Predictors of Treatment Noncompletion in a Sample of Inpatient Sex Offenders With Serious Mental Illness

Treatment noncompletion remains a recurring concern for providers of sex offender treatment. The empirical literature has repeatedly demonstrated that offenders who do not complete treatment, whether through dropout, program-initiated removal, or outright refusal, are at higher risk of sexual reoffending (1–3). Rates of sex offender treatment noncompletion range from 15% to 86% (4), depending on the specific subpopulation of offenders, method of treatment delivery, and setting in which treatment is delivered.

Various factors have been identified that predict treatment noncompletion among community and incarcerated sex offenders. Among juvenile offenders, significant factors include youthful age (5), impulsivity (5,6), prior criminal conviction (6), characteristics of criminality and delinquency (6), denial and lack of accountability (6,7), self-harm and emotional disorders (6), exposure to early sexual victimization (6), sexual maladjustment (7), and prior experiences with therapy (6). Among adult offenders, factors have been categorized as characteristic of historical risk, criminogenic need, or responsivity to treatment, as identified through multiple studies as well as meta-analyses (3). Historical risk variables that predict treatment noncompletion have included a history of violent offending (3,8,9); juvenile delinquency and adult involvement with the criminal justice and correctional systems (8,10–12); noncontact sexual offenses (3,8); sexual assaults against adult (3,11) and male (3) victims and victims the offender does not know (13); lengthier time before release (13); and a history of sexual victimization (10). Demographic variables, such as less education (10), unemployment (8), and divorce (14), have also predicted treatment noncompletion among community and incarcerated sex offenders. Predictors reflective of criminogenic need and responsivity that differentiate completers from noncompleters include a diagnosis of antisocial personality disorder or characteristics of psychopathy (9,12,15–17), sexually deviant interests and attitudes supportive of offending (3), hostility (18), denial and lack of insight (3,10,11), low motivation in treatment (3,11), deterioration or poor treatment progress (8), and behavioral problems while engaged in treatment (8,11).

Such factors have been identified in samples of incarcerated sex offenders and those in community and specialized residential settings. However, characteristics predicting treatment noncompletion by sex offenders with serious mental illness, or who receive treatment while residing in inpatient psychiatric facilities, have not been systematically reported in the literature. Sex offenders with serious mental illness who require inpatient or otherwise secure care have different needs, risks, and treatment responsivity than typical community or correctional sex offender samples (19). Such clients often require concurrent mental health treatment and psychotropic medication, case management, social skills and vocational training, and more extensive aftercare support. Also, they may have criminal backgrounds characterized by psychiatric hospitalization rather than incarceration, as well as varying rates of violence in community versus residential placements (19). Because of their increased psychiatric and behavioral needs, it is likely that their response to sex offender treatment and reasons for failing to complete treatment may differ from what has been reported in the literature. Predictors of treatment noncompletion among violent offenders with mental disorders include many of the same factors, such as antisocial personality disorder and psychopathy, although some additional factors that may be more relevant for sex offenders with serious mental illness have been identified as well, including incidents of aggression within six months prior to treatment (20).

The study aimed to evaluate predictors of sex offender treatment noncompletion in a sample of forensic psychiatric inpatients referred for community and residential sexual offenses. The primary goal was to facilitate the identification of sex offenders with serious mental illness who are at risk of treatment noncompletion, with a secondary goal of comparing relevant risk factors with those identified within the literature for other sex offender samples.

Methods

Participants

In this study, treatment noncompletion was evaluated in a sample of 156 adult male sex offenders residing in an inpatient forensic psychiatric hospital in the midwestern United States. Participants were part of a larger study evaluating sex offender treatment effectiveness (21) within the institution, and all had been deemed eligible for sex offender treatment by their individual treatment teams and facility administration as a result of a history of illegal sexual behavior in community or residential settings. Participants were expected to be a part of sex offender treatment throughout their stay, and all who were eligible for treatment were included in the study. Available sex offender treatment programming used Safe Offender Strategies (22), a treatment involving a focus on developing adaptive self-regulatory strategies that has shown promise in reducing problematic sexual and impulsive behaviors, particularly for persons with major mental illness or intellectual impairments (21,23). Groups were offered weekly for participants, with an expected duration of multiple years of treatment. Clients were assigned to groups on the basis of matched intellectual functioning and psychiatric symptom presentation, as reported by their treatment teams. Participants were offered supplemental treatment programming in accordance with facility policy, which included social skills programming, specific psychosocial interventions, and individual therapy. Successful completion of treatment programming increased participant likelihood of release and transfer, although no specific penalties were given for refusing to attend groups.

Participants were included in the analysis regardless of whether they ever participated in treatment during the study period. Participants were followed for two years on a rolling basis, with the start date being the date identified for inclusion by their individual treatment teams. Treatment completion was defined as remaining a part of sex offender treatment programming for the two-year duration of data collection, or until the point at which clients were deemed appropriate for transfer. Dates ranged from November 2007 to July 2012. Data collection was approved by the institutional review board of the University of Missouri–Columbia. Because all data were archival, no participants were contacted or interviewed as part of this study, nor was their treatment dependent on their participation in this or any other research at the facility.

Participant characteristics.

Participants ranged from 21 to 75 years of age, with a mean±SD of 42.2±12.8 years. They were somewhat ethnically diverse, with 55% (N=85) Caucasian, 40% (N=62) African American, 3% (N=4) Hispanic, 2% (N=3) self-identified as of mixed racial heritage, and 1% (N=2) who declined to affiliate themselves with a particular racial or ethnic heritage. Most participants had never been married (N=126, 81%), whereas the remainder were divorced (N=28, 18%) or currently married (N=2, 1%). Most participants had no known children (N=121, 78%).

At the time of this study, most participants were admitted under civil commitment (N=83, 53%), although these commitments often reflected initial forensic admission as incompetent to proceed that was later commuted to civil commitment by the court. The other admission types were not guilty by reason of insanity (N=45, 29%), renewable court commitments (N=2, 1%), or admissions as incompetent to proceed (N=26, 17%). Because reasons for admission with regard to offense history and admitting legal status did not vary, both civil and criminal commitments were collapsed across the sample. The average length of admission at the time of analysis was 10.5±6.7 years and ranged from one to 45 years.

Diagnostic and criminal histories.

Participants were characterized by a wide range of psychiatric disorders and comorbid psychiatric conditions. Diagnoses were coded from their most recent psychiatric annual evaluations at the time they were selected for inclusion in sex offender treatment. Diagnostic labels were categorized into groups that described major classifications of mental disorders, as listed in the DSM-IV-TR (24). This was intended to minimize diagnostic variability across clinicians and to allow for diagnostic comorbidity.

Data describing participants’ offense histories were obtained from annual social service assessments and admission reports. These reflected both incidents of formal contact with the criminal justice system (arrests) as well as documented incidents of aggression or violence that had occurred in residential or psychiatric placements. Participants’ scores on the Static-99R (25) were derived from archival client medical records. Clinicians at the facility routinely used this instrument, a static actuarial measure of future sex offense recidivism postrelease, in making conditional release or transfer decisions. Estimated risk scores were available for all 156 participants and reflected estimates of risk at the time of either release or two years of treatment completion. (For those who had refused to participate in treatment, time of treatment completion was two years from the time they had been referred to treatment.) Static-99R scores ranged from −3 to 9, with a mean of 2.72±1.98. (Possible scores range from −3 to 12, with higher scores indicating a greater risk of recidivism.) Overall, 41 participants (26%) scored within the low-risk range on this instrument. Sixty-three participants (40%) were classified in the low- to moderate-risk range. Of those remaining, 40 participants (26%) were at moderate to high risk and 12 participants (8%) were at high risk of recidivism.

Analysis

Specific predictive variables of interest were derived from the empirical literature (3) and covered factors relevant to sex offenders with serious mental illness (19), which included participant age, number of arrests, number of violent and sex offense arrests, length of admission, level of security at the start of treatment, aggression during the six-month pretreatment period, Static-99R risk level, and psychiatric diagnosis. Two forms of regression were used to analyze the data. The first analysis used linear regression (enter method) to determine the effect of these variables on percentage of groups attended over two years. The second analysis evaluated the effect of these same variables through logistic regression (enter method), with participants identified as completers or noncompleters according to whether they had attended at least 50% of available sex offender treatment groups during the two-year period. The 50% cutoff was equivalent to the cutoff the facility used as a determinant of client progress and qualification for additional privileges or assignment to client stages or levels.Two forms of regression analysis were used to account for possible differences in prediction between continuous percentage of groups attended and the standard cutoff measure used by the facility.

Results

Major classifications of disorder and incidence of diagnosis are indicated in Table 1. Most participants met criteria for multiple psychiatric disorders, with two-thirds (N=104, 67%) having two or more axis I disorders, and 53% (N=82) having both axis I and axis II disorders. Such rates are consistent with prior research involving sex offenders with serious mental illness and intellectual disability (19). A majority of participants (N=121, 78%) had at least one prior psychiatric hospitalization, and a mean±SD age of 21.1±10.5 years (range 5–62 years) at first psychiatric hospitalization.

Nearly all participants had been arrested on at least one occasion (N=146, 94%), with a mean of 5.2±6.4 arrests per person. Their arrests were for violent, nonsexual offenses (N=70, 45%, 1.8±1.0 arrests), sexual offenses (N=117, 75%, 2.6±1.3 arrests), property offenses (N=71, 46%, 2.7±3.1 arrests), and drug offenses (N=32, 21%, 2.4±1.6 arrests). Records also noted a history of physical, nonsexual aggression for 89% of participants (N=139), much of which may not have resulted in arrest or other contact with the criminal justice system.

Of the 117 (75%) participants who had been formally charged with a sexual offense, 59% (N=69) had been convicted or found not guilty by reason of insanity for a sexual offense, with an average of 1.5 convictions each. An additional 25% of the sample had never been formally charged with a sexual offense, although they had demonstrated difficulties with controlling sexual behavior as noted by family members or self-report. Table 2 shows problematic and illegal sexual behavior for both groups. Some differences were observed between groups, typically reflective of a greater likelihood of arrest associated with more severe or violent sexual behaviors. As is evident, offenders often had problematic or illegal sexual behaviors in both community and residential settings. According to self-report and other official data, participants were 24±11 years of age (range 8–62) at the time of their first sexual offense, which often involved either a sexual act against a child (N=82, 53%) or an aggressive sexual act against a nonconsenting adult (N=39, 25%). Almost equal numbers of participants had sexually offended against adults (N=108, 69%) and children (N=100, 64%), with 52 participants reporting both adult and child victims. In addition, 37% (N=57) were incest offenders.

Clients attended 70%±35% of available sex offender treatment groups during the two-year treatment period under observation (range 0%−100%). It was notable that 52% (N=81) of participants had attended 85% or more sex offender treatment groups, with 17% (N=27) refusing or excluded from all sex offender treatment groups during that time. Table 3 provides a summary of the relationships between the proposed predictive variables and percentage of groups attended.

Linear regression analysis produced a significant model (R²=.376, F=4.15, df=17 and 117, p<.001). Significant predictors of greater percentage of sex offender treatment groups attended in this model included fewer arrests, less physical aggression in the six months before treatment referral, and shorter length of admission, with infrequent attenders more likely diagnosed as having an intellectual or a developmental disorder or disability or a cognitive disorder or disability and borderline personality disorder (Table 4).

Logistic regression analysis examined factors affecting group attendance versus nonattendance, defined as a threshold of 50%. Using this threshold, 113 participants (72%) attended more than 50% of offered groups, whereas 43 participants (28%) attended 50% or less of groups offered. This analysis revealed a significant prediction model (χ²=51.98, df=17, p<.001; Nagelkerke R²=.424) that predicted 81% of cases. Significant predictors of noncompletion in this model included higher number of arrests, increased incidents of physical aggression in the six months before treatment referral, longer length of admission, diagnosis of a psychotic spectrum disorder, and lack of a diagnosis of cognitive or intellectual impairment (Table 5).

Discussion and Conclusions

Prior research has established a number of variables that significantly predict treatment noncompletion among samples of adult and juvenile sex offenders. Variables include those often associated with higher risk of reoffending, including historical variables, such as age (5) and number and type of arrests (6,8–13), victim characteristics, and sexual deviance (3), and responsivity variables reflective of disengagement in treatment (3,8,10,11), continued behavioral problems (8,11), and the like. However, these factors have been identified in samples of offenders primarily in correctional or community treatment, whereas the literature relating to sex offenders with serious mental illness suggests that other variables could be relevant in determining treatment completion and other outcomes (19,20).

In this study, linear and logistic regression both produced significant models that predicted treatment noncompletion in a sample of 156 sex offenders with serious mental illness residing within a secure psychiatric facility. Consistent with research describing general sex offender risk (3,25), some of the factors most predictive of treatment noncompletion (that is, number of arrests and degree of recent aggressive behavior) reflected overall criminality or risk, although a direct measure of risk was not immediately significant (20). The presence of additional predictors suggests that sex offender risk or associated predictors are an insufficient means of predicting treatment compliance in this population of psychiatric sex offenders.

The significant relationships between noncompletion of treatment groups and additional factors perhaps reflected the severity of the participants’ psychiatric and behavioral disorders, which presumably affected their ability and willingness to consistently participate in sex offender treatment programming. This addresses the important question raised by previous research as to whether or not psychiatric variables are also important with regard to sex offenders with serious mental illness (19,20). Diagnoses of borderline personality disorder, intellectual/developmental or cognitive disorders or disabilities, and psychotic spectrum disorders are commonly seen among inpatient offenders receiving psychiatric treatment and are believed to affect treatment willingness and progress. Such findings carry implications for treatment planning in psychiatric settings and for predicting whether persons are at risk of not completing sex offender treatment. Significant diagnostic indicators identified in this analysis are associated with behavioral inconsistency, impulsivity, and difficulties with problem solving in the moment. Such characteristics may contribute to higher rates of treatment noncompletion over time. Similarly, such diagnoses suggest interpersonal skills deficits that could limit a client’s ability to appropriately engage in treatment and maintain meaningful therapeutic relationships. Thus it would be important for treatment providers working with similar populations to increase efforts to engage such clients in treatment and consistently monitor severity and presentation of psychiatric symptoms throughout the duration of treatment to minimize treatment noncompletion.

An important limitation of the analyses, however, is that they represent the treatment experience of a single sample of sex offenders in one psychiatric facility during a two-year treatment period. Thus results may not generalize to all samples of sex offenders with serious mental illness, and additional research is clearly needed. Another limitation is the veracity of diagnostic information, which can vary depending on diagnosing clinician and available symptom evidence. Finally, this study was archival and did not incorporate interactive measures of treatment engagement or commitment to change, which could supplement our knowledge of mechanisms of treatment completion among sex offenders with serious mental illness, or measures of therapeutic alliance, which have been implicated in treatment attrition in similar samples (11,20). Even so, this study provides valuable insight into the factors related to treatment noncompletion in a psychiatric inpatient sample of sex offenders, potentially resulting in more efficient management approaches and potentially greater treatment compliance for this population. This and future research may improve clinical practice by investigating means of addressing factors related to both criminality and psychopathology among psychiatric inpatients who are sex offenders.