Time trend and spatial distribution of the cases of lethal violence against women in Brazil

Abstract Objective: to analyze the time trend and the spatial distribution of the cases of lethal violence against women in Brazil, according to age group and to race/skin color. Method: an ecological study of time series, with spatial distribution of the deaths of women victims of aggression, registered in the Mortality Information System, resident in Brazil, Brazilian geographic regions and states. Due to underreporting of deaths in some states, correction factors of the mortality rates were employed. For the trend analysis, we adopted the polynomial regression model. In addition to that, the mean rates and annual upward/downward trends were distributed considering the Brazilian federative units as analysis units. Results: the mean rate was 6.24 cases of lethal violence per 100,000 women, with a significant variation across the Brazilian regions and states. The main victims of violent death in Brazil are young, black-/brown-skinned and indigenous women, with a growing trend in these population segments. The North and Northeast regions stood out with the most significant mean annual increases (0.33; r2= 0.96 and 0.26; r2= 0.92, respectively). Conclusion: there was a stable trend regarding lethal violence against women, with significant regional differences. Young, black-/brown-skinned and indigenous women are more vulnerable to violent death in Brazil.


Introduction
Gender violence is a global social and public health problem, whose final stage is femicide, a crime that has been responsible for victimizing thousands of women every year (1) .
In recent decades, discussions on how to define and measure femicide have increased significantly, driven by the high incidence in Latin America and the recent laws implemented in several countries that have harshened its punishment (2) .
Identification of the femicide cases is extremely relevant for the statistical analysis and implementation of effective public policies to combat violence against women.
However, a number of researchers report certain difficulty sizing the femicide occurrences due to nonexistence of information sources that allow such identification (3)(4)(5) .
This scenario is aggravated in Brazil due to the Government's slowness to investigate and process femicides, in addition to not providing information on closed cases. Due to the absence of concrete data, the research studies on the theme are carried out with data on female mortality due to aggression, currently considered as an indirect femicide indicator (6) .
In 2017, the United Nations (UN) estimated a global rate of 1.3 cases of lethal violence against women perpetrated by intimate partners or other family members per 100,000 women. Africa was the region where women presented the highest risk of being killed, recording a rate of 3.1 per 100,000 in the same year, followed by the Americas (1.6/100,000), Asia (0.9/100,000) and Europe (0.7/100,000) (7) .
In this segment, more than half of the countries with the highest femicide rates are in South America, Central America and the Caribbean. Murder involving couples and murder of women preceded by rape are the most common forms of the crime (8) in Latin America.
In Brazil, the indicators of lethal violence against women draw the attention. In 2019, the rate was 3.5 cases per 100,000 women, a 17.3% reduction when compared to 2018, when 4.3 cases per 100,000 women were recorded. Although this result seems to be positive news, the analysis should be cautious because, in the same period, there was a 21.6% increase in violent deaths of women without indication of the cause, whether murder, accident or suicide (9) .
Among the main risk factors are gender inequality, previous domestic violence, family history of violence, low schooling, unemployment, poverty, young age, ethnicity and race/skin color, and access to weapons by the aggressor (10)(11) .
Recently, community violence, alcohol and drug use, adverse childhood experiences, conditions of exacerbated disparity and family violence have been described as social comorbidities that increase the probability of femicide (12) .
In order to reduce the number of cases of violence against women, some international policies were instituted, such as the Sustainable Development Goals, which listed "achieving gender equality and empowering all women and girls" as one of its targets (13) .
The Brazilian policies for the protection of women also advanced with the publication of the Maria da Penha Law, in force since 2006, which devised mechanisms to curb domestic and family violence against women (14) , although experts point out that there was no reduction in the cases of domestic violence after implementing this law (15) .
In 2015, the Brazilian State also sanctioned the Femicide Law, which amended the penal code to provide for femicide as an aggravating circumstance for the crime of murder (16) .
Due to the high rates of lethal violence against women in Brazil and to the scarcity of studies on the theme, especially after entry into force of the Femicide Law, the objective of this research was to analyze the time trend and the spatial distribution of the cases of lethal violence against women in Brazil, according to age group and to race/skin color.

Study design
This is an ecological study of time series, with spatial distribution of the rates of lethal violence against women, registered in the Mortality Information System (Sistema de Informação sobre Mortalidade, SIM).

Study locus
The data on lethal violence against women living  performance (18) . Health Information Systems (HISs) are important tools for management, as they contribute to the quality of the services, reduce costs and enable research and teaching activities (19) .

Population and data source
The population consisted of cases of lethal violence against women aged from 15 to 59 years old. This age group was defined after a previous analysis of the SIM database indicated a higher incidence in this age group.
Women aged 60 years old or more were excluded because violence against older adults follows a different dynamic. A study evidenced that, in most cases, the aggressors of children/adolescents and adult women are the partners, while in aged women they are family members (20) .
Among the age group selected for the analysis there was stratification from 15 to 39 years old, which in this study were considered young women, and from 40 to 59 years old, thus allowing a comparison of the rates between both age groups.

The indicators of lethal violence against women
were selected in the SIM using codes X85 to Y09 of the

International Classification of Diseases and Health-Related
Problems (ICD-10). These codes refer to all forms of aggression, analyzed from the gender perspective because they are considered a proxy for femicide (6) .  For the trend analysis, the polynomial regression model was used because of its high statistical power.
Thus, the rates of lethal violence against women were considered as dependent variables (Y) and the years of study as independent variables (X). The "year" variable was converted into a centralized variable in year (X-2009) and the series were smoothed using a three-point moving mean (24) .
The polynomial regression models were tested as linear (Y = β 0 + β 1 X), second order (Y = β 0 + β 1 X + β 2 X 2 ) and third order (Y = β 0 + β 1 X + β 2 X 2 + β 3 X 3 ), considering the significant trend when the estimated model obtained p-value < 0.05. For the choice of the best model, the analysis of the dispersion plot, the value of the coefficient of determination (r² -the closer to 1, the more adjusted the model is) and the analysis of residuals (assumption of true homoscedasticity) were considered. When all the criteria were significant for more than one model and the coefficient of determination was similar, the simplest model was chosen (25) . The analyses were performed using the SPSS software, version 20.1. The spatial distribution of the rates of lethal violence against women was presented in intervals, from the maximum to the minimum rates, and the maps were represented on gray scales, defining lighter colors for lower rates and darker colors for higher rates. In turn, distribution of the mean annual increases or reductions was presented using gray for the states that recorded a reduction and black for those that presented an increase.
All figures were prepared using the QGIS software,

Ethical aspects
As this is a research study with data in the public domain, waiver of analysis was requested to the     In the South region, the state of Paraná presented the highest mean for the period (7.27) and Rio Grande do Sul, the highest mean annual increase (0.14; r 2 = 0.87).

Results
Finally, in the Midwest, Mato Grosso recorded the highest mean rate in the period (9.50) and Goiás, the highest mean annual increase (0.24; r 2 = 0.50) (   It is also noted that most of the states presented a

Discussion
There is still scarcity of time trend and spatial distribution analyses of the rates of lethal violence against women that consider the regional differences in Brazil, for the rates, contributing to a more precise analysis. In addition, the SIM has expanded its coverage and shown good performance (18) , being fundamental for conducting ecological studies on the mortality of the Brazilian population in the national scope.
In addition to that, for the spatial analysis, the data were aggregated in time and space. Therefore, new studies may be carried out disaggregating the data in relation to the period and expanding the analysis units.
Another limitation concerns the nonexistence of a national database with specific femicide cases, causing the study to investigate lethal violence against women from a gender perspective. However, it is noteworthy that the analysis contributed to identifying the differences across the Brazilian territories, enabling targeting of actions to control the problem.
The debates on femicide, a crime that involves gender-motivated lethal violence against women, have gained space in recent decades influenced by the high incidence in Latin America (2) .
The increase in the number of cases exerted pressure on several countries' governments to implement specific laws that criminalize the murder of women on genderrelated grounds, although the result of the legislation is not known due to the absence of a national database (26)(27) .

Currently, all Latin American countries except Cuba and
Haiti have passed specific laws or implemented devices to fight against gender violence (28) .  (29) .
Some studies indicate that the law in question contributed to conferring visibility to the topic and encouraging women to report their aggressors, although the statistics do not indicate significant reductions in the number of cases after its implementation (30) .
In line with the literature, the data found in this Brazilian study (40) .
In general, the most expressive mean rates coincide with the socio-economically disadvantaged regions of the country: North and Northeast. On the other hand, the stable trend in the Midwest region and the reduction trend in the Southeast and South are consistent with the existence of better indicators (41) .
Other studies indicate a mean rate of 4.7 violent deaths per 100,000 women in the Northeast, while incidence was lower in Rio Grande do Sul: 3.2 per 100,000 women (42)(43) . These data confirm the variability in adolescence (45)(46) .
In Spain, a research study evidenced that economic precariousness and unemployment trigger situations of violence against women, due to the financial dependence generated (47) . This allows us to infer that women who are economically dependent on their partners tend to remain in abusive and violent relationships, increasing the risk of femicide.
Brazilian inequality is also seen in the provision of in 2020, the state has only three exclusive courts. In contrast, the Federal District has 16 exclusive courts for a female population estimated at 1,587,124 (48) .
In line with these data, the number of Specialized  (43) and from 20 to 39 years old (44) .
Violent death of young women is a reality in other countries, such as Turkey, the United States, Ecuador and a region of South Africa (35,(51)(52)(53) . These findings are alarming Recently, the persecution habit, a behavior called stalking, has gained ground in the debates, and is even criminalized in some countries. In this logic, a research study conducted with adolescents in the United States evidenced that 51.1% of the girls had already being victims of stalking or harassment (58) .
Regarding race/skin color, higher rates of lethal violence were found among black-and brown-skinned women, concentrated in the Brazilian North, Northeast, Midwest and part of the Southeast region. The high rates in this population segment are consistent with the results of another study (59) . It is also noted that, while lethal violence in white-skinned women is decreasing in most states, it is increasing among black-and brown-skinned women, reasserting data from the literature (60) . A number of researchers also draw the attention to the genocide that has been taking place in the dispute for territories, in which women from certain ethnic and race-related groups are killed as a message of terror (59) .
Another factor related to this increase is the interference of white-skinned people in indigenous communities. Situations such as the modification of the natural landscape and the introduction of alcohol and drugs contribute to making gender relations conflictive (65) .
In addition, the migration of indigenous people to urban centers, living in a situation of social vulnerability in most of the cases, favors the increase in violence (66) .
The UN highlights that institutional racism is an obstacle for black, brown and ethnic minority women to enjoy access to justice, as well as it aggravates the situation of vulnerability they experience (67) .
The World Health Organization (WHO) reiterates that violence against women is preventable and derives from social, cultural and economic factors, among others, being reduced through the implementation of effective programs and policies. In this sense, it becomes fundamental to analyze the socioeconomic level, as well as the existence of specific laws, income levels, poverty and economic deprivation, employment, schooling, access to health services and political participation (68) .
Concrete measures to support women in situations of violence prevent the occurrence of fatal crimes since, in most of the cases, death is preceded by episodes of psychological, patrimonial, physical and sexual violence (60) .
Finally, in the Health and Nursing context, knowing the regions with the highest incidence of lethal violence against women, as well as the predominant age group and race/skin color, favors the planning of comprehensive care, which aims at interrupting the cycle of violence, avoiding recurrence of fatal cases.

Conclusion
There was a stable trend in the rates of lethal violence against women in Brazil, with important regional disparities.
The North and Midwest regions recorded the highest mean rates, the North and Northeast regions had the most significant annual increase, and the Southeast was the only region that presented a significant reduction in the rates.
An upward trend in the number of cases was evidenced in young women, in the age group from 15 to 39 years old. The spatial distribution of the rates, according to race/skin color, revealed higher incidence values in black-and brown-skinned women, with an upward trend.
It is necessary to recognize the regional differences in Brazil so that the strategic planning of actions is based on the peculiarities of each location and on the risk and protection factors regarding violence against women.