Suicide While Locked Up in Texas: Risk Factors for Death by Suicide in Custody

Abstract

In the United States, suicide is a leading cause of death in prisons and jails, with incarcerated individuals being nine times more likely to die by suicide than the general population. Identifying vulnerabilities at each stage of custody (prebooking, jail, prison) and factors that increase suicide risk can improve prevention efforts. A hierarchical binary logistic regression was conducted on data from the Texas Justice Initiative’s Deaths in Custody Report. Variables included race/ethnicity, sex, age at death, days in custody, classification of crime as violent or nonviolent, and custody type of prebooking, jail, or prison. Among main effects, when compared to suicide rates in prison, jail suicide deaths were over three and a half times more likely (OR = 3.61), and the period of prebooking emerged as a period of staggering risk of suicide death, with suicides being over 5,000% more likely than at other stages of custody (OR = 50.86). When interactions were entered, Latinx individuals were at a particularly increased risk of suicide death (OR = 10.46), likelihood of suicide death decreased with each year of age (OR = .89), nonviolent offenders were just under three and a half times more likely to die by suicide when compared to violent offenders (OR = 3.45), and each stage of custody was shown to affect the relationship between age-related rates of suicide in different ways. Results call for further investigation into suicide among understudied populations in corrections, such as Latinx individuals, juveniles in the prison system, and nonviolent offenders, to identify the groups at the highest risk of premature death in correctional systems.

Keywords

forensic suicide jail prisons custody

Issues pertaining to forensic suicide prevention have been at the forefront of the public’s consciousness following a string of high-profile correctional suicide deaths. These deaths have garnered not only media interest but also increased scrutiny of the legality, ethics, and care for incarcerated individuals who experience suicidal crises. In U.S. prisons, the rate of death by suicide has risen 83% in the 20-year period since 2002, and within jails, incarcerated individuals are more likely to die by suicide than any other form of death (Office of Justice Programs, 2021). In fact, the most recent data shows the highest rate of suicide deaths in jails since data reporting began in 2001 (Dixon et al., 2020; Mumola, 2005). A growing body of international literature has increased focus on some of the most relevant issues that impact an incarcerated individual’s risk of suicide, from a history of self-injury (Favril & O’Connor, 2021), the social environment of individual correctional facilities (Stoliker, 2018) to history of restraint and solitary confinement (Kaba et al., 2014).

The existing literature also supports that history of aggression, poor physical health, components of psychopathy, and exposure to Adverse Childhood Events (ACEs) were associated with violent offense history, which, in turn, was related to increased risk of death by suicide among prisoners (Sarchiapone et al., 2009). A Bureau of Justice Statistics Report (Mumola, 2005) found that within jails, violent offenders (93 deaths per 100,000 residents) had a suicide rate over three times higher than nonviolent offenders (31 cases per 100,000 residents). Similarly, prisoners charged with a violent crime (19 deaths per 100,000 residents) had a rate of suicide deaths more than twice that of nonviolent prisoners (9 deaths per 100,000 residents). Indeed, the proportion of violent offenses within a prison environment was found to significantly predict the number of deaths by suicide (Zhong et al., 2021). However, some foundational work examining suicide within corrections would support that nonviolent offenders are the most likely to die by suicide within corrections (Hayes, 1983, 2012). As multiple studies on national samples have tended to find mixed results, there is a need for a more granular view to determine how violence within charges can affect one’s risk of death by suicide. Suicide theory may help us determine how violence could play into what we know about the development of suicide risk.

Suicide Theory

Suicide risk refers to general risk factors thought to be relevant within the progression of suicidal thoughts, to suicidal behaviors, and finally to death by suicide. Forensic suicide risk refers to specific factors relevant in a legal context that influence this risk of suicide. To understand the nature of the link between some common suicide risk factors and death by suicide within corrections, one can look to suicide theory. A foundational theory of suicide is drawn from Durkheim’s Suicide: A Study in Sociology (1897). This theory consists of two core principles. First, social factors, rather than just individual reasons, significantly impact suicide rates. Second, different types of social integration and moral regulation contribute to distinct forms of suicide. Building on this theme of suicide being highly influenced by socialization, the Interpersonal Theory of Suicide (IPTS Joiner, 2005) is a well-known theory of suicidality that posits that there are three components that lead to death by suicide. The first two, in which suicidal ideation is formed, consist of thwarted belongingness, or the feeling as if one is not accepted by others and has little social connectedness, and perceived burdensomeness, or the belief that one is a burden on others or on society. Once suicidal ideation is formed, suicidal behavior can result if Acquired Capability (AC) is present. This term describes reductions in fear and pain sensitivity sufficient to overcome self-preservation reflexes and is linked to exposures to painful and provocative events (van Orden et al., 2010). Another theory relevant to factors in correctional suicide is the 3-Step Theory (3ST; Klonsky & May, 2014). This theory, similar to the IPTS, poses three factors which are inherent in the formation of suicidal ideation and the progression to suicidal behavior. In the 3ST, pain and hopelessness together form desire to die by suicide. If the experiences of pain overwhelm resources for social connectedness, then the desire to die by suicide escalates. With the presence of AC, suicidal behavior can follow. These two theories share overlap in many areas, but both place a great deal of importance on the component of AC and suicidal behavior, which also implicates it in death by suicide. In fact, higher levels of AC have been associated with increased likelihood of both suicidal behavior and death by suicide (Smith et al., 2010).

Suicide Theory Components and Correctional Suicidality

Aggressive behavior is associated with increases in AC, and therefore, an association is inferred between aggressive behavior and violent offenses (Burke et al., 2018; Polaschek et al., 2009). Thus, it is likely that among violent offenders, an increase in aggression may lead to higher rates of suicidal behavior. In addition, violent offenders are shown to have higher exposure to painful and provocative events in the form of ACEs, higher pain tolerance, and higher rates of hopelessness (McCoy et al., 2010; Miller et al., 2014; Stinson et al., 2021), all concepts enmeshed in suicide theory and linked to increased risk of suicidality (Abramson et al., 2002; Thompson et al., 2019). However, the role of AC in forensic populations is more complicated than assumed, as a study conducted on 343 male incarcerated individuals in two Mississippi Department of Corrections facilities found that incarcerated individuals convicted of more than one violent offense had lower levels of AC compared to nonviolent offenders. In fact, those with violent offenses had reductions in AC even when compared to general population samples (Mandracchia et al., 2018).

Dispositional risk factors are also important considerations when conceptualizing suicide models and have been traditionally understudied in forensic literature (Stoliker, 2018). In Durkheim’s Theory, groups, such as those that can be drawn from demographic factors within corrections, should influence suicide risk differently. Two important variables that have emerged from prior research are race/ethnicity and age. Existing literature indicates that when examining the relationship of race/ethnicity as a dispositional characteristic, White incarcerated individuals are most at risk for death by suicide (Mumola, 2005). Whereas research focusing on adjusted rates for suicide death with respect to population levels within particular facilities has found race/ethnicity to not be a significant indicator of death by suicide (Dye, 2011). In a recent study, which included both dispositional and facility characteristics, White incarcerated individuals were at a 35% increased risk for suicidal behavior and comprised 49.1% of all suicide deaths in the prison samples, with populations of each facility ranging from as low as 9% White to as high as 89%. This same research has posited that rather than the impact of race/ethnicity itself, it is race/ethnicity relative to the majority population within a particular facility that causes suicide risk, in that a person of a racial identity that is not the majority identity of other inmates in a facility may be at greater risk (Stoliker & Galli, 2021). Further research is warranted to refine how race/ethnicity may impact overall suicide risk and interplay with individual characteristics and custodial and facility factors to help comment on these mixed findings.

Age was also found to be significantly associated with suicide risk, with each year of increased age revealing a 1.3% decrease in suicide attempts in jail suicide rates (Stoliker, 2018). This is a vast departure from findings on general population suicide risk, which indicates that middle age is the most vulnerable time for suicide risk during the lifespan (Center for Disease Control (CDC), 2023). A recent systematic review of corrections found that above the age of 25 years, there was no significant increase in suicide death (Zhong et al., 2021), again indicating that younger individuals are most vulnerable to death by suicide in forensic institutions. Along this logic, juvenile incarcerated populations are a group particular concerning for suicide risk, particularly those housed in adult prisons, as they have been found to be the most vulnerable to violence and death by suicide (Daniel, 2006; Kuanliang et al., 2008). The consistently supported risk for younger age incarcerated individuals elucidates a necessity for deeper understanding of the mechanisms by which age impacts increased suicide risk while in custody.

Biological sex has also been shown to impact rates of suicidal ideation, suicidal behavior, and suicide deaths within correctional settings. However, the directionality of this relationship is often unclear, with results often finding conflicting or mixed results. For example, in some research on corrections systems, females tend to have higher rates of suicidal ideation and more instances of suicidal behavior (McCullumsmith et al., 2013). Other studies of different correctional systems have found males to have more instances of suicidal behavior and a greater rate of death by suicide (Daniel & Fleming, 2006). Other statewide studies, such as one 5-year study conducted in the Texas Department of Corrections over 30 years ago, found that all suicide deaths were male (Anno, 1985). Others have argued that disproportionate numbers of males in the correctional system obscure that females have a higher rate of death by suicide when adjusted for discrepancies in population (Dye, 2011). Further complicating these findings in corrections are studies that have shown interactions between other demographic variables and sex (e.g., age, crime type, etc.) to be key in determining the nature of how sex affects risk of death by suicide in corrections (McCullumsmith et al., 2013). More research is necessary to determine individual and interaction effects of sex in correctional suicide deaths

Custody type is also an important factor to consider when researching risk factors for forensic suicide risk. Custody types are generally defined in stages, with the first being prebooking, which includes first contact with law enforcement, police custody, arrest, and transportation, typically to jail. Next in the process is jail custody, the initial stage of incarceration, where people await resolution of their case or serve sentences under 1 year. Lastly is prison custody, which is the facilities that handle long-term housing of incarcerated individuals post-conviction. Jail custody consistently leads state and national statistics with the highest suicide rates (Kenny et al., 2008). Interestingly, a meta-analysis comparing over 60 years of incarceration suicide data found a complex relationship between offense and custody types, with prebooking suicide deaths being almost predominately comprised those charged with violent offenses. Conversely, suicides in prisons were found to be mostly individuals convicted of nonviolent offenses; jail suicides fell somewhere between the two (Felthous, 2009).

From 2001 through 2019, Texas had the country’s second-highest jail suicide rate, with 448 deaths, and the second-highest suicide rate in prisons, with 527 deaths. A disproportionate amount of these suicide decedents were those being held on mental health charges who had committed no crime at all (Texas Justice Initiative, 2021). The state of Texas is often strict on crime and punishment, leading to unique features in its policing and correctional systems; since the reinstatement of the death penalty in 1976, it has executed 586 people, with almost five times as many executions as the next closest state, Oklahoma, with 123 (Death Penalty Information Center, 2023). In fact, Texas single-handedly executed more individuals than the next six highest states combined (Death Penalty Information Center, 2023). At the end of 2019, Texas’ rate of incarceration was 840 people per 100,000 people in its jails, making it the 10th highest state for jail population and, de facto, 10th highest in the world the state to be ranked as a country (Prison Policy Initiative, 2021). Further, Texas is 1 of 13 states that have mandatory minimum sentencing laws which impose the harshest sentencing guidelines, with no regard for mitigating circumstances. Lastly, the public defender system in Texas is grossly overwhelmed, leading to defendants often being coerced into plea agreements or receiving substandard defenses (Laughland & Mathieu-Leger, 2016). These factors likely exacerbate an already overburdened criminal justice system and contribute to an increased risk of death while in custody.

Taken together, there are variety of factors which are thought to influence the likelihood of death by suicide in each period of custody. However, often these findings, when conducted on individual facilities, tend to present findings that are mixed. For example, the nature of sex or crime type classification has been found to conflict at different points, depending on the stage of custody. As such, it is difficult to find overarching risk factors for death by suicide for individuals progressing through custody. Accordingly, we do not have a clear view of those entering custody at heightened risk of death by suicide. A custody-wide look at correctional suicide risk factors appears to be called for.

Considering existing literature on forensic suicide and understanding the distinctive criminal justice system within Texas, we aim: (a) to identify the relationship between crime type (non-violent versus violent offenses) and risk of suicide, (b) to investigate how time in custody and custody type impacts death by suicide (c) explore dispositional traits of race/ethnicity, and age and how they may increase suicide risk. First, we hypothesize crime type (violent vs. nonviolent) will impact suicide risk, with the former category having the highest rate of death by suicide. We hypothesized this due to the more serious penalties associated with violent crimes likely sparking feelings of hopelessness about the future, as well as increases in aggression, impulsivity, and pain insensitivity associated with exposure to violence. Second, consistent with higher rates of jail suicide when compared to prisons, we predict that suicide risk will be elevated in the initial stages of incarceration, with decreases each subsequent day in custody. We reason that, given the large and dramatic shift in environment as one enters custody, with little in the way of protective factors, suicide risk would peak upon first entry into custody and decrease thereafter. And given that time in custody for most individuals is associated with a similar linear progression through stages of custody (i.e., prebooking to jail to prison), we also predict a relationship exists between suicide risk and custody type, with the earliest stage of custody, prebooking carrying the highest risk for death by suicide, followed by jail custody, and lastly, prison custody. Taking into consideration prior literature, we also hypothesize that younger individuals will be most likely to die by suicide while in custody, with each day in custody decreasing the likelihood of death by suicide. Finally, we hypothesize that some interaction effects will emerge for these variables that will give us a more nuanced view of who is entering custody at the highest risk of death by suicide.

Methods

Participants and Procedures

To foster transparency, nonprofit groups such as the Texas Justice Initiative have published statewide data on all custody deaths in Texas since 2005. Data for this study were pulled from the Texas Justice Initiative’s Death in Custody Report (Texas Justice Initiative, 2021), an open-source dataset covering all deaths that occurred under law enforcement custody beginning in 2005 and continuing to the present day. The current study used data available until October 2020 (n = 10,895). Death records spanned a wide range, with the youngest decedent being 15 years old and the oldest being age 95 (M = 54.34). For manner of death, we elected to exclude deaths by drug overdose, homicide, or deaths under suspicious circumstances. This is because, as past research has shown, many deaths ruled drug overdoses may be misclassified suicide attempts (Rockett et al., 2018); as we cannot definitively classify these deaths, we decided to use a sample of natural deaths and suicides only.

For the classification of violent and nonviolent offenses, we used the Texas Penal Code’s sections 19 & 22 to classify crimes according to the statutes of Texas law (Tex. Penal Code § 19.02, 1994; Tex. Penal Code § 22.01, 2003). Violent crimes included assault, rape, robbery, deadly conduct, kidnapping, manslaughter, and murder. Nonviolent crimes included any crime which did not meet this definition. For this classification, we excluded records that fell into legal gray areas of classification, for example, felony animal abuse or child sex crimes, which are not classified as strictly violent or nonviolent and have their own categorization as “child crimes” under Texas state law, and thus were excluded due to this ambiguity.

Before conducting our analyses, we conducted a cross-tabulation of our focal predictors concerning the violent nature of the crime (violent offense or nonviolent offense) and our outcome variable, death by natural causes or death by suicide in Texas custody (n = 6,349). From our sample, 14.6% (n = 931) died by suicide, of which 64% (n = 596) were classified as nonviolent offenders. In comparison, 56% of the natural deaths in our sample were non-violent offenders. The majority of records within our sample came from prisons (n = 6,152), with county and municipal jails (n = 853) and prebooking (n = 278) comprising the rest of the death records. Initially, a category of deaths logged from private prisons (n = 35) was considered for analysis, but this subset proved to be too small to include for further consideration and was thus excluded. Table 1 includes the demographic statistics for our dataset broken down by the manner of death.

Table 1.

Demographic Statistics by Manner of Death.

Demographics	Natural Death	Suicides	Total n (% of Death by Suicide)
Race/ethnicity by manner of death
White	2,679	467	3,146 (50.2%)
Black/African-American	2,002	198	2,200 (21.3%)
Hispanic/Latinx	1,668	266	1,934 (28.5.%)
Sex by manner of death
Male	6,099	891	6,990 (92.3%)
Female	323	74	397 (7.6%)
Crime type by manner of death
Nonviolent offense	3,038	596	3,634 (64%)
Violent offense	3,314	335	3,649 (36%)
Type of custody by manner of death
Jail—county	571	282	853 (30.3%)
Police custody (prebooking)	40	238	278 (25.6%)
Prison	5,741	411	6,152 (44.1%)

Data Analytic Approach

We conducted a set of Pearson’s chi-square tests to determine whether our categorical variables of race/ethnicity, classification of crime, and type of custody significantly contributed toward the likelihood of death by suicide. We found that all variables were significant in this analysis. Table 2 contains these results. In addition, for our continuous variables, we ran two asymptotic general independence tests to determine if there was a difference in values. We found that both age at death (Z = 40.1, p < .001) and days in custody (Z = 18.4, p < .001) were shown to be significant, indicating that our variables were appropriate for further analysis.

Table 2.

Chi-Square Tests on Categorical Variables.

Predictor	χ²	df	p
Violent vs. nonviolent offense	82.8	1	<.001
Type of custody	1,900.7	3	<.001
Race/ethnicity	42.07	2	<.001
Sex	9.06	1	.002

We then conducted a hierarchical binary logistic regression with the manner of death (suicide vs. natural causes) as the outcome variable. For the first three blocks, the reference level was set for each level as the last level for that variable; for example, in the Race/Ethnicity variable, containing Black, Latinx, and White decedents, White decedents were used as the reference level due to being the last subgroup within that categorization. We conducted this regression through R. A copy of the dataset and R script can be accessed through the project’s Open Science Framework page.

Results

The results of the binary logistic regression conducted are listed in Table 3 & Table 4. Being that we did not have any particular a priori hypothesis about the relationship between race/ethnicity or sex and odds of death by suicide, race/ethnicity and sex were entered into the first block alone. A significant difference emerged between Black and White incarcerated persons in likelihood of death by suicide (p ≤ .001), such that Black individuals died by suicide about 43% less often than White individuals. This significant difference did not emerge between Latinx and White individuals. Regarding sex, a significant difference emerged between Male and Female individuals (p ≤ .001), such that Males died by suicide 37% less often than Females. Race/Ethnicity and Sex alone explained just above 1% of the variance in likelihood of death by suicide versus natural causes (ΔR² = .014).

Table 3.

Hierarchical Logistic Regression of Variables Influencing Risk of Death by Suicide in Texas Custody.

	Block 1			Block 2			Block 3
Variable	β	OR	p	β	OR	p	β	OR	p
Race/ethnicity
Black	−.57	0.57	<.001 ***	−1.31	0.27	<.001 ***	−0.57	0.56	.35
Latinx	−.07	0.93	.4	−.61	0.54	<.001 ***	−.28	0.76	.69
White	0b	—	—	0b	—	—	0b	—	—
Sex
Male	−.045	0.63	<.001 ***	.36	1.43	.06	.16	1.17	.85
Female	0b	—	—	0b	—	—	0b	—	—
Type of custody
Prebooking				3.93	50.76	<.001 ***	−.06	0.94	.96
Jail				1.3	3.68	<.001 ***	−.52	0.6	.43
Prison				0b	—	—	0b	—	—
Type of crime
Violent				−.95	0.39	<.001 ***	−.8	0.45	.16
Nonviolent				0b	—	—	0b	—	—
Days in custody				−.01	0.99	.29	<.001	0.999	.18
Age at death				−.11	0.89	<.001 ***	−.11	0.9	<.001 ***
Interactions
Race/ethnicity by type of custody
Black × prebooking							−.49	0.61	.31
Black × jail							−.65	0.52	.04 *
Latinx × prebooking							.72	2.07	.25
Latinx × jail							.54	1.72	.08
Race/ethnicity by crime type
Black × violent							−.46	0.63	.08
Latinx × violent							−.11	0.9	.69
Race/ethnicity by days in custody
Black × days in custody							<.001	1.001	.36
Latinx × days in custody							<.001	1.01	.008 **
Race/ethnicity by age at death
Black × age at death							−.01	0.99	.63
Latinx × age at death							−.03	0.97	.004 **
Race/ethnicity by sex
Black × male							−.16	0.85	.72
Latinx × male							.67	1.95	.23
Sex by type of custody
Male × prebooking							.05	1.06	.95
Male × jail							.22	1.25	.64
Sex by type of crime
Male × violent							.51	1.67	.22
Sex by days in custody
Male × days in custody							<.001	1.001	.23
Sex by age at death
Male × age at death							−.01	0.99	.58
Days in custody by type of custody
Days in custody × prebooking							−.002	0.998	.87
Days in custody × jail							−.001	0.999	.13
Days in custody × age at death							−<.001	0.999	.94
Days in custody by crime type
Days in custody × violent crime							−<.001	1.001	.2
Age of death by type of custody
Age at death × prebooking							.09	1.09	<.001 ***
Age at death × jail							.04	1.04	<.001 ***
Age at death × violent crime							−.02	0.98	.06
Type of crime by type of custody
Violent crime × prebooking							.29	1.34	.5
Violent crime × jail							.38	1.45	.15
AIC	5,581.3			3,116.1			3,060.7
R² (Nagelkerke)	.014			.539			.553
R² (Cox & Snell)	.007			.288			.296
−2 LL	−2,767.43			−1,550.65			−1,442.37
χ²	54.43		<.001 ***	2,482.9		<.001 ***	2,581.8		<.001 ***

Note. 0b signifies a reference level; thus, the term is redundant. Bolded figures indicate significant results. AIC = Akaike Information Criteria.

p < .05. **p < .01. ***p < .001.

Table 4.

Trimmed Final Model of Variables Influencing Risk of Death by Suicide in Texas Custody.

	Block 4
Variable	β	OR	p
Race/ethnicity
Black	0b	—	—
Latinx	2.34	10.46	.003 **
White	1.83	6.24	.001 **
Type of custody
Prebooking	0b	—	—
Jail	−.04	0.66	.57
Prison	.28	1.33	.68
Type of crime
Violent	0b	—	—
Nonviolent	1.24	3.45	.002 **
Age at death	−.04	0.96	.008 **
Interactions
Race/ethnicity by type of custody
Latinx × jail	.03	0.61	.95
Latinx × prison	−.88	0.41	.13
White × jail	.16	1.17	.75
White × prison	−.66	0.52	.15
Race/ethnicity by crime type
Latinx × nonviolent	−.24	1.001	.40
White × nonviolent	−.37	0.69	.15
Race/ethnicity by age at death
Latinx × age at death	−.02	0.98	.049 *
White × age at death	−.001	1.001	.95
Age of death by type of custody
Age at death × jail	−.04	0.95	.004 **
Age at death × prison	−.08	0.92	<.001 ***
Type of crime by type of custody
Nonviolent crime × jail	−.19	0.82	.64
Nonviolent crime × prison	.05	1.05	.89
AIC	3,067.5
R² (Nagelkerke)	.557
R² (Cox and snell)	.298
−2 LL	−1,511.34
χ²	2,556.6		<.001 ***

Note. 0b signifies a reference level; thus, the term is redundant. Bold indicates significant results. -2 LL = -2 Log Likelihood.

p < .05. **p < .01. ***p < .001.

Next, we entered the main effects of our focal variables. We found significant effects for each level of race/ethnicity, such that Black individuals and Latinx individuals showed a reduced likelihood of death by suicide when compared to White individuals. When compared to deaths by suicide in prison, the period of prebooking exhibited a 5,068% increase in likelihood of death by suicide. Jails, similarly, exhibited a 361% increase in likelihood of suicide death. Both effects were significant. Each year of age decreased likelihood of death by suicide by about 10%, which was a significant effect. Only Sex and Days in Custody did not exhibit a significant effect. These variables explained just under 54% of the variance in likelihood of death by suicide in custody (ΔR² = .539).

For our third block, we included interaction terms for all focal variables, which added 26 unique interaction terms. With the introduction of these interaction effects, the only significant main effect was age of death, such that the odds ratio was unchanged, wherein each increasing year of age reduced likelihood of death by suicide by 10%. We found an interaction effect between race/ethnicity and custody type such that Black individuals in jails died by suicide about 48% less often than White individuals in jails (p ≤ .04). Additionally, we found a significant interaction effect between age and race/ethnicity, within young Latinx individuals, who died 3% less often for every year of increasing age than young individuals with other racial identities (p ≤ .003). Latinx individuals had a small increase in likelihood of death by suicide for every day spent in custody when compared to both Black and White individuals, such that every day spent in custody, their likelihood of death by suicide would increase one-tenth of a percent. Taken together, this meant that young Latinx individuals who had spent a large amount of time in custody were at a notably increased risk of death by suicide. Finally, we found a significant interaction effect between age at death and type of custody, such that young individuals were at a 109% increased likelihood of death by suicide in prebooking, and a 104% increased risk of death by suicide in jail when compared to prisons. This block explained just over 55% of the variance in likelihood of death by suicide (ΔR² = .553). Figures 1 –3 contain the plotted simple slopes that help to visualize these interaction effects in this model.

Figure 1.

Interaction effects in the untrimmed interaction block for suicide versus natural death, age of death, and type of custody.

Figure 2.

Interaction effects in the untrimmed interaction block for suicide versus natural death, race, and age of death.

Figure 3.

Interaction effects in the untrimmed interaction block for suicide versus natural death, race, and days from entering custody until death.

Finally, we utilized Zhang’s (2016) model-building strategy for logistic regression to create a final trimmed block of logistic regression, as shown in Table 4. Terms were eliminated from Block 3 if they did not exhibit a significant main effect or contribute to a significant interaction effect. In addition, to obtain terms for reference level variables in the previous models, reference levels were flipped to the first level in a group, as opposed to previous blocks that used the last level. In this model, while White individuals had just over six times the likelihood of death by suicide (OR = 6.24, p = .001) when compared to Black individuals, Latinx individuals were particularly at risk of death by suicide, with just under 10 and a half times the likelihood of death by suicide (OR = 10.46, p = .003), both of which were found to be significant effects. Contrary to our expectations, we found a significant main effect for individuals charged with nonviolent offenses, such that these individuals were three and a half times more likely to die by suicide in custody (OR = 3.45, p = .002). To confirm that this finding was not in error as a result of an excessive amount of mental health wellness checks conducted by police, which are often logged as nonviolent prebooking deaths should they result in a death by suicide, we conducted a quick exploratory check. We found that even with the prebooking period removed from analyses, a logistic regression confirmed that within our dataset, violent offenders were at a reduced risk of death by suicide (OR = .474, p < .001), close to a 53% reduction in likelihood. Interaction terms found that for every increased year of age, individuals in jails were less likely to die by suicide when compared to prebooking, meaning that with younger individuals, suicide deaths in jail were more likely. This was also the case in prisons, where each increasing year of chronological age reduced likelihood of death by suicide by 8%. Additionally, we found an interaction effect between Latinx individuals and age at death, such that every increasing year of age for Latinx persons, the likelihood of death by suicide in custody decreased by 2%, a significant difference from individuals of other racial identities. This model slightly improved the amount of variance in likelihood of suicide death when compared to death by natural causes over the untrimmed model (ΔR² = .557). As indicated in Tables 3 and 4, our goodness of fit statistics, including AIC and −2LL, favored the final trimmed model. Figures 4 and 5 visualize the plotted simple slope for the final model.

Figure 4.

Interaction effects in the final model for suicide versus natural death, age of death, and type of custody.

Figure 5.

Interaction effects in the final model for suicide versus natural death, race, and age of death.

Discussion

Contrary to our hypothesis, non-violent offenders were at a three and a half times greater risk for death by suicide. This was unexpected for us, as this finding contrasts with data from past research, which found violent offenders to be at a greater risk of suicide (Favril & O’Connor, 2021; Webb et al., 2013). A study by Bukten and Stavseth (2021) supported finding of an increased risk of suicide death for those convicted of serious violent offenses and homicides. This effect, also supported in other research (Webb et al., 2012), suggests heterogeneity of suicide risk exists among violent offenders, and perhaps individuals capable of committing extremely violent crimes are those more at risk for death by suicide when compared to other violent offenders, such as those charged with assault or robbery. Research has also indicated that the presence of mental health disorders, demographics of the sample, and methodology (i.e., self-report, clinician-administered) can impact findings on suicide risk (Douglas et al., 2006).

Research on violent offending and suicide frequently explores the role of antisocial traits, behavior, and psychopathy. While aggression and psychopathy can be linked to violent offenses, some research suggests these factors may actually protect against suicide (San Segundo et al., 2018; Swogger et al., 2009). The IPTS (Joiner, 2005) proposes that an individual develops beliefs related to being a burden to their community (perceived burdensomeness) and feelings that they do not belong to their community (thwarted belongingness), which initiates suicidal thoughts and behaviors. Some have argued that violent offenders could be “immunized” from developing these thoughts due to deficits in connecting with others socially (Verona & Javdani, 2011). Douglas et al. (2006) examined psychopathy and suicide risk in a large sample (N = 1,711) and found that impulsive/behavioral features of psychopathy were correlated with increased suicide risk, but no effect was found for affective/interpersonal features of psychopathy. This lends credence to the theory that certain traits of psychopathy provide immunity and decreased risk for suicide, whereas other types of psychopathic features and offending behaviors may indicate increased risk. This argument also implies that nonviolent offenders may be higher on components of the IPTS, such as perceived burdensomeness or thwarted belongingness. Consistent with this hypothesis, prior research (Kennedy et al., 2011) has found that one’s sense of inadequacy is one of the strongest predictors differentiating nonviolent offenders from violent offenders in that nonviolent offenders have a greater sense of inadequacy. Notably, for our sample, this suggests that these externalizing factors are protective against suicide risk in corrections, as they somewhat inure the violent offender to the critical social drivers of suicide.

We also examined the influence of age and found a significant effect, such that suicide in correctional institutions tends to be a problem affecting younger individuals. This effect was maintained across every model, including the most restrictive model, showing a consistent reduction of risk of suicide death, about 10% for every year of age. In our final trimmed model, this effect was also found, but the rate of risk reduction was closer to 4%. While this is consistent with past data, which has found a disproportionately elevated risk of suicide among juveniles interred in adult correctional facilities (Daniel, 2006), it is notable that outside of custody, suicide risk spikes during middle age (CDC, 2023). Thus, there may be some combination of developmental and environmental factors within custody that lead to increased rates of suicide deaths for younger individuals. Further, interaction effects supported that the age at which someone enters custody affects risk of death by suicide differently, depending on the type of custody; young individuals in jail are at a period of increased risk for death by suicide, and their levels of risk tend to decrease with age. This effect is even more pronounced for young persons in prisons. Several possible explanations emerge in combination with our other results. First, many first-time offenders in jail are incarcerated due to drug offenses, in which early adolescent use has been shown to predict incarceration in jails four-fold compared to older adult use. Thus, a large number of nonviolent offenders in jail are there due to drug and alcohol-related charges, which have been shown to be an acute risk factor for suicidal behavior and death by suicide among adolescent and young adult samples (Pompili et al., 2009; Slade et al., 2008). Other possibilities involve both individual and developmental factors relevant to younger persons or systems-level factors relevant to forensic contexts. Individual-level factors may include deficits in future-oriented thinking (Steinberg et al., 2009), higher levels of sensitivity to social disruptions like those involved in incarceration (Schriber & Guyer, 2016), or higher pain tolerance (El Tumi et al., 2017). Systems-level factors may include impaired coping skills for incarceration-associated distress (Brown & Ireland, 2006) or difficulties in adjustment to incarceration (McShane & Williams, 1989). There may be multiple contributing factors for why those who are at an especially increased risk of suicide are entering jails young and dying by suicide within a short amount of time.

For those individuals who go on to enter the prison system, the period after they are booked may be a critical point where a young individual’s risk of death by suicide is the highest and eventually decreases with age. These results do contrast with established demographic studies, which have shown a rate of death by suicide in prisons by age to be more comparable with the general public, meaning that rates of death by suicide are higher in middle age (Zhong et al., 2021). One possible explanation may relate to behavioral and cognitive impulsivity, particularly salient to adolescents who have not finished developing the neurological networks shown to regulate inhibition and cognitive control (e.g., frontal cortex, ventromedial prefrontal cortex, etc.), which has been shown to correlate with suicidal behavior in adolescents (Liu et al., 2017; Romer, 2010). Another potential explanation for the increased rate of death by suicide by young persons in prisons could be related to environmental factors; research has suggested that younger individuals in prison report higher levels of isolation, violent assaults, and peer conflict (Kenny et al., 2008). Young people may have a harder time integrating into the prison system compared to older persons, and these difficulties may contribute toward higher rates of suicidality. Further research is needed to investigate specific factors and mechanisms contributing to elevated risk of suicide for young persons at all stages of custody.

Another unexpected finding related to the significant influence of race/ethnicity on likelihood of suicide death. For many models, White individuals emerged as a particularly high-risk group for deaths by suicide. When taken with other findings from this analysis, this profile of incarcerated individuals at the greatest risk of suicide matches the prototypical jail suicide decedent characterized in Hayes’ (1983) seminal work almost 40 years ago. It described the average decedent as a young, single, White male with no past history of offenses, currently being held on nonviolent charges who would die by suicide within 24 h. This implies a startling lack of progress made in changing the narrative of this decedent despite 40 years of evolving suicide research and changing custodial policy. However, within our final model, the group with the highest risk for death by suicide were Latinx individuals, who have been sorely neglected in available literature. Past research has shown that Latinx individuals are disproportionately represented in U.S. custody, and further, one in four incarcerated Latinx adolescents are held in adult correctional facilities, and age has been shown to predict suicidal behavior in Latinx prison samples (Arya et al., 2009; Pérez-Ramírez et al., 2021). Our results were consistent with these findings, such that young Latinx individuals are at increased risk of death by suicide in corrections. Additional research is greatly needed on the unique experiences of Latinx individuals in custody to elucidate their risk factors for suicide death.

Our next hypothesis was that the more recently an incarcerated individual had been arrested, the higher their risk of death by suicide. Our findings largely did not support this hypothesis. We did not find support that the days in custody had an effect on likelihood of suicide death when compared to natural causes; in fact, when comparing main effects, this variable was the only variable to not exhibit a significant effect on the likelihood of suicide death in pre-interaction effect blocks. It may be that, outside of the initial first days of custody, which has been established as a critical period for suicide in prior research (Hayes, 2012), the total number of days in custody may not influence the suicide risk once this period has passed.

There are also limitations to our findings for the initial stage of police custody, the prebooking stage. Within prebooking, along with arrests and detainments, there were mental health and welfare checks wherein the police would be called to check on someone in acute mental distress; these checks sometimes ended with the person in question dying by suicide and thus being logged as a suicide death within custody. This complicates criminogenic analysis as these deaths do not fit within the model of criminal offending, as these individuals had not been formally charged with any crime. Even more complicated is that Texas classifies “suicide-by-cop” deaths, in which an individual intentionally provokes police into an act of homicide, as a homicide death, meaning that some homicide deaths within prebooking may have been misclassified. Due to the initial alarming 50-fold increase in the rates of suicide deaths in prebooking custody under the model with main effects, further research should cover this period in greater scrutiny to examine how to best assure suicide prevention strategies for arrests and detainment, especially before mental health and welfare checks. Additionally, studies should conduct a deeper examination of how suicide deaths are recorded and reported by police.

This model has many potential areas of growth and limitations to consider when interpreting these results. First, our categories for race/ethnicity do not capture the diverse range of Black, Indigenous, or People of Color (BIPOC) identities entering our legal system. It will be important for future research to examine relative risks for suicide death by racial and ethnic identities. This research could inform efforts at social equity in the face of disproportionate policing of those with marginalized identities. Essentially, if we acknowledge that we have disproportionate rates of BIPOC identities within the correctional system, we should also ensure that we are equitable in efforts to normalize these suicide rates, including through research attention. Second, though we included sex within the results, the vastly disproportionate number of males in our study does mean that factors relevant to death by suicide in females may have been covered up due to the greater magnitude and pull in predominately male suicide deaths. As discussed in the results, in the first block, females were at higher risk of death by suicide, but the larger sample of male deaths by suicide may have impaired the ability to garner important information on death by suicide in the smaller female sample once additional variables were introduced. Third, in distilling our categories for death down to natural deaths and suicides, we may have missed factors that influence other causes of death within our sample, such as those influencing death by homicide or accident. Finally, our categories distilled charges down to simple violent crimes or nonviolent crimes. The reality is that this may be homogenizing what are actually quite different subtypes of criminal activity. People charged with violent sex crimes likely differ in suicide risk profile when compared to those charged with assault. Among individuals charged with nonviolent crimes, this divergence is especially apparent, as nonviolent crimes cross the spectrum from excessive parking infractions, malicious mischief, fraud, burglary, drug crimes, immigration holds, or a litany of other possible charges. Thus, future research should seek to break these broad charge classifications into more comprehensive and dynamic categorizations to study how an individual’s suicide risk within custody may be largely influenced by the reasons why they find themselves there in the first place.

In addition, a limitation of this research is the inability to integrate information about incarcerated individuals after they leave prison custody. Therefore, a natural extension of this research would be to highlight the important period of post-custody, which includes probation, parole, and recent release. This often-ignored period of readjustment represents another stage of custodial risk for suicide; a report from Britain found a 15-fold increase in rates of death by suicide by incarcerated individuals reintegrating into society after incarceration when compared to the general populace (Grierson, 2019) and a Norwegian cohort study found suicide risk doubled on the first day of release from prison (Bukten & Stavseth, 2021). This period of readjustment is likely to also be reflective of the American justice system, as epitomized in the suicide death of Kalief Browder 2 years after his release from Rikers, where he was held for 3 years awaiting trial, with long periods of time in segregation. This notable event brought calls for criminal justice reform and support from President Barack Obama (2016) in a column for the Washington Post.

Additionally, there are subgroups in need of more research on their distinctive suicide risks, such as death row inmates, those in federal or military custody, individuals in immigration detention, and a rapidly increasing population of female incarcerated individuals. Future research should consider factors influencing risk of suicide among different population groups to examine how these results generalize and what informed steps can be taken to improve care for these individuals in custody.

Understanding how criminological, custodial, and dispositional factors signal risk before suicidal behaviors are endorsed or enacted provides an important opportunity for prevention and increased support. Having easy-to-understand and easy-to-disseminate data can improve the quality of care provided by correctional staff and law enforcement. This is not to suggest that restrictive measures be placed upon those who arrive with elevated risk factors, but rather, these results could be used to improve screening and initiate additional clinical intervention. Additionally, given the paucity of research on incarcerated individuals who died by suicide, and with much of the seminal research often more than 20 years old, this study seeks to add to the growing literature and ultimately improve our understanding and care for those most at risk for death by suicide.

Footnotes

Data availability statement

This study was registered with OSF.io (), where de-identified data are available to the public.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interests with respect to the authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research and/or authorship of this article.

ORCID iD

Rocky B. Marks

Author Biographies

Rocky B. Marks, MS is a fifth-year doctoral candidate at Seattle Pacific University. His research largely focuses on issues of forensic suicidality, including factors contributing to effective suicide prevention in forensic settings. He also has an interest in neurobiological theories of suicidality. He is currently completing his predoctoral psychology internship at the Utah State Hospital on the Forensic track.

Nicole Moreira, MS is a fourth-year doctoral candidate at Seattle Pacific University. Her research focuses primarily on biopsychosocial factors contributing to psychopathy, recidivism, violent offending, and suicide risk in forensic populations. She is also a trained Designated Crisis Responder with experience serving forensic and outpatient populations.

Katherine O’Connell, MS is a 2021 Tillman Scholar and fourth-year doctoral candidate at Seattle Pacific University. She is a former counterterrorism analyst and army veteran of Operation Enduring Freedom and Operation Iraqi Freedom. Her research is largely focused on the intersectionality of trauma, biological basis, and suicidality in military and first responder populations.

Keyne C. Law, PhD is an Assistant Professor of Clinical Psychology at Seattle Pacific University. Her research focuses on studying mechanisms, processes, and systems that contribute to suicidal behavior. Specifically, she is interested in using self-report, behavioral, biological, and ecological approaches to examine out how a person becomes motivated and capable to act on suicidal thoughts.

References

Abramson

L. Y.

Alloy

L. B.

Hogan

M. E.

Whitehouse

W. G.

Gibb

B. E.

Hankin

B. L.

Cornette

M. M.

(2002). The hopelessness theory of suicidality. In Joiner

Rudd

M.D.

(Eds.), Suicide science: Expanding the boundaries (pp. 17–32). Springer.

Anno

B. J.

(1985). Patterns of suicide in the Texas Department of Corrections 1980–1985. Journal of Prison & Jail Health, 5(2), 82–93.

Arya

Augarten

Villaneuva

Villarruel

(2009). America’s invisible children: Latino youth and the failure of justice. Advance online publication. https://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.1892966

Brown

S. L.

Ireland

C. A.

(2006). Coping style and distress in newly incarcerated male adolescents. Journal of Adolescent Health, 38(6), 656–661.

Bukten

Stavseth

M. R.

(2021). Suicide in prison and after release: A 17-year national cohort study. European Journal of Epidemiology, 36, 1075–1083. https://doi.org/10.1007/s10654-021-00782-0

Burke

T. A.

Ammerman

B. A.

Knorr

A. C.

Alloy

L. B.

McCloskey

M. S.

(2018). Measuring acquired capability for suicide within an ideation-to-action framework. Psychology of Violence, 8(2), 277–286. https://doi.org/10.1037/vio0000090

Center for Disease Control. (2023). Suicide data and statistics. https://www.cdc.gov/suicide/suicide-data-statistics.html

Daniel

A. E.

(2006). Preventing suicide in prison: A collaborative responsibility of administrative, custodial, and clinical staff. Journal of the American Academy of Psychiatry and the Law Online, 34(2), 165.

Daniel

A. E.

Fleming

(2006). Suicides in a state correctional system, 1992–2002: A review. Journal of Correctional Health Care, 12(1), 24–35.

10.

Death Penalty Information Center (2023). The Death Penalty in 2022: Year End Report. https://deathpenaltyinfo.org/facts-and-research/dpic-reports/dpic-year-end-reports/the-death-penalty-in-2022-year-end-report

11.

Dixon

K. J.

Ertl

A. M.

Leavitt

R. A.

Sheats

K. J.

Fowler

K. A.

Jack

S. P. D.

(2020). Suicides among incarcerated persons in 18 U.S. States: Findings from the National Violent Death Reporting System, 2003–2014. Journal of Correctional Health Care, 26(3), 279–291. https://doi.org/10.1177/1078345820939512

12.

Douglas

K. S.

Herbozo

Poythress

N. G.

Belfrage

Edens

J. F.

(2006). Psychopathy and suicide: A multisample investigation. Psychological Services, 3(2), 97–116. https://doi.org/10.1037/1541-1559.3.2.97

13.

Durkheim

(1897). Suicide. Free Press.

14.

Dye

M. H.

(2011). The gender paradox in prison suicide rates. Women & Criminal Justice, 21(4), 290–307. https://doi.org/10.1080/08974454.2011.609400

15.

El Tumi

Johnson

M. I.

Dantas

P. B. F.

Maynard

M. J.

Tashani

O. A

. (2017). Age-related changes in pain sensitivity in healthy humans: A systematic review with meta-analysis. European Journal of Pain, 21(6), 955–964.

16.

Favril

O’Connor

R. C.

(2021). Distinguishing prisoners who think about suicide from those who attempt suicide. Psychological Medicine, 51(2), 228–235. https://doi.org/10.1017/S0033291719003118

17.

Felthous

A. R.

(2009). Introduction to this issue: Correctional mental health care. Behavioral Sciences & the Law, 27(5), 655–659. https://doi.org/10.1002/bsl.900

18.

Grierson

(2019). Freed prisoners killing themselves at a rate of one every two days. The Guardian. https://www.theguardian.com/society/2019/nov/18/freed-prisoners-killing-themselves-at-a-rate-of-one-every-two-days

19.

Hayes

L. M.

(1983). And darkness closes in . . . A national study of jail suicides. Criminal Justice and Behavior, 10(4), 461–484. https://doi.org/10.1177/0093854883010004005

20.

Hayes

L. M.

(2012). National study of jail suicide: 20 years later. Journal of Correctional Health Care, 18(3), 233–245. https://doi.org/10.1177/1078345812445457

21.

Joiner

T. E.

(2005). Why people die by suicide. Harvard University Press.

22.

Kaba

Lewis

Glowa-Kollisch

Hadler

Lee

Alper

Selling

MacDonald

Solimo

Parsons

Venters

(2014). Solitary confinement and risk of self-harm among jail incarcerated individuals. American Journal of Public Health, 104(3), 442–447. https://doi.org/10.2105/AJPH.2013.301742

23.

Kennedy

T. D.

Burnett

K. F.

Edmonds

W. A.

(2011). Intellectual, behavioral, and personality correlates of violent vs. non-violent juvenile offenders. Aggressive Behavior, 37(4), 315–325.

24.

Kenny

D. T.

Lennings

C. J.

Munn

O. A.

(2008). Risk factors for self-harm and suicide in incarcerated young offenders: Implications for policy and practice. Journal of Forensic Psychology Practice, 8(4), 358–382. https://doi.org/10.1080/15228930802199317

25.

Klonsky

E. D.

May

A. M

. (2014). Differentiating suicide attempters from suicide ideators: A critical frontier for suicidology research. Suicide and Life-Threatening Behavior, 44(1), 1–5. https://doi.org/10.1111/sltb.12068

26.

Kuanliang

Sorensen

J. R.

Cunningham

M. D.

(2008). Juvenile inmates in an adult prison system: Rates of disciplinary misconduct and violence. Criminal Justice and Behavior, 35(9), 1186–1201. https://doi.org/10.1177/0093854808322744

27.

Laughland

Mathieu-Leger

(2016). The human toll of the America’s public defender crisis. The Guardian. https://www.theguardian.com/us-news/2016/sep/07/public-defender-us-criminal-justice-system

28.

Liu

R. T.

Trout

Z. M.

Hernandez

E. M.

Cheek

S. M.

Gerlus

(2017). A behavioral and cognitive neuroscience perspective on impulsivity, suicide, and non-suicidal self-injury: Meta-analysis and recommendations for future research. Neuroscience & Biobehavioral Reviews, 83, 440–450. https://doi.org/10.1016/j.neubiorev.2017.09.019

29.

Mandracchia

Sylvara

Wikowsky

Tieman

Smith

(2018). An exploratory analysis of violent offending and the acquired capability for suicide in male prisoners. International Journal of Offender Therapy and Comparative Criminology, 62(11), 3436–3446. https://doi.org/10.1177/0306624X17740558

30.

McCoy

Fremouw

McNeil

D. W.

(2010). Thresholds and tolerance of physical pain among young adults who self-injure. Pain Research and Management, 15(6), 371–377. https://doi.org/10.1155/2010/326507

31.

McCullumsmith

C. B.

Clark

C. B.

Perkins

Fife

Cropsey

K. L.

(2013). Gender and racial differences for suicide attempters and ideators in a high-risk community corrections population. Crisis: The Journal of Crisis Intervention and Suicide Prevention, 34(1), 50–62. https://doi.org/10.1027/0227-5910/a000160

32.

McShane

M. D.

Williams III

F. P

. (1989). The prison adjustment of juvenile offenders. Crime & Delinquency, 35(2), 254–269.

33.

Miller

J. D.

Rausher

Hyatt

C. S.

Maples

Zeichner

(2014). Examining the relations among pain tolerance, psychopathic traits, and violent and nonviolent antisocial behavior. Journal of Abnormal Psychology, 123(1), 205–213. https://doi.org/10.1037/a0035072

34.

Mumola

C. J.

(2005). Suicide and homicide in state prisons and local jails. U.S. Department of Justice.

35.

Obama

(2016). Why we must rethink solitary confinement. The Washington Post. https://www.washingtonpost.com/opinions/barack-obama-why-we-must-rethink-solitary-confinement/2016/01/25/29a361f2-c384-11e5-8965-0607e0e265ce_story.html

36.

Office of Justice Programs (2021). Suicide in local jails and state and federal prisons, 2000–2019—Statistical tables. https://bjs.ojp.gov/press-release/suicide-local-jails-and-state-and-federal-prisons-2000-2019-statistical-tables

37.

Pérez-Ramírez

Barthelemy

J. J.

Gearing

R. E.

Olson

Giraldo-Santiago

Torres

L. R.

(2021). Suicide in Mexican prisons: Mental health symptomology and risk factors for suicidal ideation and suicidal behavior. The Prison Journal, 101(6), 675–698. https://doi.org/10.1177/00328855211060319

38.

Polaschek

D. L. L.

Calvert

S. W.

Gannon

T. A.

(2009). Linking violent thinking. Journal of Interpersonal Violence, 24(1), 75–96. https://doi.org/10.1177/0886260508315781

39.

Pompili

Lester

Innamorati

Del Casale

Girardi

Ferracuti

Tatarelli

(2009). Preventing suicide in jails and prisons: Suggestions from experience with psychiatric inpatients. Journal of Forensic Sciences, 54(5), 1155–1162.

40.

Prison Policy Initiative. (2021). Texas profile. https://www.prisonpolicy.org/profiles/TX.html

41.

Rockett

I. R.

Caine

E. D.

Connery

H. S.

D’Onofrio

Gunnell

D. J.

Miller

T. R.

Jia

(2018). Discerning suicide in drug intoxication deaths: Paucity and primacy of suicide notes and psychiatric history. PLoS one, 13(1), e0190200.

42.

Romer

(2010). Adolescent risk taking, impulsivity, and brain development: Implications for prevention. Developmental Psychobiology, 52(3), 263–276. https://doi.org/10.1002/dev.20442

43.

San Segundo

M. S.

Ferrer-Cascales

Bellido

J. H.

Bravo

M. P.

Oltra-Cucarella

Kennedy

H. G

. (2018). Prediction of violence, suicide behaviors and suicide ideation in a sample of institutionalized offenders with schizophrenia and other psychosis. Frontiers in Psychology, 9, 238288. https://doi.org/10.3389/fpsyg.2018.01385

44.

Sarchiapone

Carli

Giannantonio

M. D.

Roy

(2009). Risk factors for attempting suicide in prisoners. Suicide and Life-Threatening Behavior, 39(3), 343–350.

45.

Schriber

R. A.

Guyer

A. E.

(2016). Adolescent neurobiological susceptibility to social context. Developmental Cognitive Neuroscience, 19, 1–18.

46.

Slade

E. P.

Stuart

E. A.

Salkever

D. S.

Karakus

Green

K. M.

Ialongo

(2008). Impacts of age of onset of substance use disorders on risk of adult incarceration among disadvantaged urban youth: A propensity score matching approach. Drug and Alcohol Dependence, 95(1), 1–13. https://doi.org/10.1016/j.drugalcdep.2007.11.019

47.

Smith

P. N.

Cukrowicz

K. C.

Poindexter

E. K.

Hobson

Cohen

L. M.

(2010). The acquired capability for suicide: A comparison of suicide attempters, suicide ideators, and non-suicidal controls. Depression and Anxiety, 27(9), 871–877. https://doi.org/10.1002/da.20701

48.

Steinberg

Graham

O’brien

Woolard

Cauffman

Banich

(2009). Age differences in future orientation and delay discounting. Child Development, 80(1), 28–44.

49.

Stinson

J. D.

Quinn

M. A.

Menditto

A. A.

LeMay

C. C.

(2021). Adverse childhood experiences and the onset of aggression and criminality in a forensic inpatient sample. International Journal of Forensic Mental Health, 20(4), 374–385. https://doi.org/10.1080/14999013.2021.1895375

50.

Stoliker

B. E.

(2018). Attempted suicide: A multilevel examination of inmate characteristics and prison context. Criminal Justice and Behavior, 45(5), 589–611. https://doi.org/10.1177/0093854818754609

51.

Stoliker

B. E.

Galli

P. M.

(2021). Prevalence and correlates of suicidal ideation and attempt according to prisoners’ race: An exploratory analysis. Race and Social Problems, 13(4), 292–305. https://doi.org/10.1007/s12552-020-09310-3

52.

Swogger

M. T.

Conner

K. R.

Meldrum

S. C.

Caine

E. D.

(2009). Dimensions of psychopathy in relation to suicidal and self-injurious behavior. Journal of Personality Disorders, 23(2), 201–210. https://doi.org/10.1521/pedi.2009.23.2.201

53.

Texas Justice Initiative (2021). Death in custody report. https://texasjusticeinitiative.org/datasets/custodial-deaths

54.

Thompson

M. P.

Kingree

J. B.

Lamis

(2019). Associations of adverse childhood experiences and suicidal behaviors in adulthood in a U.S. nationally representative sample. Child: Care, Health and Development, 45(1), 121–128. https://doi.org/10.1111/cch.12617

55.

van Orden

K. A.

Witte

T. K.

Cukrowicz

K. C.

Braithwaite

S. R.

Selby

E. A.

Joiner

T. E

. (2010). The interpersonal theory of suicide. Psychological Review, 117(2), 575–600. https://doi.org/10.1037/a0018697

56.

Verona

Javdani

(2011). Dimensions of adolescent psychopathology and relationships to suicide risk indicators. Journal of Youth and Adolescence, 40(8), 958–971. https://doi.org/10.1007/s10964-011-9630-1

57.

Webb

R. T.

Qin

Stevens

Shaw

Appleby

Mortensen

P. B.

(2013). National study of suicide method in violent criminal offenders. Journal of Affective Disorders, 150(2), 237–244. https://doi.org/10.1016/j.jad.2013.04.001

58.

Webb

R. T.

Shaw

Stevens

Mortensen

P. B.

Appleby

Qin

(2012). Suicide risk among violent and sexual criminal offenders. Journal of Interpersonal Violence, 27(17), 3405–3424. https://doi.org/10.1177/0886260512445387

59.

Zhang

(2016). Model building strategy for logistic regression: Purposeful selection. Annals of Translational Medicine, 4(6), 111. https://doi.org/10.21037/atm.2016.02.15

60.

Zhong

Senior

Perry

Hawton

Shaw

Fazel

(2021). Risk factors for suicide in prisons: A systematic review and meta-analysis. The Lancet Public Health, 6(3), e164–e174. https://doi.org/10.1016/S2468-2667(20)30233-4