Determinants of intimate partner violence among women of reproductive age: a cross-sectional study in Türkiye

Abstract

Background

Intimate partner violence (IPV) is a major global public health issue that adversely affects women’s physical, psychosocial, and reproductive health. Although individual characteristics have been widely studied, relational dynamics and intergenerational experiences of violence remain less explored, particularly in the Turkish context. Examining these factors together may contribute to a more comprehensive understanding of IPV and inform prevention strategies.

Aim

This study aimed to examine IPV exposure among women of reproductive age in Türkiye and to identify sociodemographic, partner-related, and relational factors associated with such exposure.

Methods

A descriptive, cross-sectional study was conducted between 15 November 2024 and 15 January 2025 with 453 women aged 18–49 years who had a current partner. Participants were recruited through snowball sampling via five social media platforms. Data were collected using a researcher-developed online questionnaire comprising sociodemographic and partner-related characteristics and a single-item measure of exposure to IPV. Statistical analyses included the chi-square test, the Mann–Whitney U test, and Firth penalized logistic regression to address sparse-data bias and quasi-complete separation. Multicollinearity was assessed using variance inflation factors. Internal validation was performed using bootstrap optimism correction (B = 500) and 10-fold cross-validation repeated 10 times. Statistical significance was set at p < .05.

Results

Overall, 11.3% of participants reported IPV exposure. Firth penalized logistic regression showed that witnessing the partner’s violent behavior toward others (OR = 94.42; 95% CI: 6.06–1471.46), low relationship satisfaction (OR = 24.71; 95% CI: 4.43–137.92), a history of family violence (OR = 15.48; 95% CI: 5.21–46.02), lower income relative to expenses (OR = 19.12; 95% CI: 4.93–74.14), and increasing age (OR = 1.27 per year; 95% CI: 1.17–1.38) were independently associated with IPV exposure. The model yielded a Nagelkerke R² of 0.746 and an optimism-corrected AUC of 0.966. Given the limited number of IPV events and the wide confidence intervals for several estimates, the magnitude of the observed associations should be interpreted cautiously.

Conclusions

IPV exposure in this study was associated with individual, relational, socioeconomic, and prior violence-related characteristics. The observed associations are consistent with a multilevel understanding of IPV and highlight the relevance of considering relational and family experiences alongside individual and socioeconomic characteristics in IPV prevention and assessment.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12905-026-04946-6.

Background

Intimate partner violence (IPV) is one of the most common forms of violence against women and is recognized as a major public health problem worldwide. IPV is defined as a form of violence involving physical assault, sexual coercion, emotional abuse, and controlling behaviors perpetrated by a current or former spouse or romantic partner, resulting in physical, sexual, psychological, or economic harm [1, 2]. Globally, approximately one in three women (30%) experience physical or sexual violence at least once in their lifetime, and most of it is perpetrated by a current or former partner [2, 3]. Such violent behaviors may occur as isolated incidents or as chronic patterns of abuse that persist over many years [4].

Intimate partner violence has severe and multidimensional negative consequences for women’s physical, psychological, and reproductive health. Research indicates that IPV is associated with numerous health problems, including depression, anxiety, post-traumatic stress disorder, chronic pain, unintended pregnancies, and sexually transmitted infections [2, 3, 5–7].

The occurrence of IPV is shaped by the interaction of individual, relational, societal, and cultural factors. This multidimensional structure is commonly addressed within the framework of the ecological model, which emphasizes the need to evaluate factors at the individual, relationship, community, and societal levels [8]. The literature shows that factors such as younger age, low educational level, unemployment, economic hardship, alcohol use, relationship conflict, and a history of family violence are associated with IPV [9–11]. In particular, witnessing or experiencing violence during childhood may lead to the normalization of violence in adult relationships and increase both the risk of victimization and perpetration [5, 12–14]. This pattern is consistent with social learning theory, which addresses the intergenerational transmission of violence [15, 16].

In addition to individual and socioeconomic characteristics, relational dynamics between partners play a crucial role in the occurrence of IPV. Factors such as relationship dissatisfaction, jealousy, controlling behaviors, and power imbalances between partners are among the key determinants of IPV [8]. Furthermore, economic dependency and lack of social support may hinder women from leaving abusive relationships and increase their risk of continued exposure to violence [3].

Violence against women remains a significant social and public health issue in Türkiye. Studies indicate that approximately one-third of women in Türkiye have experienced violence at some point in their lives [17–19]. Additionally, research suggests that gender inequality, traditional family structures, and cultural norms may contribute to the persistence of IPV in Türkiye [20–22].

Although several studies conducted in Türkiye have examined the prevalence and correlates of intimate partner violence, relatively fewer studies have simultaneously evaluated individual, relational, socioeconomic, and intergenerational factors within a multivariable analytical framework informed by the ecological model. A more comprehensive understanding of how these factors are associated with IPV may help inform prevention strategies and support the development of targeted interventions. Accordingly, this study aimed to examine IPV exposure among women of reproductive age in Türkiye and to identify sociodemographic, partner-related, and relational factors associated with such exposure. In doing so, the present study gives particular attention to relational characteristics, notably relationship satisfaction and a partner’s wider pattern of violent behavior, which have received comparatively little attention in the Turkish IPV literature relative to sociodemographic correlates.

Research questions

1. What proportion of participants reported exposure to intimate partner violence?

2. What sociodemographic, partner-related, and relational factors are associated with intimate partner violence exposure?

Methods

Study design and population

This study was designed as a descriptive, cross-sectional study aimed at determining exposure to intimate partner violence (IPV) and its associated factors among women of reproductive age in Türkiye. The study was reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for cross-sectional studies.

The target population consisted of women aged 18–49 years who were residing in Türkiye and had a current partner (married or in a committed relationship) at the time of the study.

As the total size of this population could not be precisely determined, the required sample size was calculated using Cochran’s formula for an unknown population proportion, assuming a 95% confidence level, a 5% margin of error, and an expected proportion of 50% (p = .50). The minimum required sample size was calculated as 384 participants. To account for potentially incomplete questionnaires and exclusions based on the eligibility criteria, recruitment continued beyond the minimum required sample size. A total of 502 women responded to the online questionnaire. Of these, 49 were excluded because of incomplete questionnaires or failure to meet the eligibility criteria, resulting in a final analytic sample of 453 women. No missing data were present in the variables included in the analyses.

Participants were recruited using a snowball (chain-referral) sampling method, which was considered appropriate given the sensitive nature of the research topic and the difficulty of reaching women experiencing IPV through probability-based methods [2, 3]. Although recruitment was open to eligible women residing in Türkiye, the non-probability sampling approach did not produce a nationally representative or regionally stratified sample, and the geographical distribution of participants was not controlled. The sample should therefore be interpreted as a heterogeneous non-probability sample of women residing in Türkiye rather than as representative of all women of reproductive age in the country.

Inclusion criteria were: (1) being between 18 and 49 years of age, (2) residing in Türkiye during the study period, (3) having access to the internet and a social media account, (4) having a current partner (married or in a committed relationship), and (5) providing voluntary informed consent to participate.

Exclusion criteria were: (1) not meeting any of the inclusion criteria, (2) submitting an incomplete questionnaire.

Data collection process

Data were collected between 15 November 2024 and 15 January 2025 through an online survey. The questionnaire was hosted on Google Forms (Google LLC, USA), and all responses were collected electronically through this platform. The questionnaire, prepared by the researchers, was distributed via five social media platforms (Twitter^®, Instagram^®, Facebook^®, WhatsApp^®, and Telegram^®), which were selected due to their widespread use among women of reproductive age in Türkiye.

Recruitment followed a snowball sampling procedure. In the first wave, the survey link was shared with eligible women in the researchers’ social networks. These initial participants were then asked to forward the link to other eligible women in their own networks, thereby expanding the sample through successive waves until the target sample size was exceeded. No financial or material incentives were offered for participation.

Before beginning the questionnaire, all participants were presented with an invitation text explaining the purpose of the study, the voluntary nature of participation, the estimated completion time, the confidentiality and anonymity of responses, and the right to withdraw at any stage. Submission of the completed online questionnaire was accepted as indication of informed consent. The questionnaire took approximately 5 to 7 min to complete.

Data collection tools

Data were collected using a Personal Information Form developed by the researchers based on a comprehensive review of the national and international literature on intimate partner violence [8, 23, 24]. The draft questionnaire was reviewed by three experts in the field to assess the relevance, clarity, and comprehensiveness of the items. Based on their recommendations, the questionnaire was revised and finalized. The questionnaire was then pilot tested in a small group of women from the target population (n = 15) to assess clarity, comprehensibility, and feasibility prior to the main data collection. The pilot data were not included in the final analysis. The form was structured to collect study-specific descriptive information and consisted of 22 questions organized into two sections, together with a separate item assessing the primary outcome.

The first section included 13 questions assessing sociodemographic and background characteristics such as age, marital status, educational level, employment status, family type, having children, monthly household income, place of residence, parental education levels, alcohol use, smoking status, and history of family violence.

The second section included 8 questions related to partner and relationship characteristics, including partner’s age, educational level, employment status, alcohol use, smoking status, duration of the current relationship, relationship satisfaction, and witnessing the partner’s violent behavior toward others. The relationship satisfaction variable referred to the participant’s overall satisfaction with her current intimate partner relationship rather than sexual satisfaction specifically. Participants were asked to rate their overall relationship satisfaction, which was subsequently categorized as high or low satisfaction for the analysis. The item on witnessing the partner’s violent behavior toward others asked whether the participant had observed her partner behaving violently toward people other than herself.

Exposure to intimate partner violence was operationalized as a binary variable based on participants’ self-report. Participants were classified as ‘exposed to IPV’ if they responded ‘yes’ to experiencing at least one form of physical, psychological/emotional, sexual, or economic violence from their current partner during the relationship; participants who responded “no” were classified as “not exposed to IPV.” This classification was based on a single yes/no self-report item rather than a set of behaviorally specific questions. Because classification depended on whether participants themselves identified and reported having experienced such violence from their current partner during the current relationship, this measure reflects self-recognized IPV. It therefore captures a narrower construct than behaviorally specific, act-based instruments and may not include experiences that participants did not label as violence.

Data analysis

Data were analyzed using IBM SPSS Statistics version 26 (IBM Corp., Armonk, NY, USA). Descriptive statistics were presented as frequencies and percentages for categorical variables, and as means with standard deviations for continuous variables. The normality of continuous variables was assessed using the Kolmogorov–Smirnov test. Bivariate associations with exposure to intimate partner violence were examined using the chi-square test for categorical variables and the Mann–Whitney U test for non-normally distributed continuous variables.

Multivariable associations with intimate partner violence exposure were examined using Firth penalized logistic regression, which was selected to address sparse-data bias and quasi-complete separation arising from low cell frequencies in key predictor categories [25]. Unadjusted Firth logistic regression analyses were first performed for all candidate predictors to provide an overview of their individual associations with IPV exposure. The primary multivariable model included five variables representing individual (age), relational (relationship satisfaction and witnessing partner violence), intergenerational (history of family violence), and socioeconomic (income status) domains of the ecological framework. Because the participant’s age showed substantial overlap with marital status, having children, duration of the current relationship, and the partner’s age, age was retained in the model as a representative variable to reduce multicollinearity-related instability. In the primary model, variance inflation factor (VIF) values ranged between 1.01 and 1.41, and the events-per-variable ratio was 10.2 [26].

Model performance was evaluated using the Hosmer–Lemeshow goodness-of-fit test, Nagelkerke R², and classification measures including overall accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC). Internal validation was performed using two approaches: bootstrap optimism correction (B = 500 replicates) and 10-fold cross-validation repeated 10 times. Calibration was assessed using the calibration slope. The Brier score was computed as an additional measure of overall prediction error. Adjusted odds ratios with 95% confidence intervals are reported. Given the limited number of IPV events (n = 51) and the wide confidence intervals for several estimates, the magnitude of the observed associations was interpreted cautiously. Statistical significance was set at p < .05.

Ethical considerations

Ethical approval was obtained from the Non-Interventional Clinical Research Ethics Committee of Kütahya University of Health Sciences, Türkiye (Date: 24 October 2024; Approval No: 2024/12 − 06). The study was conducted in accordance with the principles of the Declaration of Helsinki. All participants were provided with a written information text at the beginning of the online questionnaire, explaining the purpose of the study, the voluntary nature of participation, the approximate completion time, the anonymity and confidentiality of responses, the right to withdraw at any stage without consequence, and the contact details of the research team. Informed consent was obtained electronically, and submission of the completed questionnaire was considered as indication of consent to participate. No personally identifying information was collected. Given the sensitive nature of the research topic, contact information for the Alo 183 Social Support Helpline (the national helpline for women, children, and families operated by the Turkish Ministry of Family and Social Services) was provided at the end of the questionnaire for participants who might need support.

Results

The relationship between participants’ sociodemographic characteristics and exposure to intimate partner violence (IPV) is presented in Table 1. Of the 453 women included in the study, 11.3% (n = 51) reported exposure to IPV, whereas 88.7% (n = 402) reported no IPV exposure. The mean age of women reporting IPV exposure was 36.12 ± 5.39 years, compared with 25.19 ± 7.68 years among those reporting no IPV exposure. Among women reporting IPV exposure, 58.8% (n = 30) were aged ≥ 35 years, 35.3% (n = 18) were aged 25–34 years, and 5.9% (n = 3) were aged 18–24 years (Table 1). In addition, 58.8% were married, 76.5% were employed, 82.4% had children, and 72.5% reported a history of family violence (Table 1).

Table 1.

Sociodemographic characteristics of participants by exposure to Intimate Partner Violence (IPV) (N = 453)

Variable	n	%	Yes (IPV exposure)	No (No IPV exposure)	Test / p
Age
18–24 years	264	58.3	3	261	χ²(2) = 68.463 p < .001*
25–34 years	87	19.2	18	69
≥35 years	102	22.5	30	72
Marital status
Not married	282	62.3	21	261	χ²(1) = 9.876 p=.002*
Married	171	37.7	30	141
Education level
High school	240	53.0	6	234	χ²(2) = 39.346 p < .001*
Bachelor’s degree	132	29.1	27	105
Postgraduate	81	17.9	18	63
Employment status
Unemployed	276	60.9	12	264	χ²(1) = 32.016 p < .001*
Employed	177	39.1	39	138
Family type
Extended family	33	7.3	0	33	χ²(1) = 3.382 p= .066
Nuclear family	420	92.7	51	369
Having children
Yes	135	29.8	42	93	χ²(1) = 73.062 p < .001*
No	318	70.2	9	309
Income status
Income less than expenses	72	15.9	18	54	χ²(2) = 16.221 p < .001*
Income equal to expenses	306	67.5	27	279
Income greater than expenses	75	16.6	6	69
Residence
Rural	48	10.6	3	45	χ²(2) = 9.361 p= .009*
District	120	26.5	6	114
City center	285	62.9	42	243
Mother’s education
Primary or below	318	70.2	42	276	χ²(2) = 4.058 p= .131
High school	90	19.9	6	84
Associate/Bachelor’s	45	9.9	3	42
Father’s education
Primary	213	47.0	39	174	χ²(2) = 20.868 p < .001*
High school	135	29.8	9	126
Higher education	105	23.2	3	102
Alcohol use
Yes	144	31.8	21	123	χ²(1) = 1.874 p= .171
No	309	68.2	30	279
Smoking status
Yes	123	27.2	15	108	χ²(1) = 0.048 p= .827
No	330	72.8	36	294
History of family violence
Yes	88	19.4	37	51	χ²(1) = 99.827 p < .001*
No	365	80.6	14	351

IPV was defined as self-reported exposure to physical, psychological/emotional, sexual, or economic violence from a current partner during the relationship. *p < .05

IPV exposure differed significantly according to age group (χ²(2) = 68.463, p < .001), with a higher proportion of IPV exposure among women aged ≥ 35 years (29.4%) and 25–34 years (20.7%) than among those aged 18–24 years (1.1%). Married women reported a higher proportion of IPV exposure than women who were not married (17.5% vs. 7.4%; χ²(1) = 9.876, p = .002). IPV exposure also differed according to educational level (χ²(2) = 39.346, p < .001), with higher proportions among women with postgraduate (22.2%) and bachelor’s-level education (20.5%) than among those with a high school education (2.5%). Employed women reported IPV exposure more frequently than unemployed women (22.0% vs. 4.3%; χ²(1) = 32.016, p < .001), and the proportion was higher among women with children than among those without children (31.1% vs. 2.8%; χ²(1) = 73.062, p < .001).

IPV exposure differed significantly according to income status (χ²(2) = 16.221, p < .001), with the highest proportion observed among women whose income was lower than their expenses (25.0%), compared with those whose income was equal to (8.8%) or greater than (8.0%) their expenses. Differences were also observed according to place of residence (χ²(2) = 9.361, p = .009); women living in city centers reported a higher proportion of IPV exposure (14.7%) than those living in districts (5.0%) or rural areas (6.2%). IPV exposure differed according to paternal education (χ²(2) = 20.868, p < .001), with the highest proportion among women whose fathers had primary-level education (18.3%), compared with high school (6.7%) or higher education (2.9%). Women with a history of family violence reported IPV exposure more frequently than those without such a history (42.0% vs. 3.8%; χ²(1) = 99.827, p < .001). No statistically significant associations were observed for family type, maternal education, alcohol use, or smoking status (p > .05) (Table 1).

The relationship between participants’ partner-related characteristics and exposure to intimate partner violence (IPV) is presented in Table 2. Among partner-related variables, the mean age of partners of women exposed to IPV (38.51 ± 5.77 years) was significantly higher than that of women not exposed to IPV (27.23 ± 8.37 years) (U = 3017.5, p < .001). The prevalence of IPV was 30.0% among women whose partners were aged ≥ 35 years, compared to 9.1% among those aged 25–34 years and 1.3% among those aged 18–24 years, with a statistically significant difference (χ²(2) = 68.229, p < .001). Women whose partners were employed reported IPV exposure more frequently than those whose partners were unemployed (15.1% vs. 2.2%; χ²(1) = 14.455, p < .001).

Table 2.

Partner-related characteristics according to exposure to Intimate Partner Violence (IPV) (N = 453)

Variable	n	%	Yes (IPV exposure)	No (No IPV exposure)	Test / p
Partner age (continuous)
IPV group	51	Mean ± SD	38.51 ± 5.77		U = 3017.5 p < .001*
Non-IPV group	402	Mean ± SD	27.23 ± 8.37
Partner age group
18–24 years	224	49.4	3	221	χ²(2) = 68.229 p < .001*
25–34 years	99	21.9	9	90
≥35 years	130	28.7	39	91
Partner employment
Unemployed	135	29.8	3	132	χ²(1) = 14.455 p < .001*
Employed	318	70.2	48	270
Partner alcohol use
Never	261	57.6	24	237	χ²(2) = 3.878 p=.144
1–5 times/month	162	35.8	21	141
Weekly	30	6.6	6	24
Partner smoking
Yes	183	40.4	21	162	χ²(1) = 0.000 p = 1.000
No	270	59.6	30	240
Relationship duration
1–5 years	285	62.9	15	270	χ²(3) = 40.394 p < .001*
6–10 years	69	15.2	18	51
11–15 years	60	13.2	6	54
≥16 years	39	8.6	12	27
Partner education
Primary/High school	90	19.9	15	75	χ²(1) = 2.647 p=.104
Higher education	363	80.1	36	327
Relationship satisfaction
High satisfaction	426	94.0	30	396	χ²(1) = 146.592 p < .001*
Low satisfaction	27	6.0	21	6
Witnessing partner violence
Yes	21	4.6	18	3	χ²(1) = 114.500 p < .001*
No	432	95.4	33	399

IPV was defined as self-reported exposure to physical, psychological/emotional, sexual, or economic violence from a current partner during the relationship. *p < .05

IPV exposure also differed significantly according to relationship duration (χ²(3) = 40.394, p < .001). The highest proportion was observed among women in relationships lasting ≥ 16 years (30.8%), followed by 6–10 years (26.1%), 11–15 years (10.0%), and 1–5 years (5.3%). Among women reporting low relationship satisfaction, 77.8% reported IPV exposure, compared with 7.0% of those reporting high relationship satisfaction (χ²(1) = 146.592, p < .001). Similarly, IPV exposure was reported by 85.7% of women who had witnessed their partner’s violent behavior toward others, compared with 7.6% of those who had not (χ²(1) = 114.500, p < .001). No statistically significant associations were observed for partner alcohol use, partner smoking status, or partner educational level (p > .05) (Table 2).

Unadjusted Firth logistic regression analyses were first conducted for all candidate variables (Supplementary Table S1). Examination of cell frequencies for key binary variables revealed sparse distributions, particularly for low relationship satisfaction and witnessing partner violence toward others. Among the 27 women reporting low relationship satisfaction, 21 reported IPV exposure and six did not. Among the 21 women who had witnessed their partner’s violent behavior toward others, 18 reported IPV exposure and three did not (Table 3). These sparse cell distributions supported the use of Firth penalized logistic regression.

Table 3.

Cell frequencies of key binary variables according to IPV exposure (N = 453)

Variable	Category	IPV exposure, Yes	IPV exposure, No	Total
History of family violence	Yes	37	51	88
	No	14	351	365
Low relationship satisfaction	Yes	21	6	27
	No	30	396	426
Witnessing partner violence toward others	Yes	18	3	21
	No	33	399	432
Low income status	Yes	18	54	72
	No	33	348	381

Sparse cell distributions were particularly evident for low relationship satisfaction and witnessing partner violence toward others. Low income status was defined as income lower than expenses

The primary multivariable Firth model included five variables representing individual, relational, intergenerational, and socioeconomic domains of the ecological framework. Participant age showed substantial overlap with marital status, having children, relationship duration, and partner age; therefore, age was retained as the representative variable in the primary model. VIF values ranged from 1.01 to 1.41, and the EPV ratio was 10.2.

In the primary Firth penalized logistic regression model, age (OR = 1.27, 95% CI: 1.17–1.38, p < .001), history of family violence (OR = 15.48, 95% CI: 5.21–46.02, p < .001), low relationship satisfaction (OR = 24.71, 95% CI: 4.43–137.92, p < .001), witnessing partner violence toward others (OR = 94.42, 95% CI: 6.06–1471.46, p = .001), and low income status (OR = 19.12, 95% CI: 4.93–74.14, p < .001) remained associated with IPV exposure after adjustment (Table 4). The confidence intervals were wide for several estimates, particularly for witnessing partner violence and low relationship satisfaction, indicating substantial imprecision in the estimated magnitude of these associations.

Table 4.

Firth penalized logistic regression analysis of factors associated with intimate partner violence exposure (N = 453)

Variable	B	SE	z	p	OR	95% CI
Constant	−12.020	1.701	−7.068	< 0.001*	–	–
Age (continuous)	0.243	0.042	5.820	< 0.001*	1.275	1.175–1.384
History of family violence	2.740	0.556	4.930	< 0.001*	15.482	5.209–46.016
Low relationship satisfaction	3.207	0.877	3.656	< 0.001*	24.714	4.429–137.918
Witnessing partner violence	4.548	1.401	3.246	< 0.001*	94.424	6.059–1471.463
Low income status	2.951	0.692	4.266	< 0.001*	19.116	4.929–74.142

Dependent variable: exposure to intimate partner violence (0 = No, 1 = Yes). B regression coefficient, SE Standard error, z z-statistic, OR Odds ratio, CI Confidence interval. Firth penalized logistic regression was used to address sparse-data bias and potential quasi-complete separation. Model performance: Nagelkerke R² = 0.746; Hosmer–Lemeshow χ²(8) = 9.737, p = .284; accuracy = 96.0%; sensitivity = 70.6%; specificity = 99.3%; apparent AUC = 0.968; optimism-corrected AUC = 0.966 (bootstrap B = 500); repeated 10-fold cross-validated AUC = 0.963; calibration slope = 1.077; EPV = 10.2. VIF values ranged from 1.01 to 1.41. *p < .05

The primary model yielded a Nagelkerke R² of 0.746 and an apparent AUC of 0.968. Internal validation showed a bootstrap optimism-corrected AUC of 0.966 (optimism = 0.002; B = 500) and a mean repeated 10-fold cross-validated AUC of 0.963 (SD = 0.040). The calibration slope was 1.077, and the Brier score was 0.031 for the apparent model and 0.033 after optimism correction.

In the sensitivity analysis, an extended Firth model additionally included education, employment status, partner alcohol use, and urban residence (Supplementary Table S2). Among these additional variables, higher education was associated with IPV exposure (OR = 8.26, 95% CI: 1.18–57.85), whereas employment status, partner alcohol use, and urban residence were not statistically significant. The extended model had an EPV of 5.7 and showed marked changes in several odds ratio estimates compared with the primary model, suggesting greater parameter instability when additional covariates were included.

Participant age was strongly correlated with partner age (r = .885, p < .001), and having children (rpb = 0.782, p < .001), and was also highly correlated with relationship duration (r = .699, p < .001), Partner age and relationship duration jointly accounted for 79.8% of the variability in participant age. These findings indicate substantial overlap among age- and relationship-related variables and support cautious interpretation of the association between participant age and IPV exposure.

Discussion

This study examined IPV exposure and associated individual, relational, and socioeconomic factors among women of reproductive age in Türkiye using multivariable analyses. The findings were interpreted within the ecological model and social learning theory to consider IPV in relation to characteristics at different levels. Overall, IPV exposure was associated with individual, socioeconomic, relational, and family-related characteristics. This pattern is consistent with the ecological model, which conceptualizes IPV within interconnected individual, relational, community, and broader societal contexts [8].

The proportion of participants reporting IPV exposure (11.3%) was lower than global and national estimates. According to the World Health Organization, approximately 27% of ever-partnered women aged 15–49 years have experienced physical and/or sexual IPV during their lifetime [2]. Higher estimates have also been reported in national studies conducted in Türkiye [17–19]. Differences in reported IPV estimates across studies may be related to variations in sampling and assessment approaches. In addition, stigma and underreporting are recognized challenges in estimating violence against women [2]. The lower proportion observed in the present study may therefore be related to both the characteristics of the study sample and the approach used to assess IPV exposure.

Several sociodemographic characteristics were associated with IPV exposure in the bivariate analyses. Higher proportions of IPV exposure were observed among older and married women and those with children. The association with age is particularly notable given the substantial overlap between participant age, partner age, relationship duration, and having children in the present study. Thus, age may partly represent broader life-course and relationship characteristics, including longer relationship duration, rather than an isolated age-related association. Previous systematic reviews have reported associations between IPV and multiple individual and relationship characteristics, including age-related and relationship factors, although the patterns vary across populations and study designs [10, 11]. The higher proportions of IPV exposure among married women and women with children may similarly be considered within the context of longer relationship histories and intersecting demographic and socioeconomic characteristics. Research across low- and middle-income countries has reported variation in IPV exposure according to multiple demographic and socioeconomic characteristics [27]. Together, these findings suggest that the observed age, marital status, and childbearing patterns may be better understood within a broader life-course and relationship context. It should also be noted that the analytic sample was predominantly young, unmarried, and without children, whereas the women who reported IPV exposure formed a smaller, older, and mostly married and parenting subgroup. Several of the observed associations therefore reflect characteristics that distinguished this subgroup from the larger sample rather than uniform risk gradients across the whole population.

The higher proportions of IPV exposure observed among women with higher education and those who were employed were notable, as education and employment are often considered within the socioeconomic context of IPV. Previous research suggests that the associations between women’s economic position and IPV may vary across sociocultural settings and household contexts [28]. Gender norms and inequalities may also be relevant to understanding differences in IPV across social contexts [29, 30]. In the present study, employment status was not statistically significant in the extended sensitivity model, suggesting that the bivariate association with employment may reflect overlapping socioeconomic or relational characteristics. The higher proportion of reported IPV exposure among more highly educated women may also be related to differences in the recognition or disclosure of violence; however, this possibility was not assessed in the present study. Recent evidence from a large low- and middle-income country sample similarly indicates that women’s education, media access, and autonomy are associated with a lower justification of intimate partner violence and may help challenge restrictive gender norms [31]. This is consistent with the possibility that higher education relates not only to a greater likelihood of recognizing and disclosing partner violence but also to less accepting attitudes toward it.

Low income status was associated with IPV exposure in the multivariable Firth model. This finding is consistent with previous research reporting associations between socioeconomic disadvantage and IPV [8, 27]. Economic hardship may accompany other adverse household circumstances relevant to women’s experiences of IPV. Previous research has also suggested that the relationship between women’s economic circumstances and IPV may vary according to gender norms and household contexts [28]. Together with previous research, the present finding highlights the relevance of socioeconomic circumstances when examining IPV alongside relational and family-level characteristics.

Father’s education, but not mother’s education, was associated with IPV exposure in the bivariate analyses. The higher proportion of IPV exposure among women whose fathers had lower educational levels may be related to broader family, generational, and socioeconomic characteristics. Family and sociocultural contexts, including gender norms and inequalities, have also been considered in relation to IPV in previous research [8, 30]. Father’s education was not included in the primary multivariable model and may represent broader family or socioeconomic characteristics not directly assessed in the present study. This bivariate finding highlights the potential relevance of family and generational contexts and warrants further investigation.

No statistically significant associations were observed between women’s alcohol use or smoking status and IPV exposure. Previous studies have reported associations between IPV and substance use, although these relationships may vary according to the population studied and the characteristics and measurement of substance use and violence [2, 5]. The broad alcohol use and smoking categories used in the present study may not have captured more specific patterns related to frequency, duration, or severity of substance use.

Several partner- and relationship-related characteristics were associated with IPV exposure in the bivariate analyses. Higher proportions of IPV exposure were observed among women with older and employed partners and among those with longer relationship durations. In the present study, partner age and relationship duration were strongly correlated with participant age, suggesting substantial overlap among these age- and relationship-related characteristics. Previous research has reported associations between IPV and multiple partner and relationship characteristics, with variation across populations and study designs [10, 11]. The observed bivariate associations with partner age and relationship duration may therefore be considered within this broader relationship and life-course context. No statistically significant associations were observed between IPV exposure and partner alcohol use, smoking status, or education.

Low relationship satisfaction was associated with IPV exposure in the multivariable Firth model. Relationship characteristics have been examined in relation to IPV in previous research, although the specific relational factors considered vary across studies [10, 11]. Research on attachment-related characteristics has also reported associations with physical IPV victimization and perpetration [32]. In the present study, the relationship between low relationship satisfaction and IPV exposure may be bidirectional: IPV exposure may be related to lower relationship satisfaction, while low satisfaction may coexist with conflictual or otherwise difficult relationship dynamics. Given the cross-sectional design, the temporal direction of this association cannot be established. Nevertheless, the observed association highlights the relevance of relationship context when examining IPV. More broadly, low relationship satisfaction may be at least as much a consequence of ongoing violence as a contributing factor, and it is therefore more appropriately regarded as a correlate or marker of a troubled relationship than as an independent determinant of IPV.

Witnessing a partner’s violent behavior toward others was also associated with IPV exposure in the multivariable Firth model. This finding may be considered within social learning theory, which proposes that behavior can be learned through observation and social experience [15]. The observed association may also indicate that violence toward others and IPV coexist within broader patterns of violent behavior. However, the present study did not assess the development or learning of violent behavior and therefore cannot establish the mechanisms underlying this association. The finding highlights the potential relevance of a partner’s violent behavior toward others when examining IPV exposure. This item asked whether participants had observed their partner behaving violently toward people other than themselves, and it may therefore reflect a partner’s broader propensity for violence rather than a mechanism specific to IPV.

A history of family violence was also associated with IPV exposure in the multivariable Firth model. Previous longitudinal and recent studies have reported associations between childhood or parental violence exposure and IPV experiences in later relationships [12, 13, 16]. Such associations have been considered within social learning and intergenerational frameworks, which propose links between earlier family experiences, learned behavioral patterns, and later relationship contexts [15, 33]. However, recent scholarship has emphasized that intergenerational “transmission” of IPV should not be understood as a deterministic or uniform process and that the pathways underlying associations across generations remain complex [14]. The association observed in the present study is consistent with the potential relevance of earlier family violence experiences to IPV exposure, while highlighting the importance of considering these experiences within broader relational and sociocultural contexts.

The magnitude of several odds ratio estimates in the primary Firth model was large, particularly for low relationship satisfaction and witnessing a partner’s violent behavior toward others. These estimates were based on sparse cell distributions, as shown in Table 3, which may have contributed to the wide confidence intervals. Firth penalized logistic regression was used to reduce sparse-data bias and address potential quasi-complete separation [25]. Nevertheless, considerable uncertainty remained regarding the precise magnitude of several associations. Accordingly, the magnitude of the odds ratio estimates should be interpreted with caution. In particular, for witnessing a partner’s violence toward others, only 18 of the 21 women who reported this experience were classified as exposed to IPV, so the corresponding estimate rests on very few participants and is highly unstable. These odds ratios should therefore be read as indicating the direction of the association only; their magnitude cannot be interpreted as a meaningful effect size, and the wide confidence intervals should be given greater weight than the point estimates.

The primary Firth model showed high discrimination within the study sample (AUC = 0.968). The optimism-corrected AUC (0.966) and repeated 10-fold cross-validated AUC (0.963) were similar to the apparent estimate, suggesting limited optimism in model discrimination. The calibration slope of 1.077 was also close to the ideal value of 1. These findings suggest that model performance was relatively stable across the internal validation procedures; however, performance in other populations requires evaluation in independent samples. The high apparent discrimination is largely attributable to a small number of strong but sparsely distributed predictors rather than to broad predictive accuracy across the sample. The model was intended to be explanatory, identifying factors statistically associated with IPV exposure, rather than to serve as a predictive or screening tool, and the reported AUC and R² should not be interpreted as evidence of clinical predictive utility.

The extended sensitivity model provided additional context for the stability of the primary model estimates. Higher education was associated with IPV exposure in the extended model, whereas employment status, partner alcohol use, and urban residence were not statistically significant. However, the lower EPV and marked changes in several odds ratio estimates indicated greater parameter instability when additional covariates were included. These findings support the use of the more parsimonious primary model as the main basis for interpretation, while also indicating that additional sociodemographic and partner-related characteristics should be examined in larger studies with more IPV events.

Overall, the findings are consistent with an ecological perspective in which IPV is associated with characteristics across individual, socioeconomic, relational, and family contexts rather than with a single factor. Age and low income status were associated with IPV exposure alongside relational characteristics, including low relationship satisfaction and witnessing a partner’s violent behavior toward others. The association with a history of family violence is also consistent with the potential relevance of intergenerational perspectives, while the finding related to witnessing a partner’s violent behavior toward others may be considered within a social learning framework. Taken together, the findings highlight the value of considering IPV within a multilevel framework encompassing individual, socioeconomic, relational, family, and broader sociocultural contexts. Such a perspective may inform comprehensive prevention approaches that consider women’s socioeconomic circumstances, relationship contexts, and family experiences.

Limitations and strengths

This study has several limitations that should be considered when interpreting the findings. First, the cross-sectional design precludes causal inferences; although the analyses identified factors associated with exposure to intimate partner violence, the temporal sequence and directionality of these associations cannot be established. Second, IPV exposure was assessed using a single dichotomous self-report item rather than a validated multi-item instrument. Although this approach provided a brief assessment of IPV exposure, it may not have fully captured the multidimensional nature of IPV, particularly psychological and economic forms that participants may not readily recognize or identify as violence. Consequently, the proportion of IPV exposure observed in this study should not be directly compared with estimates from nationally representative surveys or studies using validated multi-item instruments. Future studies should employ validated multi-item measures to provide a more comprehensive assessment of IPV. Third, the use of self-reported data may have introduced social desirability and recall biases, which are well-documented in research on sensitive topics such as partner violence. Fourth, the online snowball sampling strategy may have introduced selection bias by preferentially including women with internet access, social media use, and willingness to participate in an online survey. Women with limited digital access, lower educational levels, rural residence, or those living under controlling or abusive partner circumstances may therefore have been underrepresented. Accordingly, the study sample represents a non-probability sample and should not be considered nationally representative; the findings should not be generalized to all women of reproductive age in Türkiye. Fifth, the relatively low number of IPV events (n = 51) and sparse distributions in some variable categories resulted in wide confidence intervals for several odds ratio estimates. Although Firth penalized logistic regression was used to reduce sparse-data bias and address potential quasi-complete separation, uncertainty remains regarding the precise magnitude of several associations. Finally, the questionnaire did not assess potentially relevant factors, including childhood-specific experiences of abuse, mental health status, social support, and help-seeking behavior; therefore, the analyses could not encompass the full range of factors relevant to IPV within an ecological framework.

Despite these limitations, the study has several strengths. The simultaneous examination of individual, socioeconomic, relational, and family-related characteristics within an ecological framework provides a multilevel perspective on factors associated with IPV exposure in the study sample. The use of Firth penalized logistic regression addressed sparse-data bias and potential quasi-complete separation, while the sensitivity analysis allowed the stability of the primary estimates to be examined when additional theoretically relevant covariates were considered. Methodological transparency was further strengthened by reporting cell frequencies for key variables, multicollinearity diagnostics, and the events-per-variable ratio. In addition, bootstrap optimism correction and repeated 10-fold cross-validation provided internal validation of model discrimination. Finally, reporting in accordance with the STROBE guidelines supports transparency and reproducibility.

Conclusion

The findings of this study highlight the multifactorial context of IPV exposure and its associations with individual, socioeconomic, relational, and family-related characteristics. Age, low income status, low relationship satisfaction, a history of family violence, and witnessing a partner’s violent behavior toward others were associated with IPV exposure in the multivariable Firth model. These findings are consistent with an ecological perspective that considers IPV across multiple levels of context and with social learning and intergenerational perspectives on previous exposure to violence.

The findings underscore the importance of considering socioeconomic circumstances, relationship context, and family violence experiences in IPV prevention and support efforts. Healthcare professionals, particularly nurses and primary care providers, may have an important role in recognizing possible indicators of IPV and facilitating access to appropriate support services. Prevention approaches may also benefit from attention to healthy relationship skills, non-violent conflict resolution, and family experiences of violence. Future studies using longitudinal designs, validated multidimensional IPV measures, and more diverse samples are needed to clarify the temporal direction of the observed associations and examine their applicability across different populations. Qualitative research may further contribute to understanding women’s experiences and the relational and sociocultural contexts of IPV.

Abbreviations

CI

Confidence Interval

IPV

Intimate Partner Violence

Max

Maximum

Min

Minimum

OR

Odds Ratio

SD

Standard Deviation

VIF

Variance Inflation Factor

WHO

World Health Organization

Authors’ contributions

Author Contributions Conceptualization; N.Ç., R.A. Data; N.Ç., R.A. Formal analysis; N.Ç., R.A. Investigation; N.Ç., R.A. Methodology; N.Ç., R.A. Resources and software; N.Ç., R.A. Validation; N.Ç., R.A. Writing - original draft; N.Ç., R.A. Writing-review & editing; N.Ç., R.A.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit.

Data availability

The datasets used and/or analyzed during the current study are available from the first author upon reasonable request.

Declarations

Ethics approval and consent to participate

was obtained from the Non-Interventional Clinical Research Ethics Committee of Kütahya University of Health Sciences, Türkiye (Date: October 24, 2024; Approval No: 2024/12 − 06). Informed consent was obtained electronically via Google Forms. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable. No identifying images or personal/clinical details of participants are included in this manuscript.

Competing interests

The authors declare no competing interests.