Artigos

Introduction
One of the greatest challenges faced by the Brazilian Unified Health System is its constant need to adapt to treatment demands inherent to the aging population, changes in the morbidity and mortality profile, and the expansion and increasing complexity of the Brazilian urban structure. Contrary to what was proposed by the originator of the epidemiological transition theory, Abdel Omran, in his classic 1971 article, Brazil (and many other countries) has never followed the supposedly traditional model of the progressive replacement of certain morbid conditions by others.
We will not go into detail about the author’s original model or the critiques it has received. It is important to note, however, that in Brazil (and in several other contexts), the afore mentioned substitution of morbidity and mortality patterns ? from infectious diseases to non-communicable diseases ? never truly occurred. Rather, what has been observed is an overlap of communicable and non-communicable diseases1, along with the emergence and re-emergence of various infections, as well as the persistent health impacts of violence.
The issue of violence represents a blind spot in Omran’s formulation, for reasons that remain unclear, especially considering that the author himself experienced prolonged conflicts in his home country (Turkey). Whether due to conflicts between nations or internal urban violence within a given country, it is difficult to assume that the health impacts of violence ? in its many forms ? would be of little relevance in this supposed transition.
The first efforts coordinated by Prof. Cecilia Minayo, in the 1980s, are striking because such a relevant issue had remained overlooked within the field of collective/public health for so long and by such a wide range of authors. This researcher gradually gained recognition in a field that was, until then, neglected by the biomedical area ? despite the fact violence had been systematically addressed by modern philosophy since at least Hobbes (1588-1679) and Kant (1724-1804).
It was only in the 1990s (i.e. with a delay of a decade) that this theme entered the agenda of the Pan American Health Organization (PAHO). Minayo became a member of the working group created by PAHO to draft the resolution (CD30.R42)2 on the inclusion of violence and accidents in the health sector agenda, in 1993.
These glaring errors and omissions recall the critiques made by N. Taleb of S. Pinker’s book The Better Angels of Our Nature3, which argues for a supposed global decline in violence. In Taleb’s words, once you exclude all the statistics that show exactly the opposite, the results support the alleged decline (a debate that gained global resonance)4.
On the international level, the consolidation of Urban Health5 ? established the notion of analyzing, in an integrated way and through multiple complementary and triangulable methods, a range of issues such as racism, gender-based violence, and, most importantly for our purposes: armed violence as attacks perpetrated against communities, territories, and services. This perspective frames the issue as a form of structural violence, predominantly directed against residents of vulnerable communities, but also against the territories where they live and other areas that make up the urban social geography. It also targets the actors and institutions that seek to mitigate its effects, with particular emphasis on the health sector. Health services, in this context, find their facilities destroyed or damaged, their operations hindered, and their professionals intimidated or prevented from doing their work ? if not directly threatened.
It´s important here to distinguish contexts where civil strife and/or wars are in place from contexts such as Brazil. In the former, as comprehensively documented by Paul Farmer, thorough and purposeful destruction of health facilities and systematic killing of health professionals created what the author called “medical deserts”6.
The work of R.J. Sampson, such as his detailed analysis of the disintegration of the social fabric across large areas of Chicago, USA ? and his emphasis on contexts of high vulnerability and structural violence have solidified the integrated analysis of community, territory, and the omission and/or perverse action of governmental policies in a comprehensive manner3.
This text analyzes the issue based on previous field research conducted in a set of low-income communities in Rio de Janeiro. It draws on the methodological contributions of Judea Pearl and other authors7 in applying mediation analysis, aiming for a nuanced and flexible understanding of individual, community, and territorial determinants.
Methodology
STUDY POPULATION
The data presented here are based on the research project “Drug policy and consumption, structural violence, and access to healthcare”, a survey on violent events and their health impacts, conducted in 2022 across six communities in Rio de Janeiro. The main objective of the study was to assess the impact of shootouts on access to healthcare services and the health status of the selected communities, estimating the costs of healthcare services and treatment interruptions due to violence8. The former econometric assessment has not been explored by the current paper.
The survey employed quota sampling. The target population was defined based on mutually exclusive categories (quotas) by sex and age group (18-29 years, 30-44 years, and ?45 years old). The proportion of the target population within each quota was determined using data from the 2010 Census, resulting in a total of 1,500 interviews. Inclusion criteria included individuals who had resided in the selected community for at least one year, ?18y old, of any sex, and who agreed to participate in the study and signed the Informed Consent Form. Participants who had any condition that prevented them from understanding the nature and purpose of the study or who were unable or unwilling to sign the Informed Consent Form were excluded.
The instrument used in the survey was adapted from the study by Borges et al. (2013)9 and was divided into sections covering the respondent’s sociodemographic data, housing conditions, general health and selected health conditions, health promotion (which is not addressed by the current paper), and opinions on (in)security.
Additional data can be found in the afore mentioned report (https://www.arca.fiocruz.br/handle/icict/70171), which does not include the mediation analysis presented here, as detailed below.
Public data from the National Registry of Health Facilities (CNES), available through DATASUS, were also used to calculate the (in)completeness of Family Health teams during the study period.
Unfortunately, we did not have access to additional data at the level of individual health units, which, according to the Ethics Committee of the Municipal Health Secretariat, would have required protocol approval for each unit involved. This was not feasible due to both the time required for implementation and the need for temporal alignment between data from different sources. As a result, we limited our analysis to publicly available data.
UNITS OF ANALYSIS AND TERRITORIES
Six communities were selected for this study: three classified as less violent—Jardim Moriçaba (in the neighborhood of Senador Vasconcelos – Planning Area 5.2 – West Zone), Parque Conquista (in the neighborhood of Caju – Planning Area 1.0 – Central Zone), and Parque Bancários (in the neighborhood of Ilha do Governador – Planning Area 3.1 – North Zone); and three classified as more violent—CHP-2 – Provisional Housing Center 2 (in the neighborhood of Manguinhos – Planning Area 3.1 – North Zone), Nova Holanda (in the neighborhood of Maré – Planning Area 3.1 – North Zone), and Vidigal (located in São Conrado – Planning Area 2.1 – South Zone).
OUTCOMES
The health conditions analyzed in this article refer to self-reported diagnoses based on the following question: “Since October 2021, have you received a diagnosis or taken medication for any of the following conditions?” [listed conditions].
For the purpose of this article, the following health conditions were considered as outcomes, defined as “primary care-sensitive conditions”: hypertension, diabetes, depression, and anxiety.
The original survey included a broad range of conditions, signs, and symptoms; however, this broader set extends well beyond the scope of the present article and does not include mediation analysis. This article focuses on selected conditions and highlights the results of the mediation analysis.
The selection of “hypertension” and “diabetes” is justified by the fact that these are chronic health conditions which, if not properly managed, can lead to serious and potentially fatal complications.
Primary care plays a fundamental role in the management of these conditions due to its ability to provide a preventive approach, ensuring early detection, control, and integrated management. Through regular consultations, health education, monitoring of blood pressure and glucose levels, and support for adopting healthy lifestyle habits, primary care helps reduce the risk of complications and promotes a better quality of life for patients. Furthermore, by serving as the first point of contact within the healthcare system, it facilitates access to treatment and the coordination of specialized care when necessary, ensuring effective and patient-centered management 9,10,11,12.
Anxiety and depression are highly prevalent mental disorders that significantly affect individuals' quality of life, influencing their ability to work, maintain healthy relationships, and carry out daily activities. Primary care plays a crucial role in managing these conditions due to its strategic position as the first point of contact within the healthcare system, allowing for early identification and timely treatment. Primary care professionals are ideally positioned to offer an approach that considers the physical, psychological, and social aspects of the patient's health. In addition, they can initiate the management of these conditions by implementing brief interventions, providing psychosocial support, and referring patients to specialized services when necessary. This integrated approach contributes to reducing the stigma associated with mental illness, promotes access to treatment in its early stages, and facilitates continuous follow-up, which is essential for preventing relapses and ensuring effective and sustainable recovery13,14. Unfortunately, the mental health component falls short of its pristine, comprehensive, aims and the integration of primary care with specialized care, to be delivered at CAPS (Psychosocial Care Centers) remains elusive in several contexts. The adverse consequences of armed violence on mental health care have been emphasized by some papers, such as the article by Prata et al15.
EXPOSURE
The exposure refers to the occurrence of violent conflicts resulting from confrontations between armed factions, both among themselves and with security forces. Data on these confrontations were obtained through the “Fogo Cruzado” app, a nonprofit initiative that provides information on shootings and gunfire incidents in Rio de Janeiro. These data are publicly available and include the geolocation of events, allowing for the identification of areas most vulnerable to armed conflict. The data also undergo audits to ensure the reliability and accuracy of the information16.
The selected communities were classified as more or less violent based on the frequency of shootouts around primary healthcare units located within each territory, within a radius of up to 400 meters, in the year 2019 ? immediately prior to the COVID-19 pandemic, a global emergency that significantly impacted the functioning of healthcare units as well as the social and economic dynamics of the territories themselves.
Although the original concept issued by the United Nations Office for Disarmament Affairs mentions “Armed violence refers to the use or threatened use of weapons […]” (verbatim; our emphasis)17, threats cannot be empirically measured. They have been modeled by the Theory of Games18, which is not our aim in the present paper.
Although modern high-caliber weapons typically have a range well beyond 400 meters, using broader radio would inevitably lead to the overlap of neighboring communities and districts. Naturally, these areas may be ? and often are ? affected by such conflicts; however, no information was collected regarding their residents or services located outside the selected territories.
COVARIATES
The covariates included in the models were: age group (18-29, 30-44, and ?45y old), sex at birth (male vs. female), race/skin color (white vs. other [deliberately simplifying racial/ethnic distinctions typically aggregated by the Census]), education level (elementary, secondary, and higher education), and income (?2 minimum wages vs. >2 minimum wages).
MEDIATOR VARIABLE
As a proxy for the absence of healthcare professionals who should be part of the primary care team but were unable to work due to violence, the mediator variable was derived from the following survey question: “Have you ever heard of a situation in which a doctor and/or other healthcare professional did not come to work because of insecurity/violence in your community?” (Yes/No).
There is no doubt that this exclusive focus underestimates the broader effects of violence on healthcare services; however, it translates into objective outcomes that are easily assessed by respondents: canceled appointments, unperformed exams, lack of prescriptions, etc.
The legislation defines the procedures to be exclusively performed by physicians, as well as the diagnosis and therapeutic interventions that cannot be performed in their absence19.
ETHICAL CONSIDERATIONS
The original project, from which the present study derives as a novel and original analysis, was approved by the Research Ethics Committee (CEP) of the Rio de Janeiro Municipal Health Department – CAAE 56610422.5.0000.5279.
This article presents the full results of the mediation analyses, which ? by decision of the coordination team of the main study (https://cesecseguranca.com.br/wp-content/uploads/2023/08/RELAT%C3%93RIO_Saude-na-linha-de-tiro.pdf) ? were not included in the aforementioned report and were briefly mentioned in the box titled “Another possible analysis” (p. 53).
At the time, the study coordination team chose not to present an analysis whose complexity diverged from the objectives that guided the Report. Findings that challenge common sense and present potentially counterintuitive results may not prove useful for informing public policy or decision-making. This insight has been explored by decision science specialists since the 1980s16. More recently, the work of Centola has helped clarify that complex concepts and behaviors may not be effective in supporting decision-making or promoting behavioral or social change. Depending on the nature of diffusion processes, such findings may even produce counterproductive effects, undermining the original propositions20,21.
A scientific journal is the appropriate vehicle for disseminating complex concepts, as it provides authors with the opportunity to thoroughly substantiate and discuss their findings, as well as to refer interested readers to the relevant bibliography.
THEORETICAL MODEL
Figure 1 presents the theoretical framework used in the mediation analysis involving self-reported diagnoses, classification of communities as more or less violent, and the absence of healthcare professionals due to violence. The absence of healthcare professionals caused by violence was introduced as a mediator in the relationship between the occurrence of shootouts and the health condition diagnosis. The total effect is decomposed into a direct effect (relationship between violence and the health condition) and an indirect effect (relationship between violence and the health condition, potentially explained by the mediator variable of healthcare professional absence due to violence).
Insert Figure 1 here
STATISTICAL ANALYSIS
Absolute and relative frequencies were calculated for the covariates, health outcomes, and the mediator variable according to the classification of communities as more or less violent.
In the multivariate analysis, binary logistic regression models were employed to assess the associations between health outcomes — systemic arterial hypertension (SAH), diabetes, depression, and anxiety — and the independent variables sex at birth, age group, and community classification (more or less violent). All covariates were entered simultaneously, without the application of stepwise procedures or iterative variable selection methods. These results were presented as odds ratios with their respective 95% confidence intervals. Model fit was assessed using the Hosmer-Lemeshow test and graphical residual analysis.
A simple mediation analysis was conducted involving self-reported health diagnoses, the classification of communities as more or less violent, and the absence of healthcare professionals due to violence. The absence of healthcare professionals resulting from shootouts was introduced as a mediator in the relationship between community violence and the diagnosis of the health conditions.
In the mediation analysis, binary logistic regression models were used to estimate the indirect effect (mediated pathway), the direct effect (exposure independent of the mediator), and the total effect on the health outcomes. The total effect was decomposed into two components: the direct effect (the direct relationship between violence and each given health condition) and the indirect effect (the relationship between violence and the health condition, potentially explained by the mediating variable of healthcare professional absence due to violence).
The proportion mediated was estimated by decomposing the total effect into direct and indirect effects, using a mediation model for binary variables as proposed by VanderWeele22. The proportion mediated was calculated as the ratio of the indirect effect to the total effect, with all estimates performed on the log-odds scale.
Exploratory analyses, which are not included in this article, justify this decision and will not be detailed here. The methods originally proposed by Jenkins and subsequently updated in recent textbooks were used23. All analyses were performed using R version 4.2.0, libraries mediation, tidyverse and forecast24.
Results
Table 1 presents the distribution of covariates, health outcomes, and the absence of healthcare professionals due to violence, according to the classification of communities as more or less violent. No statistically significant differences were found for the sociodemographic variables ? namely, sex, age group, and race/skin color.
Differences were observed in the proportions related to: (i) sociodemographic factors ? education and income; (ii) outcomes ? hypertension, depression, and diabetes; and (iii) service-related variables ? presence/absence of healthcare professionals. Individuals with completed elementary education were more frequently interviewed in less violent communities (53.1%), whereas those with higher levels of education were more frequently interviewed in more violent communities (55.6%).
Individuals who reported earning up to two minimum wages were proportionally more frequently interviewed in more violent communities (51.2%) compared to less violent communities (48.8%).
Self-reported diagnoses of hypertension were more frequent among individuals residing in more violent communities at the time of the interview (56.8%), as were diabetes (59.8%) and depression (61.7%).
The absence of healthcare professionals due to violence was proportionally higher in more violent communities (72.3%) compared to less violent communities (27.7%).
Insert Table 1 here
Table 2 presents the results of the logistic regression models for health outcomes. Regarding hypertension, individuals living in more violent communities had an 18% higher likelihood of having hypertension compared to those living in less violent communities (OR:1.18; 95%CI:1.01-1.40). Male individuals had a lower likelihood of self-reported hypertension compared to females (OR:0.76; 95%CI:0.64-0.90), while individuals aged 30–44 and those over 45 had a higher likelihood of reporting hypertension compared to young adults aged 18–29 (OR:2.37; 95%CI:1.93-3.13 and OR: 5.72; 95%CI:4.45-7.50, respectively).
The characteristics associated with self-reported diabetes were being in the 30–44 age group (OR:1.83; 95%CI:1.30-2.74) and 45 years or older (OR:3.53; 95%CI:2.52-5.18). The likelihood of self-reported depression was 24% higher (OR:1.24; 95% CI: 1.01-1.53) among residents of communities with a higher frequency of shootings compared to those in communities with fewer such events. Individuals ?45y old had a 44% higher likelihood of self-reported depression (OR:1.44; 95%CI:1.11-1.88) compared to those aged 18-29.
The likelihood of self-reported anxiety was 12% higher in more violent communities, but no statistically significant association was observed. The odds ratio for men compared to women was 0.69 (95%CI:0.58-0.82), and no significant association was found with age group.
Insert Table 2 here
Figure 2 below summarizes the mediation effects of the absence of healthcare professionals due to violence in the relationship between the occurrence of shootings and health condition diagnoses. The absence of professionals significantly mediated the association between violence and diagnoses of hypertension, diabetes, depression, and anxiety, accounting for 28.7%, 26.9%, 23.1%, and 31.2% of the association, respectively (p-value < 0.05). These findings highlight that the absence of healthcare professionals due to violence negatively influences the association between health outcomes and the level of community violence.
In other words, violence affects not only the residents of these communities and their territories but also the functioning of the services located within them. The potential mitigating effect that health services ? and other public policies ? might have on these conditions is weakened when the operational capacity of these services is compromised (in this specific case, due to the absence of qualified professionals).
Insert Figure 2 here
Discussion
Sociodemographic data and health outcomes were compared across communities classified as less and more violent. The variables that guided the definition of the sampling quotas ? sex, age group, and race/skin color ? showed comparable results, which supports the adequacy of the sampling plan and its implementation.
Quota sampling falls short of the optimal procedures of fully probability sampling in terms of the reliability of statistical inference25. Recruiting respondents in mobile contexts (as intended in this study), rather than based on their households or other fixed locations, presents complex and costly challenges for implementing probability samples. In addition to quota sampling, non-conventional sampling methods ? e.g. the Starfish method ? have been proposed. However, their broad application and comprehensive evaluation regarding their contribution to optimizing statistical inference remain far from consensus and are limited to a few applications, such as a recent study conducted in Paraguay26. To the best of our knowledge, the combination of chain-referral techniques and geolocation in a single sampling method has yet to be evaluated in Brazil.
It is not possible to determine, within the context of a cross-sectional survey, the underlying reasons for the observed differences in education and income levels across the communities analyzed. As a hypothesis, one might suggest that the location and sociodemographic composition of the respective communities could be associated with the differences observed in education and income.
Only one of the four self-reported outcomes selected for analysis in this article ? anxiety ? did not show a statistically significant difference between the two subsets of communities. This finding is to be expected, given that anxiety is a highly prevalent condition, with a subjective and complex self-definition, and often intermittent in nature. The multifaceted determinants of anxiety ? which range from various organic conditions to intra- and interpersonal conflicts, as well as numerous contextual variables not limited to violence ? make it particularly difficult to propose any consistent hypothesis to explain this result27.
In contrast, diabetes, hypertension, and depression were more frequently observed among individuals residing in the more violent communities.
The initial results were obtained using multivariate models that considered only variables measured at the individual level. These findings changed substantially once the mediation analysis was introduced, which is the central focus of this article.
As shown in the theoretical model presented in Figure 2, the mediation analysis results highlight the relevance of both individual-level variables and the proxy used to assess the functioning of healthcare services in each territory. It is likely that the use of a single proxy significantly underestimates the actual role of various mediating variables, such as access (potentially hindered by violence), difficulties in scheduling regular appointments, damage to public infrastructure, and disruptions in the management of supplies and other resources due to instability and insecurity, among others. Precisely for this reason, it is noteworthy that even the use of a single proxy as a mediating variable (in a context where no additional information was available) was associated with such a pronounced effect.
The absence of healthcare professionals ? and the resulting understaffing of teams working in the territory ? proved to be a key effect modifier in the previously described associations between violence and self-reported health outcomes. All analyses showed that the mediating variable accounted for effects ranging from 20% to 30%. These substantial proportions reinforce the need to analyze both individual-level factors and service operations in each of the communities in an integrated manner. Such analyses are essential not only as diagnostic tools but also for the design and evaluation of interventions that take into account the multiple dimensions of urban health in an integrated way28.
Despite the fact data have been generated by a cross-sectional survey using a non-probability sample (i.e. far from optimal statistical inference), one of the most influential scholars in the field of causal inference observes, with his unpresuming perspective:
“For me, too, mediation was a struggle ? ultimately one of the most rewarding of my career, because I was wrong at first, and as I was learning from my mistake […]. For a while, I was of the opinion that indirect effects have no operational implications, because, unlike direct effects, they cannot defined in the language of interventions. It was a personal breakthrough when I realized that they can be defined in terms of counterfactuals and that they can also have important policy implications”7.
In the present study, individuals residing in more violent territories more frequently reported a diagnosis of hypertension. This finding is consistent with the results of a study by Tung et al.29, which investigated the association between increased crime and elevated blood pressure in Chicago, 2014-2016, in a sample of 17,783 adults. That study found a 3% increased likelihood of elevated blood pressure, an 8% increased likelihood of missed medical appointments, and a 6% increase in hospital admissions related to cardiovascular diseases. The combination of these various adverse effects results in a significant impact on morbidity, contributes to the burden on the healthcare system, compromises the quality of care and patients’ quality of life. Potentially (though not explicitly measured), it may be associated with higher mortality from a condition that, in principle, could be managed at the primary care level.
It is worth highlighting the results of a study conducted in 2016/7 in an urban area of Jamaica, which documented various health conditions among men and women exposed to violence, victimization, and conflict in their residential territories. Among these conditions were hypertension and diabetes, with findings indicating that the greater the exposure to violence, the higher the prevalence of hypertension among men, while an especially high prevalence of diabetes was observed among women30.
A study conducted with youth exposed to violence, including residents of Detroit, Michigan (USA), and youth from non-urban areas neighboring Detroit did not find an association between violence and each individual item of the depression assessment scale, but young people living in urban areas exhibited several depressive symptoms more frequently than those residing in non-urban areas31.
Another study conducted in an urban area of Southeastern Pennsylvania observed a high prevalence of violence among youth living in poorer neighborhoods of this partially urbanized region. However, it also highlighted the positive impact of strategies aimed at mitigating the effects of armed violence through the mobilization of psychosocial support services and mental health treatment32.
A study conducted in Brazil, which highlights the role of primary care teams, concluded that armed violence was identified as the main concern among both service users and healthcare professionals33. Despite the centrality and severity of this issue, the successful adoption of strategies to reduce the impacts of armed conflict was observed in territories with functioning health units, operating in accordance with the standards established by the International Committee of the Red Cross34 for regions affected by armed violence35.
Armed violence and shootouts influence the operation of healthcare activities, potentially leading to the closure of health units, the forced displacement of users to other services, and an increase in demand for care due to both the direct effects (e.g., injuries) and indirect effects (such as the development of post-traumatic stress disorders) of violence36. In some areas facing intense and protracted conflicts this could lead to the so-called “medical deserts”, a terminology coined by the World Health Organization and fully developed by Farmer.
The mediation analysis presented in this article aligns with recent studies on the adverse ? yet often overlooked ? impact of violence on healthcare professionals37,38,39. As discussed by several authors, challenges to the work of health professionals (including threats, evasion and killings) have a major impact on the proper management of several medical conditions and epidemics.
To the best of our knowledge, no other analysis had the comprehensiveness of Farmer´s analysis (in his 600-pages long book) on the mismanagement of Ebola and several other endemic and epidemic diseases in Western Africa. The combination of state-of-the-art biomedical knowledge and in-depth historical analyses over centuries offer a perspective that has been an inspiration for our work in progress.
Studies that analyze the impacts of violence on communities, territories, and services in an integrated manner are still scarce in Brazil. The need for such analyses is repeatedly mentioned by community members (as captured by a recent, yet unpublished in-depth survey carried out in Manguinhos, using vignettes provided by Prof. Claire Nee, U. of Portsmouth, UK), healthcare professionals, and managers. However, complex and multidetermined phenomena are often examined through a level of simplification that falls short of what is necessary.
Grounded in non-reductionist models, which incorporate biological and psychological dimensions, as well as population characteristics, cultural formations, and territorial contexts, it becomes possible to achieve more accurate diagnoses of the multiple effects of violence (as well as other phenomena with complex determinants), fostering public policies that may truly be considered structural.
References
1. Borges GM. Health transition in Brazil: regional variations and divergence/convergence in mortality. Cad Saude Publica 2017;33(8).
2. Queiroz C. Maria Cecília Minayo: Um olhar sociológico sobre a saúde. Revista Pesquisa FAPESP; 2024 maio; edição 339. Available from: https://revistapesquisa.fapesp.br/maria-cecilia-minayo-um-olhar-sociologico-sobre-a-saude/
3. Sampson RJ. Great American city: Chicago and the enduring neighborhood effect. Chicago: University. of Chicago Press; 2012.
4. Harris M. The Pinker Problem. Medium 2024 [cited 2022 Feb 20]. Available from: https://mikeharrisny.medium.com/the-pinker-problem-df7a40eb1fdf
5. Galea S, Vlahov D. Handbook of urban health: populations, methods, and practice. New York: Springer; 2005.
6. Farmer P. Fevers, Feuds, and Diamonds. United Kingdom: Picador; 2021.
7. Pearl J, Mackenzie D. The book of why: the new science of cause and effect. New York: Basic Books; 2018.
8. Lemgruber J, Siracusa M, Machado R, Napolião P. Saúde na linha de tiro: impactos da guerra às drogas sobre a saúde no Rio de Janeiro. Rio de Janeiro: CESec; 2023. Available from: https://cesecseguranca.com.br/livro/saude-na-linha-de-tiro-impactos-da-guerra-as-drogas-sobre-a-saude-no-rio-de-janeiro/.
9. Borges D, Cano I, Cabral C, Pinto A. O impacto da violência no acesso à saúde nas comunidades de baixa renda. 2014 [cited 2022 Feb 20]. Available from: https://igarape.org.br/wp-content/uploads/2016/04/O-impacto-da-viol%C3%AAncia-no-acesso.pdf
10. Ministério da Saúde (MS). Estratégias para o cuidado da pessoa com doença crônica: diabetes mellitus. Brasília: DAB/SAS; 2012.
11. Ministério da Saúde (MS). Estratégias para o cuidado da pessoa com doença crônica: hipertensão arterial sistêmica. Brasília: DAB/SAS; 2012.
12. Molina MRAL, Wiener CD, Branco JC, Jansen K, Souza LDMD, Tomasi E, Silva RZ, Pinheiro RT. Prevalência de depressão em usuários de unidades de atenção primária. Arch Clin Psychiatry 2012;39(6):194–7.
13. Portugal FB, Campos MR, Gonçalves DA, Mari JDJ, Fortes SLCL. Qualidade de vida em pacientes da atenção primária do Rio de Janeiro e São Paulo, Brasil: associações com eventos de vida produtores de estresse e saúde mental. Cien Saude Colet 2016; 21(2):497-508.
14. Instituto Fogo Cruzado. Dados de tiroteio [Internet]. 2024. Available from: https://fogocruzado.org.br/
15. Prata NISS, Groisman D, Martins DA, Rabello ET, Mota FS, Jorge MA, Nogueira ML, Calicchio RR, Vasconcelos RV. Mental health and primary health care: territory, violence and the psychosocial approaches challenge. Trab Educ Saude 2017; 15(1):33-53.
16. Kahneman D, Slovic P, Tversky A. Judgment under uncertainty: heuristics and biases. Cambridge: Cambridge University Press; 2018.
17. United Nations. Small arms: Armed violence [Internet]. 2024. Available from: https://disarmament.unoda.org/convarms/small-arms-armed-violence/
18. Erickson KH. Game Theory: A Simple Introduction (Simple Introductions). EUA: CreateSpace Independent Publishing Platform; 2013
19. Conselho Federal de Medicina (CFM). Parecer CFM nº 7/10: Obrigatoriedade da presença do médico na coordenação municipal do PSF [Internet]. Brasília: CFM; 2010 [citado 2025 abril 28]. Disponível em: https://sistemas.cfm.org.br/normas/visualizar/pareceres/BR/2010/7
20. Centola D. How behavior spreads: the science of complex contagions. New Jersey: Princeton University Press; 2020.
21. Centola D. Change: how to make big things happen. New York: Little, Brown Spark; 2021.
22. VanderWeele TJ. Explanation in causal inference: methods for mediation and interaction. New York: Oxford University Press; 2015.
23. Datar R, Garg H. Hands-On Exploratory Data Analysis with R. United Kingdom: Packt Publishing; 2019.
24. R Core Team. R: A language and environment for statistical computing [Internet]. Vienna: R Foundation for Statistical Computing; 2024. Available from: https://www.R-project.org/
25. Elliott MR, Valliant R. Inference for Nonprobability Samples. Statist Sci [Internet]. 2017 [cited 2024 Feb 25];32(2). Available from: https://projecteuclid.org/journals/statistical-science/volume-32/issue-2/Inference-for-Nonprobability-Samples/10.1214/16-STS598.full
26. Morel Z, Aguilar G, Samudio T, López G, Rios?González CM, Giménez L, Schaerer C, Gómez S, Báez T, Estigarribia G, Méndez J, Muñoz S, McFarland W. Intergenerational sex and early sexual debut are associated with HIV infection among transgender women in Paraguay. HIV Med 2023; 24(9):990–9.
27. Craske MG, Stein MB, Eley TC, Milad MR, Holmes A, Rapee RM, Wittchen HU. Anxiety disorders. Nat Rev Dis Primers 2017; 3(1):17024.
28. Blankenship KM, Friedman SR, Dworkin S, Mantell JE. Structural Interventions: Concepts, Challenges and Opportunities for Research. J Urban Health 2006;83(1):59-72.
29. Tung EL, Chua RFM, Besser SA, Lindau ST, Kolak M, Anyanwu EC, Liao JK, Tabit CE. Association of Rising Violent Crime with Blood Pressure and Cardiovascular Risk: Longitudinal Evidence from Chicago, 2014–2016. Am J Hypertens 2019; 32(12):1192-8.
30. Gary-Webb TL, Dyer H, Mckenzie J, Younger-Coleman N, Tulloch-Reid M, Blake A, Govia I, Bennett N, MCFarlane S, Wilks, RJ, Williams DR, Ferguson TS. Community stressors (violence, victimization, and neighborhood disorder) with cardiometabolic outcomes in urban Jamaica. Front Public Health 2023; 11:1130830.
31. Borg BA, Rabinak CA, Marusak HA. Violence exposure and mental health consequences among urban youth. Curr Psychol 2023; 42(10):8176-85.
32. Clarke AT, Grassetti SN, Brumley L, Ross KY, Erdly C, Richter S, Brown ER, Pole M. Integrating trauma-informed services in out-of-school time programs to mitigate the impact of community gun violence on youth mental health. J Prev Interv Community 2023;51(4):332-351.
33. Silva MM, Ribeiro FML, Frossard VC, Souza RMD, Schenker M, Minayo MCDS. “No meio do fogo cruzado”: reflexões sobre os impactos da violência armada na Atenção Primária em Saúde no município do Rio de Janeiro. Cien Saude Colet 2021; 26(6):2109-18.
34. Comitê Internacional da Cruz Vermelha. Acesso mais seguro para todas as Sociedades Nacionais. 2024 [cited 2024 Feb 21]. Available from: https://www.icrc.org/pt/acoes/cooperacao-com-sociedades-nacionais/acesso-mais-seguro-para-todas-sociedades-nacionais#:~:text=O%20que%20%C3%A9%20o%20Marco,e%20acesso%20%C3%A0s%20comunidades%20necessitadas
35. Santos RSD, Mourão LC, Almeida ACVD, Daher DV, Santos KMD, Brazolino LD. Nuances entre o Acesso Mais Seguro e o conflito armado no contexto da atenção básica. Cad Saude Publica 2020;36(10):e00139519.
36. Sampaio JFDS, Andrade CB. Violência armada e repercussões no cotidiano de trabalho de profissionais da Estratégia de Saúde da Família. Physis 2023; 33:e33066.
37. Doehring MC, Curtice H, Hunter BR, Oaxaca DM, Satorius A, Reed K, Beckman A, Vaughn T, Palmer M. Exploring verbal and physical workplace violence in a large, urban emergency department. Am J Emerg Med 2023; 67:1-4.
38. Robiner WN, Barnes RD, Freese RL, Palmer B, Kim MH. Homicides of psychologists, physicians, nurses, pharmacists, social workers, and other health professionals: National Violent Death Reporting System data 2003–2020. J Clin Psychol 2023; 79(12):2932–46.
39. Yusoff HM, Ahmad H, Ismail H, Reffin N, Chan D, Kusnin F, Bahari N, Baharudin H, Aris A, Shen HZ, Rahman MA. Contemporary evidence of workplace violence against the primary healthcare workforce worldwide: a systematic review. Hum Resour Health 2023; 21(1):82.