The original sample was 109,022 women between 15 and 49 years old. We selected women from each study who reported having been married, completed the intimate partner violence questionnaire, and were interviewed in a private setting without the interference of other family members or their husbands. The final sample for the analyses included 63,321 women (Final samples by country: Colombia = 24,890, Dominican Republic = 5,801, Haiti = 6,650, Honduras = 12,497, and Peru = 13,485).

• Sociodemographic characteristics of women: 1) Woman’s age in years. 2) Woman’s highest educational level was measured in four categories (No education = 0, Basic school = 1, High school = 2, and University studies = 3) 2) Women’s working status (1 = Yes, and 0 = No).

For the overall sample, the average female age was 33.1 years (33.09–33.14), and these means were similar among countries. Concerning the educational level, a higher proportion of women who finished high school were found in Colombia (Proportion = P [95% CI] = 45.7% [45.1–46.3]). The highest proportion of women with university studies was found in Colombia (P [95% CI] = 27.1% [ 26.5–27.6]) and the lowest proportion of women who finished university studies were in Haiti (P [95% CI] = 3.5% [3.1–4.0]) and Honduras (P [95% CI] = 5.0% [4.6–5.4]) (Table 1).

The proportion of poorest population was greater in Colombia (P [95% CI] = 27.0% [26.5–27.6]), and Honduras (P [95% CI] = 25.5% [24.7–26.3]) compared with the other countries. The place of residence was more frequently urban in Colombia (P [95% CI] = 72.9% [72.4–73.5) and in the Dominican Republic (P [95% CI] = 70.7% [69.5–71.8]), while the highest proportion of the population living in rural areas (P [95% CI] = 61.6% [60.7–62.4]) was found in Honduras (Table 1).

Concerning husband characteristics, the mean age was 37.4 (95% CI: 37.2–37.40) years, and this average was similar between the three countries (Table 1).

• Sample Description: We describe the sample according to demographic characteristics. Even though previous reports had described the DHS sample characteristics, we presented a descriptive analysis of the sample used for the current article because the analysis was performed in a subsamples (i.e., women between 15- to 49-years-old, married, and who completed the IPV questionnaire). We calculated proportions using percentages and 95% confidence intervals (CI). For women and husband/partner age, we used the mean and 95% CI.

To assess the typologies of IPV, we conducted an LCA exploring models ranging from 1 to 4 classes. We chose the model with three latent classes because this model was parsimonious and had a better fit. We named the three classes according to the level of violence as high-level, middle-level, and non-IPV classes.

Women in the high-level IPV class were victims of high levels of control, physical violence, and were aggressors of their partners/husbands.

Women in the high-level IPV class had a higher education than women in other classes; their husbands used more alcohol and predominantly lived in urban settings. Women in the middle-level of IPV class suffered high levels of control but low levels of physical and sexual violence. Women in the middle-level and high-level IPV classes justified IPV more, and there was a higher proportion of women without education than in the high-level IPV class. Women and their husbands/partners from the middle- and non-IPV classes had a higher proportion of no education. Finally, participants in the non-IPV class had not been victims or aggressors of IPV forms.

The typologies we obtained in our analysis differed from those described by Johnson [11] and Strauss [15]. The high-level categories could be similar to the mutual violence control type described by Johnson [11] because this class also had high levels of all forms of intimate partner violence. The non-IPV class was similar to the cluster of situational violence described by Johnson as the lower level of violence. However, in our model, women were not aggressors of their partners. These results were congruent with those shown in three systematic reviews about variables associated with IPV [23–25].

This paper is the first study to address IPV typologies in Latin-American and low-middle income countries, settings with scarce prior evidence and datasets that previously lacked research on this topic. This research provides information about different countries and describes the different identified classes according to demographic and social characteristics. We also included different forms of victimization as well as female aggression to the partner/husband, providing a more compressive depiction of the violence dynamics in partnerships.

Among the limitations of this study, we discuss the bias due to differences in the sampling, bias because of the differences in study focus, misclassification of the outcome, women underreporting domestic violence, and cultural differences regarding domestic violence among countries.

First, we consider that bias resulting from sampling differences are minimal because these surveys were all collected based on multistage sampling procedures performed similarly between countries, since Macro International uses similar sampling methodologies and women selection to guarantee comparability between countries. In addition, surveys were collected in different years; however, their intimate partner violence is not a seasonal event.Instead, IPV is a health outcome explained by structural risk factors. Also, the surveys included the same target population allowing comparisons of countries. The target populations are women aged 12–49 in all the surveys from all education and socioeconomic levels across countries.

Second, bias due to differences in the study aims and survey focus is minimal. The focus of the surveys was similar among countries; they aimed to estimate demographic indicators, women’s health behavior, intimate partner violence, pregnancy, immunizations, and practices in Sexual and Reproductive Health (Supplementary Table S1). Specifically, regarding the aim of IPV, we found that all countries’ surveys sought to describe the proportion of women victims of IPV.

The third limitation is the possible misclassification of the outcome. We considered that a misclassification bias of IPV could arise because questionnaires are not validated. However, the questionaries about IPV are similar among countries. The questions regarding IPV and demographic variables used in this analysis were comparable between counties. Macro international used the same questionaries and methodologies across countries to maintain comparability. For this reason, there are publications including multiple countries using this data [26]. The team also harmonized the questionaries by reviewing each question-and-answer category.

The fourth limitation is the potential for information bias that could arise from women not reporting their actual exposure to IPV due to fears surrounding reporting aggression, which would result in underestimating the proportion of domestic violence and could affect the identification of IPV typologies.

Finally, the cultural and social differences among the counties studied could affect IPV reporting due to more permissive attitudes or cultural acceptance of different forms of violence. Information bias can arise because the participants could be more reluctant to report being victims of domestic violence in those countries where IPV was more accepted or where women are less empowered about their rights. The underreporting of domestic violence could generate a different distribution of latent classes. Honduras has a marked patriarchal culture that increases the likelihood of indigenous populations underreporting domestic violence [27]. The Dominican Republic is a Caribbean country where gender inequities persist, as do rigid gender roles and a high level of social acceptance of IPV [28]. In Colombia, the gender power imbalance driven by social inequalities, low empowerment of women’s rights, and low education level persists [29, 30]. In Peru, low education levels, unemployment, and a history of family violence are important risk factors for IPV victims [31]. Peru has a high proportion of women from ethnic minorities, with higher levels of violence and more economic and social disparities [32]. Haiti had high levels of poverty and violence. In addition, social norms maintain gender imbalances, women’s poverty, and acceptance of IPV, leading to reduced reporting of IPV [33, 34]. For all these differences among countries, we conducted an overall latent class model and tested the multigroup latent class in which we found classes were similar between countries.

Policies and services should be prepared to identify and treat couples according to their typology, implement screening tests to classify couples according to these typologies, and perform interventions according to their type. First, policies should screen women in health and well-being services to determine which type of IPV class they belong to. Second, policies should include not only strategies to punish the IPV aggressors but also promote health and early prevention strategies. For instance, we need more strategies to empower women and shift those cultural beliefs that IPV is acceptable, thus hampering the vicious cycle that perpetuates domestic violence through generations. Also, policies should include health promotion programs that reduce the socioeconomic gaps by providing equal opportunities in employment and education for both genders. Third, policies should create and provide funding for differential interventions that consider the IVP typologies. Women in the high-level IPV class should be included in integrative interventions that remove women and their children from the violent environment and provide free access to legal and mental health services. Policies for the middle-level should include education for women and their husbands, conflict resolution, and women’s empowerment.

Women with middle-levels of IPV could benefit from an intervention that addresses IPV justification and increases women’s education level. Those couples with high levels of IPV should benefit more from extensive interventions that include counseling, legal services, and alcohol reduction. As women and their husbands/partners in this class had high education levels and were wealthy, women may benefit from interventions to empower their rights and conflict resolution. Women with high levels of IPV may also need mental health and stress management interventions to address the possible consequences of this intervention’s mental health. Finally, both high and middle-level IPW could benefit from interventions that improve communication, conflict resolution, and stress management.

We found three typologies of IPV: high-level, middle-level, and non-IPV. Policymakers should create programs to prevent and treat different forms of violence rather than treating domestic violence as a uniform phenomenon. First, according to these typologies, policies should create screening methods that detect and intervene in the population. Second, policies should include strategies to punish IPV aggressors and promote health and early prevention strategies that reduce IPV approval, increase women’s empowerment about their rights, and increase their socioeconomic and education opportunities. Police should intervene according to IPV levels. Women who had high levels of IPV could benefit from programs that include legal and mental health services, alcohol reduction, cultural change, and women’s rights empowerment. Women who are victims of middle levels of IPV could benefit from interventions to reduce the justification of IPV and increase women’s education levels. Women with middle- and high-IPV may need interventions to increase conflict resolution skills and stress management.