Abstract
Suicidality and self-harm behaviors among sex offenders remain underreported in the clinical literature and are often misunderstood in this complex population. The present study aims to identify rates of suicide attempts and self-injurious behaviors in a sample of 1,184 psychiatric inpatients, 462 of whom are sexual offenders. Between-group comparisons revealed significant differences in history of suicide attempts and self-harm behaviors, with sexual offenders evidencing greater rates of both. Significant psychiatric correlates of suicide attempts and self-harm behaviors among sex offenders varied by group and included a variety of psychiatric symptom presentations. These are compared with the general literature on suicide risk and the sex offender population. Implications for treatment of these behaviors in a sex offender population are discussed.
Suicidality and self-harm behaviors are of considerable interest to those who work in mental health or related human services professions. Implementing effective prevention strategies for those at risk for suicide or self-injury, developing treatment interventions, and managing risk of suicide in vulnerable groups are all predicated on the assumptions that we know who is most at risk and that we have clearly identified risk factors in specific populations.
Research regarding suicide, suicide attempts, and self-harm behaviors is lacking in the sex offender literature. Information about suicidality among those with histories of sexual offending primarily originates from either raw epidemiological data describing statistics on completed suicides, or data describing suicide rates among jail detainees or very specific groups of sexual offenders (e.g., those with offenses against children). For example, in two studies, men who had committed only sexual offenses—meaning that they had not committed other types of violent offenses—were 15 to 24 times more likely to have committed suicide than same-aged males in the community (Pritchard & Bagley, 2001; Pritchard & King, 2005). In a similar study, men who had committed sexual offenses against children were 40 times more likely to commit suicide than same-aged, nonoffending males in the general population (Pritchard & King, 2004). Furthermore, these men were found to be at significantly elevated risk of suicide beyond that of men with mental disorders or histories of sexual victimization (Pritchard & King, 2004). Similar epidemiological research by Brophy (2003) suggested that while the suicide risk for Irish males during a 10-year period was more than 1 in 5,000, the suicide risk was calculated at approximately 1 in 1,600 for sexual offenders with adult victims and 1 in 64 for sexual offenders with child victims. Finally, recent research by Webb et al. (2012) using epidemiological samples of Danish males and females evidenced significantly elevated suicide completion rates among men who had committed sexual offenses, though less elevated than the rates of suicide among nonsexual violent males in the sample.
Research using psychiatric samples of sex offenders is limited to one empirical study, with findings inconsistent with these epidemiological results. This study focused primarily on completed suicides among forensic psychiatric patients and revealed that mentally disordered sex offenders were significantly underrepresented among patients who had committed suicide (Haynes & Marques, 1984). However, it was noted that the majority of suicide completers in the study were psychotic and fewer of the sex offenders were diagnosed with psychotic disorders in that particular sample. This finding suggests the need for additional research examining the role of psychosis in relation to risk for suicide among sexual offenders.
Thus, although there is preliminary epidemiological data from several countries describing suicide completion among incarcerated or psychiatrically hospitalized sexual offenders, little is known regarding the likelihood of suicide attempts, the prevalence of self-injurious behavior, or important risk factors that could help identify those likely to engage in these behaviors. From a prevention and risk management standpoint, we must know this information to effectively treat such problems and prevent the occurrence of completed suicides, suicide attempts, and self-harm behaviors among the sex offender population.
On review of what is known about suicidality and self-harm in other populations, certain characteristics or risk factors stand out as relevant for sex offender groups. From epidemiological research, established risk factors for suicide include a number of demographic, psychiatric, and situational characteristics like being male (Centers for Disease Control and Prevention, 2009), having a diagnosis of depression or a psychotic spectrum disorder (e.g., Brown, Beck, Steer, & Grisham, 2000; Caldwell & Gottesman, 1990; Conwell et al., 1996; Pritchard & King, 2004; Roy, 1982), problems with alcohol or substance abuse (e.g., Conwell et al., 1996; Murphy, Wetzel, Robins, & McEvoy, 1992), history of family maltreatment (e.g., Briere & Runtz, 1986; Dube et al., 2001; Pritchard & King, 2004), recent loss or isolation from support network (e.g., Joiner, 2005), and impulsivity or aggressive behaviors (Centers for Disease Control and Prevention, 2009; Douglas, Herbozo, Poythress, Belfrage, & Edens, 2006). A history of suicidal ideation or suicide attempts is also a strong predictor of future suicide risk (Brown et al., 2000; Roy, 1982), as is recent contact with the criminal justice system (e.g., Byrne, Lurigio, & Pimentel, 2009).
Risk factors for self-injurious or self-harm behavior (i.e., behavior that is self-injurious in nature but not precipitated by a desire to die, as opposed to suicide attempts or completed suicides precipitated by an intention to end one’s life) are similar in some ways, though there are differences. Although persons with mental illness, specifically depression, or psychotic disorders are at greater risk, individuals with borderline personality disorder, other personality disorders, or problems with emotion regulation may also be at increased risk (Gratz, 2003; Linehan, 1993). Those with significant histories of trauma and maltreatment, substance use problems, and increased impulsivity are also at greater risk of engaging in self-harm behavior (e.g., Gratz, 2003; Gratz, Conrad, & Roemer, 2002; Wagner & Linehan, 1994). A review of the literature reveals limited study of self-harm or self-injurious behavior, though some recent research highlights these risk variables in the context of correctional populations, in which there is elevated risk of such behavior (Dixon-Gordon, Harrison, & Roesch, 2012).
Many of these risk factors may be disproportionately aggregated in sex offender samples. Persons with a history of sexual offending are often characterized by varying symptoms of depression and psychosis, personality disorders, problems with substance abuse, impulsivity and aggressiveness, a history of trauma and related sequelae, involvement with the criminal justice system, and a sense of hopelessness, stigma, and social isolation (e.g., Abracen, Looman, & Anderson, 2000; Ahlmeyer, Kleinsasser, Stoner, & Retzlaff, 2003; Davison & Taylor, 2001; Dunseith et al., 2004; Harsch, Bergk, Steinert, Keller, & Jockusch, 2006; Hartwell, 2004; Hoertel, Le Strat, Schuster, & Limosin, 2012; Jespersen, Lalumiere, & Seto, 2009; Kraanen & Emmelkamp, 2011; Marshall, 1989, 1993; Stinson & Becker, 2011). A recent review of suicide attempts among incarcerated sexual offenders found that many of these factors were significantly elevated in those men who had attempted suicide, in comparison with those who had not (Jeglic, Spada, & Mercado, 2013). This compounding effect may make sexual offenders a population particularly vulnerable to risk of suicidality or self-harm. We do not yet fully understand the prevalence of suicide attempts or suicidality and self-harm behaviors in this population, nor have we sufficiently identified risk factors that can aid in differentiating those most at risk in an already-high-risk sample of individuals.
The purpose of this study was to examine historical suicide attempts and self-harm behaviors in forensic psychiatric inpatients further, including a large proportion of sex offenders. Given the relative dearth of information regarding the prevalence and risk of suicidality and self-harm behaviors among sexual offenders, this study was exploratory in nature. We sought to answer the following questions:
In a psychiatric sample, how do the histories of suicide attempts compare in sex offenders and nonsex offenders? Are there differences in prevalence among sex offenders, according to type of victim, or offense?
In a psychiatric sample, how do the histories of self-harm behaviors compare in sex offenders and nonsex offenders? Are there differences in prevalence among sex offenders, according to type of victim, or offense?
Do psychiatric symptoms or disorders differentiate those who are more or less likely to have a history of suicide attempts and/or self-harm behaviors in groups of psychiatric sex offenders versus psychiatric patients who are not sex offenders?
Can relevant psychiatric symptoms or disorders be used to predict or postdict likelihood of suicide attempts or self-harm behaviors in sex offenders?
Method
Participants
Data were collected from 1,184 male patients at a Midwestern U.S. state forensic hospital. Participants were selected if they were male and had resided at the facility at any point from January 1, 2005 until December 31, 2010. Only males were selected to control for any gender bias in rates of suicide attempts and self-harm behavior, and also due to the low incidence of female clients with sexual offending admitted to the facility. Participants were admitted to the hospital for a variety of reasons, including: (a) involuntary civil commitment due to the potential for imminent danger to self or others (n = 314), (b) direct admissions from jail or the Department of Corrections for acute stabilization of mental health issues (n = 205), (c) for competency restoration treatment following an adjudication of incompetent to proceed to trial (n = 504), (d) for treatment following a finding of Not Guilty by Reason of Mental Disease or Defect (n = 167), (e) for other pretrial evaluation (n = 89), or (f) for treatment following acts of aggression in other state-run facilities in which it was decided that a higher degree of supervision and security was necessary (n = 153). As some participants were admitted multiple times during the selected time period, it was possible for them to have been admitted under different legal statuses (i.e., 1,432 admissions for 1,182 clients). The majority of admissions were for competency restoration (42.6%), though an almost equal number were admitted to the facility at some point following a suicide attempt or act of self-harm while in custody at another facility.
Participants resided in any of the three security levels of the hospital (i.e., maximum, intermediate, or minimum). The mean age of the sample was 40.4 years (SD = 13.3), and the average length of stay totaled 3.5 years (SD = 5.3), though this included persons still admitted to the facility at the time of data collection. The group was ethnically diverse, with participants of Caucasian (60%), African American (37%), Hispanic (2%), Asian, Native American, Arabic, and mixed descent (the latter four groups totaling 1%; see Table 2).
Procedure
This study examined available archival data on all 1,184 participants, including the following variables: total length of stay, history of sexually inappropriate behavior, psychiatric diagnoses, positive history of self-harm, and incidents of suicide attempts. Suicide attempts and self-harm behaviors were both historical (i.e., prehospitalization) and may have also occurred during hospitalization. Behaviors were coded as suicide attempts if the patient or other available medical documentation reported a self-injurious behavior precipitated by specific suicidal intent. For example, a client who was previously hospitalized for an attempt to hang himself while in jail was coded as positive for a history of suicide attempts. Behaviors were coded as self-harm behaviors if the patient or other available medical documentation reported a self-injurious behavior intended to cause harm but not associated with suicidal intent. In this case, a client’s record may have indicated that soon after admission, he broke a compact disc and used it to cut himself on the arms and chest, but upon examination, stated that his intent to die was low. Behaviors both prior to and during hospitalization were included. As nearly as half of the participants were initially hospitalized due to a suicide attempt or self-harm behavior while incarcerated or in custody, records were available describing the nature and severity of these injuries. For others, suicidality and self-harm data were taken from patient self-report to the psychiatrist at the time of admission and in annual evaluations of psychiatric need, and from interviews conducted with members of the client’s family by facility social workers at the time of admission. All data regarding suicide attempts and self-harm were coded as “yes/no” to increase reliability, particularly given the use of self-report. Suicide attempts were additionally coded in a separate column as “none/one/two or more” in an attempt to capture data from persons with multiple suicide attempts. No participants completed a suicide during the course of their stay.
A number of the patients in the sample had been admitted on more than one occasion, and for those patients, their length of stay was an aggregate of the total number of days that they had been admitted to the hospital during the 6-year reporting period. Participants were identified as sexual offenders for the purposes of this study if they had history of a sexual arrest or conviction (including their index offense), or if there was documentation of sexual behavior while hospitalized that would meet the legal standard for a sexual offense. As such, 462 participants were identified as sexual offenders, with the remaining 722 classified as nonsexual offenders. Additional data regarding the approximate ages of their victims were also obtained from facility records.
Participants’ diagnoses were coded from their charts, which included psychiatric admission and discharge assessments completed by facility psychiatrists. Psychiatric diagnoses were dichotomized as yes/no for the presence of a mood disorder, anxiety disorder, psychotic disorder, impulse control disorder, or paraphilic disorder. Data were coded in this way to allow for comorbidity, or to represent multiple symptom clusters where appropriate (e.g., schizoaffective disorder resulted in a “yes” for both mood and psychotic disorder diagnoses). Additionally, this coding scheme allowed for reliability despite diagnostic differences between treating clinicians (e.g., a diagnosis of paranoid schizophrenia vs. schizoaffective disorder vs. psychotic disorder not otherwise specified were all coded as “yes” for psychotic disorder, so as to minimally indicate the detected presence of psychotic symptoms). Axis II data were coded in a similar manner. Participants were identified as having a cognitive disorder (e.g., mental retardation or a developmental disability), antisocial personality disorder, borderline personality disorder, other Cluster B personality disorder, or other personality disorder. Antisocial and borderline personality disorders were chosen specifically for independent coding due to their high association with impulsive or self-harm behaviors. All data were coded from the available records by one clinician to ensure that data were reliably obtained from the same sections of the medical record, and that a standard definition could be applied to “suicide attempt” and “self-harm behavior.” However, an additional subset of 50 cases were randomly selected from the larger data set to ensure reliability of coding constructs. The coder was a graduate student who was blind to the results of the study and did not have access to the primary coder’s data. This individual was given instructions as outlined above on how to code the data. Alpha coefficients for these data ranged from 0.696 (ratings of adult victims) to 0.980 (diagnosis of a cognitive disorder). Collectively, sex offense-related variables produced an average reliability coefficient of 0.834, suicide attempts and self-harm variables had an average coefficient of 0.905, and psychiatric diagnostic data had an average coefficient of 0.865. These are described further in Table 1.
Interrater Reliability Across Coded Variables.
Note that “other Cluster B personality disorder” was excluded, as no individuals in the subsample of 50 participants were found to have this in their records by either coder.
Multivariate Analyses
All analyses were conducted using SPSS versions 17 and 20. Chi-square analyses were used to determine differences in rates of suicide attempts and self-harm behaviors between those with and without a history of sexual offending. Similar comparisons were made by dividing the sexual offenders by victim type: Child only (victim is below the age of 18), adult only, and both adults and children. A series of logistic regressions were conducted to identify the variables most correlated with a history of suicide attempts in both offenders with and without a history of sexual offenses. Length of admission and age were entered together in the first block as covariates for all regression analyses. The following variables were entered simultaneously in the second block as predictors: an Axis I diagnosis of a mood, anxiety, impulse control, or psychotic disorder, an Axis II diagnosis of borderline, antisocial, other Cluster B or any other personality disorder, or a cognitive disorder. Initially, the full regression used sex offender status as a predictor in block two, and in subsequent models, the group was split into either sex offenders or nonsex offenders, using victim type to further differentiate significant correlates in the sex offender group.
Results
Suicide Attempts and Self-Harm: Comparing Sex Offenders and Nonsex Offenders
The participants with a history of sexual offending were, on average, older (M = 41.8, SD = 13.5) and had a longer length of admission (M = 4.6 years, SD = 6.0) than participants with no known history of sexual offending (age: M = 39.5, SD = 13.1; length of admission: M = 2.7 years, SD = 4.7). Of the total sample of 1,184 participants, 416 had a history of one or more suicide attempts, as self-reported upon admission, observed while hospitalized, or verified through independent sources. A chi-square analysis revealed a small, though significant, difference in history of suicide attempts between the sex and nonsex offenders (38.5% vs. 33.0%; χ2 = 3.83, p = .05, Cramer’s V = .06). There were no significant differences between groups in the number of suicide attempts made, (F(1, 1,182) = .004, p > .05).
With regards to self-harm behaviors, results revealed a significant difference (χ2 = 13.20, p < .01, Cramer’s V = .11), with 27.1% of sexual offenders and 18.1% of nonsex offenders demonstrating a self-reported, observed, or otherwise documented history of such behaviors. When examining those participants with a history of both suicide attempts and self-harm behaviors, another significant chi-square was revealed (χ2 = 6.21, p = .01. Cramer’s V = .07). Again, a higher percentage of sex offenders (14.9%) than nonsex offenders (10.1%) had a history of both behaviors.
Suicide Attempts and Self-Harm: Comparing Sex Offenders by Victim Type
To further examine the nature of the relationship between sex offender status and suicide/self-harm behaviors, the sample was divided into four groups by victim type: nonoffenders (n = 765); including noncontact sexual offenders), adult only offenders (n = 159), child only offenders (n = 194), and individual who offended against both children and adults (n = 66). For the purposes of these analyses, noncontact sexual offenders were included in the nonoffender sample as their victims could not be readily identified for classification into the other three groups (n = 43). Additionally, nearly all of these behaviors had occurred during the course of hospitalization, further differentiating them from the other sex offenders in the sample. There were no significant differences between subgroups of sex offenders with regards to history of suicide attempts. However, significant differences between groups were observed when considering a history of self-harm behavior (χ2 = 20.83, p < .001, Cramer’s V = .13). Follow-up analyses revealed a difference between self-harm among nonsex offenders (18.4%) and adult only offenders (29.6%; χ2 = 10.10, p < .01, Φ =.104) as well as between nonsex offenders (18.4%) and offenders against children and adults (37.8%; χ2 = 14.36, p < .001, Φ = .13). A significant difference was revealed between offenders against children and adults (37.8%) and child only offenders (2.2%; χ2 = 6.30, p < .05, Φ = .16).
Finally, those with a history of both suicide attempts and self-harm behaviors were examined. The overall chi-square between the four groups was significant (χ2 = 6.21; p < .05, Cramer’s V = .10), and a significant difference was observed between nonsex offenders (10.0%) and adult only offenders (18.9%; χ2 = 9.96, p < .01, Φ = .10). No other significant differences were observed between these groups. See Figure 1 for cross-group comparisons.
Suicide attempts and self-harm frequency, percentage across groups.
Multivariate Analyses: Psychiatric Characteristics of Nonsex Offenders, Sex Offenders, and Sex Offenders by Group
As the current sample was drawn from a population of forensic inpatients in a psychiatric facility, all participants met diagnostic criteria for at least one disorder. Diagnostic comorbidity was high. Significant diagnostic differences were observed between some of the groups, as is noted in Table 2. Overall, persons with a history of sexual offending had significantly higher diagnostic rates of impulse control disorders, paraphilias (no nonsex offenders met diagnostic criteria for a paraphilia), Cluster B personality disorders, and cognitive disorder diagnoses (e.g., mental retardation, pervasive developmental disorders, dementia), whereas nonsex offenders had significantly higher rates of psychotic spectrum diagnoses. Mood and anxiety disorder diagnoses, as well as other, non-Cluster B personality disorder diagnoses, did not significantly differ. Between sex offender groups, there were significant diagnostic differences between those who offended against adults only, children only, and adults and children both. These were highly varied and are described further in Table 2.
Demographics and Chi-Square Comparisons of Psychiatric Diagnoses by Group.
p < .05. **p < .01. ***p < .001.
Tukey post hoc results demonstrate significant differences in age between nonsex offenders and sex offenders with both adult and child victims.
Tukey post hoc results demonstrate a significant difference in age between individuals with child victims and individuals with both adult and child victims.
Tukey post hoc results indicate a significant difference in length of admission between nonsex offenders and those with adult victims and those with both adult and child victims.
Tukey post hoc results indicate a significant difference between sex offenders with adult victims and sex offenders with only child victims and with adult and child victims.
Tukey post hoc analyses revealed a difference between those with child only victims and sex offenders with child and adult victims.
ns = not significant; PD = personality disorder.
Multivariate Analyses: Psychiatric Correlates of Suicide and Self-Harm in Nonsex Offenders and Sex Offenders
To further examine these groups, all the variables outlined above were entered into a logistic regression analysis, in Block 2 to determine which predicted the outcome behavior (suicide, self-harm, or both). For all the analyses described in this section, age and length of admission were entered in Block 1 as covariates. In examining suicide attempts in the sample of 1,184 psychiatric inpatients as a whole, a significant model emerged (χ2 = 91.62; Negelkerke’s R2 = .12, p < .01; see Table 3). Significant predictors included a history of sexual offending, diagnosis of a mood disorder, diagnosis of an anxiety disorder, diagnosis of borderline personality disorder, and diagnosis of any Cluster B personality disorder. A significant model was also identified for self-harm among the whole sample (χ2 = 132.76, Negelkerke’s R2 = .23, p < .01; see Table 3), with similar predictors: a history of sexual offending, diagnosis of a mood disorder, diagnosis of an impulse control disorder and diagnosis of borderline personality disorder. Finally, a significant model was identified for the sample for persons who had engaged in both suicidal and self-harm behaviors (χ2 = 87.26, R2 = .17, p < .01; see Table 3). The significant predictors included a history of sexual offending, diagnosis of a mood disorder, and diagnosis of borderline personality disorder.
Significant Predictors of Suicidality and Self-Harm Among the Whole Sample.
p < .05. **p < .01.
ns = not significant.
The sample was then split into nonsex offenders and sex offenders to examine if predictors differed for the groups, especially given that being a sex offender alone was such a strong predictor of suicidality and self-harm. Interestingly, for nonsex offenders, significant models were identified for history of suicide attempts (χ2 = 64.75, R2 = .14, p < .01), self-harm behaviors (χ2 = 66.39, R2 = .20, p < .01), and both suicide attempts and self-harm behaviors (χ2 = 78.93, R2 = .22, p < .01), with virtually the same predictors for each (see Table 4).
Significant Predictors of Suicidality and Self-Harm Among Nonsex Offenders.
p < .05. **p < .01.
ns = not significant.
However, in the case of those with a history of sexual offending, while significant models were found for all three types of behavior, the predictors were more varied (see Table 5). Among the sex offenders, significant predictors of suicide attempts (χ2 = 39.58, R2 = .13, p < .01) included a diagnosis of a mood disorder, diagnosis of an anxiety disorder, diagnosis of a psychotic disorder, antisocial personality disorder, borderline personality disorder, and diagnosis of another Cluster B personality disorder. With regards to self-harm behaviors among the sex offender group (χ2 = 57.72, R2 = .25, p < .01), fewer predictors were identified, though a significant model still emerged. Significant predictors included diagnosis of a mood disorder, diagnosis of an impulse control disorder, and borderline personality disorder. For sex offenders with histories of both suicide attempts and self-harm (χ2 = 47.17, R2 = .17, p < .01), only two predictors were noted: diagnosis of an anxiety disorder and borderline personality disorder.
Significant Predictors of Suicidality and Self-Harm Among Sex Offenders.
p < .05. **p < .01.
ns = not significant.
Multivariate Analyses: Psychiatric Correlates of Suicide and Self-Harm in Sex Offenders by Group
In the results described above, it is evident that being a sex offender has a strong relationship to history of suicide attempts and self-harm, and that among sex offenders, a diverse set of factors predict suicidality and self-harm behaviors. Thus, sex offenders were further divided into categories according to victim type to better understand these relationships (see Table 6 for side-by-side comparisons). For all models in this section, age and length of admission were entered in Block 1. For suicide attempts, significant models were found for adult only offenders (χ2 = 20.17, R2 = .20, p < .05), child only offenders (χ2 = 21.89, R2 = .20, p < .05), and offenders against both adults and children (χ2 = 24. 73, R2 = .43, p < .05). The only significant predictor of suicide attempts among adult only offenders was borderline personality disorder, whereas there were no significant predictors or correlates for child only offenders, despite the significant overall model. For those with both adult and child victims, suicide attempts were significantly related to diagnosis of a psychotic disorder and antisocial personality disorder.
Significant Correlates of Suicide Attempts and Self-Harm Behaviors Among Sex Offenders, by Victim Choice.
p < .05. **p < .01.
For self-harm behaviors, significant models were again identified for each group: adult only offenders (χ2 = 25.44, R2 = .23, p < .01), child only offenders (χ2 = 29.29, R2 = .38, p < .01), and offenders against both adults and children (χ2 = 33.84, R2 = .55, p < .01). Significant psychiatric correlates of self-harm behaviors for adult only offenders included diagnosis of a mood disorder and borderline personality disorder. For child only offenders, the only significant factor was diagnosis of an impulse control disorder. Finally, for offenders against both adults and children, the single significant predictor was diagnosis of a psychotic disorder.
For those in the sample who had engaged in both suicide attempts and self-harm behaviors, predictive models were again significant for adult only offenders (χ2 = 28.38, R2 = .21, p < .01), child only offenders (χ2 = 23.24, R2 = .38, p < .01), and offenders against both children and adults (χ2 = 21.95, R2 = .48, p < .05). As was the case with self-harm behaviors, different predictors emerged for each of the three sex offender groups. Among adult only offenders, borderline personality disorder was predictive of both suicide attempts and self-harm behaviors. For child only offenders, diagnosis of an anxiety disorder was predictive. Lastly, for adult and child offenders, only antisocial personality disorder emerged as a significant predictor in the analysis.
Discussion
The current study focused on histories of suicide attempts and self-harm behaviors in a large psychiatric sample that included persons with and without a history of sexual offending. The primary goals of the analysis were to establish rates of suicidal and self-harm behaviors in the sample, as compared with the current literature on suicidality among sex offender populations, and to identify psychiatric correlates that could be used to differentiate risk among varied members of this population. As was noted earlier in this article, the literature is fairly limited in this area, and what little there is relies on epidemiological research on completed suicides among child sexual abusers. Noted risk factors like psychiatric disorders, history of prior suicide attempts, and isolation—while significant predictors of suicide in a general population—may be more common in sex offender or psychiatric samples, and are therefore less discriminatory from the perspective of risk management and suicide/self-harm intervention.
The literature has been somewhat divided in terms of increased versus decreased likelihood of completed suicide among sex offenders. As noted initially, a series of studies using epidemiological data found that sex offenders against children were significantly more likely to have committed suicide than a comparison sample of community males (Brophy, 2003; Pritchard & Bagley, 2001; Pritchard & King, 2004, 2005), though these were studies of correctional rather than forensic patients. In comparison, one study of suicide among forensic psychiatric patients revealed that sexual offenders were disproportionately less represented among completed suicides (Haynes & Marques, 1984). In the current study, the 462 male clients identified as sexual offenders were significantly more likely to have a history of suicide attempts than the remaining 722 males who had no known history of sexual offending. However, when comparing those sex offenders with a history of suicide attempts based on victims, there were no significant differences between groups. Furthermore, in the analysis of predictive correlates, it was found that simply being a sex offender was predictive of the reported or documented occurrence of at least one suicide attempt. Thus, we can conclude that in a psychiatric sample such as this one, it is possible that sexual offenders are more likely to have evidenced a history of suicidality and suicide attempts in comparison with other psychiatric patients. This is in contrast with the findings of Haynes and Marques (1984), in that sex offenders in this sample of forensic inpatients were more likely to have made suicide attempts than nonsex offenders. This may reflect real differences in the two samples with regards to severity and comorbidity of psychiatric diagnosis, types of sexual offenses, or legal admission status. It may also reflect differences between forensic patients who attempt versus who complete suicide. There were no persons who committed suicide among our sample, and this is an area that merits further study.
When sex offenders were further divided into groups according to their victim type (excluding the small number who had committed only noncontact offenses), surprising findings emerged. Although the majority of the sex offender suicide literature points to greater rates of completed suicide among persons with child victims and lower rates among those with adult victims (e.g., Brophy, 2003; Pritchard & Bagley, 2001; Pritchard & King, 2004, 2005), these data reveal no significant differences in the rates of suicide attempts for these two groups. In fact, there is a slight (though nonsignificant) trend suggesting that the sex offender subgroup with the highest likelihood of suicide attempts is the adult only offender group. Furthermore, although the overall rates of suicidality differed between sex offenders and nonsex offenders, those with only child victims evidenced rates of suicide attempts most similar to the nonsex offenders in the sample. It is thus possible that there are important differences in suicidality and suicide attempts among psychiatric sex offenders in comparison with those in correctional samples, as had been examined in previous research, or that there are differences in those who attempt versus those who commit suicide that are being captured in the current analysis.
We were unable to find any published literature related to rates of self-harm behaviors among sexual offenders. Here, we examined the presence or absence of a history of self-harm behavior, finding that significantly more persons in the sex offender sample had engaged in acts of self-harm than the nonsex offending psychiatric comparison group (i.e., 27% vs. 18%). Similarly, more of the participants with a history of sex offending had engaged in both suicide attempts and self-harm behaviors than psychiatric controls (15% vs. 10%). And as was the case for suicide attempts, merely having a history of sexual offending was significantly predictive of self-harm behaviors in the entire sample. Within sex offender subgroups, those most likely to engage in self-harm behaviors were persons with both adult and child victims, followed by those with adult only victims. Again, those with child only victims did not significantly differ from nonsex offenders with regards to self-harm behaviors, despite their higher reported rate of completed suicides in the literature. Although suicide and self-harm are different constructs, we would expect some relationship between them, particularly given the relationship between increased suicide risk and status as a perpetrator of child sexual abuse noted in the previous literature (e.g., Brophy, 2003; Pritchard & Bagley, 2001; Pritchard & King, 2004, 2005). Once again, this may reflect real differences between psychiatric and correctional samples of sexual offenders, or differences between those who commit suicide and those who attempt suicide or engage in nonsuicidal self-injury.
Psychiatric symptoms and disorders are given great prominence in the suicidality literature as risk factors for suicide (Brown et al., 2000; Caldwell & Gottesman, 1990; Conwell et al., 1996; Pritchard & King, 2004; Roy, 1982) and self-harm (Gratz, 2003; Linehan, 1993). Given that our sex offenders in the sample showed higher rates of suicide attempts and self-harm behaviors, one would expect them to additionally show greater universal rates of psychiatric impairment. However, in this sample, this was not the case. Although sex offenders evidenced significantly higher rates of impulse control disorders, cognitive disorders, and all Cluster B personality disorders, the nonsex offenders had significantly higher rates of psychotic spectrum disorders. This is consistent with the findings from earlier research (Haynes & Marques, 1984), which found that sex offenders were less often diagnosed with psychotic spectrum disorders than other forensic psychiatric patients in the sample. Within the sex offender group, diagnostic features differed dramatically according to the nature of the offense type. Given the variance in psychiatric symptomatology, it is not singly the presence or severity of mental disorder that predicts suicidality between these groups, and the occurrence of psychiatric impairment is not evenly distributed among the different sex offender types.
These differences in psychiatric presentation likely impacted the factors used to postdict suicide attempts and self-harm behaviors. Among nonsex offenders, the more traditional predictors of suicidality and self-harm emerged—diagnosis of a mood disorder, borderline personality disorder, and diagnosis of an impulse control disorder. Among sex offenders, the psychiatric correlates were more varied depending on behavior (i.e., suicide attempts, self-harm, or both behaviors) and group (i.e., adult only offenders, child only offenders, or those with both offense types), and other variables of interest gained significance. This suggests a rich area for further study, to help us move beyond superficially linking specific psychiatric correlates with suicidality and self-harm to the development of a more complex algorithm for understanding risk in this population. In a population that may already be labeled “high risk” for suicide or self-harm due to psychiatric impairment, legal involvement, and a host of other factors, knowing that certain psychiatric diagnosis in identified offender groups are associated with greater risks of either suicide attempts or self-harm behavior can assist us in developing strategies for prevention and risk management in psychiatric, correctional, or community placements. Perhaps the strongest predictor of suicide attempts or self-harm behaviors remains a prior history of doing so, but these data give us additional information that can identify those at the highest risk for these behaviors, even among those with such histories and psychiatric problems.
In sum, these findings provide us with needed information about the risk of suicide attempts and self-harm among persons who have a history of sex offending and psychiatric disorders. As is evident from these findings, such persons are more vulnerable to suicide attempts and self-harm behaviors, and certain characteristics of their psychiatric presentation and offense type may differentially predict their suicide attempts or self-harm behaviors. Future research is needed to help us understand these and other contextual characteristics that will allow for more accurate behavioral prediction, prevention of suicide attempts, suicide completion, and self-harm behaviors, and the implementation of intervention approaches to address problems related to client suicide and self-harm.
Limitations
One important limitation of this study is its reliance on self-reported data—both with regards to history of suicide and self-harm behaviors and also psychiatric symptoms characteristic of certain diagnoses. Although many of the self-harm behaviors included in the analysis occurred within an inpatient setting and were therefore more readily verifiable, this was not the case for many of the suicide attempts. However, records noted that in many cases, a suicide attempt had led to at least one prior psychiatric admission, so this was perhaps more reliable. This limitation may also explain why the number of attempts was insignificant in the analyses, as it was easier to formally verify the presence of absence of a suicide history, while the number of attempts was more reliant on individual self-report. Similarly, psychiatric diagnosis is heavily dependent on self-reported symptoms and individual variability among clinicians. Though the authors of this study took steps to minimize this bias whenever possible, it should be again noted that conclusions have been drawn using a combination of source data, including self-reported information.
Another consideration regarding the data is that the authors of this study did not identify the intended “purpose” or potential motive for self-harm or ambiguous suicidal behaviors. In other words, we classified a behavior as self-harm, regardless of whether or not there may have been secondary gain (e.g., attention, transfer, medication, and so forth). This decision was intentional, though future research may want to consider examining the severity of the attempt. Regardless of the reason for a suicide attempt or self-harm behavior, it still represents a risk to the health and safety of the person involved. From an institutional perspective, it still represents a need for identifying persons at risk and taking measures to ensure their safety.
A final limitation is that the sample used here comes from a psychiatric setting, and the rates of suicidality and self-harm behavior may not be representative of all persons who have committed sexual offenses. Oftentimes, what precipitated hospital admission was a suicide attempt, so these persons may be at greater risk of suicidality at baseline and may not accurately represent risk characteristics among persons without similar psychiatric impairment. This potentially higher base rate in the present sample may also be the reason for differences between findings in this sample and prior work, in that other studies have focused on child sexual abusers who completed suicide attempts while awaiting trial or sentencing, whereas this study looked at persons who have attempted suicide or engaged in nonsuicidal self-harm behavior across a range of sex offender subgroups. These groups may not present in the same way, and the psychiatric admission status of those in the current sample may affect this. However, it still provides valuable information about prevalence and potentially predictive characteristics of suicide and self-harm among sex offenders who are in a psychiatric setting or who have serious mental illness, so that groups at risk may be more precisely identified.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.