Prison Lethality: Epistemic Harm and Death Connected to Brazilian Carceral Spaces

Abstract

Deaths caused by or connected to exposure to Brazilian prisons are widely acknowledged as frequent and preventable, yet official data fails to capture their scale, causes, and circumstances. To circumvent what official administrative datasets miss, this article examines an original dataset of 1077 criminal case files from 27 Brazilian state courts involving individuals who died between 2017 and 2021 after having been incarcerated. Drawing on the systematic document review of these cases, we analyze sociodemographic characteristics, health information, causes of death, and judicial responses, distinguishing between deaths occurring in custody (“internal”) and after release (“external”). Our findings reveal pervasive omissions in basic demographic and medical information, extensive use of ill-defined causes of death, and a striking absence of investigation in most cases, including deaths under direct state custody. We identify instances of obfuscation and judicial inaction that, together with the absence of reliable administrative data, are likely to sustain institutional ignorance and normalize preventable deaths. This study advances debates on incarceration and health, state accountability, and proposes the concept of prison lethality: the capacity of carceral spaces to increase people’s exposure to health risks and harms, combined with the epistemic practices that shed light on or obfuscate this capacity.

1. Introduction

Leonardo died of tuberculosis in a prison hospital unit in Rio de Janeiro. He was 31 years old, and his death certificate listed the cause of death as “pulmonary infection, bilateral pleural effusion, and progression of morbid state.” Leonardo was serving a semi-open sentence (a regime in which individuals may work or study outside but sleep inside a prison unit) at the prison Plácido de Sá Carvalho Penal Institute (IPPSC). At the time of his death, in May 2017, the IPPSC had already been under intervention by the Inter-American Commission on Human Rights for nearly a year and under the supervision of the Inter-American Court of Human Rights for more than three months.1 The involvement of the highest bodies of the Inter-American system had been requested by the Rio de Janeiro Public Defender’s Office, prompted precisely by the deplorable sanitary conditions of that overcrowded prison facility and, most of all, by the high number of deaths recorded there. Leonardo’s death could not have been more predictable or more widely forewarned. And yet, his family continues to seek compensation in the Brazilian justice system, which has failed to acknowledge its own responsibility for Leonardo’s death. Leonardo’s story is one among many of people who have died in carceral spaces in Brazil.

From extensive media coverage of violent prison uprisings met with extreme police repression and so-called “resistance killings” (Ferreira 2021; Machado et al. forthcoming), to the harrowing personal accounts of formerly incarcerated individuals and their families, documented in ethnographic accounts and other qualitative research (Godoi 2017; Mallart and Rui 2017; Biondi 2018), prisons and jails have been lethal sites in the country, with their spatiality extending far beyond prison walls and specific sentences (Godoi 2017; Mallart and Rui 2017; Biondi 2018).

Yet official data does not reflect this reality. Similar to other countries, prison mortality figures in Brazil are imprecise and unreliable, aggregated at the state level under broad categorizations of combined “natural causes,” “external causes” (“suicides”, “criminal”, and “accidents”) and “unknown causes” and disconnected from any contextual information that would allow for greater scrutiny of why someone has died. Moreover, available data is restricted to deaths that were registered to have happened inside a carceral unit or “under custody,” while “external deaths,” despite any individual’s exposure to carceral spaces, are not included in official carceral mortality statistics, nor is such exposure systematically recorded in other official mortality datasets.

In order to determine whether we could know more of these deaths and their documented causes, this paper analyses a dataset compiled for this study of 1077 deaths between 2017 and 2021 reported in court rulings from 27 Brazilian state courts and involving individuals whose criminal proceedings were still ongoing at the time of death and who had been incarcerated during those proceedings. Brazilian courts have the legal duty to oversee and decide the entirety of each criminal case, including the execution of sentences, while also monitoring carceral conditions. When a defendant dies during the course of criminal proceedings, judges must formally acknowledge the death to declare the “extinction of punishability,” which brings these deaths into the court record. Depending on the procedural stage of the case, this recognition may occur before judges responsible for the criminal prosecution or before those overseeing the execution of the sentence. Across these judicial arenas, courts retain the authority to request investigations from police or to inquire directly into the operation of correctional facilities. The cases studied offered a window into the way court authorities become aware of—and whether they act upon—the life and death of incarcerated or formerly incarcerated people under criminal prosecution or serving their sentences.

After systematically reviewing a broad set of documents contained in each court case file, we show that even official court records produced by criminal legal system authorities on each one of these lives and deaths rely on ill-defined causes of mortality and extensive missing information on the demographics and life conditions of people exposed to the criminal legal system. Courts also failed to ask for further information or trigger death investigations, revealing a broader pattern of institutional inaction.

To make sense of these findings, we propose the concept of “prison lethality” that captures two interconnected dimensions of how carcerality operates: its material harm, as prisons are likely to cause death by increasing the risks of harm and illness, and its epistemic harm, as carceral institutions produce forms of official knowledge about these risks that ultimately obscure them, a form of racialized strategic ignorance (McGoey 2012; Mueller 2018; Ferreira 2021; Ferreira et al. 2020; Freitas 2019; Vinuto and Ramos 2024) that contributes to the persistence of lethal carceral conditions and the absence of meaningful reform.

1.1. The Brazilian Carceral System

With the third largest carceral population in the world, behind only the United States and China, Brazil holds nearly one million people in prisons and jails presently. This population has more than tripled over the past twenty-five years, rising from 232,755 incarcerated individuals in 2000 (a rate of 132 per 100,000 inhabitants) to 909,067 in December 2024 (416 per 100,000) (World Prison Brief n.d.). While there are no reliable estimates of the cumulative number of individuals who have passed through the carceral system over time, available studies suggest that between one third and 40 percent of those released from prison are later reincarcerated (Viana 2023). Given this level of turnover, the number of people exposed to incarceration over the past decades is likely far greater than the prison population at any given moment and plausibly reaches into the millions.

This sharp increase has been attributed especially to punitive criminal policies, especially drug-related ones, which favor incarceration over non-custodial sentences (Machado et al. 2019) and racial profiling in ostensive policing (Sinhoretto et al. 2022), affecting mainly Black and brown, poor young adults (Ferreira et al. 2020; Santos et al. 2021).

Incarcerated individuals are detained in 1382 correctional facilities unevenly distributed across the country’s 26 states and the Federal District (Brazilian National Penitentiary Department 2025). Prison conditions are widely recognized as severely deficient: they have been declared by the Brazilian Supreme Court to constitute an “unconstitutional state of affairs” (Machado 2022) and have been the object of widespread condemnation by international human rights organizations and the Inter-American Commission and Inter-American Court of Human Rights (Godoi and Matosinhos 2021). Overcrowding is pervasive, with the prison population exceeding official capacity by an average of 40 percent nationwide, and some states operating at twice their intended capacity (Fórum Brasileiro de Segurança Pública 2025). Facilities are frequently understaffed and many provide limited, if any, access to physical and mental health care, as well as to educational and work training programs (Ferreira et al. 2020; Vasconcelos et al. 2023; Fórum Brasileiro de Segurança Pública 2025; Brazilian National Penitentiary Department 2025).

The Brazilian carceral system works under the governance of state authorities, with great variation in how prisons are run, staffed, and equipped (Fórum Brasileiro de Segurança Pública 2025; Godoi 2017). However, federal criminal law establishes the minimum contours of carceral policy across the country. Moreover, unlike in many countries where correctional agencies control most post-sentencing decisions, Brazilian law assigns central supervisory authority over sentence execution to the judiciary. Courts retain continuous authority throughout the criminal process, from trial and sentencing to the supervision of sentence execution, for as long as a sentence is being served, whether inside or outside prison walls. Within the judiciary, two sets of judges play distinct roles. Criminal trial judges determine criminal liability and impose sentences, frequently following prosecutors’ requirements. Judges responsible for the execution of sentences oversee the implementation of those sentences, including the conditions of confinement and an individual’s progression through different custodial regimes. They may, for example, request and authorize transitions from closed to semi-open or open regimes depending on legal criteria and individual circumstances. Because punishment remains legally ongoing until a formal extinction of the penalty, individuals transitioning out of prison remain under judicial supervision during semi-open and open regimes, conditional release, and other forms of supervised re-entry, extending the reach of carceral governance beyond prison walls. In addition to deciding on sentence progression, execution judges are also legally required to monitor prison conditions through periodic inspections of correctional facilities and through reporting systems coordinated by the National Council of Justice (Conselho Nacional de Justiça, CNJ), which collect information on prison and re-entry populations, conditions of confinement, and the functioning of local detention facilities.

Brazilian prisons do not operate as socially isolated institutions. As a large body of ethnographic and sociological research has shown, carceral spaces are extremely porous and remain deeply connected to the communities that surround them through dense networks of visitation, communication, and circulation of people, goods, and information (Araújo 2016; Barbosa 2013; Godoi 2017; Lago 2019; Mallart and Cunha 2019; Mallart and Rui 2017; Martins 2022; Melo 2024). Chronic understaffing and resource scarcity mean that many everyday aspects of prison governance, from the organization of internal spaces to the management of daily routines, are partially mediated through incarcerated people’s self-organization, often coordinated with families and actors in illegal markets, such as the Primeiro Comando da Capital (PCC), a prison-based organization that has developed extensive structures operating in illicit markets across the country, while being a key player in managing violence and death inside and outside prison walls (Feltran 2018; Godoi 2017; Biondi 2018; Manso and Dias 2018). Rather than functioning as closed institutions, Brazilian prisons have been described as “communicating vessels” (vasos comunicantes) in which the boundaries between inside and outside are constantly traversed by social relations, economic exchanges, and forms of governance (Godoi 2017).

1.2. Prison Exposure and Deaths in Official Datasets

Brazil maintains four national information systems that record deaths, but none were designed to systematically capture mortality among people who are or had been incarcerated. Two national sources of vital statistics produce mortality data for the entire population: the “Civil Registry Statistics” produced by the Brazilian Institute of Geography and Statistics, and the Mortality Information System (SIM), produced by the Brazilian National Healthcare System (Sistema Único de Saúde, or SUS). Both datasets rely on death certificates as their primary source of mortality data, which do not include a specific field indicating whether the deceased was or had ever been under state custody (Machado et al. 2023) or whether the death occurred in a prison facility.

Two other nationwide public databases consolidate different information about prison conditions and prison population, including carceral death numbers reported by state authorities: SISDEPEN, an initiative of the Brazilian National Penitentiary Department within the Ministry of Justice (part of the Executive Branch), and CNIEP (National Registry of Inspections in Penal Establishments), maintained by the National Council of Justice (part of the Judiciary Branch). As of now, these databases constitute the only parameters for measuring prison mortality nationwide. None of them include information on deaths of individuals who had previously been exposed to carceral spaces but were already permanently released or under some form of conditional or temporary release at the time of death. Moreover, when comparing both databases, research has found disagreement across the numbers and causes of mortality reported, reflecting differences in sources of data collection, coverage, scope of prison facilities included, frequency of database updates, and classification of deaths (Machado et al. 2023).

For instance, while SISDEPEN reports five categories for causes of death (natural/health reasons, criminal, suicide, accidental, unknown), with data collection starting in 2005 and semi-annual updates, CNIEP includes only two categories for causes of death (natural and homicide/accidental), with data starting in 2009 and monthly updates. Moreover, while SISDEPEN clearly states that it relies on data provided directly by state security and prison administration agencies, CNIEP does not disclose its data collection methodology and instead relies on prison inspection reports, which are likely provided by prison administrators or by the individual judges responsible for inspections.

Neither of the datasets allows for assessing individual-level mortality patterns within the prison population. Both rely on aggregated data at the prison-facility level, already classified and summarized into the broad categories of causes of death, and only SISDEPEN disaggregates data by gender, while other sociodemographic characteristics, such as age or race, are absent from the reports. In addition, neither dataset allows researchers to determine precisely when a death was registered or whether the cause of death was established before or after the autopsy report and the conclusion of the police investigation (Machado et al. 2023). It is also unclear whether any control mechanisms exist to correct records when the cause of death is later revised, for example, when a death initially classified as suicide is determined to be a homicide (Machado et al. 2023).

Furthermore, the categories used to classify causes of death in both datasets are overly broad and generic, limiting their capacity to capture the complexity of the cases analysed (Machado et al. 2023). In particular, deaths classified as “natural” or “unknown” may obscure systematic violence, including torture, poor prison conditions, or inadequate access to healthcare services.

Recent works have shown that, aside from the low levels of granularity and detail in these datasets, mortality rates for the prison population reported by SISDEPEN are likely underreported (Machado et al. 2023; Liu et al. 2021; Sánchez et al. 2021), or incorrectly labelled as indeterminate, a pattern that became particularly clear during the COVID-19 and the rise in “indeterminate” deaths, while “natural/health-related” deaths retained levels comparable to previous years (Machado et al. 2023). Combined, these limitations make it difficult to systematically assess deaths connected to incarceration through official mortality datasets alone and motivate the alternative research design developed in this study.

1.3. From Prison Exposure to Prison Lethality

Through the epidemiological notion of prison exposure (Massoglia 2008a, 2008b; Massoglia and Remster 2019; Massoglia and Pridemore 2015; Schnittker et al. 2022), we learn that time spent in carceral spaces can increase the risks of experiencing primary health stressors, such as witnessing or being victimized by violence or being subjected to the stressful environment of prison life; secondary stressors, such as stigma and reduced social standing as one re-enters the “free world,” as well as the higher-than-average exposure to respiratory and sexually-transmitted infections. The relationship between prison and health is relevant not only as one spends time under-custody, but also at re-entry, with a robust body of work documenting elevated risks of mortality, especially in the period immediately following release (Binswanger et al. 2013; Binswanger et al. 2007; Merrall et al. 2010), but also extending long after exposure through faster epigenetic aging (Berg et al. 2021).

Yet, scholars working in this area often rely on independent mortality counts or creatively linked public health data as ways to avoid relying exclusively on information published by correctional authorities. Official prison health and mortality data are frequently opaque, highly aggregated, and unreliable (Beale and Dolovich 2025; Fliss et al. 2024; Penal Reform International 2022; World Health Organization Regional Office for Europe 2022; Tomczak and Mulgrew 2023).

This article introduces the concept of prison lethality as a framework for understanding how carceral institutions can produce death not only through direct acts of violence, neglect, or by exposing people to stressors and infectious conditions, but also through the epistemic practices that shape how such harms are documented, classified, or obscured in official records. In this sense, prison lethality is composed of two elements: the afore-mentioned harms prison exposure may cause to individuals’ health and well-being, during and after the carceral experience, and that might increase their chances of death or illness; and the epistemic practices of the specific organizations in charge of governing and producing knowledge about the incarcerated and re-entering populations, including their health conditions and needs.

The epistemic dimension of prison lethality converses with what Tomczak and Mulgrew (2023) argue to be a need for a “new epistemology” of deaths related to incarceration, one that accounts not only for deaths that occur inside and outside prison walls, but also for the specific circumstances in which they occur—including where, how, and who may have been involved.

But prison lethality also includes in this new epistemology the dimension of what is known of the ignorance produced, arguing for the need of knowing what information is missing (omission), or obfuscated through imprecise and broad categories detached of context, and how such ignorance produced might shape institutional inaction, such as refraining state authorities from further scrutinizing policies or inquiring over their own liability in causing harm (McGoey 2012).

In this sense, the ways authorities categorize deaths, document individual circumstances, or fail to pursue further inquiry after these categorizations, are examples of practices that may influence how prison exposure unfolds and how its consequences are publicly recognized or obscured. In short, this epistemic dimension may itself act as a latent cause of the most proximal stressors that lead to material harms connected to prison exposure. Considering that incarcerated bodies are for the most part brown and Black, young, and vulnerable men and women within the Brazilian population, ignorance can further reinforce cycles of racial exclusion (Mueller 2018) and become embedded in what Mbembe (2003) describes as necropolitical forms of governance, in which institutions participate in determining whose lives are protected and whose deaths remain politically and administratively tolerable. The present study does not trace the full institutional production of such ignorance or the internal decision-making processes through which it emerges. Instead, it offers a first step in documenting the forms of ignorance that become visible in judicial records and official documentation, while also considering how the official datasets described in the previous section are themselves permeated by different forms of omission and obfuscation.

1.4. The Current Study

This study examines criminal proceedings records and describes what judicial authorities know and record about deaths of individuals exposed to carceral spaces while they remain subject to criminal proceedings. In doing so, it examines the epistemic dimension of prison lethality in an attempt to circumvent some of the problems official databases incur, especially their inability to offer more detailed and disaggregated accounts of individuals’ characteristics and causes of death or their inability to cover deaths of individuals upon re-entry.

2. Materials and Methods

We assessed a sample of 1077 deaths reported within open criminal law cases in 27 Brazilian state courts from 2017 to 20212 of people who had been incarcerated. The criminal lawsuit encompasses all the relevant official information available to the judicial system for both deciding a case and monitoring sentences and the carceral and post-release conditions of people who had been indicted and tried for a criminal offense. By law, the death of a defendant must be notified to the court, which then will “extinguish one’s criminal punishability due to death” (Article 107(1)). Article 107(1) establishes that criminal punishability is extinguished upon the death of the defendant, regardless of the procedural stage of the case or the individual’s custodial status at the time of death. As a result, courts must formally acknowledge deaths that occur not only inside prisons or jails, or while individuals are in state custody, but also deaths of individuals who have never been incarcerated or who had previously been incarcerated and later released. This includes individuals on parole or temporary release, or serving their sentence under alternative custodial regimes such as the semi-open (regime semiaberto) or open (regime aberto) regimes. In all such cases, the death must be recorded in the court file so that the judge can formally declare the extinction of punishability, after the prosecutor has been heard, and close the case, pursuant to Article 62 of the Code of Criminal Procedure. Because this procedural step is required whenever a defendant dies while still under criminal prosecution or sentence execution, Article 107(I) rulings generate judicial records of deaths connected to incarceration across both trial courts and courts responsible for sentence execution. By contrast, deaths occurring after an individual has fully completed their sentence or otherwise ceased to be subject to criminal proceedings do not generate such rulings, as there is no remaining punishability to extinguish. Consequently, Article 107(1) decisions capture deaths of individuals who remained under penal jurisdiction at the time of death but exclude deaths occurring after the complete termination of the criminal sentence.

Overall, court files include information on individuals’ demographic characteristics, criminal trajectories, and custodial histories. In Article 107(1) cases, they are also likely to include the notification of death, the death certificate, and information on the circumstances of death and its procedural handling when any investigation occurs. These records do not provide systematic information about prison conditions or institutional decision-making inside prisons. However, they do offer a window into what state institutions claim to know or not know of those who died and were or had been incarcerated, and what courts, especially execution judges and prosecutors, have further inquired about, revealing patterns of omission, obfuscation, and investigative inaction that are likely to inform these organizational practices.

With the institutional support of the National Council of Justice, which granted the research team access to its national databases of criminal proceedings and sentence execution records, and after receiving responses to freedom of information requests sent to all 27 state courts in the country, we compiled a list of 115,550 lawsuits filed between 2017 and 2021 that referenced Article 107(I), regardless of whether the defendant had ever been incarcerated. Our research team randomly selected 44 cases from each one of the country’s 27 state courts, making up a universe of cases of 1188 cases. Cases were selected based on whether the person had died while in custody, or whether the deceased had been deprived of liberty at some point during the specific criminal proceeding analysed, even if the death itself occurred outside custody. As a result, the dataset captures deaths occurring at different stages of the criminal procedure, including pretrial detention, trial, appeal, and sentence execution, both inside and outside custodial settings. But in all cases, the person had been deprived of liberty within the criminal proceeding assessed or remained under custody. After removing duplicate cases for individuals whose deaths had been notified in more than one criminal lawsuit and cases for which complete documentation was not digitally accessible for consultation, we reached a final count of 1077 cases,3 each one accounting for a single death.

We reviewed and analyzed each case’s full set of judicial documents, which ranged from the investigation, trial, and sentencing stages to post-conviction and sentence supervision phases. From each case, we collected information on sex4, race, age at the time of death, educational attainment, family characteristics (marital status and number of children), income, prior involvement with the criminal legal system, and the circumstances of death, including where the death occurred, the causes recorded in death certificates and medical examinations, and whether any investigation into the cause of death was initiated.

To ensure consistency in data extraction, we developed a structured coding instrument implemented through Google Forms. Three legal researchers (all holding law degrees) coded an initial random sample of 100 cases (excluded from the final analytical sample) to test internal consistency and identify potential interpretive disagreements. Following this pilot phase, these researchers worked in collaboration with five additional researchers (also trained in law) to code the full dataset. Interpretive disagreements were resolved through team discussion during the coding process, and, in a final stage, the research team jointly reviewed the coded dataset to identify inconsistencies and correct systematic errors.

Causes of death were transcribed verbatim from the death certificates contained in the case files. Although death certificates do not report the International Classification of Diseases (ICD) code, they typically describe causes of death using language aligned with ICD-10 sections. A physician specializing in correctional healthcare and a health law expert independently coded all recorded causes of death according to ICD-10 classifications, and any discrepancies were resolved through cross-checking and discussion between the two researchers.

All data was organized into two groups, “internal deaths” and “external deaths,” defined by two complementary criteria: the custodial arrangements to which the deceased was subjected at the time of death and the physical location where the death occurred. Almost all deaths were deemed internal or external based on the custodial arrangement, unless the custodial arrangement was “semi-open,” where the individual works or studies outside but lives and sleeps in custodial spaces. In these cases, we adopted a conservative approach that classified the death as “internal” or “external” based on the place of death.

Internal deaths involve individuals who were full-time in custody in a prison or jail, or in “semi-open” custody but whose deaths occurred inside the carceral establishment.5 Cases involving individuals who, already ill or injured, died in hospitals shortly after being transferred from prison units, or who died in transit while being transported from prison to another location, were classified as “internal deaths.” SISDEPEN considers as deaths “in the custody of the facility” those related to “all instances (…) of mortality of individuals in the custody of the facility, even if the death occurred outside the facility (in a healthcare unit, for example).”6 For this reason, spaces that in some way functioned as extensions of carceral establishments were also considered “internal,” such as healthcare units, whether part of the Brazilian healthcare system, or administered by the prison system, spaces visited during sentence enforcement, and transport vehicles transferring individuals between detention facilities and other locations. Following these criteria, we found 104 internal deaths within our sample.

External deaths, in turn, refer to cases involving individuals who, although having been previously incarcerated in the course of the criminal case under assessment, were released at the time of their death, living outside custodial spaces, such as under house arrest, in open-sentence regimes, or on provisional or temporary leave from correctional units. For those who were under “semi-open” custody, but whose deaths occurred outside the carceral establishment, their deaths were conservatively classified as external. Considering their previous incarceration in these cases sheds light on the circumstances of deaths among people who, after institutionalization, passed away while still linked to the criminal legal system and under court supervision within the specific criminal proceeding being assessed, but who died outside of custody. These cases allow us to see what types of causes are likely to be more frequently at re-entry, and whether courts further investigate them, tracing their potential connections to previous carceral exposure. Certainly, courts have more limited control over these deaths and more limited avenues for action, but because these individuals remained under judicial supervision while serving sentences, these deaths could still draw criminal courts’ attention and trigger further scrutiny. We found 973 cases classified as external deaths within our sample.

3. Results

Table 1. General Descriptives of Deaths in Criminal Cases.

	External Deaths	Internal Deaths
Cases	973 cases	104 cases
Sex (%)
Male	95.9	97.12
Female	3.29	1.92
Other gender identities	0.00	0.96
Missing from documentation	0.72	0.00
Age of Death
Youngest	17	16
Median	29	33
Oldest	88	78
Missing from documentation	41 cases	5 cases
Nationality (%)
Brazilian	99.28	98.08
Other nationalities	0.32	0.96
Missing from documentation	0.41	0.96
Race/ethnicity (%) *
Black	4.32	2.88
Brown	58.27	66.35
White	17.88	25.00
Indigenous	0.21	0.96
Asian	0.10	0.00
Disagreement	9.86	4.81
Missing from documentation	9.76	4.76
Marital Status (%)
Single or divorced	70.91	66.35
Married or partner	25.69	30.77
Missing from documentation	3.39	2.88
Children (%)
No children	29.18	25.96
1–2 children	28.16	28.84
3–4 children	10.89	9.61
More than 5 children	2.98	4.80
Missing from documentation	28.77	30.76
Education (%)
Illiterate	1.85	1.92
Incomplete primary education	27.03	31.73
Completed primary education	10.48	11.54
Incomplete secondary education	7.19	7.69
Completed secondary education	7.71	3.85
Incomplete higher education	0.41	0.96
Completed higher education	1.23	1.91
Missing from documentation	44.09	40.38
Occupation (%)
Taking care of home/family	0.21	0.00
Retired	2.88	1.92
Self-employed	49.33	45.19
Unemployed	14.08	11.54
Student	5.04	2.88
Other	0.31	0.00
Pensioner	0.10	0.00
Military police officer	0.10	0.00
Missing from documentation	27.95	38.46
Monthly income (%)
No income	1.13	0.96
$19–$94 USD ($101–$500 Reais)	0.21	0.96
$95–$189 USD ($501–$1000 Reais)	0.82	0.96
$189–$283 USD ($1001–$1500 Reais)	2.88	1.92
$283–$377 USD ($1501–$2000 Reais)	0.92	0.00
$378–$472 USD ($2001–2500 Reais)	1.54	0.00
More than $472 ($2500 Reais)	1.23	0.96
Missing from documentation	87.36	92.31
Comorbidities (%)
Yes	8.94	1.92
No	1.03	27.88
Missing from documentation	90.03	70.19

Source: Original. * Percentages for racial and ethnic groups are calculated without cases for which there was disagreement. Disagreement determined when reports for initial and final race or ethnicity disagreed, without considering cases in which the information for one of the two reports was missing.

shows the sociodemographic and racial characteristics of the deceased, together with what is documented about their family, education, and comorbidities. As we can see, after combining percentages for both internal and external deaths, the vast majority of the deaths studied were “external” (973 external and 104 internal deaths), involving men (96.73%) self-identified or identified as “brown” (58.12%),7 of a median age of 29 at the time of death (ages between 16 and 88). For those for whom we had information, most were low-income (70.23%), living on less than $283 USD a month, unemployed or self-employed (88.37%), with incomplete primary education (70.96%), no children (40.65%), and single or divorced (70.47%).

Table 2. Frequency of crimes per carceral space of death.

	External Death	Internal Death
Cases (%)	973 cases	104 cases
Crimes against property	43.68	42.31
Drug-related offenses	20.76	20.19
Firearms and disarmament offenses	14.49	10.58
Crimes against the person	12.33	25.00
Other crimes	10.38	12.50
Traffic-related offenses	7.19	0.96
Domestic and family violence	5.86	0.96
Crimes against public administration	2.88	1.92
Crimes against sexual dignity	1.64	11.54
Crimes against the public order	1.23	0.00
Missing from documentation	0.62	0.00
Previous contact with CLS
Adult criminal legal system	55.50	57.69
Missing from documentation	17.99	13.46
Youth justice system	5.55	4.81
Missing from documentation	88.18	8846

Source: Original.

shows the types of crimes individuals were accused or sentenced for, and whether they had any previous contact with the carceral system as adults or youth. As we can see, most of the deceased were accused or sentenced for crimes against property (43.81%) and/or drug-related offenses (20.74%), and a majority had prior contact with the adult criminal legal system (68.37%). Yet this portrait is deeply fragmented.

Table 1 and Table 2 also show systematic gaps in the records that significantly limit what can be known about the lives and deaths of those who died while still entangled with the criminal legal system.

First, basic sociodemographic information is missing from court records, despite criminal procedural law and administrative regulations requiring this information to determine sentence terms and duration, possibilities for parole, or probation, and access to maternal and reproductive care, and social benefits for the incarcerated person’s family.

Within each case’s extensive documentation, it was impossible to identify the deceased person’s monthly income in 87.36% of external death cases and 92.31% of internal death cases. Formal education level was missing in 44.09% of external deaths and 40.38% of internal deaths. Information on the number of children and prior employment status was absent in approximately one-third of both external and internal death cases. Even prior contact with the criminal legal system, an element central to sentencing decisions and early-release assessments, was missing in 17.99% of external death proceedings and 13.46% of internal death proceedings.

Race and ethnicity present an additional layer of opacity. Racial classification was entirely omitted across all procedural forms, including indictments, prison or jail admission records, health documentation, and death certification, in nearly 10% of external death cases and 5% of internal death cases. When race or ethnicity was recorded, inconsistencies across documents were common, with contradictory classifications appearing in 9.76% of external death proceedings and 4.76% of internal death proceedings.

Health-related information is more strikingly absent. Over 90% of external-death proceedings and approximately 70% of internal-death proceedings contained no information on the individual’s comorbidities, as health-related questions in prison or jail admission and in-custody records were routinely left blank.

Table 3. Death information and investigation.

	External Death	Internal Death
Cases	973 cases	104 cases
Place of death (%) *
Health facility	33.61	44.23
Street	43.58	3.85
Correctional unit	0.10	49.04
Outdoors	2.26	0
Home	10.28	0
Missing from documentation	10.17	2.88
Who reported the death? (%)
Court personnel	47.89	21.15
Prosecutor	8.94	1.92
Defense lawyers	12.44	9.62
Police	7.19	2.88
Correctional Dept	2.88	40.38
Family	2.77	0.00
Other state agencies	5.76	14.42
Missing from documentation	12.13	9.62
Death investigation (%)
No	11.10	9.62
Yes	10.28	20.19
Missing from documentation	78.62	70.19

Source: Original. * Health facility includes forensic medical complex or psychiatric treatment hospital and overall healthcare units; Street includes urban street deaths, and deaths in religious or commercial establishments, and deaths that occurred “in transit”; Outdoors includes deaths in rivers, seas, lakes, woods, farms, ranches, rural areas and rural properties; Correctional unit include prisons, jails, and other units for semi-open and open sentence programs.

describes the documented place of death, who reports the death to the judicial authorities, and whether the death was investigated. Most external deaths were reported to have occurred on the street (43.58%) or in healthcare facilities (33.61%). Internal deaths were more often declared to have occurred in healthcare facilities (44.23%) or within correctional units themselves (49.04%). Even so, the place of death could not be determined in 10.17% of external deaths and 2.88% of internal deaths, underscoring ongoing gaps in basic factual documentation. Most notably, there was no information indicating that any investigation into the circumstances of death had occurred in 78.62% of external death cases and in 70.19% of internal death cases, i.e., even when deaths occurred while individuals remained under direct state custody.

Regarding the causes of death,

Table 4. All causes of death per carceral space of death.

External (973 Cases)	%	Internal (104 Cases)	%
X93: Assault by handgun	36.6	I00–I99: Circulatory system diseases	24.0
X99: Assault by sharp object	34.1	R99: Ill-defined causes of mortality	18.3
S27: Intrathoracic injury	28.8	X99: Assault by sharp object	17.3
S06: Intracranial injury	22.6	S27: Intrathoracic injury	16.3
I00–I99: Circulatory system diseases	21.9	J98.9: Respiratory disorder, unspecified	15.4
R58: Hemorrhage	21.8	T71: Asphyxiation	14.4
R57: Shock	14.1	A41: Other sepsis	12.5
R99: Ill-defined causes of mortality	11.6	J09–J18: Influenza and pneumonia	12.5
A41: Other sepsis	6.5	X91: Assault by hanging	10.6
J98.9: Respiratory disorder, unspecified	5.8	R57: Shock	8.7
J09–J18: Influenza and pneumonia	4.9	R58: Hemorrhage	8.7
D50–D89: Blood and immune system disorders	4.9	S06: Intracranial injury	8.7
Y34: Unspecified cause	4.2	B34.2: Coronavirus infection	4.8
V99: Unspecified transport accident	3.8	N00–N99: Genitourinary diseases	4.8
K00–K93: Digestive system diseases	3.0	K00–K93: Digestive system diseases	3.8
N00–N99: Genitourinary diseases	3.0	V99: Unspecified transport accident	3.8
C00–D48: Neoplasms	2.9	A15–A19: Tuberculosis	2.9
B34.2: Coronavirus infection	2.9	C00–D48: Neoplasms	2.9
V01–V98: Transport accidents	2.8	X93: Assault by handgun	2.9
N17–N19: Renal failure	2.7	D50–D89: Blood and immune system disorders	1.9
T71: Asphyxiation	2.2	E10–E14: Diabetes mellitus	1.9
X91: Assault by hanging	1.7	G00–G99: Nervous system disorders	1.9
G00–G99: Nervous system disorders	1.6	K70–K77: Diseases of liver	1.9
F10: Due to alcohol	1.5	N17–N19: Renal failure	1.9
K70–K77: Diseases of liver	1.5	Y20: Undetermined hanging	1.9
E10–E14: Diabetes mellitus	1.2	A00–B99: Certain infectious and parasitic diseases	1.0
R68.8: Other general symptoms	1.2	A30: Leprosy [Hansen disease]	1.0
A15–A19: Tuberculosis	0.8	B20–B24: HIV disease	1.0
B20–B24: HIV disease	0.8	E87: Fluid/electrolyte disorders	1.0
F17: Due to tobacco	0.7	F17: Due to tobacco	1.0
U04: SARS	0.7	T29-T32: Burns/corrosions	1.0
E66: Obesity	0.6	U04: SARS	1.0
A00–B99: Certain infectious and parasitic diseases	0.6	W00–W19: Falls	1.0
E87: Fluid/electrolyte disorders	0.6	W74: Drowning, unspecified	1.0
T29–T32: Burns/corrosions	0.6	W87: Electric current exposure	1.0
W74: Drowning, unspecified	0.5	X84: Self-harm, unspecified	1.0
W00–W19: Falls	0.4
S18: Neck amputation	0.3
F00–F99: Mental and behavioural disorders	0.3
F19: Due to multiple drug use	0.3
R64: Cachexia	0.3
X99/Y00: Sharp/blunt assault	0.3
W87: Electric current exposure	0.2
E00–E90: Endocrine, nutritional, and metabolic diseases	0.2
E86: Volume depletion	0.2
E88: Other metabolic disorders	0.2
F12: Due to cannabinoids	0.2
S42–S92: Fractures (limbs)	0.2
A27: Leptospirosis	0.1
B55: Leishmaniasis	0.1
B58: Toxoplasmosis	0.1
L93: Lupus erythematosus	0.1
M10: Gout	0.1
N49: Inflammatory male genital disorders	0.1
Q00–Q99: Congenital malformations	0.1
T36–T50: Drug poisoning	0.1

Source: Original.

indicates that external deaths were overwhelmingly attributed to violent trauma, particularly assault by a handgun (36.6%) or assault by a sharp object (34.1%). Internal deaths, by contrast, were more frequently attributed to circulatory system diseases (24.0%), but also included a substantial proportion linked to violence and trauma, such as assault by a sharp object (17.3%), assault by hanging (10.6%), and intrathoracic injury (16.3%).

Yet, in both types of cases, a significant proportion of causes described in death certificates were unspecified, indeterminate, or ill-defined. We found a high frequency of death certificate language matching ICD ‘R’ codes, i.e., symptoms, signs, and ill-defined conditions that cannot constitute underlying causes of death (França et al. 2020; Aquino et al. 2025). These terms were often used as the sole reason to describe 134 causes of death (11.6% of external deaths, 18.3% of internal deaths) or combined with other causes in 492 cases (46% of external deaths and 34.61% of internal deaths), as shown in

Table 5. Most common combinations of causes of death per carceral space of death.

External (973 Cases)	%	Internal (104 Cases)	%
Ill-defined causes of mortality	11.6	Ill-defined causes of mortality	18.3
Intracranial injury|Assault by handgun	4.5	Asphyxiation|Assault by hanging	5.8
Intracranial injury|Assault by sharp object	4.3	Circulatory system diseases|Shock|Assault by sharp object	3.8
Intrathoracic injury|Assault by handgun	2.9	Circulatory system diseases	2.9
Intrathoracic injury|Assault by sharp object	2.4	Respiratory disorder, unspecified	2.9
Hemorrhage|Intrathoracic injury|Assault by handgun	2.3	Hemorrhage|Assault by handgun|Assault by sharp object	2.9
Hemorrhage|Assault by handgun|Assault by sharp object	2.0	Intracranial injury	2.9
Assault by sharp object	2.0	Other sepsis	1.9
Circulatory system diseases	1.7	Coronavirus infection|Influenza and pneumonia|Respiratory disorder, unspecified	1.9
Circulatory system diseases|Shock|Intrathoracic injury|Assault by handgun	1.7	Circulatory system diseases|Shock|Intrathoracic injury|Assault by sharp object	1.9
Hemorrhage|Assault by handgun	1.7	Influenza and pneumonia	1.9
Intracranial injury|Assault by handgun|Assault by sharp object	1.7	Intracranial injury|Intrathoracic injury|Assault by sharp object	1.9
		Intracranial injury|Assault by sharp object	1.9
		Intrathoracic injury|Asphyxiation	1.9
		Asphyxiation	1.9

Source: Original.

below. Unspecified or incomplete causes are also common—for instance, sepsis or unspecified respiratory disorder is appointed a cause of death in close to 5% of internal death cases. Table 5 presents the most frequent combinations of causes of death recorded in the criminal proceedings.

Rather than reflecting discrete and mutually exclusive categories of mortality, the causes assigned across cases reveal a pattern of overlapping, incomplete, and often ambiguous classifications, an obfuscation that affects deaths categorized as violent, natural, self-inflicted, and accidental alike. Both internal and external deaths frequently included assault-related causes. Of the 973 external deaths, 586 cases (60.2%) involved some form of assault, most commonly by firearms or sharp objects. Notably, 29 of the 104 internal deaths (27.8%) also included assault codes’ language among their listed causes. These findings indicate that violent causes of death are not confined to deaths occurring outside custodial spaces, but remain present—even if at lower proportions among deaths occurring while individuals are under direct state custody.

At the same time, deaths classified as “natural” are marked by substantial diagnostic imprecision. Circulatory system diseases constitute the most frequently recorded category among internal deaths (24.0%), followed by respiratory conditions and generalized infectious processes. However, these diagnoses are often presented in broad terms, such as cardiogenic shock, cardiac arrhythmia, or cardiorespiratory arrest, and as stand-alone causes of death, without supporting clinical detail, medical history, or temporal information regarding symptom onset, treatment, or deterioration. In approximately 5% of internal death cases, sepsis and respiratory failure appear either as standalone causes or as terminal events without any preceding diagnosis, comorbidity, or causal explanation documented in the record.

Deaths potentially attributable to suicide also illustrate the limits of classification. In only one case, the specific cause of death mentioned suicide or self-inflicted injury. Asphyxiation is present in 15 internal death cases (14.4% of cases) and 21 external death cases, often combined with hanging or intrathoracic injury. The absence of investigative documentation, psychiatric history, or contextual information makes it difficult to determine whether these deaths were classified following forensic inquiry or merely inferred from bodily presentation. Similarly, accidental deaths, such as fires, falls, electrocution, and other unintentional injuries, appear sporadically and without consistent narrative explanation. These events are frequently coded through injury mechanisms alone, without information regarding environmental conditions, institutional responsibility, or prior risk assessments.

In 76% of external cases and 61% of internal cases, the criminal file contains no additional information about the circumstances of death beyond the death notification, which in most cases is provided through the death certificate. When information about the circumstances is provided, it varies in scope and authority: in some instances, it comes from the police authority responsible for investigating the case, while in others it is provided by healthcare personnel describing the patient’s condition prior to death. There is no clear pattern regarding which authority provides such contextual information or the level of detail included.

4. Discussion

The database analyzed in this study is not statistically representative of all prison deaths or of population-level mortality in Brazil. It is instead a judicially produced dataset anchored in criminal proceedings triggered by Article 107(I) of the Brazilian Penal Code. Future research should expand the scope of this database by increasing the number of cases and weighting them by state prison populations and broader demographic distributions. While these features of the dataset limit its generalizability, the data assessed does offer a unique vantage point into how deaths are often processed within the legal system itself. Across the 1077 cases analyzed, we observe a pattern of omission and obfuscation characterized by pervasive forms of missing and ill-defined information. This pattern is not confined to demographic variables or health histories but extends to the recording of causes of death and to the presence, or absence, of any institutional response following a death. The frequent use of ICD “R” codes’ language (symptoms, signs, and ill-defined conditions), either alone or combined with other causes, exemplifies how official records within these judicial proceedings often do not provide sufficient explanation for deaths reported. This makes it difficult not only to assess individual cases, but also to shed light on how national databases such as SISDEPEN aggregate these cases into broad categories like “natural,” “homicide,” “accident,” “suicide,” or “unknown.”

We plan to systematically compare cause-of-death classifications derived from judicial records with those found in the databases described earlier. Such comparisons are particularly relevant for external deaths, where the patterns of ignorance observed here may reflect broader bureaucratic limitations in civil registration and medical certification systems—limitations that are well documented in population-level mortality research, especially with respect to garbage codes and non-standardized ICD use (Aquino et al. 2025; França et al. 2020).

Still, the combinations of causes presented in Table 4 and Table 5 illustrate the problems of the aggregated data in the available datasets. A substantial number of deaths, both internal and external, include indicators of violent trauma, particularly assaults involving firearms or sharp objects. Even among internal deaths, where individuals remained under direct state custody, violent causes appear with notable frequency. Yet, these deaths are likely underreported and ultimately folded into highly aggregated categories within the general datasets that obscure the role of carceral exposure in producing them. For instance, the SISDEPEN dataset shows that, between 2017 and 2021, violent internal deaths represented around 16% of all internal deaths during this period (Brazilian National Penitentiary Department 2025). In our dataset, restricting the analysis to internal deaths described as caused by assault by hanging, assault by handgun, or assault by sharp object, the percentage of violent deaths occurring in the same period is 28%.

Violent deaths have long been associated with incarceration in Brazil. This association is rooted in a historical trajectory that links punishment, racialized control, and the production of death. As Flauzina (2008) argues, Brazil’s criminal legal system is founded on violence, with racism as a central organizing principle, an inheritance from colonial slavery that has never been fully dismantled. During the recent military dictatorship (1964–1985), deaths in spaces of confinement were closely linked to torture and political repression, with prisons functioning as sites of disappearance and extrajudicial killing (Arns 2022). The murder of journalist Vladimir Herzog,8 staged as a suicide, remains a paradigmatic example of how institutional violence was concealed through bureaucratic manipulation.

These practices did not disappear with democratization. The 1992 Carandiru massacre stands as a defining moment in which the logic of extermination was made explicit with several problematic effects in terms of memory, truth, and justice to this day (Machado et al. forthcoming). Since the 1980s, lethal violence among incarcerated and formerly incarcerated individuals has also become a dominant public narrative, particularly as interpersonal conflicts inside prisons escalated into deadly encounters both inside and outside prison walls. More recently, the emergence of new prison-based organizations involved in illegal markets has intensified inter- and intra-prison conflicts as well as violent confrontations with law enforcement. These dynamics spill over into urban violence while also producing episodes of mass violence within prisons, such as the massacres of early 2017 in the North and Northeast regions (Biondi 2018; Manso and Dias 2018; Feltran 2018). In these accounts, deaths are frequently framed as the result of incarcerated people’s own actions toward one another, a framing that absolves the state of responsibility even when authorities respond with extreme or disproportionate force.

At the same time, so-called “natural deaths” can play a central role in sustaining prison lethality. Our findings show that internal and external deaths are often classified under natural causes, particularly circulatory and respiratory conditions, sepsis, and pneumonia. These classifications reveal more about these deaths than the “natural” label that is used in the official databases. However, because of the lack of further context and the framing of these causes as bodily processes, they might conceal long processes of neglect and deterioration connected to the exposure to stress, overcrowding, poor sanitation, delayed diagnosis, and inadequate access to care in prison settings. When framed as natural or as the result of a biological process alone, these deaths are likely normalized and rendered legally unremarkable, even though they could be tied to conditions of confinement and institutional failure during incarceration spells.

Judicial actors play a critical role in this process. As Duarte et al. (2021) show, courts frequently deny state liability in lawsuits filed by incarcerated and formerly-incarcerated people and their family members by asserting that adequate care was provided or that no causal link can be established between incarceration and illness individually, arguments that resonate with the absence of detailed information about the circumstances of death documented in most of the proceedings examined in our study.

Suicide presents another dimension of prison lethality marked by epistemic fragility. While suicide prevention has become a central concern for prison administrations, courts, and post-release care in countries such as the United Kingdom, France, and Portugal (Janca et al. 2023; Liebling 2001; Miller et al. 2024; Pratt et al. 2009; Zhong et al. 2021), in Brazil, the debate remains incipient. The data examined here reveal the difficulty of identifying suicide with a high degree of confidence, as deaths involving asphyxiation or hanging are often ambiguously classified and rarely investigated. This ambiguity further illustrates how classificatory practices can deflect responsibility rather than generate understanding. In a similar vein, deaths attributed to accidents also underscore how infrastructural neglect and unsafe life conditions inside prisons can be lethal, while remaining largely invisible in official accounts.

Finally, the issue of disappearance represents perhaps the most extreme form of epistemic erasure. Episodes such as the uprisings at Pedrinhas, in 2010, Alcaçuz, in 2016, and Monte Cristo, in 2017, involved not only mass deaths but also dozens of missing individuals. None of the official databases available records disappearances, and judicial cases under Article 107(1) will not be informed of disappearances for the purposes of punishability extinction. In fact, as documented by Prando (2021), prison administrations and the court system classified the missing as “fugitives,” a designation that shifts blame onto the victims and effectively halts investigation. The practice exemplifies how administrative labeling can function as a technology of disappearance that can allow the state to manage death by denying its occurrence.

The pervasive absence of investigations, observed in the majority of both internal and external death cases, reveals a criminal legal system that did not consistently exercise its authority to question, clarify, or intervene in the cases studied, despite having the authority and power to do so granted by criminal law. In this descriptive analysis, we identify omission, obfuscation, and inaction, but cannot determine the precise organizational, political, or professional incentives that produce them and lead courts to often not inquire further and rely on the opaque information presented to them.

Research on police killings in Brazil and the failure to hold police officers accountable in the vast majority of cases offers a potential hypothesis to the problem—it is possible that courts do not investigate and inquire further about the records they receive and endorse because they both have “normalized” the death, and especially violent death, of a “disposable population” (Mallart and Godoi 2017; Willis 2015) while also being removed from the initial labeling of that death, done by prison administrators, medical examiners, registries, police officers, who will control the initial categorization (Brinks 2007). In this sense, courts might have internalized a specific cultural understanding of what these deaths mean, while also facing fewer incentives to question the authorities who provide the information that informs their work. Ignorance, in this sense, would be produced by this initial categorization and maintained by courts, with the potential effect of shielding state institutions from accountability while rendering carceral death administratively ordinary. In future work, we hope to examine whether this explanation can be extended to deaths related to carcerality or whether other ways of conceptualizing ignorance and liability also need to be considered.

Beyond its empirical contributions, this study is likely to offer guidance for policy reform. Improving data policies and the production of accurate, reliable, and comprehensive information on deaths of people in contact with the criminal legal system, both inside and outside prison, should be understood as a foundational step toward monitoring prison conditions and assessing the broader health and mortality impacts of incarceration. Building on existing systems such as SEEU, CNIEP, and SISDEPEN, reforms should prioritize the adoption of disaggregated, standardized, and shared classificatory categories, with more granularity and disaggregation at the individual level, together with the systematic recording of missing and unidentified persons. At the court level, norms that make investigations mandatory, rather than left to the discretion of judicial authorities, whenever a person has been incarcerated in the course of criminal proceedings, should be adopted, together with the active mobilization of different sources of knowledge beyond official documentation. Much of what we know about Leonardo’s case was derived from independent investigations conducted by academic and civil society organizations, together with the work of the public defenders’ office before the Inter-American Court of Human Rights. Indeed, journalists, civil society actors, family members, friends, and public defenders should be an integral part of the sources consulted by courts if they are to increase epistemic justice in efforts to understand death and its connections with carceral spaces.

5. Conclusions

In this paper, we describe how deaths of people in contact with the criminal legal system are documented, classified, and ultimately rendered legible within judicial records in Brazil. By focusing on how people and causes of death are recorded outside the aggregated databases available, we show how even in this more granular and official documentation, omissions, obfuscation, and institutional inaction are likely shaping what can be known about deaths connected to carceral spaces. We advance a first step toward a larger project of thinking epistemically about how we know causes of death, deaths themselves, and spaces of punishment, adding an epistemic dimension to the “prison exposure” literature, a dimension that may itself produce harm and function as a latent contributor to the material harms associated with carceral spaces. Although the paper focuses on Brazil, we believe the work carried out here is likely relevant beyond this context, and may be particularly applicable to countries whose penal systems remain shaped by histories of slavery, authoritarian rule, and racialized punishment.