A retrospective, cross-sectional survey on ANC seeking behaviour was conducted. The survey contained mostly closed-ended questions, allowing for quantitative analysis and was adapted from an earlier study. The questionnaire contained sections on socio-demographics, pregnancy discovery, ANC attendance (including reasons for late attendance), the nature of the pregnancy and ANC experiences, including a specific question on the screening tests and services provided at the first full screening visit. Gestational age at first clinic attendance was self-reported. We distinguish between the timing of the first clinic visit and women’s first screening visit.

The data and findings from the study are representative of adolescent mothers who used the public healthcare system to give birth in the Cape Town Metropolitan area in the three sampled facilities. All adolescent mothers aged between 16 years and 18 years of age who used the healthcare facilities to give birth between December 2018 and early April 2019 (from Monday to Saturday) were asked to respond to survey questions given that the day and months of birth are random, this sampling approach allowed us to obtain a simple random sample within the geographic areas covered. A total of 202 women were surveyed.

For this analysis, late ANC access was defined as access after the first trimester. The first trimester corresponds to the WHO definition of timely ANC access.^24,28 This contrasts with the definition of 5 months (20 weeks) used by the South African Department of Health to monitor timely access for early antiretroviral therapy (ART) initiation purposes.^25,29

The study was conducted in three public healthcare facilities in metropolitan Cape Town, in the Western Cape province of South Africa. The Western Cape is one of nine provinces in South Africa and is therefore a subset of the broader South African healthcare system. Each province in South Africa has its own provincial government responsible for healthcare within the province. This provincial autonomy means that the finding from this study may not necessarily apply to the broader South African context.

Two district hospitals, one in the Northern sub-district and the other in Eastern sub-district, and a Midwife Obstetric Unit (MOU) in the Eastern sub-district were chosen. In the South African healthcare system, uncomplicated or lower-risk births are performed in MOUs, while higher risk, assisted natural or caesarean deliveries are performed in district hospitals.

The choice of area was based on the average late ANC access in these areas and builds on the sampling frame of a previous study conducted among non-adolescent mothers (18+).²⁷ In the previous study to which this study is a follow-on, healthcare facilities had been selected to focus on geographic areas with high late ANC access rates.

In the Western Cape, it is recommended that pregnant women access ANC prior to 20 weeks of gestation, with a 6-week interval between visits until week 28. The next ANC visit is then scheduled at 34 weeks, and subsequently based on the individual risk profile.³⁰ Metropolitan Cape Town was selected because of its relatively poor ANC performance compared with other districts in the Western Cape. In 2018–2019, approximately 66.15% of women presenting at Cape Town facilities did so prior to 20 weeks of gestation.³¹ No distinction is made between adolescent or adult pregnancies in these records.

The selected age category falls outside the WHO definition of adolescents, which ranges from ages 10 years to 19 years.³² The sample was limited to adolescents older than 16 because of ethical considerations. The South African Department of Health’s ethical guidelines allows a waiver of parental consent for only children aged 16 and older for sensitive topics such as sexual activity for which parental involvement in data collection may influence their responses.³³ Ethical clearance for this study was therefore limited to adolescents aged 16–18 years. The upper bound of 18 was chosen as adolescents aged 19 had been interviewed in the same facilities in a previous study and results on this has been reported elsewhere.²⁷

Given the relative rarity of adolescent births and administrative report issues with adolescent births and timing of ANC access, it was not possible to accurately power the study sample to be representative of adolescent births in the two sub-districts in which birth facilities were located. The sample can be considered an indicative cross-sectional sample of adolescent births during the relevant period (November 2018 to April 2019) at these facilities. All adolescent mothers who gave birth at these facilities on Mondays to Saturdays were approached by enumerators to be surveyed and almost all agreed to participate in the survey.

Adolescent mothers were surveyed by enumerators using a structured questionnaire (mostly close-ended questions) at the facilities shortly after delivery, at their bedside in the post-labour ward.

Quantitative surveys were conducted between November 2018 and April 2019 following pilot testing at one facility. An amended version of a questionnaire that had been field-tested in a different context in a previous study³⁴ was used. An initial sample of five survey interviews were used to pilot the survey and no adjustments were necessary.

The choice to survey mothers at the delivery facility rather than ANC facility was because of the large proportion of adolescent women not attending ANC prior to birth and concerns of losing the most vulnerable from the sample if interviewed after discharge from the facility.

The questionnaire was based on the survey instrument used to study late ANC attendance among adult (age 18+) mothers²⁷ but extended to include questions on contraception knowledge and use before conception. Gestational age both at first clinic attendance and at first full screening was self-reported.

Given an anticipated low response rate on the household income question, we elected to use an asset index as a measure of socio-economic status (SES).³⁵ This was performed by using data on ownership of 14 durable assets by the household collected through the questionnaire. Wealth status was classified according to girls’ and women’s position in the index, classified as either Top (60%) or Bottom (40%).

The data were analysed using univariate, bivariate, and multivariate methods. Statistical differences (comparing) in bivariate analysis between women who accessed ANC before or after the first trimester were identified through the chi-square test (χ²) run as a command in Stata. Multivariate analysis of the relationship between late ANC attendance and the selected covariates were estimated through linear probability models using ordinary least square (OLS) analysis. In Appendix 1, we report the results of the analysis drawing on a larger sample which includes respondents who did not attend ANC at all (n = 195). p-values < 0.05 were considered statistically significant, although p-values < 0.1 are also reported but not considered statistically significant for the purpose of this study.

All collected data were captured in SurveyCTO and analysed using Stata 16 (StataCorp [2019] Stata Statistical Software: Release 16. StataCorp LLC, College Station, Texas, United States).

Ethical clearance to conduct this study was obtained from the University of Stellenbosch, Research Ethics Committee: Human Research (Humanities) (No. SU-HSD-002848).

In addition, to ensure that the survey was sensitive to the concerns and circumstances of adolescent experiences, we first conducted focus group discussions (FGD) on the advice of the ethics committee, the results of which are not reported extensively here. At the FGDs, some adolescent mothers from vulnerable backgrounds voiced that once pregnant they do not always receive support from their parents and the broader community. They expressed that their pregnancies bring shame to their families. Some experienced difficulties during their pregnancy, therefore having someone to talk to, even a survey enumerator, would have helped them to voice and process their experiences. Based on the results of the FGDs, the ethics committee approved the study protocol.

Informed consent was obtained prior to survey completion from all respondents, and each respondent received a small gift valued at R25.00 (approximately $1.75) to thank them for their time. All interviews were conducted by enumerators in either English or IsiXhosa. Interviewers were trained to be sensitive to private questions and to treat respondents kindly and respectfully.

A total of 202 adolescents provided responses to the survey. Approximately four out of five of the adolescents sampled (n = 1677, 82.7%) reported seeking care for their pregnancy at a public healthcare facility (Figure 1), 3% (n = 6) of adolescents sought ANC at a private health facility and were therefore excluded from most of the analysis, and 14.4% (n = 29) reported not seeking ANC at all. Of the 167 respondents who sought care at a public facility, 166 confirmed that they received a comprehensive screening (aligned with the elements of basic ANC at the time needed to be delivered at the first full ANC visit), while one respondent was unclear and did not provide enough information to be included.

We restricted our analysis to adolescents who attended public ANC clinics and reported full information (n = 166) or did not access ANC at all (n = 29). Descriptive socio-demographic data on the sample who attended public ANC clinics is provided later in this section.

Most adolescents in our restricted sample (66.15%, n = 129/195) reported attending their first ANC visit only after their first trimester (3 months) or not at all (Table 1). While almost 30.0% (n = 59) attended their first ANC booking visit between 3 months and 5 months of gestation, more than 20.0% (15.9% + 4.1% + 0.51%) only attended at or after 5 months gestation (Table 1). On an average, adolescents attended 6.6 ANC visits (standard deviation [s.d.] = 2.6) and the average gestational age at their first full ANC visit was 23.6 weeks (s.d. = 12.5).

Among adolescent women who attended ANC late (> 3 months, excluding those who did not attended at all), the majority (45/99 or 45.5%) reported not seeking care earlier because they were unaware, they were pregnant (Figure 2). Another prominent reason includes delaying or putting off the appointment (29.3%). A lack of knowledge and information regarding ANC also contributed to late ANC-seeking behaviour. Among late care seekers, some reported not thinking early care was necessary (20.0%) or required (19.2%) or did not know where to seek care (9.1%).

Adolescents who attended ANC late (> 3 months) were asked whether there were any personal factors that would have made them go sooner (Figure 3) and overwhelmingly (52.53%) reported that they would have accessed care earlier if they found out they were pregnant sooner. Other prominent factors included if they were told to go sooner (22.22%), if they thought it was necessary (21.21%), or if they thought that their baby had a problem (12.12%).

Adolescents who did not attend ANC at all (N = 29) were also questioned (Figure 3). More than two-thirds (68.97%) reported that they did not think it was necessary, while one third (31.3%) of the late attenders reported that they did not know they were pregnant until quite late.

Table 2 contains the results from the bivariate analysis and multivariate logistics regression results for late ANC attendance (> 3 months), respectively. We distinguish between two sets of results, the characteristics of (1) those who accessed care late (> 3 months) versus early (≤ 3 months), and (2) those who did not access care at all (N = 29) and those who accessed public ANC (N = 166).

Respondents were asked if they had planned their current pregnancy and, if not, whether they were happy to be pregnant. Nearly half (45.13%) of the respondents reported that their pregnancy was unintended and that they were unhappy about it. Adolescent women with unintended and unwanted pregnancies were significantly more likely to attend ANC after 3 months as opposed to 3 months or earlier (43.4% vs. 26.9%, p = 0.03). Alarmingly, 93.1% of adolescents who did not attend ANC at all reported that their pregnancy was unplanned and that they were unhappy about it. This is significantly (p = 0.000) higher than the sample of women who attended public ANC (36.75%).

Approximately 1 in 5 women in our sample self-reported consuming alcohol during their pregnancy. These women were significantly (p = 0.018) more likely to attend their first ANC visit after 3 months of gestation (25.25%) than at or before 3 months of gestation (10.45%).

Majority (93.37%) of the adolescents lived within a 30-min travelling distance of the clinics. This figure is skewed towards early ANC attenders. While 40.3% of early attenders lived less than 15 min from a clinic, 23.23% of late attenders lived within the same approximate distance. Among late attenders, 68.69% and 8.08% of adolescents reside within 16 min – 30 min and 31 min – 89 min from their closest clinic, respectively. This is 55.22% and 2.99% for early attenders. The differences are statistically significant (p = 0.042).

Women aged 18 were significantly more likely to access ANC early (48.8% vs. 17.24%, p = 0.000), while women aged 17 (25.3% vs. 31.3%) and 16 (25.9% vs. 51.72%) were significantly less likely to access ANC. Adolescents who reported being in school were more likely to be non-attenders (52.41% vs. 82.76%, p = 0.002). Education is likely to be determined by age among adolescents.

The relationships between late (> 3 months) ANC-seeking behaviour and our range of socio-economic and demographic variables are expressed in Table 3. This analysis only includes respondents who attended public ANC facilities and provided information on our covariates (N = 166). The main predictors of late care attendance include pregnancy intention, alcohol consumption and travel mode. Reporting that your pregnancy is unwanted is a significant and positive correlate of late ANC attendance (coefficient = 0.220, p < 0.05), even after controlling for covariates (Table 3).

Similarly, reporting the consumption of alcohol during pregnancy is positively and significantly associated with late ANC attendance (coefficient = 0.173, p < 0.05), but the variable is insignificant in a multivariate regression (see columns six to eight of Table 3). Adolescents from wealthier households are less likely to access ANC late, but this difference is only significant at a 10% level where we observe the difference in behaviour among those in quintile 2 (coefficient = −0.245, p < 0.1) and quintile 3 (coefficient = −0.213, p < 0.1) compared with those in quintile 1 (see column eight of Table 3). Adolescents who reported that their mode of transport to the health facility was walking were significantly less likely to access care late (coefficient = −0.172, p < 0.05) (see column eight of Table 3).

In the appendix (Appendix 1), we report the results of the analysis drawing on a larger sample, which includes respondents who did not attend ANC at all (n = 195), opposed to the 166 respondents in Table 3. The dependent variable in these regressions is equal to one if the adolescent attended ANC late or not at all. In these new multivariate analyses, we include a binary variable controlling for non-attendance, and exclude travel mode and time as this was not captured for non-attenders. The results for this larger sample remain similar in size and significance. Reporting that a pregnancy was unwanted (coefficient = 0.158, p < 0.05) and reporting the consumption of alcohol during pregnancy (coefficient = 0.145, p < 0.05) are significant and positive correlates of late ANC attendance (Additional File 1).

This study used self-reported data on timing of access to ANC. This measure may suffer from recall and social desirability bias. It would therefore be important in future to obtain clinical estimates of gestation at the time of ANC access. This study analysed data for mothers aged 16–18 years. Antenatal care attendance among mothers younger than 16 years may take place even later in pregnancy. The role of unwanted pregnancies in the timing of ANC access among this group of mothers should be further explored in studies with appropriate ethical design and clearance processes.

Because of an inability to obtain proxy parental consent from the ward councillor for the area in which one of the four facilities were based, this facility had to be dropped from the study. This area was subject to violent service delivery protests at the time the survey was conducted. While adolescents living in this area are likely to be exposed to more everyday incidents of violence, there is insufficient data to support the idea that they would be more socio-economically vulnerable than other adolescents. However, the omission of this facility may have created an inadvertent study bias. We were unable to conduct any supplementary or alternative research in the area because of a lack of permission for data collection in the area.