Spatial variation and factors associated with women’s injury due to intimate partner violence in Madagascar: Evidence from the 2021 demographic and health survey

Abstract

Background

Intimate partner violence (IPV) is a violation of human rights that affects the physical, mental, and social well-being of women. It reduces their productivity and quality of life and results in injuries. Despite its serious impact, there is limited evidence about the magnitude, spatial variation, and determinants of women’s injury in Madagascar.

Objective

This study aimed to find the spatial variation of women’s injury due to IPV and its determinants in Madagascar.

Design

Community-based cross-sectional.

Methods

This study was conducted using the Madagascar Demographic and Health Survey 2021. A total of 1515 women (weighted) aged 15-49 who experienced IPV were included. The data were cleaned and analyzed using STATA version 17. ArcGIS version 10.8 software was used for mapping the special variation, hot spot analysis, and spatial autocorrelation. Spatial cluster detection of injury was conducted through SaTScan version 10.1 software. A multi-level binary logistic regression model was used to determine factors associated with injury.

Results

Women’s injury due to IPV showed geographical variations in Madagascar. The prevalence of injury was 37.4% (95% CI, 34.97, 39.85). The strongest hot spots were identified in the central and northern regions of Madagascar. The multi-level logistic regression analysis identified several factors associated with women’s injury due to IPV, including old age category [AOR = 1.49 (95% CI, 1.003, 2.21)], husband drinking alcohol [AOR = 2.36 (95% CI, 1.64, 3.39)], the husband’s jealousy or anger when she talks to other men [AOR = 2.08 (95% CI, 1.47, 2.95)], the husband’s attempts to limit her contact with family members [AOR = 2.16 (95% CI, 1.50, 3.11)], and high community poverty [AOR = 1.47 (95% CI, 1.01, 2.14)].

Conclusion

This study identified significant geographical variation in the IPV-related injury of women in Madagascar. In addition, husband’s alcohol drinking, old age, community poverty and marital control behaviors were risk factors. Therefore, interventions that address the spatial variation of women’s injury, community-based awareness and behavioral change programs that reduce marital-controlling behaviors, promote and strengthen social networks and support for old-age women, reduce poverty and alcohol-use reduction strategies are advised.

Introduction

Intimate partner violence (IPV) is any behavior that includes the purposeful use of physical force, sexual coercion, controlling behaviors, and psychological abuse by current or former intimate partners. ¹ IPV is a major public health issue and violation of women’s human rights. ² It affects their physical, social, and sexual health ³ and results in injury, mental suffering, ⁴ and reduced quality of life. ⁵ IPV has no boundaries that affect all individuals regardless of age, gender, socioeconomic status, or cultural background.^6,7 The effect is more severe among pregnant women, affecting their and the fetus’s health. ⁸

Globally, the prevalence of IPV is high among women. A recent global study shows that IPV varies globally from 16% in Central Europe to 49% in Oceania. ² The lifetime prevalence of sexual IPV and/or physical violence among ever-partnered women was 29.8% in America, 24.4% in Europe, 37.7% in Southeast Asia, 36.6% in Africa, ⁴ 45.6% in sub-Saharan Africa, ⁹ and 40.18% in East Africa. ¹⁰ A study in 46 low- and middle-income countries showed less than five percent in Armenia and Comoros to more than 40% in Afghanistan. ¹¹ In addition, the prevalence among ever-married women varied from 18% in Cambodia to 48% in Zambia. ¹² In East Africa, IPV ranged from 30.03% in Mozambique to 46.34% in Kenya. ¹⁰ In Madagascar, IPV is a common and serious public health problem. The 2021 Madagascar Demographic and Health Survey evidenced that physical, sexual, or emotional IPV affected twenty-seven percent of ever-partnered women. In addition, 20% of women were harmed by at least three husbands’ controlling behaviors and 50% of them never sought help or told anyone about their physical or sexual violence.^13,14 Literature demonstrates that IPV also occurs among men, ¹⁵ but it is less prevalent than that against women, which varies from 13.7% to 70.9% in low- and middle-income nations. ¹⁶

The magnitude of IPV is higher in African countries due to social acceptance, poor socioeconomic status, low levels of education and a lack of women’s decision-making power. ¹⁷ IPV reduces women’s productivity, increased use of health care and social services and affects country development at large. ¹⁸ A lifetime experience of IPV accounts for an average of 17 days of lost women’s productivity ¹⁹ and costs about USD 102 million in loss of working productivity for ever experience of violence and about USD 303.3 million in costs for service provision. ²⁰ It leads to overuse of healthcare services and an increase in medical costs. ²¹

Women exposed to IPV are more affected by physical injuries and indirect health consequences, including child loss, chronic morbidity, ²² induced abortions, ²³ unintended pregnancy, preterm delivery, low birth weight, anxiety, ²⁴ depression, ²⁵ post-traumatic stress disorder, emotional distress, and suicidal behaviors.^22,23,26 IPV exposed women to chronic pain, gastrointestinal disorders, neurological problems, and respiratory conditions. ²⁶ In addition, it leads to unemployment, job uncertainty, absenteeism, tardiness, work disruptions, and diminished job effectiveness. ²⁷ Moreover, IPV is associated with sleep problems, eating disorders, substance abuse, risky sexual behavior, HIV, and other transmitted infections.^4,28

Intimate partner violence has adverse effects that range from immediate injuries to chronic health effects and even death. ¹⁹ It poses a serious threat to women’s physical health, contributing to increased mortality, poorer overall health, and injuries. ²⁹

Repeated violence reduces women’s physical resilience and increases injuries and disability. ³⁰ Women are shoved, punched, kicked, strangled, and abused by their husband or intimate partner with the intention of causing pain, injury, and emotional suffering. ³¹ This results in physical injuries, including bruises, fractures, and life-threatening traumatic brain injuries. ³² The prevalence of IPV-induced injuries is substantial among reproductive age women. A WHO report indicates that the level of injury among ever-abused women varies from 19% in Ethiopia to 55% in provincial Peru. More than 20% of women had been injured more than five times due to severe physical violence. ²³ In addition, the WHO report 2013 showed that the proportion of women with IPV-induced injury was 48.1%. ⁴ In the United States, every year, several million people, the majority of women, sustain acute physical injuries. ³³ Of 3.83 million female patients, 4.3% were victims of violent injury. ³⁴ Another study in the USA showed that 256 women were intentionally injured, 434 had contusions, 41 had fractures and dislocations and 89 had lacerations. ³⁰ In addition, the prevalence of IPV-induced injury was 13.4% in Portugal, ³⁵ 25% in Cambodia, ³⁶ and 79.9% in South Korea. ³⁷ In Canada, IPV caused injuries for 51% of Indigenous and 39% of non-Indigenous women. ³⁸ In China, five bone fractures were found out of 293 married women. ³⁹ In Malaysia, 87.7% of women experience soft tissue injuries ⁴⁰ ; in Israel, 52.4% had contusions, 20% lacerations and 3.4% fractures ⁴¹ ; in Colombia, 18.3% had bruises and 12.4% had open wound injuries ⁴² ; and in Ethiopia, 48.5%. ²⁸

In Madagascar, women aged 15 to 49 have experienced emotional, physical, or sexual abuse perpetrated by a husband or partner at some point, and 37% were affected by injuries due to sexual or physical violence at any time.^13,14

Although a study showed the high prevalence of IPV in East Africa, there remains a limited study on the spatial variation and determinants of IPV-induced injury in Madagascar. ¹⁰ There is a paucity of epidemiological studies using spatial and multilevel analyses to identify the geographical distribution of injury hotspots and risk factors of IPV-induced injury among reproductive-age women in Madagascar and in Africa at large. In addition, while the Madagascar DHS report includes IPV-induced injuries such as cuts, bruises, and sprains, the spatial variation of overall injury and associated factors has not been studied. Notably, this study fills this research gap and is important for targeted intervention among women and supporting the Sustainable Development Goal (SDG) 5.2, which aims to eliminate all forms of violence against women and empower women and girls. ⁴³ Indeed, this study contributes to the growing body of literature on IPV-induced injury among reproductive-age women and will be used for future longitudinal and qualitative studies in Madagascar. Furthermore, women’s involvement is crucial for countries’ development and maintaining and empowering them is mandatory for sustainable development. Therefore, this study aimed to determine the overall prevalence, spatial distribution, and determinants of women’s injury due to IPV in Madagascar. The findings will provide important input for policymakers to design targeted interventions and shape health programs for women.

Methods

Study design, period and data source

This study used a cross-sectional study design based on the 2021 Madagascar demographic and health survey (DHS) data. This data source provides the latest evidence on women’s injuries due to IPV and its associated factors. From this data source, the women’s data (IR file) was used to determine the overall prevalence, the spatial distribution, and determinants of women’s injury due to IPV.

Study area

This study was conducted in Madagascar, located in the Indian Ocean off the southeastern coast of Africa. It is geographically found between latitudes 12°S and 26°S and longitudes 43°E and 51°E. ⁴⁴ Its capital city is Antananarivo and, in 2020, it had a total population of 28 million. It is the fifth-largest island in the world with a land area of about 587,041 km². ⁴⁵

Study population, sample size, and sampling procedure

The DHS data used a two-stage stratified cluster sampling method. In the first stage, enumeration areas were selected using probability proportional to size, stratified by region and urban–rural residence. Second, a fixed number of households were systematically chosen from each selected cluster. ⁴⁶ As this study used secondary data, no prior sample size calculation was performed. The DHS sample size was determined using standard producers to produce a reliable estimate of health outcomes that represents national data. For this study, a total of 1515 weighted samples were determined. Since DHS follows a robust sampling methodology and includes a large sample of women, the study has sufficient power to detect an association between IVP-induced injury and its associated factors.

In this study, ever-in-union women who were selected and interviewed for the domestic violence module and experienced one or more types of physical or sexual violence by their current or most recent husband or intimate partner were included. Women not selected for the domestic violence module, selected but privacy not possible, never in a union, or ever in a union but not experiencing physical or sexual violence by their husband or intimate partner were excluded. The Madagascar DHS women’s data contains a total of 18,869 samples. Among them, 7,321 were selected for the domestic violence module but only 7,308 were interviewed. From selected and interviewed women, 5,966 of them were ever in union and 1,336 were never in a union. From ever in union, 1,574 women were experiencing one or more types of physical or sexual violence by their husbands or intimate partners. Finally, a total of 1515 weighted women were included in this study (Figure 1).

Figure 1.

Flow diagram for sample selection of reproductive age women from the Madagascar DHS 2021 dataset for analyzing injury due to intimate partner violence, 2025.

Study variables

Outcome variable

Injury due to IPV among women aged 15-49 was the outcome variable. Injury due to IPV was determined from any of these injuries (d110a = 1 or d110b = 1 or d110d = 1). This includes ever having bruises; ever having eye injuries, sprains, dislocations, or burns because of a husband/partner; and ever having wounds, broken bones, broken teeth, or other serious injuries. To determine the overall injury, we combined each of the three injury types and gave a value of zero for no injury and 1 for injury. Finally, a weighted proportion of injury per cluster was used for spatial analysis.

Independent variables

The predictor variables for this study were selected by a thorough review of existing literature and the DHS data set. After reviewing and collecting crucial variables related to the objective of this study, we classified them as individual-related factors such as age, current marital status, husband control behavior, husband alcohol drinking, maternal occupation, wealth index, attitude toward wife beating, sex of household head, media exposure, family size, access to healthcare, coverage by health insurance, current abstaining, currently/formerly in union and women’s highest educational level. community-level factors, including residence (urban or rural), Altitude, community poverty level and region.

Operational definition and measurement of variables

Husband/intimate partner: It refers to the respondent’s current or most recent husband of ever-married women and the current intimate partner of never-married women, as defined in the DHS domestic violence module. ⁴⁷

IPV is a purposeful sexual, physical, or emotional violence perpetrated by a current or former partner.^1,4 In this study IPV was computed from the DHS dataset that ever in union women or women with intimate partners experienced either any of the physical violence (d105a–f, j in 1:4) or any of the sexual violence (d105h, i, k in 1:4) indicated by Table 1.

Table 1.

Physical and sexual intimate partner violence items for women ever in a union or with an intimate partner in Madagascar DHS data, 2021.

Physical violence (any of d105a – f, j in 1:4)	a. Ever been pushed, shook or had something thrown by husband/partner
	b. Ever been slapped by husband/partner
	c. Ever been punched with fist or hit by something harmful by husband/partner
	d. Ever been kicked or dragged by husband/partner
	e. Ever been strangled or burnt by husband/partner
	f. Ever been attacked with knife/gun or other weapon by husband/partner
	j. Ever had arm twisted or hair pulled by husband/partner
Sexual violence (any of d105h, i, k in 1:4)	h. Ever been physically forced into unwanted sex by husband/partner
	i. Ever been forced into other unwanted sexual acts by husband/partner
	k. Ever been physically forced to perform sexual acts respondent didn’t want to

IPV-induced injury: It was estimated from women who experienced any of the injuries, such as ever having bruises; ever having eye injuries, sprains, dislocations, or burns; and ever having wounds, broken bones, broken teeth, or other serious injuries due to physical or sexual violence by husband/partner. It’s dichotomized as injury (yes) and no injury (no).

Attitude towards Wife Beating: It was assessed by five items that women and men who agree that a husband is justified in hitting or beating his wife for the following reasons: if the wife goes out without telling her husband, if the wife neglects the children, if the wife argues with her husband, if the wife refuses to have sex with her husband, if the wife burns the food, or if she asks her husband to use a condom if he has a sexually transmitted infection (STI). ⁴⁶

Marital Controlling Behavior: It was assessed with “Yes” or “No” items if women had ever experienced a specific type of husband or intimate partner controlling behavior in the 12 months before the survey (Table 3).

Table 3.

Marital control behaviors of study participants, Madagascar DHS 2021.

Variable	Weighted frequency (n)	Percent (%)	Injury
			Yes (%)	No (%)
Husband jealous or angry if she talks to other men
Yes	969	63.96	433 (28.58)	536 (35.38)
No	546	36.04	133 (8.78)	413 (27.26)
Husband/partner frequently accuses her of being unfaithful
Yes	620	40.92	307 (20.26)	313 (20.66)
No	895	59.08	260 (17.16)	635 (41.92)
Husband/partner tries to limit her contact with her family
Yes	259	17.10	145 (9.57)	114 (7.53)
No	1,256	82.90	421 (27.79)	835 (55.11)
Husband/partner does not permit respondent to meet female friends
Yes	467	30.83	251 (16.57)	216 (14.26)
No	1,048	69.17	316 (20.85)	732 (48.32)
Husband/partner insists on knowing where she is always
Yes	906	59.80	385 (25.41)	521 (34.39)
No	609	40.20	182 (12.01)	427 (28.19)

Husband or partner drinks alcohol was assessed using the DHS variable and categorized as Yes or No.

Women’s educational attainment was evaluated using DHS data and categorized as no education, primary, secondary, and higher education.

Media exposure: was determined by combining scores of women listening to the radio, watching TV, using the internet, or reading a magazine, and women who have at least one media exposure were categorized as having media exposure and the rest no media exposure. ⁴⁸

Community poverty: It was defined as the proportion of women in poorer and poorest households within each cluster and divided as high and low. ⁴⁹

Number of under-five age children in household: this was obtained from the DHS data and categorized as no under-five children and at least one under-five child.

Sex of the household head: It was assessed by categorical variable to show whether the household is headed by a male or female.

Women currently working: This was used to assess their employment status and categorize them as currently working (yes) and not currently working (no).

Accessing healthcare: This variable was computed from four DHS variables. This includes getting permission to go, getting money needed for treatment, distance to health facility, and not wanting to go alone. Women who experienced at least one of these factors were categorized as accessing poor health care and those reporting no problem were classified as having good access to health care. ⁵⁰

Multilevel binary logistic regression analysis

The DHS used a multistage sampling method across many levels of hierarchy. These methods naturally introduce a correlation among observations within the same cluster. This may give a biased parameter estimate and violate the independence assumption in traditional regression models. To account for this bias from individual and community-level variation, we applied a multilevel binary logistic regression model. This model is appropriate due to the outcome variable (women’s injury) was dichotomous. Random effect models such as the Intraclass Correlation Coefficient (ICC), Median Odds Ratio, and Proportional Change in Variance (PCV) were used to estimate the variation in injury due to IPV.^51,52 The ICC was computed to quantify the community-level variation. In addition, the PCV was calculated to evaluate the contribution of independent variables in explaining variability across clusters: PCV = [(VA − VB)/VA]100

Where VA = variance of the initial model, and VB = variance of the model with more terms.

Furthermore, we fitted the following four models (null model, Model II, Model III, and Model IV) sequentially. Model I (Null model): without explanatory variables, to assess baseline cluster-level variation of IPV-related injury. Model II: We fit the model by adjusting for individual-level variables only. Model III: We fit the model for community-level variables only and Model IV (final model): We adjusted for both individual- and community-level factors simultaneously to identify factors that affect IPV-induced injury. We applied deviance (-2 log likelihood) for model comparison, with the lowest deviance indicating a better model fit. Variables with a P-value ≤ 0.25 in bivariable analysis were selected as candidates for multivariable model analysis. ⁵³ In addition, Akaike’s Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used to compare models and lowest values indicate a good fit. ⁵⁴ Finally, variables’ Adjusted Odds Ratios (AORs) with corresponding 95% Confidence Intervals (CIs) (p<0.05) in the final model were used to determine factors associated with injury and for the interpretation of the strength of association. The multilevel modeling technique was used to account for the hierarchical structure of DHS data and reduce bias linked with single-level regression models.

Data collection and tools

We conducted a cross-sectional secondary data analysis of the Madagascar DHS 2021 women recode file from the Domestic Violence module (DVM). The DHS uses standardized, pretested questionnaires administered through face-to-face interviews for both households and individual women. The DVM in DHS contains physical and sexual forms of IPV and the corresponding injury items. These domestic violence questions and injury items were used to identify women who reported IPV-related injury. To maintain the privacy and safety of respondents, the DVM was administered in private for only one eligible woman per household. The data collectors received training on the DVM, interviewing techniques, confidentiality, and respondent safety. In addition, informed consent was gained from all respondents before administering the domestic violence questions. For analysis, we included women aged 15–49 who were selected for and completed the DVM of the women’ recode file.

Statistical analysis

Before analysis, the raw data were weighted using DHS sample weights (d005/1,000,000). This helps to correct for unequal probability of selection and to ensure a national representativeness of women’s data. The dataset was cleaned; missing values were managed; and duplicates and outliers were removed. In data cleaning, we determine and compute the target variable, and we select, rename, and recode the independent variables. For predictor variables with missing values less than 5% were handled using standardized data management techniques, and predictors with greater than 5% missing values were not included in the analysis. The multicollinearity was checked before analysis using the Variance Inflation Factor (VIF), and no significant multicollinearity was found (mean VIF = 1.25, all VIF < 2). In addition, the data’s hierarchical structure and the distribution of random effects were assessed before model fitting. To analyze the prevalence of women’s injury from IPV, we first generated descriptive statistics and tabulated the proportion of combined injury and selective independent factors. We performed this using Stata version 17 for tabulation and Microsoft Excel 16 for producing proportions. The result was displayed in tables and figures. This study followed the reporting items of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. ⁵⁵

Global spatial autocorrelation

ArcGIS version 10.8 software was utilized to determine the spatial distribution of IPV-induced injury among women in Madagascar. To determine whether IPV-induced injury among reproductive-aged women was dispersed, clustered, or randomly distributed in Madagascar, the Global Moran’s I statistic was used. A Moran’s I value close to –1 indicates a dispersed pattern, values near +1 suggest a clustered pattern, and a value around 0 reflects a random distribution. ⁵⁶ Further, we used the Getis-Ord Gi* statistics and corresponding z-scores and p-values for identifying statistically significant hot spot and cold spot areas. This approach helps us to detect geographic concentrations of IPV-induced injury and identify regions with disproportionately high or low burdens.

Spatial SaTScan statistics

SaTScan version 10.1 software was used for local cluster detection of IPV-related injury. A spatial Bernoulli-based model was applied to identify statistically significant clusters with high or low rates of women’s injury due to IPV. In this model, women who reported injury resulting from IPV were considered cases, while those who did not report IPV-related injury were treated as controls. The Bernoulli model scanned the geographic area of Madagascar using a moving circular window, which systematically varied in size and location while excluding areas outside the country’s administrative boundaries. ⁵⁷ The maximum spatial cluster size was set to be less than 50% of the total population at risk. This allows detection of both small- and large-scale clusters while preventing excessively large and non-informative clusters. The log-likelihood ratio (LLR) test was used to evaluate each potential cluster and to compare the observed and expected number of cases inside the scanning window. The cluster with the highest LLR value was defined as the primary cluster, while other statistically significant clusters were ranked as secondary clusters. Statistical significance was determined using 999 Monte Carlo replications, with p-values assigned to both primary and secondary clusters.

Ethical consideration

This data was obtained from the Demographic and Health Surveys (DHS) Program through an online request at https://www.dhsprogram.com/. We used the most recent Madagascar DHS, which is publicly available and fully anonymized and contains no personal identifiers. For this study, no ethical approval was required since we used secondary data analysis. Prior to the commencement of this study, we were granted a formal data access authorization letter from the DHS Program.

Results

Sociodemographic characteristics and marital control behaviors of study participants

This study included a total of 1515 (weighted) women exposed to IPV. Most of the participants (38.68%) were aged 35–49 years and 37.49% of them were aged 25–34 years. More than fifty percent (51.82%) of women had attained primary education, 34.52% had secondary education, and 11.55% had no formal education. Most of the participants (77.62%) live in rural areas; 72.48% of participants had access to media and 72.41% of husbands were reported as alcohol drinkers. Around 96% of participants were not covered by health insurance. With respect to marital status, 63.23% were married, 12.54% were living with a partner, 16.77% were separated, 3.5% were widowed, and 3.96% were divorced (Table 2). In terms of marital control behavior, 63.96% of women reported that their husband/partner was jealous or angry if they talked to other men. Around forty percent of them reported being accused frequently of unfaithfulness. Moreover, 17.1% of them were limited in contact with their families and 59.8% of women reported that their husbands insisted on always knowing their whereabouts (Table 3).

Table 2.

Sociodemographic characteristics of study participants in the analysis of injury due to IPV from the total sample, Madagascar DHS 2021.

Variable and category	Weighted frequency (n)	Percent (%)	Injury
			Yes (%)	No (%)
Age
15-24	361	23.83	122 (8.05)	239 (15.78)
25-34	568	37.49	209 (13.79)	359 (23.70)
35-49	586	38.68	235 (15.51)	351 (23.17)
Current marital status
Married	958	63.23	350 (23.10)	608 (40.13)
Living with partner	190	12.54	55 (3.63)	135 (8.91)
Widowed	53	3.50	30 (1.98)	23 (1.52)
Divorced	60	3.96	29 (1.91)	31 (2.05)
No longer living together/separated	254	16.77	109 (7.20)	145 (9.57)
Highest educational level (women)
No education	175	11.55	52 (3.43)	123 (8.12)
Primary	785	51.82	315 (20.79)	470 (31.02)
Secondary	523	34.52	190 (12.54)	333 (21.98)
Higher	32	2.11	9 (0.59)	23 (1.52)
Residence
Urban	339	22.38	155 (10.23)	184 (12.15)
Rural	1,176	77.62	411 (27.13)	765 (50.49)
Accessing Healthcare
Big problem	1,100	72.61	412 (27. 20)	688 (45.41)
Not big problem	415	27.39	154 (10.16)	261 (17.23)
Distance to health facility
Big problem	501	33.07	164 (10.83)	337 (22.24)
Not big problem	1,014	66.93	402 (26.53)	612 (40.4)
Covered by health insurance			​	​
No	1,463	96.57	550 (36.30)	913 (60.27)
Yes	52	3.43	17 (1.12)	35 (2.31)
Number of children under age five in the household
No children	520	34.32	192 (12.67)	328 (21.65)
At least one child	995	65.68	374 (24.69)	621 (40.99)
Husband/partner drinks alcohol
Yes	1,097	72.41	457 (30.17)	640 (42.24)
No	418	27.59	308 (20.33)	110 (7.26)
Currently abstaining
Yes	96	6.34	43 (2.84)	53 (3.5)
No	1,419	93.66	523 (34.52)	896 (59.14)
Respondents currently working
No	179	11.81	67 (4.42)	112 (7.39)
Yes	1,336	88.19	499 (32.94)	837 (55.25)
Media exposure			​	​
No media exposure	417	27.52	125 (8.25)	292(19.27)
Have media exposure	1,098	72.48	441 (29.11)	657 (43.37)
Sex of household head
Male	1,141	75.31	416 (27.46)	725 (47.85)
Female	374	24.69	150 (9.90)	224 (14.79)
Wealth index
Poor	436	28.78	157 (10.36)	279 (18.42)
Middle	329	21.72	106 (7.0)	223 (14.72)
Rich	750	49.5	304 (20.07)	446 (29.43)
Justifies wife beating
Yes	789	52.08	289 (19.08)	500 (33.0)
No	726	47.92	277 (18.28)	449 (29.64)

Distribution of IPV-Induced injury by sociodemographic and marital control behaviors

Relative to the total sample, the proportion of IPV-induced injury was 13.79% and 15.51% among women aged 25–34 and 35–49 years, respectively. Similarly, from a total sample, 27.13% of rural residents, 36.3% of women not covered by health insurance and 30.17% of women who had husbands who consumed alcohol experienced IPV-induced injury. In addition, injury was common among women who were currently working (32.94%), had media exposure (29.11%) and were currently married (23.1%) (Table 2). Regarding marital control behaviors, from a total sample, 28.58% and 20.26% of women experience injury due to their husband/partner being jealous or angry if they talk to other men and frequently accusing them of being unfaithful, respectively. Furthermore, 25.41% of women affected by injury due to their husband/partner insisted on always knowing their whereabouts (Table 3).

Prevalence of injury among reproductiveage women in Madagascar

From a total of 7,306 women selected and interviewed by DVM, 81.7% (n = 5,966) who were ever in a union or had an intimate partner were eligible (Figure 1). Among these eligible women, the prevalence of weighted physical and sexual violence from ever in union women was 23.57% and 10.57%, respectively. In addition, 27.4% of women experienced either physical or sexual IPV (Figure 2). Among women who experienced IPV (n = 1,574 unweighted), 38.12% (n = 600) reported at least one injury resulting from IPV. The weighted prevalence of injury due to physical IPV and sexual IPV were found to be 43.02% and 38.45%, respectively. In addition, the overall weighted prevalence of injury due to IPV was 37.4% (n = 566) out of 1,515 IPV survivors (95% CI, 34.97, 39.85), and it is spatially distributed in Madagascar (Figure 3). In addition, the proportion of sub-injuries was bruises, 31.59%; wounds, broken bones, broken teeth, or serious injury, 5.86%; and eye injuries, sprains, dislocations, or burns, 24.52% (Figure 4).

Figure 2.

Prevalence of IPV among reproductive-age women in Madagascar DHS, 2021.

Figure 3.

Prevalence of IPV-induced injury among reproductive-age women who experienced IPV, Madagascar DHS, 2021.

Figure 4.

Different types of injuries due to IPV among reproductive-age women in Madagascar, DHS, 2021.

Spatial analysis of injury due to intimate partner violence

Spatial autocorrelation of Moran’s I overall result

The result of Moran’s I statistic (Moran’s Index = 0.0622, z-score = 2.47, and p = 0.013) found that the spatial distribution of injury due to IPV in Madagascar was not random. Rather, the distribution of injury was clustered in specific geographic areas. The Moran’s value (0.0622) indicates the areas with high IPV-induced injury tend to be found near other high-value areas and low values near low values. In addition, a P-value of 0.013 indicates the probability that the observed special pattern occurred by random is only 1.3% (Figure 5).

Figure 5.

The Moran’s autocorrelation results for injury due to IPV in Madagascar, DHS, 2021.

Hotspot and cold spot analysis of injury due to IPV in Madagascar

The hotspot analysis showed that the strongest hotspots, indicated by red color, were found in the central and northern regions, mainly in Boeny, Betsiboka, Analamanga, Itasy, Vakinankaratra, Amoron’i Mania, and Vatovavy. In addition, the moderate hot spots indicated by orange (95% confidence) and lower hot spots marked by light orange (90% confidence) were found in the central corridor of Madagascar. On the other hand, the strongest cold spots (99% confidence), marked by dark blue, are found in the southern part, mainly Androy and Atsimo-Andrefana, and northeastern regions such as Sofia, and Ambatosoa. Another cold spot with 95% and 90% confidence was found in Menabe (west) and the southeast part of Fitovinany. Non-significant areas indicated by gray points showed no statistically significant clustering (Figure 6).

Figure 6.

Hot spot and cold spot analysis of injury due to IPV in Madagascar, 2025.

SaTScan analysis and cluster detection of injury due to IPV among women in Madagascar

The spatial cluster detection analysis was conducted and identified statistically significant clusters of injury in Madagascar. Primary clusters marked by a pink window with the highest log-likelihood ratio (LLR = 57.76–67.90) were identified in southern and southeastern regions of Madagascar. This area included Androy, amoron i mania, vakinankaratra, vatovavy, Anosy, and Atsinanana. A large secondary cluster with a high risk of injury was located in the northwestern regions of Boeny and Betsiboka and nearby areas in the northwest (LLR = 36.46–57.76). Moreover, additional clusters with LLR = 19.49–36.46 were identified in Menabe and Vakinankaratra in the west-central area, in the northern regions of Diana and Sava (LLR = 14.33–19.49), and in Fitovinany in the southeast-central part of Madagascar (LLR = 9.73–14.33). On the other hand, the Melaky, Bongolava, and Atsimo Andrefana regions showed scattered cases with no statistically significant clusters (Figure 7). Most likely (primary) clusters, including 147 location IDs, were identified at coordinates 20.098473° S and 48.070992° E. In this primary cluster, a total of 743 cases of women with IPV-induced injury out of 1471 women were found. Reproductive-age women living in this cluster had 1.53 times higher risks of experiencing IPV-induced injury than women residing outside the cluster. In addition, many secondary clusters that included 116 location IDs were identified. The first secondary cluster was found at the 23.755535° S, 46.565162° E coordinates. In this cluster, a total of 86 injury causes were identified among 530 women. Another secondary cluster with a high risk of injury was located at 16.579370° S, 46.304741° E. This cluster consists of a total of 99 injury causes out of 120 women. Women living in this cluster had a 2.24 times higher risk of injury than in other areas (LLR = 51.67). Indeed, two secondary clusters located at coordinates 17.221026 S, 48.031868 E, and 13.843909 S, 48.072173 E, contain a total of 30 women each and all of them were at risk of injury by 2.67 (RR = 2.67) (supplemental file 1).

Figure 7.

Spatial cluster detection of injury due to IPV in Madagascar, 2025.

Factors associated with injury due to IPV among reproductive-age women in Madagascar

In the multilevel analysis, old age category [AOR = 1.49 (95% CI, 1.003, 2.21)], having a husband who drinks alcohol [AOR = 2.36 (95% CI, 1.64, 3.39)], having a husband who is jealous of the wife talking with other men [AOR = 2.08 (95% CI, 1.47, 2.95)], and trying to limit her contact with her family [AOR = 2.16 (95% CI, 1.50, 3.11)], and high community poverty [AOR = 1.47 (95% CI, 1.01, 2.14] were risk factors. In addition, women living in rural areas were less affected by IPV-induced injury [AOR = 0.64 (0.44, 0.94)].

Accordingly, women whose husbands drank alcohol were 2.4 times more likely to suffer IPV-related injury than their counterparts. In addition, women aged 35–49 had 1.5 times higher odds of experiencing IPV-related injury compared to those aged 15–24 years. Husband-controlling behaviors, such as the husband being jealous or angry if she talks to other men, were associated with twofold higher odds of IPV-induced injury than their counterparts. Similarly, women with high community poverty had 1.5 times higher odds of IPV injury than those with low community poverty.

Moreover, women whose husbands attempted to limit their contact with family members had two times higher odds of experiencing IPV-induced injury compared with women whose husbands did not limit their wives’ contact with their family members. Women residing in rural areas had 36% lower odds of IPV-induced injury than their counterparts. In this study a total of four models were fitted to evaluate the determinants of IPV-induced injury. The null model showed significant clustering of injury (23.53%) and the inclusion of individual-level variables in model II improved model fit (AIC = 1834.554), while community-level variables showed limited improvement (AIC = 1922.08). The full model (Model IV) had the lowest AIC and deviance was selected as the best fitting model. The proportional change in variance (PCV) in the final model showed that the included individual and community level variables explained 6.% of between cluster variation in IPV-induced injury. Furthermore, the median odds ratio (MOR) in the full model was 2.53. This indicates the persistence of between-cluster variation such that if two women with similar individual characteristics were randomly selected from different clusters, women in the higher-risk cluster would have 2.53 times higher odds of experiencing IPV-induced injury than women in the lower risk of cluster (Table 4).

Table 4.

Multi-level analysis of factors associated with injury among women in Madagascar, 2025.

Variable and category	Model I (null)	Model II AOR (95%CI)	Model III AOR (95%CI)	Model IV AOR (95%CI)
Age
15-24	​	1	​	1
25-34	​	1.18 (0.85, 1.65)	​	1.17 (0.84, 1.63)
35-49	​	1.53 (1.03, 2.26)	​	1.49 (1.003, 2.21)*
Number of under-five children in the household
0	​	1	​	1
≥1	​	1.14 (0.82, 1.60)	​	1.10 (0.79, 1.53)
Residence
Urban	​	​	1	1
Rural	​	​	0.59 (0.41, 0.85)	0.64 (0.44, 0.94)
Time to get to water source (minute)
≤30 min	​	​	1	1
>30 min	​	​	1.12 (0.79, 1.58)	1.16 (0.82, 1.62)
Community poverty
low	​	​	1	1
High	​	​	1.19 (0.83, 1.70)	1.47 (1.01, 2.14)
Cluster altitude in meter
<500	​	​	1	1
500-1500	​	​	1.48 (1.06, 2.06)	1.29 (0.91, 1.83)
>1500	​	​	1.87 (1.11, 3.16)	1.71 (0.99, 2.96)
Distance to health facility
No big problem	​	1	​	1
Big problem	​	0.80 (0.57, 1.12)	​	0.81 (0.57, 1.13)
Having radio
Yes		1		1
No		1.01 (0.75, 1.35)		1.04 (0.77, 1.40)
Husband alcohol drink
Yes	​	2.33 (1.62, 3.33)**	​	2.36 (1.64, 3.39)**
No	​	1	​	1
Respondents current working
No	​	1	​	1
Yes	​	1.05 (0.67, 1.64)	​	1.05 (0.67, 1.65)
Current abstaining
No	​	1	​	1
Yes	​	1.14 (0.52, 2.49)	​	1.13 (0.51, 2.48)
Own land with other
No	​	1	​	1
Yes	​	0.78 (0.59, 1.02)	​	0.79 (0.60, 1.05)
Husband jealous or angry if she talks to other men
Yes	​	2.10 (1.48, 2.97)**	​	2.08 (1.47, 2.95)**
No	​
Husband tries to limit her contact with her family
Yes	​	2.21 (1.54, 3.18)**	​	2.16 (1.50, 3.11)**
No	​
Husband insists on knowing where she is always
Yes	​	1.17 (0.82, 1.66)	​	1.16 (0.82, 1.65)
No	​
Measures of variation of clustering
ICC	23.53	23.16	22.05	22.35
Variance	1.012268	0.9914126	0.9304605	0.9470302
MOR	2.61	2.59	2.51	2.53
PCV	-	2.09%	8.08%	6.45%
Model fit statistics
AIC	1928.671	1834.554	1922.08	1833.442
BIC	1939.394	1909.613	1959..61	1935.309
Deviance	1924.67136	1806.55398	1908.08	1795.44244

Note. ICC = Inter cluster correlation coefficient; MOR = Median odds ratio; AOR = Adjusted odds ratio; PCV = Proportional change in variance; CI = Confidence interval; *P-value < 0.05; **P-value < 0.001.

Discussion

Intimate partner violence is a multifaceted public health issue that affects women’s physical, mental, and social well-being, reduces quality of life, and exposes them to injuries.^2–4 Therefore, this study aimed to determine the prevalence, spatial variation, and determinants of women’s injury resulting from IPV in Madagascar. As a result, the prevalence of injury was 37.4% (n = 566) (95% CI, 34.97–39.85), spatially distributed in Madagascar. This high proportion of injury may be attributable to poverty, cultural barriers, and low social support. ⁵⁸

The prevalence of injury found in this study was lower compared with studies conducted in Ethiopia (48.5%), ²⁸ South Korea (79.9%), ³⁷ Malaysia (87.7%), ⁴⁰ and Canada, where spousal violence caused 51% of injury for Indigenous and 39% for non Indigenous women. ³⁸ This may be explained by variation in study design and data source. The study in Madagascar was a national survey, whereas studies in South Korea and Malaysia were hospital-based studies and included victims seeking medical care that may have had higher injury prevalence. In addition, disparity in socioeconomic and cultural factors and injury measurement tools may cause this variation. However, the prevalence of injury in this study was higher than in the study in Cambodia, where 25% of women were injured by physical or sexual IPV. ³⁶ This high proportion of injury may arise from socioeconomic and cultural differences and perceptions of injury among respondents. In this study, there were high marital control behaviors and acceptance of wife-beating that indicate the social norms tolerate IPV. ⁵⁹ In addition, many study participants did not access healthcare and reported it as a big problem and the presence of low health insurance coverage may restrict access to timely medical care after IPV. ⁶⁰ Moreover, most of the women held primary education and a small proportion of women attained higher education. This may limit bargaining in relationships and awareness of rights and access to support services. ⁶¹ More than 10% of women had no formal education, which may have exposed them to limited access to information, economic instability, and inability to protect themselves against IPV and its related injury. ⁶² Indeed, in this study other contextual factors such as media exposure, women’s occupation, being currently married, male-headed households, and husbands’ alcohol consumption are high sources of injury.

The spatial analysis demonstrated that the level of injury was not randomly distributed; rather, it showed a spatial clustering pattern. The strongest hot spot areas were found in the central and northern regions, and the moderate and lower hot spots were found in the central corridor of Madagascar. This indicates that injury is not uniformly distributed and that Central Madagascar needs urgent interventions. Cold spots were found in the north and south coastal regions. In addition, the spatial cluster detection identified a primary cluster with the highest log-likelihood ratio in the southern and southeastern regions of Madagascar, such as amoron i mania, vakinankaratra, vatovavy fitovinany, Anosy, and Atsimo Atsinanana. This indicates that women found in these areas were at significantly greater risk of injury compared to other parts of the country. Moreover, many secondary clusters were observed and a large cluster with high risk was in the northwestern regions of Boeny and Betsiboka. Main hotspots with a clear geographic heterogeneity of injury were found in the southern and northwestern parts of Madagascar. This disparity may be due to cultural practices and weaker access to support services that need region-specific interventions. This may include empowering women and expanding health and social support services.

The multilevel analysis of this study found that husband alcohol consumption, old age category, community poverty, and marital controlling behaviors such as the husband trying to limit her contact with her family and being jealous or angry if she talks to other men were found to be risk factors for women’s IPV-induced injury.

The findings of this study demonstrated that husband’s alcohol drinking was a robust risk factor for IPV-induced injury. This finding is consistent with studies conducted in the United States, ³⁰ China, ³⁹ and India. ⁶³ The possible reasons include alcohol consumption affecting cognitive functions, reducing self-control, and causing an inability to resolve conflict peacefully that aggravates IPV and may cause injury.^64,65 Women whose husbands consumed alcohol were at a higher risk of experiencing physical, emotional, and sexual violence that increased the chance of injuries.^66,67 Frequent alcohol consumption may increase aggressive behavior, cause social and marital issues and engage in verbal or physical abuse against their spouse.^68,69 Indeed, alcohol intoxication may cause husbands to have poor judgment, irritable and aggressive toward their wives, and result in IPV and injury.^70,71 A spouse who drinks alcohol returns home late, disrupts his family, and attempts to force sexual relations with his wife, which may aggravate injury. ⁶⁸

The result of this multi-level analysis found that community poverty was a predictor of IPV-induced injury. The result is consistent in a systematic review and meta-analysis in Ethiopia. ⁷² The possible justification is that women with poor wealth index are at a higher risk of IPV that may further result in injury. ⁶⁴ In addition, economic hardship heightened household stress, conflict, and IPV that increased injury. Poverty in a poor wealth index may limit women’s autonomy, affect their decision-making, and reduce their ability to leave abusive relationships. ⁶⁵ Women in poorer wealth index may not financially well-off and are subordinates of their husbands. They may get money from their husband for financial assets and accept and see violence as no problem.^73,74

This study found that women in rural areas experienced less IPV-induced injury than their urban counterparts. The possible reasons include that, despite rural residents experiencing more or similar levels of IPV, they may underreport IPV and IPV-induced injury due to social norms that tolerate violence and normalize its consequences. ⁷⁵ They may underreport IPV-induced injury due to shame, fear of disclosure, and limited awareness.^76,77

In addition, the result of this study found that marital control behavior, such as the husband being jealous or angry if she talks to other men, is another risk factor for women’s injury. This finding is backed up by studies conducted in India ⁶³ and Uganda, ⁷⁸ which reported that jealousy and control behaviors increased the risk of IPV. The possible reason may be jealousy or anger; behavior may be a manifestation of coercive control that limits wives’ external or social relationships. Indeed, a wife talking to other men may threaten her husband’s authority, which may increase IPV and the risk of injuries. ⁶³

.

Moreover, the finding of this study demonstrated that husband-controlling behavior, such as trying to limit her contact with her family, was a risk factor for IPV-related injury. This is in agreement with studies in Uganda ⁷⁸ and sub-Saharan Africa that show women’s controlling behaviors are associated with IPV and may lead to injuries. ⁷⁹ Limiting women’s contact with her family isolates them socially, reducing the opportunities to seek help or protection from IPV. ⁸⁰ Moreover, controlling behaviors may limit the autonomy of women, erode their social capital, and cause a power imbalance in the relationship that increases the risk of conflicts, IPV, and injury. ⁷⁹

Furthermore, the results of this study found that women at old age were more affected by IPV-induced injury than their counterparts. This finding is corroborated by studies done in Ethiopia ⁸¹ and among US emergency department workers. ⁸² The possible justification is that women of advanced age have a high household burden, such as workload, household hardship, and increased arguments with their husbands, that may increase marital disputes and injury risks. ⁸¹ Women in advanced age had longer relationships and a higher chance of experiencing injuries from IPV than younger women. ⁶⁴ Old-age women are naturally more prone to injury and ailment.^83,84 In addition, older women gain weaker social support networks that may trap them in abusive relationships and limit their ability to seek medical care or legal protection. ⁴¹ Moreover, they may experience seclusion and financial exploitation, which may affect their health and lead to injuries. ⁸⁴

Implications of the study

The findings have valuable implications for women’s health, country development, research, and policy. The findings can help to improve women’s health and to increase their productivity and country development. By finding the level of injury, spatial variations, and determinant factors, the study helps to call for actions for reducing and preventing IPV and injury. In addition, the observed spatial clustering further indicates the need for geographically targeted interventions, directing resources to hotspot areas where women are disproportionately affected. In healthcare settings, the findings will be used to early identify and screen women at risk of IPV-induced injuries and to provide timely psychological and medical interventions. The findings also provide an opportunity for women's affairs offices and social workers to work together to protect the rights of reproductive-age women, empower them, and reduce IPV-induced injuries. For researchers the findings can be used to further study contextual factors of IPV-induced injuries and develop prevention strategies. Furthermore, the findings support the achievement of Sustainable Development Goal (SDG) 5.2, which aims to eliminate all forms of violence against women and girls.

Limitations

This study is one of the few studies that explores the prevalence, spatial variation, and determinants of women’s injury due to IPV in Madagascar. The study used recent nationally representative data andused multilevel and spatial analysis to identify individual and community-level factors, geographic clusters, and hotspot areas. However, the study hides the causal link of risk factors and IPV-related injury due to the cross-sectional nature.In addition, the response may be biased due to its self-reported nature, and there may be underreporting of data due to recall bias, stigma, or social desirability bias.

Conclusion

This study found a high prevalence of injury with significant geographical variation among reproductive-age women in Madagascar. In addition, husband alcohol drinking, old age, community poverty, and the husband being jealous of the wife talking with other men and trying to limit her contact with her family were risk factors. Therefore, interventions that address the spatial variation and determinants of women’s injury are required. Provide targeted interventions and resources to hotspot areas where women are disproportionally affected by IPV-induced injury. In addition, community-based awareness and behavioral change programs that reduce and address marital-controlling behaviors, promote and strengthen social networks and support for old-age women, reduced poverty, and alcohol-use reduction strategies are advised.

ORCID iDs

Anmut Endalkachew Bezie https://orcid.org/0009-0003-0993-7676

Halid Worku Jemil https://orcid.org/0009-0004-8849-7177

Gosa Mankelkl https://orcid.org/0000-0002-7834-8838

Ethical considerations

This data was obtained from the DHS Program through an online request at https://www.dhsprogram.com/. The data is publicly available, fully anonymized, and contains no personal identifiers. Therefore, ethical approval and consent to participation were not required. Prior to the commencement of this study, we were granted a formal data access authorization letter from the DHS Program.

Data Availability Statement

All data generated during this study are included in the manuscript and the dataset utilized in this study is available upon request from the DHS Program (https://dhsprogram.com/).