Sub-national analysis of intimate partner violence and termination of pregnancy among ever-in-union women in South Africa

Abstract

South Africa has one of the highest intimate partner violence (IPV) prevalence in Africa. Studies have linked IPV with negative health and socio-economic outcomes, especially during pregnancy which often led to termination of pregnancy (TOP). However, sub-national differences in the relationship between IPV and TOP have not been adequately examined in South Africa. The study aimed to examine subnational variations of the association between IPV and TOP in South Africa. Data were extracted for 2354 ever-in-union women aged 15–49 years with valid and complete information on the domestic violence module in the 2016 South African Demographic and Health Surveys dataset. The exposure variable was IPV, operationally defined as ever-experienced at least one of the physical, sexual, emotional violence, and partner controlling behaviour IPV forms. The logistic regression analysis was employed, to examine the association between IPV and TOP, while the odds ratios including their 95% confidence interval were reported. Overall, 15.3% had ever terminated pregnancy ranging from 9.5% in Northern Cape to 21.2% in Kwazulu-Natal. About 55.8% of the women had experienced IPV. The IPV, was strongly associated with about a 2-fold increase in TOP at national-level (aOR = 1.62; CI: 1.27–2.07). At subnational-level, women who had experienced IPV were associated with as high as a 2- and 8-fold significant increase in Limpopo and Kwazulu-Natal. The tendency of having a pregnancy terminated was significantly higher among ever-experienced physical violence women resident in Northern Cape (aOR = 4.26; CI: 1.28–14.16) and Kwazulu-Natal (aOR = 4.99; CI: 1.60-15.58) relative to their counterparts who had never experienced physical violence. Among women who had ever experienced emotional and partner control, Northern Cape and North West, and Kwazulu-Natal residents were at higher risk of pregnancy termination, respectively. There was evidence of varying strong associations between IPV, including its forms, and pregnancy termination in multiple provinces. Context-specific maternal health interventions targeted at curbing intimate partner violence may substantially reduce the burden of pregnancy termination in South Africa and other similar settings.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-29356-w.

Introduction

Violence especially against women is human rights abuse with long-term health concerns. A common form of such abuse is the intimate partner violence (IPV), usually suffered by women^1,2. By definition, IPV is a behaviour exhibited within an intimate relationship that causes physical, sexual, or psychological harm, including acts of physical aggression, sexual coercion, psychological abuse, and controlling behaviour³. Studies have shown that IPV has adverse effects on the physical, psychological, and reproductive health of those who experience it^3,4. Intimate partner violence against women can happen at any phase of life, including during pregnancy where it may lead to either spontaneous or induced pregnancy loss. Evidence further shows that IPV is linked to miscarriage and stillbirth through physical trauma, psychological stress or pregnancy complications, and is also associated with increased odds of unintended pregnancies that often result in induced abortion^1,5–7.

In this study, capturing the full spectrum of pregnancy loss potentially associated with IPV, termination of pregnancy (TOP) refers to a pregnancy terminated either through a miscarriage, induced abortion, or stillbirth as used in demographic and health survey⁸. Notably, a high TOP’s prevalence which may largely be unsafe with associated complications has been linked with IPV experience^3,9–12. About 77% of all TOPs in sub-Saharan Africa were classified as unsafe, constituting the highest proportion globally¹³. The region also accounted for 62% of all abortion-related deaths¹⁴. In 2019, the risk of maternal death from an unsafe TOP was estimated at 185 per 100,000 live births in the region¹³. Although the southern Africa region has the least unsafe TOP, maternal mortality in South Africa remains unacceptably high at an estimated 139 per 100,000 live births in 2015^15,16.

Access to quality sexual and reproductive healthcare such as safe TOP services is a key strategy to attain the third Sustainable Development Goals (SDGs) target, especially for improving maternal health. As one of the five leading causes of maternal deaths and associated complications¹⁷, unsafe TOP remains a preventable risk factor that undermines progress toward achieving SDGs 3 and 5, especially in sub-Saharan Africa. South Africa is among the five sub-Saharan African countries with liberal, permissive and right-based abortion laws¹⁸. While laws and policies can shape the quality and scope of healthcare accessible to women, their effective implementation is paramount to advancing sexual and reproductive health outcomes.

The capacity of a nation’s commitment often tends to be more than the program implementation efforts, especially in sub-Saharan Africa. Even though the Choice of TOP Act (CTOPA) was promulgated in 1996, evidence suggests that the CTOPA implementation has been suboptimal in South Africa^15,19–21. Women rarely seek care due to associated barriers of TOP-related services, including inaccessible information on where to obtain an abortion, inadequate awareness of safe legal care options, stigma experienced at health facilities, and a shortage of trained providers^15,22. Regrettably, besides the reported impact of IPV on pregnancy termination, these barriers to accessing safe TOP services persist and often resort to illegal and unsafe practices in South Africa²¹.

Although IPV is a global phenomenon, it is more prevalent in the sub-Saharan Africa region with disparities across countries^1,2,23. In a nationally representative study, one in three women in South Africa had experienced physical IPV at some point in their current relationship²⁴. During pregnancy, violence from an intimate partner can lead to various negative consequences which can result in a threatening situation for the survival of both the mother and the foetus²⁵. The burden of violence during pregnancy and how it leads to TOP is often underestimated because cases are rarely reported²⁵. In communities characterised by patriarchal norms, partner controlling behaviour may undermine women’s reproductive autonomy by limiting their decision-making on fertility and contraceptive use. Such constraints can increase the likelihood of unintended pregnancies and, consequently, pregnancy termination^22–24.

Studies have explored TOP^15,26,27; however, sub-national differences in the relationship between IPV and pregnancy termination among ever-in-union South African women have rarely been reported. As national analysis often conceals the sub-national level scenarios, the current study explored the association between IPV and TOP across South Africa provinces. As an administrative unit, the provincial government holds decision-making authority and is better positioned to apply province-specific findings. The findings may highlight the importance of context-specific planning for robust maternity services that is crucial for attaining the SDGs targets 3.1 and 5.2. Respectively, the SDGs targets were set out as “By 2030, reduce the global maternal mortality ratio to less than 70 per 100,000 live births” and “Eliminate all forms of violence against all women and girls in the public and private spheres, including trafficking and sexual and other types of exploitation”²⁸. Two research questions were consequently addressed: (i) What is the prevalence of TOP among women who had ever experienced IPV across the provinces? (ii) Are there sub-national variations in the association between IPV experience and TOP across provinces?

Methods

Data source and study design

The study utilised data extracted from the 2016 South Africa Demographic and Health Survey (SADHS)⁸. The survey was based on a nationally representative sample carried out by Statistics South Africa with technical assistance from the Inner-City Fund (ICF) International. The survey used a two-stage stratified cluster sampling design. Of 11,083 households interviewed, there were 8,514 women aged 15–49 who participated in the survey. The detailed sampling procedure has been reported previously⁸.

Study population and variables

Of 4,357 women who were interviewed using the domestic violence questionnaire, 2,354 women who were ever-in-union (married or cohabiting) and had complete information on IPV and pregnancy termination were included in this study. Figure 1 presents the study population selection details.

Fig. 1.

Analytical study population’s selection procedure. A flowchart showing the study population’s selection details.

Outcome variable

The study outcome variable was TOP. The item was obtained from the question asking respondents to indicate whether they had ever terminated a pregnancy either through a miscarriage, induced abortion, or stillbirth⁸. The dichotomised responses were “No” and “Yes”; these are coded “0” and “1”, respectively.

Exposure variable

The main exposure variable in this study was IPV. The exposure was assessed using binary indicators of physical, sexual, and emotional violence, as well as controlling behaviour perpetrated by the women’s intimate partners. Consistent with prior studies, controlling behaviour is included in the operational definition of IPV, as its combination with other forms of violence serves as a proxy for coercive violence. It is an important construct in IPV measurement that capture social aspects of gender inequalities^10,24,29. These indicators were derived from questions asking respondents whether they had ever experienced physical, emotional, or sexual violence, or controlling behaviour exercised by their partner.

The physical violence questions were whether the woman’s partner ever: pushed, shook, or threw something at her; slapped her; struck her with his fist or something harmful; kicked or dragged her; strangled or burnt her; threatened her with a knife, gun, or other weapon; and twisted her arm or pulled her hair. The questions on sexual violence included whether the woman’s partner ever: physically forced her into unwanted sex; forced her into other unwanted sexual acts; and forced her to perform sexual acts she did not want. On emotional violence, the questions asked were whether the woman’s partner ever: humiliated her; threatened her with harm; insulted or made her feel bad; and refused to give her money. About the partner controlling behaviour, each woman was asked whether their partner gets jealous or angry if she talks to other men, does not permit her to meet her female friends, accuses her of unfaithfulness, tries to limit her contact with family, and insists on knowing where she is.

The responses for each of these questions were ‘yes’ and ‘no’; coded as “1” and “0”, respectively. A woman who had ever experienced physical, sexual, emotional violence, or controlling behaviour perpetrated by her partner was operationally defined as having at least one positive response (yes) to questions related to any form of violence or controlling behaviour, respectively. The IPV experience was categorised as “ever” if they had ever experienced at least one of the IPV forms; this was coded as “1” if yes, and “0”, otherwise.

Covariates

The main covariate for this study was the province/region of residence. Other covariates controlled for were selected with the guidance of literature based on their associations with TOP^10,11,30. These include current age, education, employment, household wealth status, cohabiting status, age at first birth, contraceptive use, parity, media exposure, ethnicity, and place of residence. For this analysis, the following variables were regrouped. current age (0 = 15–24, 1 = 25–34, 2 = ≥ 35), education (0 = At most primary,1 = Secondary, 2 = Tertiary), cohabiting status (0 = Formerly, 1 = Currently), employment status (0 = Not working, 1 = Working), contraceptive use (0 = Never, 1 = Ever), parity (0 = None, 1 = 1–2, 2 = ≥ 3) and ethnicity (1 = Black/African, 2 = Non-Black). The media exposure variable was generated and recoded as 1 = Exposed if a household had access to at least radio, television, or newspaper; otherwise, 0 = Not exposed. The wealth index variable was derived from the generated weighted factor score using principal component analysis. These scores were coded as 0 = Poor, 1 = Middle, and 2 = Rich wealth tertiles.

Statistical analysis

Data were weighted, due to the complex sampling design used during original data collection, to adjust for differences in population sizes of each region in South Africa. This is necessary to enhance the generalisability of the study’s findings. Descriptive statistics were used to describe TOP by the exposure variable (including its four forms) and other selected covariates. The chi-square (Wald) test statistic was used to assess the association between the outcome variable and the individual covariates. Thereafter, both crude and adjusted binary logistic regression models were separately conducted in each of the nine regions/provinces to investigate the association between IPV and TOP. The (adjusted) odds ratios ((a)OR) and their corresponding 95% confidence interval (CI) are reported. Data were analysed using STATA 17 SE at a 5% significance level.

Ethical approval and consent to participate

The present study analysis utilised a secondary dataset with all identifier information removed, freely available for use in the public domain and it requires no ethics approvals. Ethical approval for the parent study was obtained from the National Ethics Committee in South Africa and the ICF Institutional Review Board⁸. Informed consent was sought from each participant. Respondents were informed of their right to stop the interview at any point if they do not wish to continue. Interviews were thereafter conducted in the language most preferred by respondents. The full details can be found at https://dhsprogram.com. Meanwhile, the Demographic and Health Surveys Program authorised the utilisation of the dataset for the present analysis. All methods for data collection and data analysis were carried out following relevant guidelines and regulations on the protection of participants’ data.

Results

Ever-in-union women’s characteristics

Women’s mean (± sd) age was 35.8 (± 7.8) years (Table 1). Most women were from the Gauteng region (32.9%), Black/African ethnicity (82.7%), and residents in a rural area (71.8%). The majority of the women were formally in union (86.1%) and 71.2% had secondary education. Nearly half of the women were currently unemployed (44.5%) and 54.6% never used contraceptives (Table 1).

Table 1.

Distribution of pregnancy termination according to intimate partner violence, including its four forms across selected covariates.

Characteristics	Physical violence		Sexual violence		Emotional violence		Controlling behaviour		IPV		Total sample
	n	(p) % TOP	n	(p) % TOP	n	(p) % TOP	n	(p) % TOP	n	(p) % TOP	n (%)	(p) % TOP
Age		(0.850)		(0.156)		(0.643)		(0.084)		(0.037*)		(0.021*)
< 25	44	13.0	10	28.6	46	20.6	116	10.8	122	10.7	192 (9.7)	11.6
25–34	141	17.2	32	8.8	156	18.0	470	15.1	498	14.7	890 (38.2)	12.6
≥ 35	171	18.3	48	26.6	249	22.9	628	19.9	686	20.4	1272 (52.1)	18.0
Education		(0.109)		(0.230)		(0.514)		(0.657)		(0.628)		(0.478)
At most primary	74	27.2	16	60.2	79	28.1	203	20.2	220	20.4	378 (13.5)	17.8
Secondary	257	15.8	71	12.6	329	20.1	878	16.3	936	16.4	1659 (71.2)	14.5
Higher	25	4.4	3	41.8	43	17.0	133	17.7	150	18.1	317 (15.2)	16.6
Employment		(0.995)		(0.979)		(0.281)		(0.308)		(0.533)		(0.694)
Working	211	17.2	47	19.8	260	23.1	715	18.4	764	18.0	1338 (55.6)	15.7
Not working	145	17.1	43	19.4	191	17.8	499	15.1	542	16.0	1016 (44.5)	14.8
Wealth status		(0.911)		(0.271)		(0.134)		(0.067)		(0.083)		(0.150)
Poor	165	18.4	45	29.6	194	18.6	496	17.3	530	17.8	863 (33.4)	16.9
Middle	125	15.8	32	13.8	160	15.4	476	12.9	505	13.1	868 (33.3)	12.5
Rich	66	16.4	13	6.9	97	30.2	242	22.2	271	21.4	623 (33.3)	16.5
Cohabiting		(0.522)		(0.363)		(0.783)		(0.646)		(0.772)		(0.506)
Formerly	289	17.9	62	16.0	343	20.4	1013	16.7	1088	17.0	2015 (86.1)	15.0
Currently	67	14.2	28	28.9	108	22.3	201	18.7	218	18.2	339 (13.9)	17.1
Age at first birth		(0.572)		(0.734)		(0.305)		(0.333)		(0.165)		(0.083)
No birth	28	10.4	5	8.1	33	38.0	79	17.1	88	16.8	178 (7.3)	13.2
< 20	183	17.5	53	22.3	217	18.0	552	14.9	590	14.4	985 (40.1)	12.9
≥ 20	145	18.1	32	17.5	201	22.0	583	19.0	628	19.6	1191 (52.5)	17.4
Contraceptive		(0.007*)		(0.024*)		(0.236)		(0.546)		(0.269)		(0.236)
Ever	188	9.8	47	7.0	250	18.1	643	16.2	694	15.5	1261 (54.4)	14.1
Never	168	24.1	43	37.0	201	24.4	571	18.1	612	19.0	1093 (45.6)	16.7
Parity		(0.367)		(0.771)		(0.384)		(0.960)		(0.972)		(0.751)
0	28	10.4	5	8.1	33	38.0	79	17.1	88	16.8	178 (7.3)	13.2
1–2	177	20.4	40	20.1	217	19.4	612	17.4	650	17.5	1179 (52.4)	15.6
≥ 3	151	14.4	45	20.0	201	20.6	523	16.6	568	16.8	997 (40.3)	15.3
Media exposure		(0.678)		(0.599)		(0.975)		(0.226)		(0.246)		(0.081)
Not exposed	12	21.7	4	10.8	13	20.5	47	31.8	49	30.9	83 (3.1)	29.4
Exposed	344	17.0	86	20.0	438	20.9	1167	16.5	1257	16.7	2271 (96.9)	14.8
Ethnicity		(0.847)		(0.905)		(0.058)		(0.033*)		(0.035*)		(0.277)
Black	305	17.0	82	19.7	376	18.1	1094	15.9	1168	16.0	1935 (82.7)	14.8
Non-Black	51	18.3	8	17.7	75	32.7	120	27.5	138	26.5	419 (17.3)	17.8
Residence		(0.308)		(0.096)		(0.553)		(0.124)		(0.130)		(0.243)
Rural	216	15.4	54	13.4	259	22.0	688	15.6	738	15.8	1410 (71.8)	14.7
Urban	140	20.9	36	30.7	192	18.9	526	20.2	568	20.0	944 (28.2)	16.9
Region		(0.027*)		(0.160)		(0.149)		(0.035*)		(0.006*)		(0.047*)
Western Cape	40	15.3	3	33.4	50	23.3	85	17.8	95	17.0	221 (13.6)	15.9
Eastern Cape	69	20.4	8	47.5	84	19.4	163	17.9	177	20.1	268 (10.1)	18.1
Northern Cape	30	23.7	8	0.0	33	24.3	76	14.4	83	13.9	231 (2.2)	9.5
Free State	44	4.5	19	14.7	62	9.6	139	10.0	149	9.4	244 (5.3)	9.9
Kwazulu-Natal	22	44.0	4	30.9	25	34.4	104	33.2	107	32.8	208 (10.9)	21.2
North West	38	31.2	15	35.5	45	40.4	160	18.6	168	18.7	262 (7.5)	17.6
Gauteng	29	11.4	11	2.3	30	17.2	147	13.0	155	13.4	289 (32.9)	13.6
Mpumalanga	57	5.3	12	14.8	69	11.2	175	12.0	193	10.9	308 (8.5)	13.7
Limpopo	27	10.2	10	35.5	53	23.3	165	18.1	179	18.4	323 (9.1)	14.5
Total	356	17.2	90	19.6	451	20.9	1214	17.1	1306	17.2	2354	15.3

n – unweighted number of women; % – weighted percentage; IPV – intimate partner violence (at least one of the four IPV forms); p – p-value; *significance at 5%.

Prevalence of intimate partner violence

Collectively, over a half (55.8%) of the women had experienced IPV, ranging from 40.4% in Northern Cape to 69.2% in North West (Table 1; Fig. 2). Women who had experienced partner-controlling behaviour (51.8%) constituted the highest percentage of respondents, followed by emotional (17.9%) and physical (14.3%) IPVs, while those who had experienced sexual violence constituted the least (4.2%). The pattern was similar across the nine provinces (Fig. 3).

Fig. 2.

Prevalence of intimate partner violence and termination of pregnancy among ever-in-union women by region. A map showing the percentage distribution of women who had experienced IPV (left panel) and who had terminated pregnancy (right panel), created using ArcGIS software version 10.8.1, Esri Inc.

Fig. 3.

Pattern of intimate partner violence among ever-in-union women by region. A graphical chart showing the percentage distribution of women who had experienced IPV, including its four forms (physical, sexual, emotional violence and partner-controlling behaviour) by province.

Pattern of termination of pregnancy by region of residence and other covariates

Overall, 15.3% of the women had ever terminated pregnancy (Table 1). With the highest percentage for women aged ≥ 35 years, pregnancy termination significantly increased with age among the respondents (18.0%; p = 0.021) and among women who had ever experienced IPV (20.4%; p = 0.037). The TOP prevalence was significantly associated with region of resident (p = 0.047), ranging from 9.5% (Northern Cape) to 21.2% (Kwazulu-Natal) (Table 1; Fig. 2). With a 17.2% prevalence among women who were victims of IPV, region of residence was independently associated with TOP (p = 0.006), ranging from 9.4% (Free state) to 32.8% (Kwazulu-Natal). The proportions of pregnancy termination among the women were relatively similar across the four IPV forms. A similar pattern of TOP prevalence and its association with region of residence (p < 0.05) was observed among those who had reported physical (17.2%) and controlling behaviour (17.1%) IPV forms. Notably, the highest proportion of TOP was observed among women who had ever experienced emotional (20.9%), slightly followed by sexual (19.6%) IPV (Table 1). Besides the women’s age and region of residence, only contraceptive use and ethnicity were significantly associated with TOP (p < 0.05). Specifically, the prevalence of pregnancy termination was common among non-contraceptive users relative to contraceptive users who had experienced physical (24.1 vs. 9.8%) and sexual (37.0 vs. 7.0%) violence. Non-black women who had experienced IPV (26.5 vs. 16.0%) and those who had experienced partner controlling behaviour (27.5 vs. 15.9%) were more prone to pregnancy termination compared to their counterparts (Table 1).

Association between intimate partner violence and termination of pregnancy across the regions of residence

According to the women experience of IPV and its four forms, the crude odds ratios of pregnancy termination are presented in Table 2 and the adjusted odds of pregnancy termination are presented in Table 3; Fig. 4 (detailed adjusted odds for the covariates considered are presented in Supplementary Tables S1 – S5).

Table 2.

Unadjusted odds ratio of pregnancy termination among ever-in-union women exposed to intimate partner violence and its four forms.

Province	Physical violence	Sexual violence	Emotional violence	Controlling behaviour	IPV
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Western Cape	1.07 (0.43–2.64)	2.53 (0.22–28.62)	2.15 (1.00-4.63)*	1.65 (0.81–3.37)	1.50 (0.74–3.05)
Eastern Cape	1.19 (0.60–2.38)	2.79 (0.64–12.10)	1.21 (0.63–2.32)	0.92 (0.49–1.73)	1.36 (0.69–2.67)
Northern Cape	3.34 (1.17–9.51)*	na	3.83 (1.40-10.48)*	1.76 (0.70–4.45)	1.89 (0.75–4.75)
Free State	0.35 (0.08–1.54)	1.65 (0.45–6.08)	0.87 (0.33–2.27)	1.24 (0.54–2.85)	1.02 (0.44–2.36)
Kwazulu-Natal	3.37 (1.29–8.79)*	1.67 (0.17–16.51)	2.15 (0.82–5.63)	6.32 (2.49–15.99)*	7.48 (2.77–20.19)*
North West	1.62 (0.71–3.71)	2.59 (0.84–7.98)	3.12 (1.50–6.47)*	1.06 (0.55–2.05)	1.15 (0.58–2.26)
Gauteng	1.43 (0.51–4.02)	0.65 (0.08–5.24)	1.02 (0.33–3.10)	0.85 (0.43–1.69)	0.96 (0.48–1.89)
Mpumalanga	0.49 (0.19–1.31)	1.15 (0.24–5.41)	1.11 (0.53–2.31)	0.99 (0.52–1.86)	0.82 (0.43–1.56)
Limpopo	0.68 (0.20–2.35)	3.97 (1.08–14.63)*	1.38 (0.64–2.97)	1.80 (0.96–3.37)	2.02 (1.05–3.87) *
Pooled sample	1.19 (0.88–1.62)	1.70 (1.02–2.84)*	1.61 (1.23–2.10)*	1.43 (1.13–1.80)*	1.52 (1.20–1.92) *

*Significant at 5%; na – not computed due to teenage pregnancy not reported; OR – Odds ratio; 95% CI – 95% confidence interval; IPV – intimate partner violence.

Table 3.

Adjusted odds ratios for the association between IPV, including its four forms, and TOP among ever-in-union women in South Africa.

Characteristic	Model a	Model b1	Model b2	Model b3	Model b4
	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)
IPV
Never (R)	1
Ever	1.62(1.27–2.07)a
Physical
Never (R)		1
Ever		1.24(0.91–1.70)
Sexual
Never (R)			1
Ever			1.79(1.06–3.01)c
Emotional
Never (R)				1
Ever				1.63(1.24–2.14)a
Controlling behaviour
Never (R)					1
Ever					1.54(1.21–1.96)a
Age
< 25 (R)	1	1	1	1	1
25–34	1.65(0.95–2.85)	1.61(0.93–2.78)	1.60(0.93–2.77)	1.64(0.95–2.83)	1.64(0.95–2.84)
≥ 35	2.08(1.20–3.58)b	2.04(1.18–3.51)c	2.02(1.17–3.47)c	2.05(1.19–3.53)c	2.08(1.21–3.59)b
Education
At most primary (R)	1	1	1	1	1
Secondary	0.90(0.65–1.24)	0.91(0.66–1.25)	0.89(0.65–1.23)	0.90(0.65–1.23)	0.90(0.65–1.23)
Higher	1.06(0.66–1.68)	1.07(0.67–1.71)	1.06(0.67–1.69)	1.06(0.67–1.70)	1.06(0.67–1.69)
Employment
Working (R)	1	1	1	1	1
Not working	0.95(0.74–1.21)	0.94(0.74–1.21)	0.94(0.74–1.20)	0.95(0.74–1.21)	0.95(0.74–1.21)
Wealth status
Poor (R)	1	1	1	1	1
Middle	0.94(0.71–1.25)	0.95(0.71–1.25)	0.95(0.72–1.26)	0.96(0.72–1.27)	0.94(0.71–1.24)
Rich	0.82(0.56–1.20)	0.80(0.54–1.17)	0.80(0.54–1.17)	0.81(0.55–1.18)	0.82(0.56–1.20)
Cohabiting
Formerly (R)	1	1	1	1	1
Currently	1.14(0.83–1.57)	1.17(0.85–1.61)	1.14(0.83–1.58)	1.09(0.79–1.51)	1.15(0.83–1.58)
Age at first birth
No birth (R)	1	1	1	1	1
< 20	0.69(0.44–1.10)	0.70(0.44–1.11)	0.70(0.44–1.11)	0.70(0.44–1.11)	0.69(0.43–1.09)
≥ 20	0.87(0.55–1.36)	0.88(0.56–1.37)	0.88(0.56–1.37)	0.88(0.56–1.38)	0.86(0.55–1.35)
Contraceptive
Ever (R)	1	1	1	1	1
Never	1.16(0.91–1.47)	1.16(0.91–1.47)	1.16(0.91–1.47)	1.17(0.92–1.49)	1.15(0.91–1.47)
Ethnicity
Black (R)	1	1	1	1	1
Non-Black	1.46(1.05–2.04)c	1.31(0.94–1.81)	1.32(0.95–1.82)	1.29(0.93–1.79)	1.45(1.04–2.02)c
Residence
Rural (R)	1	1	1	1	1
Urban	1.15(0.88–1.51)	1.16(0.88–1.51)	1.16(0.88–1.51)	1.14(0.87–1.50)	1.16(0.88–1.52)
p	0.000	0.026	0.012	0.001	0.001
n	2354	2354	2354	2354	2354

^a p < 0.001; ^b p < 0.01, ^c p < 0.05; (a)OR – (adjusted) odds ratio; CI–confidence interval; R - reference category; IPV – intimate partner violence; parity variable was excluded due to its collinearity with age at first birth; media exposure variable was excluded because, on average, about 96% of the respondents had access to media. Model a – model association between IPV and TOP while adjusting for the listed covariates. Model b1-b4 – model association between physical, sexual, emotional and controlling behaviour IPV forms, respectively.

Fig. 4.

Risk of pregnancy termination among ever-in-union women exposed to IPV, and the four IPV forms. A graphical display of the adjusted odds of pregnancy termination among women had experienced IPV, including its four forms.

Odds of pregnancy termination across the provinces was significantly higher among women who had ever experienced IPV (OR = 1.52; CI: 1.20–1.92), including among those who had experienced sexual, emotional and controlling behaviour IPV, compared to those who never experienced IPV (Table 2). The association of IPV and its forms with pregnancy termination relatively varied across the regions. Kwazulu-Natal (OR = 7.48; CI: 2.77–20.19) and Limpopo (OR = 2.02; CI: 1.05–3.87) women who had experienced IPV were about seven and two times more likely to terminate pregnancy relative to their counterparts who had never had IPV, respectively. Relative higher odds of TOP were observed among women who had experienced physical violence and controlling behaviour in Kwazulu-Natal. Similar odds of TOP were found among women who had experienced sexual violence in Limpopo and among emotional abused women in Northern Cape. Relatively, these odds of TOP remained unchanged but became non-significant among women who had experienced sexual IPV in Limpopo and among those who were emotional violated in Western Cape, after controlling for the included covariates, respectively (Supplementary Table S3, Table S4).

Other associations remained similar even after controlling for other included covariates (Table 3). Specifically, at the national level, women who had experienced IPV (aOR = 1.62; CI: 1.27–2.07) were found to have higher odds of pregnancy termination (Table 3; Fig. 4). At provincial level, Kwazulu-Natal (aOR = 7.60; CI: 2.61–22.18) and Limpopo (aOR = 2.25; CI:1.12–4.51) residents who had ever experienced IPV were more prone to terminate pregnancy, respectively. While the risk of TOP among women who had experienced IPV in Kwazulu-Natal was comparable to that of their counterparts in Western Cape, Eastern Cape, Northern Cape and Limpopo, it was significantly higher than that observed in other provinces (Fig. 4).

The significant association was also replicated among those who had experienced sexual (aOR = 1.79; CI: 1.06–3.01), emotional (aOR = 1.63; CI: 1.24–2.14) and controlling behaviour (aOR = 1.54; CI: 1.21–1.96) forms of IPV (Fig. 4). In the case of the women who had ever experienced physical IPV, the tendency of having a pregnancy terminated was significantly higher among women who dwelled in Northern Cape (aOR = 3.71; CI: 1.28–14.16) and Kwazulu-Natal (aOR = 4.99; CI: 1.60-15.58) relative to their counterparts who had never experienced physical violence. The odds of TOP among women who had physical IPV in Kwazulu-Natal was comparable to those observed in other provinces, except for Free State. For emotional IPV, the risk of pregnancy termination among women in Northern Cape was comparable to that observed in other provinces, with the exception of the Free State. Specifically, Northern Cape (aOR = 5.63; CI: 1.74–18.15) and North West (aOR = 3.01; CI:1.35–6.74) women who had experienced emotional violence were six and three times more likely to terminate pregnancy, respectively, compared to those who had never experienced emotional violence in both settings. On the other hand, relative to women who had never experienced partner controlling behaviour in Kwazulu-Natal, women who had ever experienced partner controlling behaviour were more likely to terminate pregnancy (aOR = 6.35; CI: 2.31–17.49). The significant association was stronger that that observed at the national level (Fig. 4).

Notably, association between physical IPV and TOP was statistically non-significant at national level. Neither the IPV nor any of its forms was significantly associated with pregnancy termination in Western Cape, Eastern Cape, Free State, Gauteng or Mpumalanga provinces. Women age and ethnicity had strong mediating effects on TOP at national level (Table 3). At the sub-national level, contraception had a strong mediating effect on TOP, particularly showing decreased odds of TOP in Free State and increased odds in North West (Supplementary Tables S1 – S5).

Discussion

The present study investigated association between IPV and TOP across South Africa provinces among ever-in-union women, using 2016 SADHS nationally representative data. At national level, approximately three in twenty women reported having ever terminated a pregnancy. Yet, this prevalence masked substantial sub-national variation, with Kwazulu-Natal recording the highest and Northern Cape the lowest. While the national prevalence is comparable with a prevalence of 16% reported for sub-Saharan African countries^31,32, it is higher than previously reported 11% for South Africa³² and 9% for Mozambique³³. Such disparities underscore the importance of examining sub-national contexts rather than relying solely on national estimate. Kwazulu-Natal disproportionately high burden of pregnancy-related deaths in South Africa may partly explain the province’s higher prevalence of TOP¹⁶. On the other hand, Northern Cape’s lower prevalence could reflect differences in health-seeking behaviours and limited TOP service availability, which may reduce utilisation and increase the likelihood of under-reporting³⁴.

Largely unsafe pregnancy termination remains a preventable risk factor of maternal death especially in sub-Saharan Africa, South Africa inclusive^{15,19,21,35,36}. Yet, women exposure to IPV are at a heightened risk for unintended pregnancies and the subsequent decision to terminate, as IPV has been implicated as an important risk factor for pregnancy termination^3,4,10,11. Concurrently, this study confirmed the pervasive nature of IPV in South Africa, with about half of ever-in-union women reporting ever experienced some forms of IPV, with the highest prevalence reported in North West and the least in Northern Cape province. While the reported national-level prevalence is comparable to 56.9% found in Ethiopia²⁴, it is higher than 43% found in Zimbabwe, Zambia and Kenya, respectively^37,38. Also, it is substantially higher than nearly one-fourth reported for South Africa in sub-Saharan Africa studies^37,39,40. While the difference perhaps may partly be due to the inclusion of controlling behaviour in the IPV assessment, it could also point to a systemic issue, deeply rooted in institutions, cultures and traditions that transcends demographic and geographic boundaries^15,41,42.

The pattern of IPV was broadly consistent across the nine provinces - controlling behaviour was most common, followed by emotional violence, while sexual violence was least reported. This pattern aligns with findings among Armenian and Ethiopian women^10,24, but contrasts with contexts where emotional and sexual violence are more prevalent³⁸. Such contrasts underscore how cultural and societal structure shape the perpetration and reporting of IPV. The relatively high prevalence of controlling behaviour and emotional violence may suggest the normalisation of partner dominance and psychological, while the comparatively low reporting of sexual violence may be linked to cultural norms that normalise sexual coercion in marital relationships and discourage disclosure^11,38. Understanding IPV within these contexts is important, as overlapping forms of IPV may intersect and heighten their adverse effects on women’s health and well-being^24,29. These findings therefore highlight the need to conceptualise IPV not only as individual experiences but also as products of cultural and societal environments that shape women’s willingness to report abuse.

In this study, an association between IPV and TOP was observed. The IPV, including sexual, emotional or controlling behaviour IPV form, was significantly associated with about a 2-fold increase in pregnancy termination among women across all provinces. This finding is comparable with evidence from previous studies conducted elsewhere^5,7,12,24,39. However, it contrasts a study among adolescent girls and young women in sub-Saharan Africa that reported no such association in South Africa³⁹. The contrasting results may possibly be explained by methodological differences, particularly focusing on ever-in-union women and the inclusion of controlling behaviours in the IPV assessment in the present study. The finding demonstrates that IPV, regardless of its form, undermines reproductive autonomy and contributes to unintended pregnancies.

Yet sub-national differences were evident. The strongest associations between IPV and TOP were found in KwaZulu-Natal, while provinces such as the Free State showed no significant associations. Limpopo and KwaZulu-Natal women who had experienced IPV were between two and eight times more likely to terminate a pregnancy, with particularly high risks in KwaZulu-Natal where IPV and TOP were notably prevalent. KwaZulu-Natal also demonstrated the strongest associations between different forms of IPV (such as physical and controlling behaviour) and TOP, a finding supported by earlier research^10,23,31. These findings are not surprising given the contextual realities: KwaZulu-Natal has long struggled with maternal health challenges, while Limpopo’s entrenched patriarchal norms and health system constraints may exacerbate the reproductive consequences of IPV^16,43,44. Physical and emotional violence showed particularly strong effects in Northern Cape; only emotional violence was substantially associated with TOP in North West, with women at least three times more likely to terminate pregnancies compared with those not emotionally abused. This finding suggests that emotionally manipulative relationships may be contributing to covert forms of violence that affect women’s reproductive choices.

The higher rates of pregnancy termination among women who had experienced sexual and emotional violence suggest a significant impact of these forms of IPV on women’s reproductive choices and autonomy. Across all provinces but Gauteng and Mpumalanga, though the association was not significant, women who had experienced sexual violence were more likely to terminate pregnancy compared with non-sexually abused participants. Conversely, the association was significantly associated with about 2-fold increase in pregnancy termination among South African women. This association aligns with similar studies conducted among Armenian and Nigerian women^10,12. This may indicate that the trauma of sexual violence could lead to higher rates of unwanted pregnancies and subsequent terminations. Furthermore, the absence of significant associations in Western Cape, Eastern Cape, Free State, Gauteng and Mpumalanga may be attributable to underreporting. However, it may also suggest the presence of mitigating factors, such as better service coverage or stronger social support systems. In particular, Statistics South Africa reported that Western Cape and Gauteng women had the most medical aid coverage in 2019^15,41.

In all, the findings suggest that IPV, regardless of forms, places women at higher risk of pregnancy termination, though the magnitude of the association varies across provinces. The consistent national-level association underscores IPV as a structural determinant of reproductive health, while the strong provincial effects in KwaZulu-Natal and Limpopo highlight the need for context-specific interventions. In high-burden provinces, integrating IPV prevention with reproductive health services could reduce both unintended pregnancies and unsafe pregnancy terminations. In provinces with weaker associations, strengthening reporting mechanisms and improving culturally appropriate reproductive health care may be more critical. Furthermore, mediating effects such as contraceptive use and age were important considerations, corroborating previous studies^{12,13,32,33,45}. Non-contraceptive users exposed to violence were especially vulnerable to terminating pregnancies, reflecting both a lack of reproductive autonomy and gaps in access to family planning. Older women also appeared to face heightened risks, potentially due to the pervasive nature of controlling behaviours within intimate relationships^10,24. These findings reinforce global evidence that access to contraception and reproductive health education can buffer the adverse effects of IPV.

Collectively, these results highlight the need for region-specific prevention and intervention strategies, with special focus on addressing IPV as a determinant of women’s reproductive autonomy and maternal health outcomes. Tailored interventions, including improved access to contraceptives and targeted reproductive health education, may significantly reduce the burden of unsafe or induced pregnancy terminations in South Africa.

Study limitations

The study shed light on the subnational differences in the relationship between intimate partner violence and pregnancy termination in South Africa. In addition, the usage of a large and nationally representative dataset allows the study findings to be generalisable across the country. Furthermore, the inclusion of controlling behaviour is a key construct in IPV measurement that provided a more comprehensive measure of women lived experiences of violence. And as a proxy of coercive control, capturing the broader dynamics of power and domination within relationships, controlling behaviour enhances the construct validity. This study thus remains on the frontline to assessing the relationship between IPV and TOP among ever-in-union women across provinces in South Africa. The present study, however, acknowledges some limitations. First, the cross-sectional study design limits the potential of making causal inferences. Second, the actual burden of IPV and TOP might have been underestimated due to women’s self-reported data without any means of verification which could be a potential avenue of recall and social desirability biases, with underreporting particularly likely for sensitive issues such as sexual violence and abortion. Furthermore, the study limited the choices of independent variables, and excluded never-married women, which may underestimate the IPV–TOP association. Lastly, provincial-level analyses were constrained by small sample sizes for certain IPV forms, limiting statistical power to detect significant associations. Despite these limitations, the study provides robust evidence of a strong and consistent association between IPV and pregnancy termination in South Africa and highlights important provincial differences that have direct policy relevance.

Conclusions

This study quantified the association between IPV and pregnancy termination across the different subnational populations in South Africa. The findings have shown that intimate partner violence plays an important role in the provincial differences observed in the risk of pregnancy termination. Both IPV and pregnancy termination are relatively common across the provinces. In particular, Kwazulu-Natal had the highest proportion (followed by Eastern Cape and North West) of pregnancy termination. TOP was relatively common and comparable across the subnational levels. IPV is associated with a 2-fold risk of pregnancy termination in multiple provinces. These results suggest a considerable sub-national difference in the risk of TOP among women who had ever experienced IPV across provinces.

By and large, these findings underscore the urgent need for comprehensive policies and programs addressing the interplay between IPV and reproductive health in South Africa. By recognizing the subnational differences and the role of IPV in shaping women’s reproductive choices, stakeholders can develop effective province-specific interventions that enhance women’s safety and autonomy. It is therefore anticipated that interventions addressing intimate partner violence will likely lead to more effective control of illegitimate termination of pregnancy in South Africa, and subsequently, enhance the SDGs targets 3.1 and 5.2 attainment²⁸.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

IPV

Intimate partner violence

TOP

Termination of pregnancy

(a)OR

(adjusted) Odds ratio

CI

95% confidence interval

SDGs

Sustainable Development Goals

TFR

Total fertility rate

SADHS

South Africa Demographic and Health Survey

ICF

Inner-City Fund

EAs

Enumeration areas

CTOPA

Choice of TOP Act

Author contributions

RFA conceived and designed the study. RFA analysed the data and interpreted the analysed data. RFA drafted the original manuscript. RFA and MEP reviewed and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

Data availability

Data sharing does not apply to this article as no new data were created or analysed in this study. The data is publicly available and openly accessible at: https://dhsprogram.com/.

Declarations

Competing interests

The authors declare no competing interests.