A cross-sectional study design was conducted in which an investigator-administered questionnaire was used among pregnant women attending an antenatal clinic in the Sedibeng district of Gauteng province.

The antenatal clinic was at Johan Heyns Community Health Centre (CHC), which is in an industrial city of the Sedibeng district in Gauteng province and serves a mixed suburban, informal settlement and rural population. This CHC was chosen because it has a large and well-established antenatal clinic as well as a mental health care service with doctors, nurses, social workers, and psychologists available to aid patients needing escalation of care. The CHC has facilities for uncomplicated deliveries. Women identified at the antenatal clinic as having high-risk pregnancies are referred to the regional hospital for further antenatal and obstetric care. As part of routine antenatal care, women attending the antenatal clinic are screened for mental health issues by a professional nurse using a standard set of screening questions in the government-issued maternity case record. A 3-item tool validated in South Africa has been implemented as standard antenatal screening for perinatal common mental illnesses and suicidality.¹⁴ This tool screens for symptoms of anxiety, depression and self-harm thoughts over a 2-week period. Women answering yes to two or more questions are referred to a registered counsellor as per the tool recommendations.

The study population comprised pregnant women following up at Johan Heyns antenatal clinic. To be enrolled in the study, participants needed to be 18 years of age or older and have a basic proficiency in English, as all questionnaires were administered in English. Participants were recruited through convenience sampling on the day of their antenatal clinic follow-up. Convenience sampling was chosen for its simplicity, cost-effectiveness and practicality in a resource-limited setting. All women waiting for their antenatal appointment were informed of the study aims and objectives and invited to participate. Women willing to participate were taken to a private interview room in the clinic, consent was obtained, and study questions were administered. Exclusion criteria were women who were too physically unwell to participate and women without capacity to consent to the traumatic questions being asked in the questionnaire due to health concerns at the time of interview.

Calculation of sample size was conducted using Glenn Israel tables.¹⁵ The tables provide sample sizes needed for a confidence interval (CI) of 95% and a population variability of 0.5 (i.e. the maximum degree of variability is assumed). A 10% leeway on precision was applied as this is the first study of its kind to be conducted in this setting. As Sedibeng district has an estimated population of 1 039 908 people from the 2022 census,¹⁶ 100 recorded responses would be necessary to obtain a CI of 95% with 10% precision.

Women were informed of the study, provided with an information leaflet, and invited to participate while in the antenatal waiting room by Megan Fyffe. Those who wanted to participate were taken to a private consultation room and given an opportunity to ask questions before signing the consent. Socio-demographic, antenatal, medical and substance use data were collected and the PCL-5 was administered by Megan Fyffe. Antenatal data were captured from the maternal health record and during the interview. Participants were monitored for any signs of distress during the interview process. All women who exhibited distress, screened positive for PTSD, or were using substances were referred to an on-site mental health professional. A deliberately low threshold for referral was maintained given the vulnerability of the study population and the possibility of complex biopsychosocial causes of the distress.

The interviews were conducted in English. Participants were assured of anonymity, with each interviewee assigned a study number and all data collected were stored separately from any identifying information. Participants could choose to skip questions on the socio-demographic questionnaire and remain in the study. An effort was made to ensure a supportive and safe environment for women choosing to participate.

Data from all questionnaires were captured on a Microsoft Excel spreadsheet and imported into IBM SPSS version 29 for statistical analysis. Descriptive statistics including numbers, frequency distributions and means were used to summarise participants’ socio-demographic data and responses to PCL-5. Frequencies and percentages were presented for categorical variables as overall and stratified by PTSD symptoms (negative vs. positive).

Cronbach’s alpha was used to test for the reliability and consistency of the items of the PTSD scale. The tables stratified by PTSD symptoms were presented with denominators, referring to the number of participants who responded to a question and/or item. For continuous measures, means and standard deviations (s.d.) as well as medians and interquartile ranges were presented. A bar chart was used to report PTSD severity among participants with positive PTSD symptoms while a box and whisker plot provides graphical representational was used to report the means and s.d. for positive PTSD symptoms as well as the clusters of participants with positive PTSD symptoms.

To determine the statistical difference between participants with negative and positive PTSD symptoms, Chi-Square test or Fisher’s test was used where appropriate. For statistical comparison in continuous measures, medians were compared using a non-parametric test (Kruskal–Wallis test). Generalised linear model, using a log binomial distribution link function, was used to assess factors associated with positive PTSD symptoms. This model was selected because the proportion of participants with positive PTSD symptoms was > 10%. No variables were significant at univariate analysis; therefore, all variables in the univariate analysis were included in the multivariate model, and a backward selection procedure was used to reach the final model. Results for the model are presented as relative risks (RR), 95% CIs and p-values. Statistical significance was defined as p-value ≤ 0.05. Statistical analysis was conducted using SAS Enterprise Guide 7.1 (SAS Institute Inc., Cary, North Carolina, United Sates).

An effort was made to ensure that consent to participate was voluntary. A distress protocol was put in place, and on-site access to mental health professionals was confirmed. All data were anonymised with the use of separately stored study numbers for each participant. Ethical consent for the study was approved by the Faculty of Health Sciences Human Research Ethics Committee of the University of the Witwatersrand (M240414MED24-03-477), and permission to conduct the study was provided by the Acting Chief Director of the Sedibeng District Health Services and the Operational Manager of Johan Heyns Clinic.

The study sample comprised 98 women. Although 100 women initially participated in the study, one was excluded when it became apparent that she was under 18 years old. Another woman withdrew from the study as she experienced marked distress during the PCL-5 questionnaire, did not wish to continue and was referred to the mental health care services.

Socio-demographic characteristics are described in Table 1. The median age of participants was 26 years with 47 participants being between 25 and 34 years of age. Most participants were black African (84%), while white, mixed race and Indian and/or Asian population groups formed 10%, 5% and 1% of the sample, respectively. Just over a third (35%) of participants were single, and a high proportion (67%) had a monthly income of less than R5999. Only four participants reported earning over R10 000.00, while four preferred not to disclose their income status to the interviewer. A third (33%) of participants received a childcare support grant, while two participants received disability grants, both for bipolar disorder. Income insecurity was further underscored with 66 participants being unemployed.

With respect to living conditions, 46% of participants indicated they resided in rented rooms, and a large majority (66%) of participants lived in households with 2–5 people, while 20% lived alone. Educational attainment was limited, with only 40 participants having achieved matric and five having attended special education schools.

Maternal health factors and substance use history are outlined in Table 2. Half of the participants involved were primigravida and most (69%) were in their second or third trimester. Complications in previous pregnancies were reported by 16 participants. Eight participants reported having had previous contact with mental health services: five had received at least one counselling session (one after experiencing a stillbirth, one for suicidality and three for undisclosed reasons); and three had received psychiatric care (for PTSD, bipolar disorder and depression, respectively). The mental health screening tool in the maternity health record was completed in all participants, with one participant scoring 3, four scoring 2, one scoring 1 and 92 scoring 0. However, there was no documentation on whether counselling had been offered or a referral provided.

Twenty-four participants had comorbid medical conditions, two of whom had two conditions (one with hypertension and HIV and the other with hypertension and diabetes). Thirty-seven participants reported using at least one substance during their pregnancy. Tobacco was the most used substance, either smoked or used in a dry form as ‘snuff’. We separated the two forms of use as there is less understanding regarding the harmful effects of snuff use in pregnancy. Of the participants who reported using more than one substance, tobacco and alcohol were the most frequently used combination.

Fifty-three participants reported having either personally experienced or witnessed a traumatic event. In 14 women, the traumatic event was linked to previous obstetric or gynaecological complications or neonatal demise. Four women reported experiencing IPV and three had a history of childhood trauma (one of whom was currently exposed to IPV). One woman reported exposure to criminal violence and another to workplace-related conflict. The remaining participants reporting a history of trauma did not wish to disclose its nature.

The results of the PCL-5 questionnaire are presented in Table 3, including the total score and the number of participants who endorsed at least one item in each cluster (i.e. who rated their experience as moderately severe for one or more items in each cluster). A Cronbach’s alpha coefficient is a measure of internal consistency and a measure of scale reliability. A value of 0.97 indicates good internal consistency and reliability for the PCL-5 within our population sample.

The PCL-5 scores ranged from 0 to 71 with a mean score of 13.8 (s.d. = 18.1), and 16 of the 98 participants scored 31 or more. However, many women endorsed individual items, indicating the presence of PTSD symptoms even if the total score was less than 31. A score of 2 or more for an item (at least a moderate severity of the symptom) was reported by 37 participants for Cluster B (intrusion), 24 for Cluster C (avoidance), 36 for Cluster D (cognition and mood changes) and 32 for Cluster E (arousal and reactivity) symptoms.

Twelve participants met DSM-5 criteria for a diagnosis of PTSD, one of whom had a total score of only 26. Of the five participants with total scores of 31 or more who did not meet DSM-5 criteria, four had not endorsed two or more Cluster E (arousal and reactivity) symptoms and one did not endorse any Cluster B (intrusion) symptoms, which were all experienced but only mildly.

All four participants who had been exposed to IPV met DSM-5 criteria for PTSD, with total PCL-5 scores of 68, 61, 41, and 26. One participant with gynaecological trauma (a failed termination of pregnancy for the current pregnancy) met DSM-5 criteria for PTSD, with a PCL-5 score of 43, and one who had experienced a neonatal death secondary to eclampsia did not meet DSM-5 criteria, although she had a PCL-5 score of 35. None of the other women who provided details about the experienced traumatic event received a provisional diagnosis of PTSD on either scoring method. Of interest, only one of the eight women with prior contact with mental health services had PTSD (one of the women exposed to IPV who had previously received treatment for depression).

Among the 16 participants who screened positive for at least moderate PTSD when using the total score, the mean PCL-5 score was 49.2 (s.d. = 12.7), as illustrated in Figure 1. Of the clusters, Cluster D symptoms (cognition and mood changes) tended to be the most severe and Cluster C (avoidance) the least.

No statistically significant differences were found between participants with and those without PTSD across sociodemographic, maternal health or substance use variables (Table 4). Although not significant, participants who were 18–24 years had a higher proportion of positive PTSD symptoms compared to negative PTSD symptoms (56.2% vs. 39.0%, p = 0.1097). All 16 participants who had positive PTSD symptoms were black people. A high proportion of participants who had positive PTSD symptoms attended post-high school (62.5%), had a monthly income of < R5999 (68.8%), did not receive a disability grant (62.5%) and were not currently employed (75.0%).

While there were no differences between the two groups in terms of each substance used, drinking two units of alcohol per day was significantly more common among those with PTSD compared to those without (25.0% vs. 4.9%, p = 0.0072) (Table 5).

Multivariate analysis revealed alcohol use as the only factor significantly associated with having at least moderate PTSD (Relative Risk 2.7, 95% CI 1.1–7.0, p = 0.0364) (Table 6).

A total of 53 (54%) participants exhibited some distress during or after the interview and were referred to a mental health professional. In addition to those referred for management of their PTSD symptoms, others were referred for assistance with poor social support, family stressors and persistent substance use. While several referrals were to a registered counsellor or social worker, 30 were referred to the specialist mental health services.

This study has several limitations. The design of the study only allowed for general estimates of PTSD prevalence; thus, a temporal relationship or causality may not be inferred from the results. Another limitation was the presence of selection bias which meant that those participants who could not participate in the study were not screened and therefore may be at risk of PTSD and its associated complications. Similarly, only women who attended the clinic during data collection for their antenatal check-up were screened, leaving out those women who either did not visit the clinic or those who do not attend any antenatal clinic. Finally, the sample population was drawn from a peri-urban district in the most industrialised province of South Africa, and therefore, the findings may not be generalisable to the rest of the country.

Further research is needed to explore means of strengthening maternal mental health care. However, increasing the capacity of antenatal nursing staff through additional training and improved human resources is likely to help. In particular, the addition of registered counsellors and/or social workers to antenatal clinics and midwifery units would be beneficial.