Abstract:

Femicide has received relatively little research attention, despite its severity. Estimates of femicide depend on the strategies used to define it within the sociocultural and political context. This study aimed to assess intentional homicides of women, focusing on femicides, highlighting the characteristics and risk factors. This was a cross-sectional study based on daily surveillance of homicides in the press and mortality records. The study considered fatal victims of assault in women over 11 years of age in Manaus, Amazonas State, Brazil, in 2016-2017. Classification of femicide was based on Brazil’s Law n. 13,104/2015. Relative risk was estimated by Poisson regression, and a hierarchical model was used to include variables in the models. Analyses were performed in the R statistical package. Of 138 fatal victims of assault, 52 were cases of femicide, or 37.7% (CI: 29.4-45.5). Each unit addition of age reduced the risk of femicide by 3% (CI: 0.95-0.99). Risk of femicide was 40% lower (CI: 0.40-0.90) in women with up to seven years of schooling, when compared to those with eight years or more. Women killed by bodily force showed 5.5 times higher risk (CI: 2.6-11.3) of femicide, compared to those killed with firearms. Relative risk of femicide was 1.4 (CI: 1.1-2.7) in women killed in daytime, compared to those killed at night. The proportion of femicide in this study was lower than in previous estimates in Brazil, and the local burden of urban crime appears to explain part of this discrepancy. This study showed that age, schooling, use of bodily force, and time of day when the assault occurred are associated with femicide.

Keywords:
Domestic Violence; Homicide; Educational Status; Rape; Risk Factors

Introduction

Interpersonal violence results in a high burden of morbidity and mortality and thus poses a major public health problem ¹1. Stöckl H, Devries K, Rotstein A, Abrahams N, Campbell J, Watts C, et al. The global prevalence of intimate partner homicide: a systematic review. Lancet 2013; 382:859-65.. There were an estimated 390,000 intentional homicides in the world in 2016. Interpersonal violence is the principal cause of years of potential life lost due to premature death in men in Latin America and the Caribbean ²2. Naghavi M, Abajobir AA, Abbafati C, Abbas KM, Abd-Allah F, Abera SF, et al. Global, regional, and national age-sex specific mortality for 264 causes of death, 1980-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017; 390:1151-210.. Although there are fewer homicides in women (about 25% as many as in men), there is an important difference in their patterns, especially in homicides committed by the intimate partner ³3. Jewkes R. Intimate partner violence: causes and prevention. Lancet 2002; 359:1423-9.. About half of the homicides in women are perpetrated by the intimate partner, compared to only 6% of the homicides in men ⁴4. United Nations Office on Drugs and Crime. Global study on homicide 2013: trends, contexts, data. Vienna: United Nations Office on Drugs and Crime; 2014..

Given this specificity, intentional homicides of women have been viewed differently, allowing their classification as femicide versus non-femicide. Femicide means the intentional murder of a woman for gender reasons ⁵5. World Health Organization. Understanding and addressing violence against women: intimate partner violence. Geneva: World Health Organization; 2012.. Still, given the difficulties in classifying such a complex phenomenon ⁶6. Vives-Cases C, Goicolea I, Hernández A, Sanz-Barbero B, Gill AK, Baldry AC, et al. Expert opinions on improving femicide data collection across Europe: a concept mapping study. PLoS One 2016; 11:e0148364., some authors have defined femicide more widely, including any intentional murder of women ⁷7. Garcia LP, Freitas LRS, Silva GDM, Höfelmann DA. Estimativas corrigidas de feminicídios no Brasil, 2009 a 2011. Rev Panam Salud Pública 2015; 37:251-7.^,88. Martins-Filho PRS, Mendes MLT, Reinheimer DM, do Nascimento-Júnior EM, Vaez AC, Santos VS, et al. Femicide trends in Brazil: relationship between public interest and mortality rates. Arch Womens Ment Health 2018; 21:579-82.. Femicide rates vary according to the sociocultural and political context in which women live ⁶6. Vives-Cases C, Goicolea I, Hernández A, Sanz-Barbero B, Gill AK, Baldry AC, et al. Expert opinions on improving femicide data collection across Europe: a concept mapping study. PLoS One 2016; 11:e0148364.^,99. Reddy R. Gender, culture and the law: approaches to 'honour crimes' in the UK. Fem Leg Stud 2008; 16:305-21.^,1010. Alvazzi del Frate A. When the victim is a woman. In: Global Burden of Armed Violence 2011. Geneva: Geneva Declaration Secretariat; 2011. p. 113-44.^,1111. Walby S, Towers J, Balderston S, Corradi C, Francis B, Heiskanen M, et al. The concept and measurement of violence against women and men. Bristol: Policy Press; 2017.. The magnitude of femicide estimates also depends intrinsically on the strategy used to define and count the deaths, which may or may not bear a relationship to gender issues. More flexible definitions tend to overestimate femicides, while more conservative methodologies underestimate or even overlook them ⁸8. Martins-Filho PRS, Mendes MLT, Reinheimer DM, do Nascimento-Júnior EM, Vaez AC, Santos VS, et al. Femicide trends in Brazil: relationship between public interest and mortality rates. Arch Womens Ment Health 2018; 21:579-82.^,1212. Carcedo A. No olvidamos ni aceptamos: femicidio en Centroamérica, 2000-2006. San José: Asociación Centro Feminista de Información y Acción; 2010.^,1313. Waiselfisz JJ. Mapa da Violência 2015: homicídios de mulheres no Brasil. Brasília: Organização Pan-Americana da Saúde/Secretaria Especial de Políticas para as Mulheres/Ministério das Mulheres, da Igualdade Racial e dos Direitos Humanos/Faculdade Latino-Americana de Ciências Sociais; 2015..

Higher femicide rates are seen in developing countries, mainly in Latin America and the Caribbean ¹⁴14. Pan American Health Organization. Strategy and plan of action on strengthening the health system to address violence against women. https://www.paho.org/hq/dmdocuments/2015/alessandra-guedes-strategy.pdf (acessado em 10/Nov/2018).
https://www.paho.org/hq/dmdocuments/2015... . Data from the gender equality observatory for Latin America and the Caribbean show that approximately 3,000 women in the region were victims of femicide in 2017. El Salvador had the highest rate (10.2 per 100,000 women), and Brazil had an intermediate rate (1.1 per 100,000 women), compared to the other countries included in the survey ¹⁵15. Comisión Económica para América Latina y el Caribe. Observatorio de Igualdad de Género de América Latina y el Caribe. Notas para la igualdad nº 27; 2018. https://oig.cepal.org/sites/default/files/nota_27_esp_0.pdf (acessado em 18/Nov/2018).
https://oig.cepal.org/sites/default/file... . Despite the problem’s importance, studies on intentional homicides of women (and specifically on femicides) are still scarce in the literature ⁴4. United Nations Office on Drugs and Crime. Global study on homicide 2013: trends, contexts, data. Vienna: United Nations Office on Drugs and Crime; 2014.^,1616. Weil S. Making femicide visible. Curr Sociol 2016; 64:1124-37..

The issue is thus current and requires local and regional policies to produce responses for the 2030 Agenda for Sustainable Development, which proposes the elimination of all forms of violence against women, thereby promoting gender equality. The aim of this study is to assess intentional homicides of women in the city of Manaus, the largest metropolis in the Amazon Region, with a focus on femicide, highlighting its characteristics and associated factors.

Material and methods

This was a cross-sectional study based on information obtained through daily monitoring of homicides in the digital and print press in Manaus, as well as death records provided by the Amazonas State Health Department (Susam). Manaus, capital of the state of Amazonas, Brazil, is the largest economic metropolis in the Amazon region ¹⁷17. Cruz SHR, Castro ER, Sá MER. Grandes projetos urbanos em metrópoles amazônicas: modernização e conflito. Novos Cadernos NAEA 2016; 14:89-116., with a population of nearly 2.1 million in 2016, representing more than half of the state’s population (Brazilian Health Informatics Department. http://www2.datasus.gov.br/DATASUS/index.php?area=0206, accessed on 05/May/2018). Although the gross domestic product (GDP) of Manaus is quite high, the income distribution is very unequal, with the sixth worse position among Brazil’s state capitals and a Gini index of 0.63 ¹⁷17. Cruz SHR, Castro ER, Sá MER. Grandes projetos urbanos em metrópoles amazônicas: modernização e conflito. Novos Cadernos NAEA 2016; 14:89-116..

Homicides were defined here as death records available on death certificates in women over 11 years of age in Manaus in 2016 and 2017 and classified as codes X85-Y09, referring to assault, under the International Classification of Diseases, 10th revision (ICD-10) ¹⁸18. Organização Mundial da Saúde. Classificação estatística internacional de doenças e problemas relacionados à saúde (CID-10). São Paulo: Edusp; 1999.. We also considered probable homicides the violent assaults that resulted in deaths of women over 11 years of age during the same period and that were not classified by Susam as assault, according to the ICD-10, but which received news coverage in more than three different sources in the digital and print press in the state of Amazonas. The underlying cause of death in these homicides was then coded by an experienced classifier with training at the Brazilian Center for Classification of Diseases. The total deaths in both groups were then considered as the final sample of “fatal victims of assault”, which were in turn classified as femicide versus non-femicide. The study approach to identification of “femicide” among the cases classified as intentional homicides of women is similar to the criminological methodlogy that considers this type of victimization as a specific crime within the wider range of homicides ¹⁹19. Corradi C, Marcuello-Servós C, Boira S, Weil S. Theories of femicide and their significance for social research. Curr Sociol 2016; 64:975-95.. The classification of femicide cases was thus based on Brazil’s Law n. 13,104/2015²⁰20. Brasil. Lei nº 13.104, de 9 de março de 2015. Altera o art. 121 do Decreto-Lei nº 2.848, de 7 de dezembro de 1940 - Código Penal, para prever o feminicídio como circunstância qualificadora do crime de homicídio, e o art. 1º da Lei nº 8.072, de 25 de julho de 1990, para incluir o feminicídio no rol dos crimes hediondos. Diário Oficial da União 2015; 10 mar., which defines it as a type of first-degree murder, included in the list of heinous crimes and that involves domestic and family violence, as well as vilification or discrimination ²¹21. Campos CH. Feminicídio no Brasil: uma análise crítico-feminista. Sistema Penal & Violência 2015; 7:103-15.. In other words, in femicide, the woman is murdered precisely because she is a woman.

This classification was performed independently by two criminal lawyers with experience in information on violence and health, based on the victim’s and the assailant’s antecedents, the circumstances of the death, and the context related to each case. This information allowed the classifiers to determine, based on Law n. 13,104/2015 and the existing jurisprudence, whether or not the assault was associated with gender violence. Agreement between the two classifiers was assessed with the Gwet AC1 index ²²22. Gwet KL. Computing inter rater reliability and its variance in the presence of high agreement. Br J Math Stat Psychol 2008; 61:29-48., and in a second phase, discordant cases were resolved by consensus between the two individuals.

The study’s online search explored news in specialized sites in the city of Manaus. In addition, two print tabloid-type newspapers specializing in police and crime news were reviewed daily. Mainstream broadsheet newspapers with different sections were excluded because they did not prioritize wide news coverage of homicides. Google (https://www.google.com.br/) search engines were also used as an additional resource with the following terms: “woman”, “murdered”, “assassinated”, and “Manaus” ²³23. Ferrara P, Caporale O, Cutrona C, Sbordone A, Amato M, Spina G, et al. Femicide and murdered women's children: which future for these children orphans of a living parent? Ital J Pediatr 2015; 41:68..

Data were collected with a previously elaborated form. Records were reviewed weekly by the principal investigator (J.D.Y.O.) in 2016-2017. In March 2018, three months after conclusion of the weekly monitoring of deaths, all the captured records were reviewed in order to avoid duplication and inconsistencies in the data.

The target variables were: assailant’s sex; time of the assault (daytime - 06:00 AM to 5:59 PM; nighttime - 6:00 PM to 05:59 AM); number of injuries (1-3, 4 or more); report of alcohol use by the victim; report of sexual violence; area of the body injured (head, neck, chest, upper limbs, lower limbs); type of weapon used by the assailant (firearm, cold steel weapon, or bodily force); whether the assailant was a current or former boyfriend or husband of the victim; site of the assault (residence or public byway/other); and day of the week (Monday through Friday or Saturday/Sunday). Underlying cause of death, age in complete years, race or skin color (white, black/brown), and schooling (0-7 years, 8 years or more) were collected directly from the death certificates.

To estimate relative risk (RR) of femicide, we used Poisson regression with type-HC2 consistent covariance matrix estimator ²⁴24. Cribari Neto F, Ferrari SL, Cordeiro GM. Improved heteroscedasticity consistent covariance matrix estimators. Biometrika 2000; 87:907-18.. The simple regression analysis prioritized covariables with p-value < 0.1 in order to avoid discarding potentially important regressors. Multiple regression analysis was performed with the hierarchical approach ²⁵25. Victora CG, Huttly SR, Fuchs SC, Olinto M. The role of conceptual frameworks in epidemiological analysis: a hierarchical approach. Int J Epidemiol 1997; 26:224-7.. The theoretical model was defined “a priori” and included four levels that oriented the hierarchical sequence for introduction of variables in the models. The first level included age and race or skin color; the second was schooling; the third was days of the week of the assault, time of day of the assault, and site of the assault; the fourth level included alcohol use by the victim, number of injuries on the victim, and type of weapon used by the assailant. Backward selection was used, retaining variables with p-value < 0.05 in each of the levels. The preselected variables at each level were retained in the subsequent models, regardless of their p-value, and were also considered risk factors for femicide. We also tested the inclusion of possible interactions in the final model, considering p-value < 0.05. Data were analyzed with the R statistical package, version 3.3.2 (http://www.r-project.org).

The study’s protocol was approved by the Institutional Review Board of the National Institute for Amazonian Research (INPA) (protocol n. 1.422.840).

Results

In 2016 and 2017, the study identified 130 homicides and 10 more “probable homicides”, totaling 140 cases of “fatal victims of assault”. However, in two cases, due to insufficient data for determining the circumstances of the victims’ death, it was not possible to determine whether or not femicide had occurred. In the first stage of classification, 138 “fatal victims of assault” were coded as victims of femicide or otherwise, and agreement between the classifiers was 0.79 (CI: 0.69-0.89). Next, the discordant cases were resolved by consensus, reaching a total of 52/138 cases of femicide, or 37.7% (CI: 29.4-45.5). As shown in Table 1, 29% of the total deaths resulted from the victims’ direct involvement with the drug traffic (drug users or drug dealers), and another 26% were due to passive actions (victims of deadly robbery, retaliation against civilians not involved in the drug traffic motivated by complaints against drug dealing in the neighborhood, and stray bullets), or the victims’ active involvement in criminal activities (standoffs with the police).

As shown in Figure 1, when comparing areas of the body most affected during these homicides, victims of femicide suffered more injuries on the neck (31.3%) than victims of non-femicide violent deaths (13%) (p-value < 0.05). Meanwhile, victims of other violent deaths showed more head injuries (36.2%) compared to victims of femicide (19.8%) (p-value < 0.05). No relevant differences were seen in the proportions of injuries on the chest, upper limbs, or lower limbs (p-value > 0.05).

Median age (26 years) was lower in victims of femicide than victims of other violent deaths (32 years). Femicide was proportionally higher in victims with eight or more years of schooling, those killed during the daytime (morning/afternoon), victims of sexual abuse, those killed in their own homes, those killed by a current or former boyfriend or husband, on weekends, those killed with bodily force, those without a direct relationship to the drug traffic, and those with report of alcohol use before the crime (Table 2).

As shown in Table 3, in the multivariate analysis, the covariables age, schooling, time of day of the assault, and type of weapon were associated statistically with femicide (p < 0.05). On average, each unit increase in age resulted in a 3% reduction in the risk of femicide (CI: 0.950-0.988). On average, this risk was 40% lower (CI: 0.40-0.98) in victims with up to seven years of schooling compared to those with eight or more years of schooling. Risk of femicide was 5.5 times higher (CI: 2.6-11.3) in victims killed with bodily force, compared to those killed with firearms. Risk of femicide was 4.2 times higher (CI: 2.0-8.6) in victims killed with cold steel weapons or other weapons compared to those killed with firearms. Women attacked during the daytime had 1.4 times higher risk of femicide (CI: 1.1-2.7) than those attacked at night (6:00 p.m. to 6:00 a.m.).

The final model tested the interactions between age and schooling, between time of day and type of weapon, and between age and type of weapon, all with p-value > 0.05.

Discussion

In Manaus, approximately 38% of murders of women were femicides, while the remaining deaths mainly resulted from the women’s involvement in the illegal drug trade or from passive or active actions by the victims in criminal activities. Age, schooling, time of day when the assault occurred, and use of bodily force and cold steel weapons or other weapons were associated with femicide.

As far as we know, this is the first study based on all homicides that assesses femicide and risk factors in Brazilian women.

The proportion of femicides in our study was similar to that found in a study in Taiwan, based on autopsies, in which the proportion of femicides was 31% ²⁶26. Fong W-L, Pan C-H, Lee JC-I, Lee T-T, Hwa H-L. Adult femicide victims in forensic autopsy in Taiwan: a 10-year retrospective study. Forensic Sci Int 2016; 266:80-5.. It is also similar to the Brazilian national pattern (around 25%), since according to data from the gender equality observatory for Latin America and the Caribbean, there were 1,133 femicides in Brazil in 2017 ¹⁵15. Comisión Económica para América Latina y el Caribe. Observatorio de Igualdad de Género de América Latina y el Caribe. Notas para la igualdad nº 27; 2018. https://oig.cepal.org/sites/default/files/nota_27_esp_0.pdf (acessado em 18/Nov/2018).
https://oig.cepal.org/sites/default/file... , and according to the Mortality Information System, there were approximately 4,800 homicides of women that same year (Brazilian Health Informatics Department. http://www2.datasus.gov.br/DATASUS/index.php?area=0205, accessed on 08/Oct/2018). Meanwhile, the point estimate of femicides in Manaus is considerably lower than in other Brazilian studies, although the upper limit of the estimate reaches a proportion of nearly 46%. Importantly, previous studies in Brazil used different criteria to define femicide, for example, defining all homicides of women as femicides, and considering as femicide executions associated with the drug traffic or situations that involved disproportional use of lethal means by the assailant ⁷7. Garcia LP, Freitas LRS, Silva GDM, Höfelmann DA. Estimativas corrigidas de feminicídios no Brasil, 2009 a 2011. Rev Panam Salud Pública 2015; 37:251-7.^,88. Martins-Filho PRS, Mendes MLT, Reinheimer DM, do Nascimento-Júnior EM, Vaez AC, Santos VS, et al. Femicide trends in Brazil: relationship between public interest and mortality rates. Arch Womens Ment Health 2018; 21:579-82.^,2727. Margarites AF, Meneghel SN, Ceccon RF. Feminicídios na cidade de Porto Alegre. Quantos são? Quem são? Rev Bras Epidemiol 2017; 20:225-36.. Not every homicide is a femicide; especially in areas with high levels of interpersonal violence, cases such as deadly robbery (robbery leading to death) or those involving ordinary delinquency may produce wide variation in the estimates of femicide ¹²12. Carcedo A. No olvidamos ni aceptamos: femicidio en Centroamérica, 2000-2006. San José: Asociación Centro Feminista de Información y Acción; 2010..

Approximately 29% of the victims of other violent deaths were associated with reports of direct involvement with the drug traffic. In general, these victims were drug users and were killed because of unpaid debt to dealers, women who were dealers themselves and were killed in turf wars, or wives or girlfriends of dealers who were killed during settling of scores between factions. Another 26% of the victims had their deaths associated with passive or active actions in criminal activities. The victimization of women by homicide linked to urban violence appears not to be exclusive to Manaus and is relatively common in other large Brazilian cities like Porto Alegre (Rio Grande do Sul State) and Recife (Pernambuco State), among others ²⁸28. Souza KOJ. A pouca visibilidade da mulher brasileira no tráfico de drogas. Psicol Estud 2009; 14:649-57.^,2929. Portella AP, Galvão C, Abath M, Ratton Jr. JLA. Análise configuracional de homicídios: velhas e novas situações de violência letal contra as mulheres em Recife. Dilemas - Revista de Estudos de Conflito e Controle Social 2011; 4:403-39.^,3030. Gomes APPF. Como morre uma mulher? Configurações da violência letal contra mulheres em Pernambuco [Dissertação de Mestrado]. Recife: Universidade Federal de Pernambuco; 2014.^,3131. Meneghel SN, Margarites AF. Feminicídios em Porto Alegre, Rio Grande do Sul, Brasil: iniquidades de gênero ao morrer. Cad Saúde Pública 2017; 33:e00168516.^,3232. Wanzinack C, Signorelli MC, Reis C. Homicides and socio-environmental determinants of health in Brazil: a systematic literature review. Cad Saúde Pública 2018; 34:e00012818..

The high number of violent deaths of women in Manaus is not surprising, considering that from 2007 to 2014 the female incarceration rate in Amazonas more than doubled, from 16.6 to 37.4 per 100,000, with the drug traffic and robbery accounting for approximately 83% of the convictions as of November 2016 ³³33. Cardoso CM. Histórias das mulheres privadas de liberdade em Manaus: vidas marcadas pela pobreza, violência e abandono [Dissertação de Mestrado]. Manaus: Universidade Federal do Amazonas; 2017.. The upward trend in female incarceration rates in Amazonas is consistent with the Brazilian and global patterns. Brazil is one of the five countries of the world that most incarcerates women ³⁴34. Walmsley R. World female imprisonment list. Women and girls in penal institutions, including pre-trial etainees/remand prisoners 2014. London: World Prison Brief/Institute for Criminal Policy Research, University of London; 2015.^,3535. Zackseski C, Machado BA, Azevedo G. Dimensões do encarceramento e desafios da política penitenciária no Brasil. Revista Brasileira de Ciências Criminais 2016; 126:291-331..

As for sites of injuries in victims of femicide, there was a low proportion of deaths involving lesions on the head and a high proportion of deaths by firearms among the victims of non-femicide violent deaths (data not shown). Meanwhile, injuries on and around the neck were nearly three times more frequent in femicide victims. Gunshots, especially to the head, are highly lethal ³⁶36. Alvis-Miranda HR, Rubiano AM, Agrawal A, Rojas A, Moscote-Salazar LR, Satyarthee GD, et al. Craniocerebral gunshot injuries: a review of the current literature. Bull Emerg Trauma 2016; 4:65. and tend to involve premeditated attacks, normally associated with illegal activities such as executions on orders from drug dealers ³⁷37. Falbo GH, Buzzetti R, Cattaneo A. Homicide in children and adolescents: a case-control study in Recife, Brazil. Bull World Health Organ 2001; 79:2-7.^,3838. Seleye-Fubara D, Etebu E, Bob-Yellowe E. Pathology of firearm mortalities in the Niger Delta region of Nigeria: a study of 136 consecutive autopsies. Med Sci Law 2009; 49:51-5.. In other studies on femicide, the neck was also one of most frequently affected areas ²⁶26. Fong W-L, Pan C-H, Lee JC-I, Lee T-T, Hwa H-L. Adult femicide victims in forensic autopsy in Taiwan: a 10-year retrospective study. Forensic Sci Int 2016; 266:80-5..

A study of female homicide victims over 14 years of age in South Africa found that 62% of victims of deadly assault showed high blood alcohol levels, and half the victims had a blood alcohol concentration of 11% (110mg per 100mL of blood) ³⁹39. Mathews S, Abrahams N, Jewkes R, Martin LJ, Lombard C. Alcohol use and its role in female homicides in the Western Cape, South Africa. J Stud Alcohol Drugs 2009; 70:321-7.. In a review study that assessed the association between intimate-partner physical or sexual violence and alcohol consumption by women, Devries et al. ⁴⁰40. Devries KM, Child JC, Bacchus LJ, Mak J, Falder G, Graham K, et al. Intimate partner violence victimization and alcohol consumption in women: a systematic review and meta analysis. Addiction 2014; 109:379-91. demonstrated a positive association between alcohol use and victimization by intimate-partner physical or sexual violence. Although report of alcohol consumption did not remain in the final adjusted model, it was nearly 30% in victims of femicide, three times higher than in the group of victims of non-femicide violent deaths (9.5%). Still, we cannot overlook that the records are subject to underestimation, since these data are not from autopsies, unlike other studies.

Although site of the crime was not selected for the final model, nearly 50% of victims of femicide were killed in their own homes, while the vast majority (81.4%) of non-femicide violent deaths occurred in public. This result is largely expected, since studies on femicide ²⁷27. Margarites AF, Meneghel SN, Ceccon RF. Feminicídios na cidade de Porto Alegre. Quantos são? Quem são? Rev Bras Epidemiol 2017; 20:225-36.^,4141. Mouzos J. Femicide: the killing of women in Australia 1989-1998. Canberra: Australian Institute of Criminology; 1999. and on deadly assaults perpetrated by intimate partners show that the domicile is the most frequent site of fatal assault ⁴²42. Bailey JE, Kellermann AL, Somes GW, Banton JG, Rivara FP, Rushforth NP. Risk factors for violent death of women in the home. Arch Intern Med 1997; 157:777-82..

Median age of femicide victims was relatively low and much lower than victims of other violent deaths, corroborating a study in Porto Alegre ²⁷27. Margarites AF, Meneghel SN, Ceccon RF. Feminicídios na cidade de Porto Alegre. Quantos são? Quem são? Rev Bras Epidemiol 2017; 20:225-36. and differing from findings in other countries ²³23. Ferrara P, Caporale O, Cutrona C, Sbordone A, Amato M, Spina G, et al. Femicide and murdered women's children: which future for these children orphans of a living parent? Ital J Pediatr 2015; 41:68.^,2626. Fong W-L, Pan C-H, Lee JC-I, Lee T-T, Hwa H-L. Adult femicide victims in forensic autopsy in Taiwan: a 10-year retrospective study. Forensic Sci Int 2016; 266:80-5.. In Italy, for example, more than half of the victims were over 45 years of age ²³23. Ferrara P, Caporale O, Cutrona C, Sbordone A, Amato M, Spina G, et al. Femicide and murdered women's children: which future for these children orphans of a living parent? Ital J Pediatr 2015; 41:68.. In addition, our final model showed a protective effect against femicide as women’s age increased, suggesting that younger victims are more vulnerable and that femicide may negatively influence premature mortality indicators in these women.

Although physical and sexual violence are known to be the most frequent forms of intimate partner assault against women ⁴⁰40. Devries KM, Child JC, Bacchus LJ, Mak J, Falder G, Graham K, et al. Intimate partner violence victimization and alcohol consumption in women: a systematic review and meta analysis. Addiction 2014; 109:379-91.^,4343. Palermo T, Bleck J, Peterman A. Tip of the iceberg: reporting and gender-based violence in developing countries. Am J Epidemiol 2013; 179:602-12., studies on homicide victimization rarely consider investigating and recording sexual assault. Some authors contend that the low frequency of its occurrence and difficulties in operationalizing the category of sexual homicide tend to vary widely between countries and authors, which could explain its relative invisibility ⁴⁴44. Chan H-C, Heide KM. Sexual homicide: a synthesis of the literature. Trauma Viol Abuse 2009; 10:31-54.^,4545. Abrahams N, Mathews S, Lombard C, Martin LJ, Jewkes R. Sexual homicides in South Africa: a national cross-sectional epidemiological study of adult women and children. PLoS One 2017; 12:e0186432.. In our study, 19.2% of femicides involved a report of sexual violence, similar to a study in North Carolina (United States), on teenage victims of femicide, 8.9% of whom had been raped before the fatal assault ⁴⁶46. Coyne-Beasley T, Moracco KE, Casteel MJ. Adolescent femicide: a population-based study. Arch Pediatr Adolesc Med 2003; 157:355-60.. Sexual violence prior to femicide can also be viewed as an indication that these deaths were related to stigmatized occupations such as sex work ²³23. Ferrara P, Caporale O, Cutrona C, Sbordone A, Amato M, Spina G, et al. Femicide and murdered women's children: which future for these children orphans of a living parent? Ital J Pediatr 2015; 41:68.^,2727. Margarites AF, Meneghel SN, Ceccon RF. Feminicídios na cidade de Porto Alegre. Quantos são? Quem são? Rev Bras Epidemiol 2017; 20:225-36.^,4747. Stevens A, Berto D, Frick U, Kerschl V, McSweeney T, Schaaf S, et al. The victimization of dependent drug users: findings from a european study, UK. Eur J Criminol 2007; 4:385-408.^,4848. Der Ghougassian K, Otamendi A, Fleitas Ortíz de Rosas D. Violencia íntima, femicidios y armas de fuego en Argentina. URVIO: Revista Latinoamericana de Estudios de Seguridad 2015; (17):11-35. or even illicit drug use, in which sexual favors are occasionally exchanged as a form of payment ⁴⁹49. Scivoletto S, Tsuji RK, Abdo CHN, Queiróz S, Andrade AG, Gattaz WF. Relação entre consumo de drogas e comportamento sexual de estudantes de 2º grau de São Paulo. Rev Bras Psiquiatr 1999; 21:87-94.^,5050. Zeferino MT, Fermo VC, Fialho MB, Bastos FI. Semelhanças e contrastes nos padrões de uso de crack em Santa Catarina, Brasil: capital vs Meio Oeste. Ciênc Saúde Colet 2017; 22:97-106.. Therefore, as in other studies, sexual violence appears as an important component of femicide ⁵¹51. McFarlane J, Malecha A, Gist J, Watson K, Batten E, Hall I, et al. Intimate partner sexual assault against women and associated victim substance use, suicidality, and risk factors for femicide. Issues Ment Health Nurs 2005; 26:953-67..

Victims with up to seven years of schooling showed 37% fewer femicides when compared to those with eight or more years of schooling. These results are similar to a study in Haiti, where more educated women showed a higher risk of physical and sexual violence than those with less schooling ⁴³43. Palermo T, Bleck J, Peterman A. Tip of the iceberg: reporting and gender-based violence in developing countries. Am J Epidemiol 2013; 179:602-12.. Our results may reflect the empowerment of women with more schooling, who would tend to react more to psychological abuse, verbal and physical assault, and male possessiveness in a scenario where power relations between genders are still quite asymmetrical.

As for the method used by the assailant, our study showed that risk of femicide was 5.5 times greater in victims killed with bodily force and 4.2 times higher in those killed with cold steel and other weapons, compared to those killed with firearms. The latter were used in fewer than 20% of the femicides, contrary to the means used in other violent deaths, in which firearms reached nearly 80% of the cases. Studies in different contexts concerning the principal means used by assailants to kill their victims have shown that the use of firearms varies upwards or downwards when compared to Manaus, suggesting wide variation in this indicator, depending on the respective context ²³23. Ferrara P, Caporale O, Cutrona C, Sbordone A, Amato M, Spina G, et al. Femicide and murdered women's children: which future for these children orphans of a living parent? Ital J Pediatr 2015; 41:68.^,2727. Margarites AF, Meneghel SN, Ceccon RF. Feminicídios na cidade de Porto Alegre. Quantos são? Quem são? Rev Bras Epidemiol 2017; 20:225-36.^,4141. Mouzos J. Femicide: the killing of women in Australia 1989-1998. Canberra: Australian Institute of Criminology; 1999.^,4646. Coyne-Beasley T, Moracco KE, Casteel MJ. Adolescent femicide: a population-based study. Arch Pediatr Adolesc Med 2003; 157:355-60.^,5252. Frye V, Hosein V, Waltermaurer E, Blaney S, Wilt S. Femicide in New York City: 1990 to 1999. Homicide Stud 2005; 9:204-28.^,5353. Toprak S, Ersoy G. Femicide in Turkey between 2000 and 2010. PLoS One 2017; 12:e0182409..

The emerging literature on femicide is still incipient and explores a limited number of risk factors, leaving room for more in-depth assessment of such factors as the time of day or night when the assault occurs. In our study, women killed during the daytime showed 69% higher risk of being victims of femicide when compared to those killed at night, which may be related to the scene of the crime, the modus operandi, and the strategies and motivation that led the assailant to commit the crime.

Many issues concerning the determinants, characteristics, and even the definition of femicide still remain unclear. However, in the midst of so many uncertainties, it seems certain that men are the principal assailants. This study, as in various other areas of the world ³3. Jewkes R. Intimate partner violence: causes and prevention. Lancet 2002; 359:1423-9.^,4141. Mouzos J. Femicide: the killing of women in Australia 1989-1998. Canberra: Australian Institute of Criminology; 1999.^,5252. Frye V, Hosein V, Waltermaurer E, Blaney S, Wilt S. Femicide in New York City: 1990 to 1999. Homicide Stud 2005; 9:204-28.^,5353. Toprak S, Ersoy G. Femicide in Turkey between 2000 and 2010. PLoS One 2017; 12:e0182409.^,5454. Azziz-Baumgartner E, McKeown L, Melvin P, Dang Q, Reed J. Rates of femicide in women of different races, ethnicities, and places of birth: Massachusetts, 1993-2007. J Interpers Viol 2011; 26:1077-90.^,5555. Ayala Quintanilla BP, Taft A, McDonald S, Pollock W. An examination of femicides in Peru between 2009 and 2014. Int J Gynecol Obstet 2016; 134:342-3., found that in most femicides the perpetrator is the current or former boyfriend or husband, as opposed to other perpetrators, whether or not they are known to the victim.

The study’s results highlight nuances that have received little attention concerning the magnitude of femicide and the associated risk factors, as well as reflections concerning the possible consequences for the victims’ families, especially the children ⁵³53. Toprak S, Ersoy G. Femicide in Turkey between 2000 and 2010. PLoS One 2017; 12:e0182409.^,5555. Ayala Quintanilla BP, Taft A, McDonald S, Pollock W. An examination of femicides in Peru between 2009 and 2014. Int J Gynecol Obstet 2016; 134:342-3., since many of these murders occur at home, in the environment shared by the family and in the children’s presence ²³23. Ferrara P, Caporale O, Cutrona C, Sbordone A, Amato M, Spina G, et al. Femicide and murdered women's children: which future for these children orphans of a living parent? Ital J Pediatr 2015; 41:68..

The study’s principal limitation is information bias, since data extracted from websites and print newspapers are subject to imprecisions and errors, as well as editorial judgment by these media ⁵⁶56. Gillespie LK, Richards TN, Givens EM, Smith MD. Framing deadly domestic violence: why the media's spin matters in newspaper coverage of femicide. Violence Against Women 2013; 19:222-45.. On the other hand, the data retrieval rate was high in our analyses, especially on sociodemographic data and underlying cause of death, which were collected directly from the death certificates.

Although classification of deaths as femicides and non-femicides was performed independently by two criminal lawyers and their level of agreement was considered good, the discordant cases may reflect different legal interpretations of the same cases or the inherent limits of the amount of data extracted from news stories in the local press.

Although local press coverage aims to disseminate information to the general readership (i.e., not for academic purposes), given the difficulties in performing a study based on multiple sources, the strategy adopted here appears to have been useful for quick, low-cost assessment of femicides.

The study underlines femicide as a relevant source of premature mortality in women and portrays the various contexts in which this type of violence occurs, in this case in a large metropolis in a developing country with high rates of inequality and urban violence, which have a more devastating impact on the number of homicides in women, whether as victims or as active agents in criminal activities.

While revealing the need for more effective responses for the prevention of femicide, especially perpetrated by the victim’s current or former partner, the study also highlights the need for more studies on the relationship between fatal victimization and sexual violence in order to shed light on the determinants, the extent of the problem, and above all its visibility vis-à-vis government and society, which is essential for guaranteeing the right to life and the reduction of gender inequities.

Acknowledgments

The authors wish to thank the Amazonas State Health Surveillance Foundation for allowing access to the electronic death records and Naiade Gonçalves and Jéssica dos Santos Rios for their assistance with the legal classification of femicides.

References

Publication Dates

History