It does matter where you come from: mothers’ experiences of childbirth in midwife obstetric units, Tshwane, South Africa

The Tshwane District Health Service provides health care to a population of 3.3 million and is categorised as one of the least deprived districts in South Africa, ranking in the top socio-economic quintile. The district recorded around 50,000 deliveries per annum in public facilities in the years 2014 and 2015. During the same period the district recorded a delivery-in-facility rate of 96.7% [22]. Around 18% of the district deliveries took place in the 10 MOUs and another 13% of women in labour were transferred from these MOUs to hospitals for care during delivery [22].

MOUs are located in either community health centres or larger clinics and attend to low-risk deliveries as part of the free primary health care system in South Africa. MOUs are able to provide basic emergency obstetric care [23], except for the removal of retained products of conception and assisted deliveries. Seven of the Tshwane MOUs are situated in urban areas, and 3 units are based in semi-rural settings. The latter facilities are located much further from referral sites offering caesarean sections (up to 70 km), with ambulance turn-around times of 1 h and longer. At the time of the study the midwife teams in each MOU consisted of two to four midwives per shift, depending on the facility’s available human resources and the number of deliveries per month. Each shift had a midwife specialist or advanced midwife on duty, who holds an additional qualification in midwifery and is registered with the South African Nursing Council [24].

In 2014, the South African National Department of Health launched the MomConnect mHealth initiative using mobile phones to register pregnancies and interact with the registered women, with a opt-in platform that encourages women to rate the services at the facilities [25]. Tshwane District received 63 antenatal-care-related or drug-related complaints from mothers between 2014 and 2016, but no complaints regarding mothers’ intra-partum care or narratives of distress.

Data were collected by means of an anonymous, self-administered questionnaire with 32 structured and open-ended questions of which seven were socio-demographic items. The survey tool elicited data on the main concepts of respectful care. Sixteen items reflected women’s’ self-report of the clinical care they received and their experience thereof. The nine items on client satisfaction included aspects of communication and satisfaction. Items all required a yes/no/unsure response or a response on a four-point Likert-type survey scale. Many of the questions included in our survey had been used in previous sets of validated maternal experiences surveys and covered domains related to the lack of consented care, communication and feedback processes, pain relief and respectful care aspects [15, 18, 19, 27]. Text boxes included after satisfaction questions allowed participants to supply feedback and descriptions of poor service or abusive behaviour and to report unfulfilled expectations. The questionnaire was also made available in Setswana, the predominant local language, after translation from English to Setswana, followed by back-translation into English and the resolving of interpretation issues. The questionnaire was pilot-tested with 30 mothers to confirm the appropriateness of questions and ease of comprehension. (The questionnaire is attached as Additional file 1).

The design of the sample for this survey was based on historical population data on annual deliveries at each MOU (range: 390 to 1502 in 2015). A planned sample of 800 respondents was envisaged, but 653 questionnaires were received back. Factors that impeded data collection included an unexpected drop in the number of deliveries in the district during the first 3 months of 2016 and several external service-delivery strikes that hampered access to the semi-rural areas.

Mothers completed questionnaires during the same period at various sites. University students fluent in the local vernaculars were trained as research assistants. They signed a confidentiality clause and assisted mothers with the completion of the questionnaire, only on request, in a private space. A sequential sample was drawn consisting of mothers returning for follow-up visits to primary health care consultation rooms in the three-days-to-six-weeks postnatal period. This gave mothers enough time to reflect on their care. Data were collected at a venue separate from the labour ward to minimise potential interference from MOU staff [28]. All mothers who had read the information leaflet and were willing to participate in the study were screened for eligibility. No mothers eligible for inclusion refused to participate. To qualify for inclusion, a mother had to be older than 17 years, have delivered in one of the 10 Tshwane MOUs and have returned for her postnatal visit during the prescribed period. Mothers younger than 17 years or those who had delivered in a hospital were excluded. Mothers completed the questionnaire voluntarily and anonymously.

The completed questionnaires were collected, reviewed and coded. Data were captured in password-protected Excel files. The data were then crosschecked, cleaned and analysed with the aid of SAS version 9.4 [29].

Descriptive statistical measures such as means, frequencies and proportions were calculated and age, level of education, language, province or country of birth and length of residence in the Tshwane District were categorised to facilitate data interpretation. In order to establish the significance or importance of factors and attributes or items in the study that contributed to the acceptability or unacceptability of the treatment of mothers in the MOUs during childbirth, an analysis of variance approach was followed, with the inclusion of a covariate (ANCOVA). Due to the complexity of the analyses (multiple effects/factors), the classical approach of applying non-parametric procedures to ordered data, was not followed, but a transformation of the data as described below. For purposes of comparison, weighted means according to the number of 2015 deliveries based on scores on a Likert-type scale were calculated for categories of socio-demographic variables, namely age of mother, level of education, first language, province or country of birth, and length of residence in Tshwane. Missing values of categorical variables were replaced by hotdeck imputation [30], using simple random sampling with replacement of the units to produce complete data for a multivariate ANCOVA. The number of children a woman had given birth to was used as the covariate. The following six items relating to mothers’ experiences during childbirth were selected as dependents and a series of ANCOVAs were performed:

The items were generally transformed from nominal [Yes, Unsure, No] measures to a Likert-type [0, 1, 2] scale. The ties present in the Likert scores were resolved by adding a small random univariate term from the [-0.000005; 0.0000005] interval and the resultant values were then normalised using the BLOM transformation [31]. Means for each item were calculated with scores of dependents as follows: Yes = 0, Unsure = 1, No = 2. The higher the means, the less positive the experience of the mother on an aspect of care; the lower the means, the more positive the experience. Means for a response were calculated for each of the categories of the selected socio-demographic variables, as well as the standard error based on this specific response for all categories of the particular demographic variable under consideration. The category corresponding to the maximum mean and all categories within one standard error of this maximum mean were considered as those contributing to the worst treatment of mothers in MOUs (see Additional file 2).

Open-ended responses related to the six items mentioned above were collated and some striking statements were selected to illustrate a particular issue.

The questionnaire was completed by 653 mothers. The number of respondents per MOU ranged from 29 to 102, sampled in accordance with the number of deliveries performed in 2015. Table 1 contains a summary of the socio-demographic characteristics of respondents of the 10 MOUs. Their mean age was 27.0 years, with half the respondents in the age category 25 to 34 years. The mean number of children per mother was 2.1. More than half of the mothers (51.6%) originally came from other provinces or neighbouring countries. Many of the respondents (64.3%) had been living in Tshwane for 5 years or more, while 13.0% had been in the district for less than 1 year. Some South African first languages were classified into a single cluster according to mutually understood language families, namely Sotho and Nguni. Two unrelated local African languages, Xitsonga and Tshivenda, were clustered together because they are spoken in neighbouring areas in the north of the country (Limpopo Province). The remaining two official South African languages, Afrikaans and English, were conveniently clustered together. Almost all mothers had some school education; over 50% had completed Grade 12 and some also had post-school education.

Respondents spent a mean time of 7.0 h in the MOU before their babies were born. Table 2 gives an overview of mothers’ experiences of childbirth in different areas: communication, labour experiences, clinical care and perceptions of respectful care. Midwives greeted less than half of women in labour by name when receiving their maternity file on arrival at a facility. Consent to perform a physical examination was obtained from less than half of the women. Only 18.6% of respondents who requested this were allowed or offered a birthing partner during labour; 58.7% indicated that they preferred not to have a family member or partner with them because of current cultural custom. Around half of the mothers noted that all staff spoke to them courteously (“nicely”) (54.3%) and treated them with a lot of respect (47.9%), and 54.9% of mothers were completely satisfied with their birthing treatment.

Pain relief medication was not available in four of the five MOUs during the study. One MOU with 102 respondents had pain medication on hand. It was offered and administered to 58 women (56.9%) in the sub-sample.

Some socio-demographic variables were strongly linked to shortcomings during women’s childbirth and treatment. Table 3 visualises data – according to different levels of significance (p = <0.5; <0.1; <0.01) – on the respondent groups that reported the worst treatment for the six items selected for analysis. (For the results of the total analysis, see Additional files 2 and 3.) Fig. 1 provides an example of the calculation of the means of the class variable of age as a significant socio-demographic variable linked to the care women reported having received. The darker bars represent the categories contributing to the worst treatment of mothers, i.e. those categories within one standard error of the maximum mean. (Additional file 2 provides the same information in graphic form for all the variables.)

The following points summarise the main results for each of the six items (areas of care) included in the analysis:

Overall, the following groups of mothers reported to have had more negative experiences of treatment during labour: teenage and young mothers; mothers with no schooling or only primary education; mothers from other countries who speak a foreign language; and mothers with less than 20 years’ residence in Tshwane. Level of education and province or country of birth were the variables with the least influence on women’s experiences during childbirth. The results show that class variables of age, language and period of residence in Tshwane were significant predictors of the level of disrespect shown in communication and care and client dissatisfaction with treatment (p = <0.01), as seen in Table 3.

Completion of open-ended responses varied from 41 to 403 for individual items. In total, only 55 (9%) of the total number of comments (n = 595) were positive. Several examples of poor communication, abusive and hurtful examinations, withholding of care and disrespectful care were given in these responses. Some of the marginalised mothers called for a change in midwives’ attitudes and compassion, requesting equal treatment:

“We are also human!”.

“I am not happy about the way I had to deliver my baby!”.

“They must stop violence, treat us with respect, even if we are a teenage mother!”.

“They must stop telling mothers they are too young or too old to give birth”.

“They left me unattended and I gave birth on my own”.

They described their experiences of physical abuse during examinations and delivery as follows:

“They must be gentle when examining us, they don’t care whether they hurt us or not!”

“Sisters should make mothers aware that they are about to be examined – not to be penetrated with fingers while unaware!”

“They are harsh and refused to give me something for pain.”

“Some were annoyed with the mothers and ill-treating them.”

“They must stop the violence and not treat us like animals!”

A strength of this study is the fact that all 10 MOUs in one district were surveyed regarding experiences of care during childbirth. The limited period of 9 weeks for data collection provided the opportunity to obtain a comprehensive snapshot of care experiences during labour. The open-ended questions collected responses to clarify aspects of dissatisfaction and abuse in MOUs.

Limitations of the study relate to logistics that constrained data collection, including distances between MOUs, as well as external service-delivery protests, which limited access to some MOUs on certain days. A sequential sample is a form of convenience sampling, so only limited claims can be made with regard to generalisation and representativeness. Although our survey was conducted outside labour wards, fear and mistrust of the providers could have influenced respondents’ recall of negative events or triggered the coping mechanisms they employed to protect themselves from recalling the birthing.