residential and semi-industrial urban area of Hanover Park in Cape Town, South Africa. Hanover Park has a population density of 35,000 in an area of approximately two square miles, which is the highest population density in Cape Town [31]. It is considered one of the most violent communities in Cape Town, with high levels of poverty, gang-related violence, alcohol and drug abuse, and physical and sexual abuse [32, 33]. Residences comprise of overcrowded, public apartment units erected in the Apartheid era, with poor infrastructure such as toilets and plumbing. There are also small free-standing houses and an increasing number of informal dwellings. High levels of adult illiteracy are prevalent within the community as less than 20 % of the adult population completed high school. This has resulted in low rates of regular employment and income [33].

At the time of this study, there was no mental health service in Hanover Park for pregnant women in particular (personal communication with the Hanover Park Community Health Centre Manager, 01 May, 2010). Even though these patients were theoretically able to access general mental health outpatient and social work services at the Hanover Park Community Health Centre (CHC), referral to these overburdened services almost never took place. Psychiatric emergencies were referred to the CHC’s casualty unit or to tertiary level hospitals.

Sample

Participants were recruited by systematic sampling of every third woman arriving at the Hanover Park MOU for her first antenatal visit. This strategy was used after considering the average number of women that attended the MOU for antenatal care and the duration of time spent on their screening. This information was then used to estimate the average number of women screened on a daily basis and for calculating the sample number. This technique ensured that participant recruitment and data could be feasibly gathered with a cross-section of women that presented themselves at the MOU all through the day. Women recruited did not undergo any initial medical evaluation. Consent was sought from women that were 18 years and above, pregnant, and willing to participate in the study. A total of 376 women were interviewed.

Measures

Demographic and socioeconomic information was collected using a questionnaire that assessed the age, language, education, marital status, socioeconomic status (SES), obstetric information, whether the pregnancy was planned, wanted, as well as past psychiatric history. Information on participants’ monthly income was collected and converted to United States Dollars.¹

To measure the socio-economic status of the women recruited, an asset index was constructed using information on household ownership of electronic equipment (e.g., microwave, washing machine, television and fridge), transport (cars, motorbike, bicycle), sources of energy (electricity), and bank accounts (including credit card). The use of asset indexes in the measure of household socioeconomic status was chosen over the use of total household income because income data collection has numerous methodological imperfections in low-income settings. These include the difficulty with collecting such data, problems of recall bias, and its lack of sensitivity to non-cash income [34–37]. While constructing the index using a principal component analysis, the first component factor.² was used to represent the asset index. The study sample was then stratified into 4 quintiles (i.e., least poor, poor, very poor and poorest). Food insecurity was examined using the revised six-item Household Food Security Survey Module (HFSSM) [38]. This tool collected information from a period six months prior to the survey. The scale uses a 6-question checklist to ascertain food insecurity and food insufficiency. A score of 0–1 is considered food secure, 2–4 food insecure, while 5–6 is labelled as food insufficient (a more severe form of food insecurity) [38, 39].

The perceived social support from family, friends and a significant other was assessed using the Multidimensional Scale of Perceived Social Support (MSPSS). The MSPSS displays good psychological measurement precision in different study samples, displays a good internal reliability, and a strong factorial validity [40]. This measure has been used previously in South African populations [15, 41]. The tool consists of scores which range between 12 and 84 for the whole instrument and between 4 and 28 for three sub-scales. While there are no exact points for measurement, higher scores on the scale indicate an increase in perception of support.

To assess for the presence of risk factors for psychological distress during pregnancy, the Risk Factor Assessment (RFA) was used. The RFA was developed by the Perinatal Mental Health Project [42], based on their local clinical practice and the common risk factors for perinatal psychological distress and depression identified in the literature [43, 44]. It contains of a list of 11 items, each measuring the presence or absence of one risk factor, with a yes or no response option [45].

The revised conflict tactic scales (CTS2) was used to assess intimate partner violence (IPV) amongst pregnant women in peri-urban settlements in South Africa. The CTS2 is a condensed form of the original conflict tactics scale (CTS) and are used in low-income countries which are resource-constrained to screen for inter-personal violence amongst women [46]. This tool has a good reliability and has been used in cross-cultural studies in South Africa [47, 48].

This study used a widely accepted diagnostic tool, the Expanded Mini-International Neuropsychiatric Interview (MINI Plus) Version 5.0.0 to assess information on common mental disorders (major depressive episode (MDE), and anxiety disorders), suicidality, alcohol abuse and drug use [49]. The MINI Plus is relatively quick and easy to administer, and contains components for the major axis-I psychiatric conditions in DSM-IV TR. This diagnostic tool has been validated for application in South Africa [50] and is available in English and the local languages spoken by the women attending the MOU [51, 52]. Alcohol and drug use data were each collected separately but due to low statistical power emanating from low reporting of usage, the data were combined to form an alcohol and other substance (AOD) use measure. AODs terminology has been used extensively in literature to connote alcohol and other drugs use [53, 54].

Training

A clinical psychologist, with research experience, trained and supervised a research assistant and mental health officer that were responsible for recruitment and screening, and the diagnostic interviewing respectively. The mental health officer holds a 4-year undergraduate degree in psychology and is registered with the Health Professions Council of South Africa as a counsellor. The mental health officer was trained to administer the diagnostic interview and to provide counselling to women who indicated mental distress in the interview process or who qualified with any MINI-defined diagnosis. A referral protocol between the research staff, MOU and CHC was established for women that presented more severe mental health problems and required higher-level psychosocial interventions. A pilot study was conducted to assess how feasible and acceptable the research protocol was for both participants and clinic staff.

Data collection

Data were collected at the Hanover Park MOU between 22 November 2011 and 28 August 2012. Retrospective HIV status data was collected from the women’s hospital records. HIV testing was a routine part of the antenatal appointment which occurred after the study interviews were complete.

Data analysis

Univariate, bivariate, and multivariable analyses was performed using Stata v13.1. The univariate analysis produced descriptive statistics which described the data using sample statistics. Significant associations between alcohol and other drug use (AODs), demographic factors, socioeconomic factors, and psychosocial risk factors, Major Depressive Episode (MDE), anxiety disorder diagnoses and suicidal ideation were examined using non-parametric, Wilcoxon sum of rank, Fisher exact and two sample t-tests. A univariate analysis was conducted for the assets owned and other factors that were included in the asset index. To ensure that results were not skewed in the principal component analysis, all assets that were identified (or not) by the majority of the respondents were not included in the analysis. The 95 % confidence interval was selected as the level of significance.

Bivariate analyses were conducted to examine the associations between predictor variables. Variables which were significant at a probability value (p value) equal to or less than 0.05 were selected and included. Internal consistency and scale reliability within assessment tools were assessed using the Cronbach’s α Statistics [55]. Logistic regression was conducted to identify associations between the outcome of interest and key predictors, with results presented as odds ratio (OR). The dependent variable was AODs use and the independent variables included age, education level, food insecurity and insufficiency, asset index, employment status, life events, interpersonal violence, planned pregnancy and relationship types, MDE and anxiety diagnoses, suicidal ideation. Mini logistic models were used to test the associations between the dependent and independent variables and all the preliminary variables were included in an OLS (ordinary least squares) model to test for multicollinearity. Likelihood ratio tests were used to assess nested models and to test for significance. All items with a statistically significant bivariate association at p  ≤  0.05 were included in the final multivariable model.

Description of the pregnant women sampled

Although 42 % of women sampled were employed, 23 % belonged to the lowest socioeconomic status as defined by their asset index. Also, out of the 159 employed women, 30 % were food insecure and 10 % food insufficient.

Result of the Cronbach’s α tests show that the MSPSS tool (Cronbach’s α 0.89), the CTS2 tool (Cronbach’s α 0.85), and the HFSSM tool (Cronbach’s α 0.83) exhibit good internal consistency and reliability when used on the study sample. The multicollinearity test among independent variables also indicates that there was no multicollinearity within the regression model (VIF < 10).

Women in a relationship were slightly more likely to use AODs than those who were single (OR 1.01, 95 % CI 1.06–1.12), when the type of relationship was examined, women that were in a relationship but not married were more likely to use AODs than those who were married (OR 1.52, 95 % CI 0.52–1.67). To understand further how these predictors associate with AODs, women were asked about their relationship circumstances. Even though women who were living with their partners were less likely to use AODs (OR 0.48, 95 % CI 0.44–0.91), women who partly cohabited with their partners were more likely to use AODs (OR 1.55, 95 % CI 0.47–2.06). In line with this, as women perceived more social support, based on increases in their MSPSS index, they were less likely to use AODs (0.99, 95 % CI 0.95–1.02).

Women who had planned their pregnancy were less likely to use AODs than those that had not planned their pregnancy (OR 0.30, 95 % CI 0.11–0.80). Life experiences were explored as a predictor of AOD among sampled pregnant women. Women who experienced intimate partner violence (IPV) based on the CTS2 score were more likely to use AODs (OR 1.12, 95 % CI 0.38–2.23). Also, women who experienced difficult life events (loss of employment, injuries, loss of relative, financial crisis, theft, loss of a steady relationship, etc.) were more likely to use AODs than those that have not experienced difficult life events (OR 1.64, 95 % CI 0.26–1.72).

Finally, results indicate that women who were experiencing a MDE were more likely to use AOD (OR 1.25, 95 % CI 0.42–1.69). Also women that had a current anxiety diagnosis were more likely to have used AODs than those without an anxiety diagnosis (OR 1.35, 95 % CI 0.53–1.38). Anxiety diagnoses included panic disorder, agoraphobia, social phobia or social anxiety disorder, any specific phobia, obsessive compulsive disorder, post-traumatic stress disorder, and generalised anxiety disorder. Women who indicated suicidal ideation were also more likely to use AODs (1.17, CI 0.45–3.04). Those who experienced past mental health problems were two times more likely to use AODs than those who had not experienced mental health problems in the past (OR 12.13, 95 % CI 1.81–2.61).