A Preliminary Analysis of Sexual Recidivism and Predictive Validity of the Static-99R in Men Discharged From State Hospitals Pursuant to California’s Sexually Violent Predator Act

Abstract

We examined the recidivism rates and the predictive validity of the Static-99R in 335 men who were detained or civilly committed and released from California State Hospitals pursuant to the Sexually Violent Predator (SVP) Act, and followed up for approximately 21 years from date of hospital admission. In all, 8.7% were arrested or convicted for a new sexual offense during the total follow-up (N = 335) and 7.8% over a fixed 5-year follow-up (n = 205). The Static-99R demonstrated small in magnitude discrimination for sexual, violent, and general recidivism (area under the curve [AUC]/C = .56 to .63). Calibration analyses, conducted through expected/observed (E/O) index, demonstrated that the Static-99R overpredicted sexual recidivism, irrespective of whether the Routine or High Risk/Need norms were used. Observed recidivism rates were lower than predicted by Static-99R scores and may be the result of the sample’s older age at release, lengthy hospitalization, or other factors.

Keywords

Sexually Violent Predator sex offender civil commitment Static-99R recidivism

A Brief History of Sex Offender Civil Commitment (SOCC)

Twenty states, plus the District of Columbia (i.e., federal government), in the United States have SOCC laws, commonly known as Sexually Violent Predator (SVP) or Sexually Dangerous Person (SDP) statutes. These laws permit post-prison involuntary hospitalization of mentally ill, higher risk persons until they are deemed safe enough for return to the community (for reviews, see Felthous & Ko, 2018; Hoberman & Jackson, 2016). These laws generally share three requirements: (a) conviction of at least one statutorily defined sexually violent offense, (b) a current mental disorder or abnormality linked to sexual offending, and (c) likelihood to engage in future sexually violent acts.

SOCC laws are controversial and are often the subject of ethical, legal, and clinical challenges (Sreenivasan et al., 2010). Critics allege that the true intent of SOCC is to prolong the incarceration of individuals adjudicated for sexual offenses beyond their maximum prison term (Janus, 2006; La Fond, 2005; Morse, 1998). Another criticism is that the deprivation of liberty is thinly legitimized through the provision of treatment to reduce the likelihood of reoffending, although treatment efficacy and amenability are not required. Arguably however, civil commitment program components, such as individualized assessment and treatment, consideration of less restrictive alternatives, annual reevaluations, viable pathways for potential release, all seek to strike a fair balance between an individual’s liberty and the interests of society (Page & Burbank, 2016; Ra, 2007). Despite the controversies, the U.S. Supreme Court has upheld the constitutionality of SOCC laws to apply to the subgroup of sexual offenders whose serious mental disorder makes them likely to sexually reoffend (Kansas v. Crane, 2002; Kansas v. Hendricks, 1997).

Precursors of contemporary civil commitment laws in California, along with 26 other states, existed in the early 1900s and were known as sexual psychopath laws (Lieberman & Siegel, 1957; Rapaport & Lieberman, 1956). These laws were enacted in response to high-profile sex crimes and the perceived belief that certain persons with sexual offense histories could not control their impulses. Consequently, the law considered them too ill for punishment and required placing them in locked institutions for treatment. Criticized for their lack of success and ambiguous commitment criteria, for example, requiring lack of sexual control yet excluding criminally insane as well as psychopathic offenders, sexual psychopath laws eventually fell out of use or were repealed.

In 1944, California replaced its sexual psychopath law with an alternative sentence known as the Mentally Disordered Sex Offender (MDSO) law (Monahan & Davis, 1983; Oliver, 1982; Steadman et al., 1982). Individuals who met the criteria and agreed to treatment were treated at a secure psychiatric hospital. In 1982, the MDSO law was also repealed in the wake of criticisms of treatment efficacy and legal loopholes confounding the effectiveness of its discharge process (D’Orazio et al., 2009). This left no legal basis to civilly commit high-risk offenders in California until the current SVP law was introduced in 1996. Although historical and current civil commitment laws require predisposition for sexual violence resulting from mental abnormality, contemporary SOCC laws are contrasted with precursor laws in three important ways: (a) they are applied after criminal sentences are served, (b) discharge decisions are not made by the institutions but rather by the court, and (c) sexual offense treatment participation is not mandatory.

The California SVP Program

The SVP law represents California’s effort to treat and manage its highest risk persons with predatory sexual offense histories and mental conditions. The commitment process involves prison screening, independent forensic evaluations, a probable cause hearing and a trial, culminating in either civil commitment or release. Within 6 months of the expiration of the individual’s prison sentence, those convicted of a statutorily defined sexually violent offense are screened for possible referral to the Department of State Hospitals (DSH) for full evaluation. Two evaluators contracted or employed by the DSH are appointed to independently conduct evaluations addressing each of the three commitment criteria. The evaluators are not provided the reports or findings of other evaluators assigned to the case nor do they represent the opinion of any entity. If both evaluators find that the person meets the criteria for SVP commitment, the district attorney decides whether to proceed with a petition for civil commitment. If both evaluations find all criteria are not met, the person is released, typically to parole supervision. In cases of an “initial split”, where one evaluator finds commitment criteria have been met but the other does not, the person is evaluated by two additional independent evaluators. If, after the second pair of evaluations, three evaluators agree that the criteria have not been met, or the evaluators “split” again, the individual shall be released from prison. However, if three evaluators agree that the criteria have been met, a referral shall be made to the district attorney for civil commitment.

If the district attorney decides to seek civil commitment, the individual is entitled to a trial where it must be proven beyond a reasonable doubt that the person meets the SVP criteria. A judge or jury may serve as the trier of fact. Civilly committed individuals can be held indefinitely in civil confinement or in a conditional release program, until such time that the court finds that he or she is no longer “likely” to sexually re-offend by nature of a mental disorder. In California, the “likely” standard is defined in People v. Superior Court (Ghilotti, 2002) case law. The likely standard does not have a precise mathematical definition but rather is defined as a substantial, serious, and well-founded risk to reoffend in a sexually violent, predatory manner without appropriate custody and treatment.

Less than 1% of all registered sexual offenders in California are committed as “Sexually Violent Predators” (D’Orazio et al., 2019). From the California SVP law’s inception in 1996 through 2018, approximately 19,000 sexual offenders were evaluated for SVP commitment. Of these, only 12% met SVP criteria, and only 5% (n = 936) were ultimately fully committed as SVPs (D’Orazio et al., 2019). All California SVPs are adult males with the exception of one female. The average length of stay for detainee confinement due to a Probable Cause finding is about 6 years, although 25% have been confined for more than 10 years pending their commitment trial.

California State Hospital Sex Offender Recidivism Studies

Recidivism is the primary outcome of interest considering that the main purpose of SOCC laws is to protect the public through confinement and rehabilitation of dangerous sex offenders. Despite its importance, the extent that civil commitment reduces recidivism is unknown. Few peer-reviewed studies have investigated the recidivism rate and the predictive validity of risk assessment tools with this population.

To the best of our knowledge, two studies have investigated the recidivism rate of individuals hospitalized in California prior to current sex offender civil commitment laws. Frisbie and Dondis (1965) published the first report on recidivism rates of civilly committed individuals (N = 1,921) who were released from state hospitals pursuant to the sexual psychopath law. Approximately, 20% of the sample committed a new sex offense, within 1–6 years following release. Recidivism was mostly likely to occur within the first year (8%) with a gradual decline thereafter through the sixth year (1.4%). Compared with the non-recidivists, recidivists were more likely to have a diagnosis of pedophilia, stay longer in hospital, and none indicated “marital status and fatherhood.” The second study was conducted by Sturgeon and Taylor (1980) on 260 individuals discharged pursuant to the MDSO law. Nearly 30% of the released individuals were convicted of nonsexual crimes, and 15% were convicted of a new sex crime during the 5-year follow-up period. Recidivists who had received a negative recommendation from staff (unamenable to treatment) were twice as likely to reoffend sexually, whereas individuals who completed treatment were convicted of fewer new sex offenses.

While not a civil commitment study, the Sex Offender Treatment and Evaluation Project (Marques et al. 2005), is relevant to this discussion. Marques et al. compared recidivism rates of volunteers who were recruited from state prisons to spend the last two years of their sentence in a state hospital to two untreated control groups. No significant differences in the rates of sexual reoffending over an 8-year follow-up were found among the Treatment Group (22 %), Volunteer Control Group (20%), and a Nonvolunteer Control Group (19 %).

To date, no recidivism research has been published on the California SVP population. The only known information comes from an unpublished report by Jesus Padilla, a forensic evaluator, in court deposition and memo formats (Padilla & Russell, 2004). The sample consisted of 68 civilly committed individuals who were discharged and followed up to six years, with a mean time of 2.39 years. Using conviction rate as the measure of recidivism, 4% were convicted of a new sex offense, 4% for a violent offense, and 28% for any offense.

Similar to California, most states with SOCC laws do not routinely collect recidivism information on their programs. Washington (Washington State Institute for Public Policy, 2007), New York (New York State Office of Mental Health, 2009), New Jersey (Mercado et al., 2013), Virginia (Joint Legislative Audit and Review Commission, 2012), and Texas (Texas Health and Safety Code, 2009) have published government or other non-peer-reviewed reports on recidivism. Florida State has published peer-reviewed data (DeClue & Rice, 2016; Wilson et al., 2012), and a study from Minnesota has used a statistical model to estimate recidivism rates (Duwe, 2013). Recidivism rates from a sample of high-risk individuals screened for civil commitment but ultimately not committed and released without supervision in Texas have been examined as well (Boccaccini et al., 2009). Recidivism rates differed widely from state to state, ranging from 0% (Texas Health and Safety Code, 2009) to 23% (Washington State Institute for Public Policy, 2007). Comparing these findings must be done with caution as operationalization and methodological differences are present in recidivism definition and sources, follow-up time, sample size and characteristics. Ultimately, meaningful recidivism data require routine release of convicted persons to the community; however, the release rates of individuals from SOCC programs are highly variable, ranging from hundreds to just a few releases (Schlank, 2018). Thus, the recidivism rates of the fraction of individuals who are released are unlikely to be representative of the majority of SOCC individuals who remain in custody.

Assessing “Likelihood” of Sexual Recidivism

In conducting SVP civil commitment evaluations, experts employ various risk assessment tools designed to assess risk and estimate the likelihood of sexual recidivism (Helmus & Babchishin, 2017). The Static-99R is the most commonly used and researched sexual violence risk assessment tool (Phenix et al., 2016) and is routinely used in SOCC evaluations as one source of determining the individual’s risk to sexually reoffend (Kelley et al., 2018). Although the reliability and the relative accuracy of Static-99R scores in discriminating recidivists from nonrecidvists (i.e., discrimination or relative risk) have been demonstrated in numerous studies, there is less reliability in the accuracy of the risk estimates for total scores (Helmus, 2009). In short, further research is needed to establish that total Static-99R scores predict recidivism in other jurisdictions at the rates provided by the tool’s norms (i.e., calibration or absolute risk). One recent exception is Boccaccini et al. (2017), who examined the calibration properties of Static-99R in a Texas Department of Criminal Justice sample of 34,687 men convicted of sexual offenses; Static-99R routine norms overestimated rates of sexual recidivism by approximately twofold in this Texas sample as a whole and, specifically, for categorical risk Levels III and IVa. Furthermore, there are no contemporary civil commitment samples represented in the Static-99R norms. Finally, SOCC by definition should be very selective in committing only those who truly pose an exceptional risk. One way this could be determined is if Static-99R scores of the SOCC population are higher than other groups.

The Static-99R has been studied in California and has shown to have good interrater reliability and large in magnitude discrimination. Hanson et al. (2014) found that Static-99 and Static-99R scores for a sample of 476 individuals with sexual offense histories released from the department of corrections and followed for 5 years were strongly related to subsequent sexual recidivism. A subsequent study has found similar patterns for Static-99R predictive validity (Lee et al., 2016) and, with some exception, good calibration with California’s ethnically diverse probation and parolees (Lee & Hanson, 2017). The degree to which these findings from the California community supervision sample apply to California’s SVP population is unknown.

Present Study

Our first research question examines the base rates of recidivism for men who were detained or civilly committed to state hospitals between 1996 and 2017, pursuant to the California SVP Act, and then subsequently discharged to the community. We investigated sexual, violent, and general recidivism rates after community discharge. Given the nature of the sample, we anticipated higher recidivism rates compared with other samples from California. Our second research question extends previous lines of research to evaluate the predictive properties of Static-99R in assessing risk for sexual recidivism. The Static-99R is almost ubiquitous in its application to inform legal determinations of SVP status by the courts. Despite the high-stakes psycho-legal application of the tool, very little is known about its predictive properties when it has been used in the SVP context, particularly for those judged to meet the criteria at some legal threshold (i.e., full commitment vs. probable cause). To this end, we compared the performance of Static-99R with previous validation studies using discrimination and calibration statistical models. We hypothesized that the Static-99R would evidence discrimination and calibration properties commensurate with previous findings (Hanson et al., 2014), when examined in those discharged to the community after being detained or committed pursuant to the SVP law.

Method

Participants

Participants consisted of 335 men who were admitted to California State Hospitals pursuant to California’s SVP Act, scored on the Static-99R, and subsequently discharged to the community between 1996 and 2017. The men were drawn from a larger pool of 1,491 consecutive admissions under the SVP Act during this time frame. All persons were convicted of at least one sexually violent offense against one or more victims and had served a prison sentence prior to being hospitalized. At the time of admission to the hospital, subjects were either civilly committed (46.0%) or classified as detainees, meaning Probable Cause was determined (54.1%) by the California Supreme Court. The mean age at admission was 48.7 years (SD = 10.8), the mean age at discharge was 55.8 years (SD = 11.3), and the mean length of hospitalization was 7.1 years (SD = 4.6). The sample was predominantly of Caucasian ancestry (51.0%), but also included those of African American (32.8%), Hispanic (12.2%), Native American (3.0%), or other ethnic backgrounds (0.9%). Approximately, 30% of the sample had a history of special education, 52.2% had a history of childhood sexual abuse, and 59.9% had a history of head trauma. The study was approved by the State Hospital Research Committee as well as the Committee for the Protection of Human Subjects, which serves as the institutional review board (IRB) for all California state agencies.

Admission and Discharge Rates

Information on admission and discharge dates was obtained through a hospital electronic system used for census tracking. This system contained records of all patients admitted to and discharged from California State Hospitals. The first admission date, January 1, 1996, and the most recent discharge date, December 31, 2017, were used for all patients whose legal status was designated as SVP (i.e., “fully committed”) or as SVP Probable (SVPP) Cause (i.e., “detainee”). SVP and SVPPs both are initially declared to meet the SVP criteria at some legal threshold and admitted to a period of secure hospitalization pursuant to California’s SVP Act before being released. In some instances, SVPP cases are later declared to be SVP (i.e., to meet full criteria) prior to eventual release. In other instances, SVPP cases are released if they do not meet the SVP criteria; the frequency with which this occurs is not known given that we used admission status to designate the patients. It is possible, however, that detainees were more heavily represented in the present sample of discharged cases, insomuch as they would be more likely to be released (e.g., as not having met criteria). Of all 1,491 admissions from 1996 through 2017, 110 died in the hospital, 431 were discharged, and 949 were still hospitalized. The single female case was removed from the sample. Discharge was defined as unconditional (n = 376), conditional (n = 32) release to the community (i.e., through the CONREP), or transfer to jail or prison for a new offense (n = 23). Of those transferred to jail or prison, 10 were released to the community and 13 remained in custody. The final sample consisted of 335 community discharged cases with Static-99R scores and recidivism information.

Procedure

Demographic and clinical information, along with diagnoses, and risk assessment scores were extracted from SVP forensic evaluations, which consisted of precommitment evaluations and updates. As described earlier, these evaluations are conducted by at least two licensed psychologist or psychiatrist evaluators pursuant to the California Welfare and Institutions Code section 6601 to determine whether the person meets the SVP criteria. If the commitment trial does not occur in the first year of detainment after probable cause is found, the evaluations are updated according to statute or court order. The two most recent evaluations proximal to the discharge date were used in data collection.

Recidivism data were collected from the California Department of Justice (DOJ), except in two cases where information was brought to our attention through the media, resulting in the identification of two additional recidivists. Data collection, coding, and entry was done by the first author and regularly discussed with the first three coauthors. Information from evaluations was first entered into a fixed word document or excel coding fields and then opened in the program, Statistical Package for the Social Sciences (SPSS), for analysis. Given the categorical and objective nature of the data being coded, missing data were coded accordingly and not replaced through using other sources or retroactive scoring. In situations when an evaluator’s score was missing from the most recent evaluation, the score was extracted from a prior evaluation when referenced.

Measures

The variables extracted and analyzed for this study from the evaluations included education level, history of special education, childhood sexual abuse, head trauma, Diagnostic and Statistical Manual of Mental Disorders (DSM) diagnoses, and Static-99 and Static-99R scores. Education level, history of special education, childhood sexual abuse, and head trauma were coded using pooled information from two evaluations, whereas diagnostic information and risk instrument scores were selected randomly from one of the two evaluation reports, using an excel randomize function. Ethnicity, date of birth, and hospital admission and discharge dates were extracted through hospital records.

Demographic and clinical characteristics

Education level was coded at an ordinal level, whereas the other variables were coded categorically as yes, no, or uncertain. With few exceptions, we relied exclusively on evaluators’ opinions to confirm history of special education, childhood sexual abuse, and head trauma. Childhood sexual abuse was defined as sexual fondling, touching of sexual organs or being forced to touch another person’s, or sex with an adult before the age of 18 years, especially if a considerable age difference was evident. Head trauma was defined using a broad range of brain injuries, with and without loss of consciousness or memory. These variables were coded as “uncertain” unless they were not unequivocally addressed in the evaluations.

Diagnosis

Diagnoses were coded in the exact format that they were listed in the evaluations, almost always according to the DSM-III-R, DSM-IV, DSM-IV-TR, or DSM-5 criteria. In the present analysis, we combined alcohol and all drug classes into an overarching Drug/Alcohol Use Disorder category. Other specified paraphilic disorders consisted of specifiers that reflected sexual arousal to coercion and arousal to pubescent children.

Static-99/Static-99R

Static-99 and Static-99R are 10-item actuarial risk assessment tools designed to estimate the likelihood of recidivism among male adults with a history of sexual offense arrests and/or convictions (Harris et al., 2003; Helmus et al., 2012; Phenix et al., 2016). The items assess demographics, sexual and nonsexual criminal history, and victim characteristics. Static-99R is identical to Static-99, except for the age item, which expanded from 2 to 4 scoring options, with successively lower risk points assigned for increased aging. The total score for both versions is calculated through summing all the items, resulting in a total score ranging from 0 to 12 for Static-99, and from −3 to 12 for the Static-99R. Higher scores are associated with higher rates of sexual reoffending. Given that our follow-up overlapped with the Static-99 and 99R developmental time line, the sample included subjects with both Static-99 (n = 136) and Static-99R (n = 210) scores. During analysis, any Static-99 scores were converted to Static-99R scores for consistency and generalizability of findings; for all cases, the age item was scored using time of release from the hospital.

Recidivism

Recidivism was operationally defined as any new arrest or conviction in California following hospital discharge to the community. Rehospitalization from conditional release for a small number of cases was counted as a new arrest; however, these were typically for technical violations where a police officer had detained the person for rehospitalization from conditional release (e.g., having incidental contact with a child and not reporting, possession of legal pornography). Conditional release technical violations, resulting in arrest and revocation were counted as general recidivism. Each offense was sorted by its legal title and coded under one of the following three recidivism variables: sexual (e.g., indecent exposure, lewd and lascivious act with a child), violent (e.g., assault, kidnapping to commit burglary, including sexual offenses), or general (e.g., nonviolent offenses such as contempt of court, theft, failure to register, and sexual and violent offenses).

Arrest information for two cases that had come to our attention through anecdotal sources and were confirmed through public court records were also included. The follow-up consisted of the point of institutional discharge to the end of study. We examined total follow-ups and a fixed 5-year period within the community time frame.

Data Analytic Plan

The analyses proceeded in several phases. The first level of analysis consisted of descriptive statistics for race, education level, historical, hospital information, diagnosis, and risk assessment scores. The frequency of sexual, violent, and general recidivism for the total time frame and the fixed 5-year follow-up were computed. The second level of analysis examined discrimination (relative risk) properties of Static-99R scores for sexual, violent, and general community recidivism through receiver operator characteristic (ROC) curve and Cox regression survival analyses. ROC generates an area under the curve (AUC) statistic with values that range from 0 to 1.0, representing the probability that a randomly selected recidivist for a given outcome has a higher score than a randomly selected nonrecidivist; values of .56, .64, and .71 correspond to small, medium, and large effect sizes, respectively (Rice & Harris, 2005). AUCs were conducted for 5-year fixed and unfixed follow-ups for community recidivism. Cox regression is a survival analytic technique that examines predictor associations with binary recidivism over time, adjusting and controlling for individual differences in follow-up time. Cox regression generates a hazard ratio (e^B), representing the proportionate increase in the hazard of an unwanted event (e.g., recidivism) for every one-unit increase in the predictor variable (e.g., Static-99R score), with values above 1.0 representing positive predictor-criterion associations. We also computed time dependent AUC values, Harrell’s C, using the information from Cox regression, which provides an index of predictive accuracy analogous to the AUC as a function of time at risk. Harrell’s C was computed using R 4.0.0 package “Survival” (Therneau, 2020).

Finally, to examine calibration (i.e., the fit between observed and expected values), we computed the expected/observed (E/O) index (Hanson, 2017), a ratio statistic that compares the number of recidivists expected from a comparison or normative sample, divided by the number of recidivists observed in the study sample for a given score (e.g., individual Static-99R scores) or group of scores (e.g., Static-99R risk bands). The normative reference groups used for E/O index analyses (expected rates) were the 2016 Static-99R routine and high risk-need norms (Phenix et al., 2016). Values of 1.0 represent perfect calibration, that is, observed rates associated with risk scores in a study sample align perfectly with those expected from a normative or comparison sample, while values substantially above 1.0 indicate overprediction by the normative sample and values well below 1.0 indicate underprediction. The E/O index is significant when the 95% confidence interval does not overlap with 1.0, computed as follows (Rockhill et al., 2003):

95 % CI of \frac{E}{O} index = (\frac{E}{O}) e^{(+ / - 1.96 \sqrt{\frac{1}{O}})}

Results

Sample Description: Clinical and Risk Characteristics

Table 1 reports demographic, diagnostic, and hospitalization variables for the study sample. In elaboration of the brief description provided in the “Method” section, this is an older, reasonably well-educated sample, although nearly one fifth of the men had a history of special education, over one quarter experienced previous head trauma, and nearly half had reported previous sexual abuse. Consistent with the civil commitment status of the sample, approximately half the men qualified for a diagnosis of pedophilia (the most common paraphilic diagnosis), and more than a third qualified for other specified paraphilia. More than half the sample met the criteria for a substance use disorder, and there was a high base rate of personality disorder (PD), with the most common being antisocial PD. Smaller minorities of the sample had non-substance-related major mental health diagnoses (e.g., mood disorder, psychotic disorders). Table 2 reports Static-99R scores and breakdown according to risk level using the five-tiered scheme. About two thirds of the men scored either Above Average (Level IVa, 34.4%) or Well Above Average (Level IVb, 30.4%) risk. A sizable minority of the sample scored average risk (Level III), whereas very few cases were Below Average (Level II) or Very Low (Level I) risk.

Table 1.

Demographic, Diagnostic, and Hospitalization Variables for Study Sample (N = 335).

Variable	N	%	M	SD
Race
African American	110	32.8	—	—
Caucasian	171	51.0	—	—
Hispanic	41	12.2	—	—
Native American	10	3.0	—	—
Other	3	.9	—	—
Education level
Nursery to eighth grade	13	3.8	—	—
Some high school	57	17.0	—	—
High school Diploma/GED	107	31.9	—	—
Some college	87	26.0	—	—
Vocational/trade School	38	11.3	—	—
College degree	33	9.9	—	—
Historical
Special education	58/189	30.7	—	—
Sexual abuse	151/289	52.2	—	—
Head trauma	109/182	59.9	—	—
Hospital information
Age at admission	—	—	48.7	10.8
Age at discharge	—	—	55.8	11.3
Length of Hospitalization	—	—	7.1	4.6
Diagnostic
Paraphilic disorders
Pedophilic disorder	153	45.7	—	—
Other specified Paraphilic Disorder, Arousal to Coercion	138	41.2	—	—
Exhibitionistic disorder	16	4.8	—	—
Sexual sadism disorder	10	3.0	—	—
Voyeuristic disorder	5	1.5	—	—
Frotteuristic disorder	2	0.6	—	—
Fetishistic disorder	1	0.3	—	—
Substance useDisorders	194	57.9	—	—
Personality disorders
Antisocial PD	128	38.2	—	—
Narcissistic PD	5	1.5	—	—
Other specified PD	57	17	—	—
Mood disorders	25	7.5	—	—
Learning disabilities	9	2.7	—	—
Schizophrenia/other psychotic	4	1.2	—	—
PTSD	2	.6	—	—
Cognitive disorders	3	.9	——	—

Note. PD = personality disorder; PTSD = posttraumatic stress disorder.

Table 2.

Static99-R Descriptive Statistics.

Static-99R risk level	n (%)	M	SD	Range
Very low	3 (0.9)	−2.00	0.0	0
Below average	14 (4.2)	−0.29	0.47	−1 to 0
Average	101 (30.1)	2.43	0.70	1 to 3
Above average	116 (34.6)	4.45	0.50	4 to 5
Well above average	101 (30.1)	6.89	0.97	6 to 10
Overall	335	4.32	2.20	−2 to 10

Recidivism Information

The mean community follow-up post-release was 7.6 years (SD = 5.0) while the mean length of hospitalization, as noted above, was 7.0 years (SD = 4.4). Base rates of community recidivism were comparatively low, particularly for sexual recidivism at 8.7% (29/335) overall and 7.8% (16/205) at 5-year follow-up. Higher observed rates were noted for violent recidivism at 15.8% (53/335) overall and 17.6% (36/205) at 5-year follow-up, and general recidivism at 31.6% (106/335) overall and 37.1% (76/205) at 5-year follow-up.

Predictive Properties of Static-99R

Discrimination

Static-99R had predictive associations with sexual recidivism at small magnitudes (AUC = .62, C = .59) over the total follow-up (Table 3). The Static-99R also evidenced significant small prediction magnitudes for violent and general recidivism. The results were reaffirmed through Cox regression survival analysis; for the hazard ratio magnitudes for every 1-point increase in Static-99R score, there was an approximate 20% increase in the hazard of sexual recidivism, the same (i.e., 20%) increase in the hazard of general violence, and 16% increase in the hazard of general (i.e., any) recidivism. AUCs for Static-99R prediction of fixed 5-year outcomes was smaller than when employing the overall follow-up, and significant only for general recidivism.

Table 3.

Static-99R Discrimination Analyses: ROC and Cox Regression Survival Analyses for Prediction of Sexual, Violent, and General Community Recidivism.

Recidivism criterion	ROC overall		ROC 5-year		Cox regression
Recidivism criterion	AUC	95% CI	AUC	95% CI	C	95% CI	B	SE	p	e ^B	95% CI
Sexual	.62*	[.52, .73]	.56	[.42, .70]	.59	[.48, .69]	.18	.09	.041	1.20	1.01, 1.43
Violent	.63**	[.55, .71]	.56	[.47, .66]	.60	[.52, .67]	.18	.07	.006	1.20	1.05, 1.36
Any	.61**	[.55, .67]	.59*	[.51, .67]	.58	[.53, .64]	.15	.05	.001	1.16	1.06, 1.27

Note. For overall follow-up, including Cox regression, N = 335. For fixed 5-year follow-up, N = 205. ROC = receiver operator characteristic; AUC = area under the curve; CI = confidence interval.

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

Calibration

The calibration properties of Static-99R were examined through computation of E/O indices relative to the 2016 routine and high risk-need normative samples. As seen in Table 4, the E/O index values indicated that the Static-99R 5-year normative sample overpredicted recidivism by two- to threefold for the California SVP sample overall and at specific risk thresholds. The most substantial departure in calibration was in the Level IVb group, in which 5-year sexual recidivism was significantly overpredicted by either the routine or high risk-need norms, followed by the IVa group in which sexual recidivism was significantly overpredicted by the high risk-high need, but not the routine, norms. The relevance by the disparities in calibration is underscored by the fact that nearly two thirds of the sample was Above or Well Above Average risk at the time of discharge, and in principle, the high risk-need norms would have greatest relevance owing to this being a sample that was detained or civilly committed according to the level of sexual dangerousness they posed. Calibration appeared reasonably strong for the Level III group but could not be evaluated for the Level I and II groups, owing to a lack of recidivists.

Table 4.

Calibration Analyses: E/O Index Comparing California Sample With 2016 Static-99R Normative Samples (Routine and High Risk-Need) on Rates of 5-Year Sexual Recidivism.

Static-99R common language metric			Study n	Expected				Observed
Static-99R common language metric				Routine		High risk-need		California SVP		E/O Index (95% CI)
	Risk category	Total score		%	n	%	n	%	n	Routine		High risk-need
I	Very low	−3, −2	3	1.3	0.04	—	—	0	0	—	—	—	—
II	Below average	−1, 0	6	2.8	0.2	7.2	0.4	0	0	—	—	—	—
III	Average	1 to 3	54	6.8	3.7	12.7	6.9	9.3	5	0.74	[0.31, 1.77]	1.38	[0.57, 3.32]
IVa	Above average	4, 5	74	13.4	9.9	19.6	14.5	5.4	4	2.35	[0.88, 6.26]	3.63	[1.36, 9.67]
IVb	Well above average	6 to 12	68	27.4	18.6	30.7	20.9	10.3	7	2.67	[1.27, 5.60]	2.99	[1.43, 6.27]
	Overall		205	17.3	35.5	22.3	45.7	7.8	16	2.22	[1.36, 3.62]	2.86	[1.75, 4.67]

Note. Significant E/O index values in bold font. E/O = expected/observed; SVP = Sexually Violent Predator; CI = confidence interval.

Discussion

This study examined the discrimination and calibration properties of the Static-99R and the release outcomes from all consecutive SVP community discharges in the state of California between 1996 and 2017 who were rated on the tool. As SOCC cases, the sample is preselected and could largely be characterized as Above Average or Well Above Average risk on the Static-99R—owing to the density of sexual and nonsexual criminal history—with a high base rate of risk-relevant diagnoses, including paraphilic disorder (e.g., especially pedophilia), substance use, and antisocial PD. The sample, however, is neither routine nor orthodox high risk/need per the Static-99R norms. One of the better parallels would be the Bridgewater Massachusetts Treatment Center of civilly committed men, preselected for treatment, with a history of violent sexual offenses, both in terms of Static-99/R profile and diagnostic presentation (see Knight & Thornton, 2007). Another unique feature of this sample is that, demographically, this was an older group of men at 55 years of age at release on average from the hospital and had been hospitalized an additional 7 years following the completion of their prison sentences. As a point of contrast, the average age at release was 40 years (SD = 12) across the 24 samples employed to generate Static-99R norms and, hence, more than a full standard deviation younger (Helmus et al., 2012). These sample features likely had implications for the predictive properties of Static-99R and the generalizability to non-released patients and SVP samples in other jurisdictions.

Discrimination and Calibration Properties of Static-99R on a Released SOCC Sample

Static-99R demonstrated broadly small in magnitude discrimination in the prediction of the three recidivism outcomes. Although Static-99R scores significantly discriminated sexual recidivists from non-recidivists, prediction magnitudes were smaller than reported in the meta-analytic literature, both in terms of AUCs and C values (Helmus et al., 2012), as well as the hazard ratios (Hanson et al., 2013). With regard to the latter, the Static-99R normative sample has stronger prediction of sexual recidivism, with a hazard ratio of 1.39 in addition to a higher 5-year base rate. It is not entirely clear what may account for lower levels of discrimination in the present sample. With a mean Static-99R score of 4.3, even with a 12-point range and SD around 2 (i.e., 68% of the sample had scores between 3 and 7), range restriction may have contributed to AUC magnitudes that are somewhat lower than in the published meta-analytic literature. Indeed, Howard (2017) cautions that sample composition and fluctuations in the variability of risk scores can greatly influence AUC magnitudes, sometimes by as much as “10% to 20% over chance between subpopulations” (p. 103). In addition to range restriction, several other factors could impact discrimination magnitude and/or power, such as the low 5-year fixed sample size, the possibility of out-of-state reoffending, the effect of lengthy total detainment duration, treatment, and intensive supervision and monitoring for the committed group but not the detainees to name but a few.

Moreover, the lower base rates of sexual recidivism, in turn, provide a lower starting point for calibration and mean that norms from a reference sample will overpredict the actual rates. In this study, this disparity was most pronounced as risk level increased (i.e., Levels IVa and IVb). This was reflected in the significant E/O index on the recidivism rates of the total sample alone and at the two highest risk levels in which the routine and high risk-need normative samples had two- to threefold higher expected rates of recidivism.

Implications for SVP Evaluators

From the perspective of an SVP evaluator, the lack of calibration of the Static-99R for a released SVP sample highlights the need for further examination of the impact of treatment, protective factors, and other intervening factors, such as lengthy total institutionalization periods and post-release offense suppression due to conditional release and parole supervision conditions. As noted previously, there are many factors that could contribute to lower than expected base rates for a patient population scoring above or well above average on a static tool. One other possibility that has particularly important implications for SVP evaluators is unmeasured dynamic risk factors. These men no doubt changed on risk-relevant propensities (e.g., substance problems, attitudes, emotional regulation, and sexual self-regulation), be it through treatment, aging, or other influences; risk-relevant propensities could also be managed through close and structured supervision, monitoring, and coordinated implementation of release plans (e.g., for work, residence). This study, however, did not include a dynamic measure, given that sexual offense specific dynamic measures had either not yet been developed or they were not regularly implemented at the time of these men’s evaluations.

Nonetheless, these considerations strongly indicate that valid risk assessments must go beyond the use of a static actuarial tool to include dynamic risk factors, that is, the long-term psychological vulnerabilities that drive past sexual offending and predispose future offending. Andrews and Bonta’s (2010) principles of risk (i.e., match treatment intensity to risk level), need (i.e., prioritize dynamic risk factors for intervention), and responsivity (i.e., tailor services to unique client features using cognitive-behavioral therapy methods) underscore the potential utility of including a dynamic measure in sexual violence risk assessments. Research demonstrates that dynamic risk measures, even a single initial administration, can add incremental predictive validity to static measures in the prediction of sexual recidivism (Hanson et al., 2015; Olver & Eher, 2019). Dynamic measures not only capture risk-relevant propensities not fully captured by established static measures, but they can also assess changes in risk from treatment or other change agents, which can strengthen release and reintegration decisions.

Limitations, Conclusion, and Future Directions

This study is a first glance at men who were discharged pursuant to California’s SVP Act and the prediction performance of Static-99R for community recidivism. The sample represents a small group of men from the population of men in California convicted for sexual offenses, who were referred for civil commitment and by virtue of this fact represent a preselected group of individuals. Although this study addresses an underresearched topic and provides important information about the release outcomes and predictive properties of actuarial tools with SOCC samples, there are several limitations with implications for future research.

First, within this preselected group of cases, the study sample consisted of men held under two very different standards of proof - beyond a reasonable doubt (the highest standard of proof) and probable cause (a lower standard of proof). This means individuals were included in the study if they were found to meet SVP criteria under a probable cause threshold but then later released as having not met SVP criteria beyond a reasonable doubt, as well as individuals who at the outset were civilly committed at trial beyond a reasonable doubt. The legal status of each patient was difficult to determine given that those individuals committed before the law changed in 2006 were committed for two-year periods and thus went back and forth between probable cause and full commitment status depending on the results of “re-commitment” evaluations. Contrast this group with others in the sample who were committed for an indeterminate length of time after 2006 when the law was changed. As such, the sample consisted of men with mixed legal status - probable cause and civilly committed - who were either released unconditionally, or released conditionally.

Second, and relatedly, a limitation inherent in any outcome research on an SOCC sample is that the release rates are low, and discharged patients may not represent the large majority of cases who are not released. SVP patients are discharged when they have not fully met, or no longer meet, commitment criteria; discharge also requires in part being assessed as a manageable release risk. Thus, the men who were discharged to the community were likely to have reduced their risk through various means (e.g., treatment, aging), or simply not meeting criteria. For a comparatively lower risk sample of discharged SVP cases, the release outcomes and Static-99R’s predictive properties for the present study sample may not generalize to the nearly three quarters of likely higher risk SVP patients who were not released during the approximate two-decade timeframe.

A third limitation is that we did not have the information on individual cases who were discharged as probable cause (i.e., who never did meet criteria) versus those who were fully committed. It is likely that fully committed cases were at higher risk (at least initially) and, in turn, could have either different release outcomes or that the Static-99R may perform differently with this group. One way to control this would be to conduct prediction analyses that were stratified by commitment status, either in separate groups or as a stratum variable in Cox regression survival analysis. Future research on SOCC samples is advised to code and control for commitment status to provide more accurate estimates of release outcomes and the psychometric properties of structured forensic tools with these populations.

Fourth, there are limitations in the recidivism data. This study, despite efforts made, did not include national recidivism data, and rather the data were constrained to California. Our only out-of-state recidivism data sources were passively received through media sources that were later confirmed. There was also substantial variability in follow-up time, ranging from several months to more than two decades. For these reasons, the current calibration findings cannot be accurately compared with any sample that used national recidivism sources such as the Static-99R norms or the California probation/parole sample. True base rate uncertainty is further compounded, given that most sexual offenses go unreported.

A final limitation involved the substantial variability of treatment circumstances of the patient sample, which will also impact recidivism base rates and generalizability. For one, the study combined treated and untreated (e.g., treatment refusals) patients. Furthermore, the treated group varied considerably in terms of treatment intensity, content, and modality. Specifically, treatment duration may have ranged from a brief period to the entire length of hospitalization. Treatment content may have varied from specialized sexual offense treatment to recreational activities and psychosocial groups for managing anger, coping with stress, overcoming substance abuse, and other mental health conditions. In addition, with respect to modality, this study overlapped with a time frame where the hospital sexual offense treatment program was transitioning from relapse prevention to an integrated good lives/self-regulation model; discharged treated patients may be in the former, latter, or both programs. Regrettably, records with respect to treatment specifics were insufficient to explore these important treatment variables in a meaningful way.

We urge caution among relevant stakeholders in the application and interpretation of our findings. Whereas it is possible that current Static-99R norms overestimate sexual recidivism among released SVP patients, the limitations and study features described above create more questions than answers. Potential future research directions should examine the role and relevance of clinical and legal variables that may have influenced observed base rates (e.g., out-of-state offending, effects of treatment, lengthy total detainment, re-offense impact of CONREP and parole supervision, and detained versus fully committed outcomes). An additional implication concerns the accuracy and utility of SVP evaluations using the Static-99R without the inclusion of a dynamic risk measure. These results highlight the need for a dynamic tool that can capture changes in risk with a variably treated and supervised, aging SVP population; all of which do not seem to be sufficiently captured by the Static-99R alone. Finally, future SVP research should consider more detailed background information on social and criminal history, perpetrator and victim demographics, and treatment information among other key areas. In conclusion, the present findings demonstrate that the sexual recidivism base rate on community discharged California SVP patients is low, and when combined with their higher static risk, has implications for the discrimination and calibration properties of the tool with this sample. Although further research is needed, the current study has important implications for research and practice with California’s SOCC population to promote accurate risk assessment, effective risk management, and successful community reintegration to prevent future sexual violence.

Footnotes

Author’s Note

The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views or opinions of any federal, state, or county government entity, including California Department of State Hospitals or California Health and Human Services Agency.

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.

ORCID iDs

Allen Azizian

Mark E. Olver

References

Andrews

D. A.

Bonta

(2010). Rehabilitating criminal justice policy and practice. Psychology, Public Policy, and Law, 16, 39–55. https://doi.org/10.1037/a0018362

Boccaccini

M. T.

Murrie

D. C.

Caperton

J. D.

Hawes

S. W.

(2009). Field validity of the STATIC-99 and MnSOST-R among sex offenders evaluated for civil commitment as sexually violent predators. Psychology, Public Policy, and Law, 15(4), 278–314.

Boccaccini

M. T.

Rice

A. K.

Helmus

L. M.

Murrie

D. C.

Harris

P. B.

(2017). Field validity of Static-99/R scores in a statewide sample of 34,687 convicted sexual offenders. Psychological Assessment, 29, 611–623. https://doi.org/10.1037/pas0000377

Declue

Rice

(2016). Florida’s released “Sexually Violent Predators” are not “high risk.” Open Access Journal of Forensic Psychology, 8, 22–51.

D’Orazio

Arkowitz

Adams

Maram

(2009). The California Sexually Violent Predator statute: History, description, and areas of improvement. California Coalition on Sexual Offending.

D’Orazio

Azizian

Olver

(2019). Recidivism and other outcomes among sexual offenders committed as “Sexually Violent Predators” in California [Conference session]. Presentation at the Association for the Treatment of Sexual Abusers 38th Annual Research and Treatment Conference, Atlanta, Georgia.

Duwe

(2013). To what extent does civil commitment reduce sexual recidivism? Estimating the selective incapacitation effects in Minnesota. Journal of Criminal Justice, 42(2), 193–202.

Felthous

A. R.

(2018). Sexually Violent Predator law in the United States. East Asian Archives of Psychiatry, 28, 159–173.

Frisbie

L. V.

Dondis

E. H.

(1965). Recidivism among treated sex offenders. California Mental Health Research Monograph, No. 5.

10.

Hanson

R. K.

(2017). Assessing the calibration of actuarial risk scales: A primer on the E/O index. Criminal Justice and Behavior, 44, 26–39. https://doi.org/10.1177/0093854816683956

11.

Hanson

R. K.

Babchishin

K. M.

Helmus

Thornton

(2013). Quantifying the relative risk of sex offenders: Risk ratios for Static-99R. Sexual Abuse, 25, 482–515. https://doi.org/10.1177/107906321246906

12.

Hanson

R. K.

Helmus

L. M.

Harris

A. J. R.

(2015). Assessing the risk and needs of supervised sexual offenders: A prospective study using STABLE-2007, Static-99R, and Static-2002R. Criminal Justice and Behavior, 42, 1205–1224. https://doi.org/10.1177/0093854815602094

13.

Hanson

R. K.

Lunetta

Phenix

Neeley

Epperson

(2014). The field validity of Static-99/R sex offender risk assessment tool in California. Journal of Threat Assessment and Management, 1(2), 102–117.

14.

Harris

Phenix

Hanson

R. K.

Thornton

(2003). Static 99: Coding rules revised 2003. Solicitor General Canada.

15.

Helmus

L. M.

(2009). Re-norming Static-99 recidivism estimates: Exploring base rate variability across sex offender samples (Master’s thesis). ProQuest Dissertations and Theses database (UMI No. MR58443).

16.

Helmus

L. M.

Babchishin

(2017). Primer on risk assessment and the statistics used to evaluate its accuracy. Criminal Justice and Behavior, 44, 8–25. https://doi.org/10.1177/0093854816678898

17.

Helmus

L. M.

Thornton

Hanson

R. K.

Babchishin

K. M.

(2012). Improving the predictive accuracy of Static-99 and Static-2002 with older sex offenders: Revised age weights. Sexual Abuse, 24, 64–101. https://doi.org/10.1177/1079063211409951

18.

Hoberman

H. M.

Jackson

R. L.

(2016). Civil commitment of select recidivistic sexual offenders deemed likely to sexually reoffend. In Phenix

Hoberman

H. M.

(Eds.), Sexual offending: Predisposing antecedents, assessments and management (pp. 779–821). Springer.

19.

Howard

P. D.

(2017). The effect of sample heterogeneity and risk categorization on area under the curve predictive validity metrics. Criminal Justice and Behavior, 44, 103–120. https://doi.org/10.1177/0093854816678899

20.

Janus

E. S.

(2006). Failure to protect: America’s sexual predator laws and the rise of the preventive state. Cornell University Press.

21.

Joint Legislative Audit and Review Commission. (2012). Review of the civil commitment of Sexually Violent Predators. Report to the Governor and the General Assembly of Virginia. http://jlarc.virginia.gov/pdfs/reports/Rpt423.pdf

22.

Kansas v. Crane, 534 US 407 (2002).

23.

Kansas v. Hendricks, 521 US 346 (1997).

24.

Kelley

S. M.

Ambroziak

Thornton

Barahal

R. M.

(2018). How do professionals assess sexual recidivism risk? An updated survey of practices. Sexual Abuse, 32(1), 3–29. https://doi.org/10.1177/1079063218800474

25.

Knight

R. A.

Thornton

(2007). Evaluating and improving risk assessment schemes for sexual recidivism: A long-term follow-up of convicted sexual offenders (Document No. 217618). Submitted to the U.S. Department of Justice.

26.

La Fond

J. Q

. (2005). Sexually violent predator laws. In La Fond

J. Q.

(Ed.), Preventing sexual violence: How society should cope with sex offenders (pp. 127–165). American Psychological Association.

27.

Lee

S. C.

Hanson

R. K.

(2017). Similar predictive accuracy of the Static-99R risk tool for White, Black, and Hispanic sex offenders in California. Criminal Justice and Behavior, 44(9), 1125–1140. https://doi.org/10.1177/0093854817711477

28.

Lee

S. C.

Restrepo

Satariano

Hanson

R. K.

(2016). The predictive validity of Static-99R for sexual offenders in California: 2016 update. http://saratso.org/pdf/ThePredictiveValidity_of_Static_99R_forSexualOffenders_inCalifornia_2016v1.pdf

29.

Lieberman

Siegel

B. A.

(1957). A program for “sexual psychopaths” in a state mental hospital. The American Journal of Psychiatry, 113, 801–807.

30.

Marques

J. K.

Wiederanders

Day

D. M.

Nelson

van Ommeren

(2005). Effects of a relapse prevention program on sexual recidivism: Final results from California’s Sex Offender Treatment and Evaluation Project (SOTEP). Sexual Abuse, 17(1), 79–107.

31.

Mercado

C. C.

Jeglic

Markus

Hanson

R. K.

Levenson

(2013). Sex offender management, treatment, and civil commitment: An evidence based analysis aimed at reducing sexual violence. https://www.ojp.gov/pdffiles1/nij/grants/243551.pdf

32.

Monahan

Davis

S. K.

(1983). Mentally disordered sex offenders. In Monahan

Steadman

H. J.

(Eds.), Mentally disordered offenders. Perspectives in law & psychology (vol. 6, pp. 1–302). Springer.

33.

Morse

S. J.

(1998). Fear of danger, flight from culpability. Psychology, Public Policy, and Law, 4(1–2), 250–267.

34.

New York State Office of Mental Health. (2009). 2008 Annual report on the implementation of mental hygiene law article 10: Sex offender management and treatment act of 2007.

35.

Oliver

A. D.

(1982). The sex offender: Lessons from the California experience. International Journal of Law and Psychiatry, 5(3–4), 403–411.

36.

Olver

M. E.

Eher

(2019). Predictive properties and factor structure of the VRS-SO: Results from an Austrian sample. European Journal of Psychological Assessment, 36(5), 748–757. http://doi.org/10.1027/1015-5759/a000551

37.

Padilla

Russell

(2004). Preliminary analysis of recidivism data. Unpublished Memorandum to Melvin Hunter, ED: Atascadero State Hospital.

38.

Page

C. C.

Burbank

B. E.

(2016). Sexually violent predator cases: A prosecutor’s perspective. In Phenix

Hoberman

H. M.

(Eds.), Sexual offending: Predisposing antecedents, assessments and management (pp. 823–834). Springer.

39.

People v. Superior Court (Ghilotti), 27 Cal. 4th 888, 902-05. (2002).

40.

Phenix

Helmus

L. M.

Hanson

R. K.

(2016). Static-99R & Static-2002R evaluators’ workbook. www.static99.org

41.

E. P.

(2007). The civil confinement of sexual predators: A delicate balance. Journal of Civil Rights and Economic Development, 22(1), 335–372.

42.

Rapaport

Lieberman

(1956). The sexual psychopath in California. California Medicine, 85(4), 232–234.

43.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29(5), 615–620.

44.

Rockhill

Byrne

Rosner

Louie

M. M.

Colditz

(2003). Breast cancer risk prediction with a log-incidence model: Evaluation of accuracy. Journal of Clinical Epidemiology, 56, 856–861. https://doi.org/10.1016/S0895-4356(03)00124-0

45.

Schlank

(2018). Sex offender management practices in the United States. Sexual Offender Treatment, 13(1/2), 1–6.

46.

Sreenivasan

Frances

Weinberger

L. E.

(2010). Normative versus consequential ethics in Sexually Violent Predator Laws: An ethics conundrum for psychiatry. Journal of the American Academy of Psychiatry and the Law, 38(3), 386–391.

47.

Steadman

H. J.

Monahan

Hartstone

Davis

S. K.

Robbins

P. C.

(1982). Mentally disordered offenders: A national survey of patients and facilities. Law and Human Behavior, 6(1), 31–38.

48.

Sturgeon

V. H.

Taylor

(1980). Report of a five-year follow-up study of mentally disordered sex offenders released from Atascadero State Hospital in 1973. Criminal Justice Journal, 4, 31–63.

49.

Texas Health Safety Code Ch, 841, § 841.001. (2009). Texas Civil Commitment-Outpatient Sexually Violent Predator Treatment Program (OSVTP).

50.

Therneau

T. M.

(2020). A package for survival analysis in S (R package version 3.1-11) [Software]. https://CRAN.R-project.org/package=survival

51.

Washington State Institute for Public Policy. (2007). Comparison of state laws authorizing involuntary commitment of sexually violent predators: 2006 update. Revised.

52.

Wilson

R. J.

Looman

Abracen

Pake

D. R.

(2012). Comparing sexual offenders at the Regional Treatment Centre (Ontario) and the Florida Civil Commitment Center. International Journal of Offender Therapy and Comparative Criminology, 57(3), 377–395. https://doi.org/10.1177/0306624x11434918