Predicting Institutional Sexual Misconduct by Adult Male Sex Offenders

Abstract

Although the Static-99R has been found to be a robust measure of long-term risk to reoffend among adult male sex offenders, few studies have investigated the relationship between Static-99R scores and institutional (i.e., prison) behavior. The current study sought to address this gap in the research by testing the ability of the Static-99 and Static-99R to predict five types of institutional misconduct: (a) sexual, (b) violent (nonsexual), (c) nonviolent (nonsexual), (d) drug-related, and (e) any nonsexual. Results indicate that the Static-99/99R may be useful as predictors of institutional misconduct and, therefore, as a risk classification measure for prisons to protect both staff and inmates and improve institutional environments.

Keywords

sex offenders incarcerated misconduct risk prediction

Sex offender risk assessment has made important advances in the past several decades, having been informed by robust meta-analytic findings about the most important risk factors for future sexual offending among identified individuals (Hanson & Bussière, 1998; Hanson & Morton-Bourgon, 2005). Moreover, in the past 15 years, there has been a proliferation of risk measures to predict sexual recidivism, many of which have been found by research to be both reliable and valid (Hanson & Morton-Bourgon, 2009). These risk measures have been cross-validated across different jurisdictions, correctional and forensic mental health samples, and using different outcome variables (e.g., arrest, charge, conviction).

Officially detected recidivism by identified sex offenders after release from custody is the most common outcome in follow-up research. However, recent reports documenting that approximately 1 in 10 prisoners has been sexually victimized in custody indicates that sexual recidivism in the form of institutional sanctions for proscribed sexual behavior are also of concern (Beck & Johnson, 2012). Sexual misconduct within correctional institutions creates safety concerns for both inmates and staff, degrades institutional milieu, and may result in increased medical or legal expenses. As such, correctional personnel and mental health professionals have increasingly been called upon to assess the likelihood of inmates engaging in this behavior while incarcerated. Accurate prediction of which inmates will engage in misconduct is essential to ensure that a correctional facility’s limited resources are directed at those most in need of security and programmatic resources. Moreover, accurate assessment of institutional infractions is essential for effective risk-management and prevention strategies.

It should be noted that a distinction can be made here between sexual behavior involving consenting partners that is prohibited (e.g., consensual sexual activity with another prisoner) and sexual behavior involving a nonconsenting person (e.g., sexual assault of another prisoner or of a staff member). Being able to identify those incarcerated offenders who are at greatest risk of engaging in institutional sexual misconduct would be of great value to addressing the sexual victimization of prisoners and maintaining a safe and secure institutional environment.

Studies of Institutional Misconduct

The prediction of institutional misconduct has been difficult for correctional personnel, as there are few empirical studies to guide correctional practice. The limited research that has been conducted typically involves forensic patients who are civilly confined in mental health facilities, with fewer studies involving samples of adult male inmates. Studies examining institutional misconduct with adult, male inmates have found an association between misconduct and age, education level, psychological well-being, and criminal history (Cunningham, Sorensen, Vigen, & Woods, 2011; Morris, Longmire, Buffington-Vollum, & Vollum, 2010; Steiner & Wooldredge, 2009). Specifically, younger offenders, inmates with lower education levels (perhaps a proxy for intelligence), and inmates who are suffering from mental health problems are more likely than their counterparts to engage in misconduct while incarcerated. Prisoners with gang affiliations, prior prison terms, prior violent-offense arrests, and shorter criminal sentences are also at increased risk to participate in misconduct (see Cunningham et al., 2011). Psychopathy, which includes personality characteristics such as irresponsibility, antisociality, and lack of regard for others, has been shown to correlate positively with institutional infractions for both forensic patients and adult inmates (Vitacco, Gonsalves, Tomony, Smith, & Lishner, 2012).

In addition to the studies on factors related to institutional misconduct, researchers have also attempted to evaluate actuarial tools to predict such behavior (Cunningham et al., 2011; Hastings, Krishnan, Tangney, & Stuewig, 2011; McDermott, Dualan, & Scott, 2011; Newbury & Shuker, 2012). The Violence Risk Appraisal Guide (Quinsey, Harris, Rice, & Cormier, 2006) has shown some success in predicting violent institutional misconduct for male (but not female) inmates, even after controlling for psychopathy (Hastings et al., 2011). In addition, the Classification of Violence Risk (Monahan et al., 2005) has demonstrated moderate predictive accuracy for institutional aggression with samples of forensic patients (McDermott et al., 2011). These studies, however, have focused on general offenders and general institutional misconduct, not specifically sexual offenders and sexual misconduct.

Studies of Incarcerated Sex Offenders

Although studies predicting the institutional misbehavior of sex offenders are scarce, there have been a few. The most notable of this research is a series of studies conducted using samples of sex offenders incarcerated in Texas (i.e., Buffington-Vollum, Edens, Johnson, & Johnson, 2002; Caperton, Edens, & Johnson, 2004; Edens, Buffington-Vollum, Colwell, Johnson, & Johnson, 2002). The authors in this series used two psychological measures, the Psychopathy Checklist–Revised (Hare, 1991) and the Personality Assessment Inventory (PAI; Morey, 1991), to predict various types of institutional misconduct, including physical aggression, verbal aggression/acts of defiance, and nonaggressive infractions. Results of these studies indicated that the Antisocial Features scale of the PAI could differentiate between sex offenders who engaged in institutional misconduct and those who did not.

While results of these studies found both measures to be predictive of the various outcomes, the studies do have some limitations. For example, neither of the measures included in the studies were designed specifically for use with sex offenders. Furthermore, only Caperton et al. (2004) included a specific measure of sexual misconduct as an outcome, and none of the PAI scales included in that study were found to significantly predict it. Although this finding may have been due to low analytic power (the study included only 137 participants), it leaves open the question of what, if anything, predicts institutional sexual misconduct.

Current Study

In this study, we examined the ability of the Static-99 (Hanson & Thornton, 1999) and Static-99R, which were originally developed to predict sexual recidivism (new charges or convictions) among identified adult male sex offenders who are at risk in the community, to predict sexual misconducts while individuals were still in custody. The Static-99/99R are currently being used by many correctional institutions for treatment placement and other risk-management decisions. If these instruments can also predict institutional misconduct, they offer a cost-effective approach to address the security issues caused by the aggressive and/or sexually inappropriate behaviors that result in institutional infractions. Finding a predictive relationship for the Static-99/99R would also suggest that the same factors that underlie sexual recidivism in the community can also explain misconducts, which has implications for our theoretical understanding of misconducts and for programming decisions to reduce risk while in custody.

We predicted that the Static-99/99R would be significant predictors of institutional sexual misconducts even with two conditions that separate studies of institutional behavior from community behavior: (a) the time frame is shorter, on average, looking at behavior in custody versus post-release, and (b) the outcome is operationally and conceptually different from recidivism in the community, because institutional misconducts would mostly be committed against nonpreferred targets (i.e., there are no children and mostly men in the institutions). In addition to the Static-99/99R, we were also able to examine prior criminal history, prior institutional misconduct history, and criminal versatility as potential predictors of institutional sexual misconduct.

Method

Sample

The sample for the study consisted of 4,211 adult male sex offenders reviewed for possible civil management (i.e., those convicted of a sexual or sexually-motivated felony) in New York State. Only adult male offenders were included in the analyses, as the Static-99R was developed for use with this population and is not necessarily appropriate for use with female or juvenile sex offenders (Harris, Phenix, Hanson, & Thornton, 2003). Of these 4,211 offenders, 508 (12.1%) were excluded from the analyses for missing values on one or more of the study variables (other than Nonconsensual Sexual Ticket, as discussed below). Attrition analyses identified no significant differences between the 3,703 offenders with complete data and the 508 offenders with missing data on any of the study variables (all p values ≥ .14).

Characteristics of the 3,703 offenders in the final sample can be found in Table 1 and are separated into three groups: (a) offenders without a sexual misconduct ticket,(b) offenders with any sexual misconduct ticket (whether consensual, nonconsensual, or unknown), and (c) offenders with a nonconsensual sexual misconduct ticket. As a whole, offenders in the sample had an average age of 32.74 years (SD = 11.66) at the time of conviction on their index offense. Age at index conviction was used in the present analysis because date of original prison admission for the index offense was missing for many offenders. Furthermore, date of conviction and date of prison admission tend to be highly correlated and very similar. For example, the difference between the two dates for those offenders for whom both dates were available averaged less than 4 months, with a median of less than 2 months. In terms of race and ethnicity, the present sample was 49.2% White (n = 1,821), 34.2% Black (n = 1,267), and 15.5% Hispanic (n = 574). According to New York State penal codes, the index sexual offense for the majority of offenders was either rape (n = 1,436; 38.8%) or sexual abuse (i.e., sexual contact not involving intercourse; n = 1,060; 28.6%), with the majority of the remaining offenders having been convicted of a criminal sexual act (discussed below; n = 743; 20.1%) or a course sexual misconduct against a child (n = 192; 5.2%).

Table 1:

Sample Characteristics.

	Sexual Ticket Group
	None (n = 3,420; 92.4%)		Any (n = 283; 7.6%)		Nonconsensual (n = 137; 3.7%)
Variables	M (SD)	n (%)	M (SD)	n (%)	M (SD)	n (%)
Age at conviction (in years)^a,b	33.13 (11.71)		28.12 (9.90)		29.27 (10.44)
Static-99R (range = −3 to 12)^a,b	2.58 (2.26)		4.12 (2.12)		4.42 (2.01)
Static-99 (range = 0 to 12)^a,b	2.97 (2.00)		4.41 (2.10)		4.82 (2.13)
Race/ethnicity^a,b
White/non-Hispanic		1,718 (50.2)		103 (36.4)		43 (31.4)
Black/non-Hispanic		1,125 (32.9)		142 (50.2)		75 (54.7)
Hispanic		540 (15.8)		34 (12.0)		17 (12.4)
Other/missing		37 (1.1)		4 (1.4)		2 (1.5)
Prior criminal history
Arrests^a,b	4.47 (4.93)		5.28 (6.51)		5.96 (6.38)
Sexual arrests	1.34 (0.99)		1.44 (0.92)		1.49 (0.92)
Violent felony (including sexual) arrests^a,b	1.44 (1.31)		1.93 (1.62)		2.28 (1.92)
Variety of offense arrests^a,b,c	1.61 (1.83)		2.30 (2.05)		2.62 (2.07)
Criminal sexual act arrests^a,b	0.48 (0.63)		0.63 (0.70)		0.60 (0.69)
Sexual disciplinary tickets prior to index offense^a,b	0.009 (0.04)		0.89 (3.01)		1.31 (4.12)
Total disciplinary tickets prior to index offense^a,b	0.94 (4.39)		3.23 (10.69)		4.33 (13.41)
Index sexual conviction
Rape		1,317 (38.5)		119 (42.0)		67 (48.9)
Sexual abuse		995 (29.1)		65 (23.0)		33 (24.1)
Criminal sexual act^d		673 (19.7)		70 (24.7)		28 (20.4)
Course sexual conduct against a child		185 (5.4)		7 (2.5)		2 (1.5)
Sexually motivated burglary or robbery		93 (2.7)		10 (3.5)		3 (2.2)
Sexually motivated assault		78 (2.3)		3 (1.1)		0 (0.0)
Other sexually motivated offenses^e		79 (2.3)		9 (3.2)		4 (2.9)
Victim variables^f
Unrelated^a,b		2,524 (73.8)		245 (86.6)		118 (92.2)
Stranger^a,b		716 (20.9)		107 (37.8)		55 (43.0)
Male^a,b		604 (17.7)		93 (32.9)		40 (31.2)
Nonsexual disciplinary tickets
Violent (nonsexual)^a,b		902 (26.4)		182 (64.3)		95 (69.3)
Drug^a,b		362 (10.6)		85 (30.0)		49 (35.8)
Nonviolent^a,b		2,583 (75.5)		269 (95.1)		129 (94.2)
Any^a,b		2,673 (78.2)		273 (96.5)		131 (95.6)

Significant difference between those offenders with no sexual ticket and those with any sexual ticket (i.e., consensual or nonconsensual). ^bSignificant difference between those offenders with no sexual ticket and those with a nonconsensual sexual ticket. ^cVariety is a count variable of eight types of arrest: assault, burglary, theft, public order, criminal mischief, custody, marijuana, and robbery. ^dAccording to New York State penal law, criminal sexual act offenses are nonconsensual oral sex and sodomy offenses. ^eIncludes sexually motivated homicides, possessing/promoting child pornography, kidnapping, incest, and prostitution offenses involving minors. ^fAs defined in the Static-99 coding guide.

Data

Data for the project were provided by the New York State Office of Mental Health (OMH). OMH reviews all offenders with a qualifying sexual offense (see the Sex Offender Management and Treatment Act, 2007) for possible civil management. These reviews begin approximately 6 months prior to an offender completing his or her sentence for a sexual offense and being released. OMH conducts these reviews using detailed file information on the offenders.

Independent Variables

Static-99 and Static-99R scores

As part of their reviews, OMH scored each offender in the present study on the Static-99 (Hanson & Thornton, 1999). The instrument consists of 10 static (i.e., unchanging) risk factors for sexual recidivism that are summed to generate scores ranging from 0 to 12, and in a recent meta-analysis, the instrument was found to have validity (receiver operative characteristic [ROC] = .67) in the prediction of sexual recidivism (Hanson, Helmus, & Thornton, 2010). Previous studies have found the Static-99 to have high levels of interrater reliability (kappas and interclass correlation coefficients ≥ .80; Harris et al., 2003), which are similar to the levels of interrater agreement found by OMH in an internal analysis of its own Static-99 use (kappas ≥ .82 for all 10 Static-99 items and the total score).

Despite the Static-99 having been coded for all offenders in the study sample as part of their OMH review, two issues arose regarding use of these scores in the current study. First, according to the most recent coding manual (Harris et al., 2003), institutional violations are to be considered when scoring the Static-99. As such, during the review for possible civil management, OMH staff factored institutional violations into the scoring of offender Static-99s when appropriate. This presented a problem for the current study, as using those scores would artificially inflate the relationship between the Static-99 and institutional misconduct in the present analysis. To avoid this artificial inflation, OMH staff went back and adjusted the Static-99 scores of any offenders whose scores were affected by institutional violations, to reflect what scores those offenders would have had at the start of their index incarceration. These adjustments were made on the basis of the same objective file information about historical variables that was used to originally score the Static-99, and the rescoring adhered to the strict Static-99 coding rules (Harris et al., 2003).

The second issue concerning use of the Static-99 scores in the current analysis was that, in late 2009, the developers of the Static-99 released an updated version of the instrument called the Static-99R. The two instruments consist of the same 10 static risk items, with 9 of the items (and their coding) being the exact same for each. The only structural difference between the two is that the Static-99R has a more detailed coding for the offender age item, which results in the instrument generating scores ranging from −3 to 12. By retroactively recoding the Static-99 offender age item for offenders in the study sample, therefore, Static-99R scores for all the offenders could be generated, and the present analysis was able to test the ability of both instruments to predict institutional sexual violations. As the results of the analyses for the Static-99 and the Static-99R were very similar, however, and as OMH recently shifted to scoring offenders on the Static-99R, the current study presents and discusses the predictive accuracy of only the Static-99R, while analytic results for the Static-99 are only mentioned in notes. The average Static-99R score for offenders in the present analysis was 2.69 (SD = 2.29).

Prior number of institutional disciplinary tickets

When incarcerated offenders violate institutional rules, they are issued disciplinary tickets. These tickets can be for any type of violation, whether violent (e.g., fighting) or nonviolent (e.g., possession of contraband drugs). To generate a measure of prior institutional rule-violating behavior, a count was created of all institutional disciplinary tickets (including sexual tickets) an offender had been issued on all incarcerations prior to his index incarceration. Offenders in the present study averaged 1.11 (SD = 5.18) prior institutional disciplinary tickets.

Prior number of institutional sexual disciplinary tickets

While the above measure included all institutional disciplinary tickets (including sexual) an offender had accumulated during prior incarcerations, a separate measure of only prior institutional sexual disciplinary tickets was also included in the analyses. Offenders in the present study averaged 0.07 (SD = 0.87) prior institutional sexual disciplinary tickets.

Prior number of sexual arrests

Sexual offenses were defined for the analysis as any offense for which an offender could be required to register as a sex offender under New York State law. Offenders in the present study averaged 1.35 (SD = 0.98) prior sexual arrests.

Prior number of criminal sexual act arrests

According to New York State penal law, all instances of nonconsensual oral sex or sodomy are classified as criminal sexual acts. Given that the study was of the behavior of incarcerated male offenders, a measure of criminal sexual act arrests was included in the analyses. It should be noted that, as registerable sexual offenses, these arrests were included in the count of the total prior sexual arrests mentioned above. Offenders in the present study averaged 0.49 (SD = 0.64) prior criminal sexual act arrests.

Prior number of violent felony arrests

A summed variable counting up the number of violent (including sexual) felonies each offender had committed in his past was created for the present study. Offenders in the present study averaged having committed 1.47 (SD = 1.35) violent felonies in their past.

Variety of offending types

To measure each offender’s criminal versatility, a variable was created counting eight different types of crime in his past (including his index offense). Specifically, the variety of the offending variable used in this study was a count of how many of the following eight types of crimes an offender had been arrested for: (a) assault, (b) robbery, (c) burglary, (d) theft, (e) public order (e.g., loitering, harassment), (f) custody (e.g., escape, absconding from supervision), (g) criminal mischief (e.g., property damage, graffiti), and (h) anything marijuana-related. This criminal versatility variable has been found in prior research to add incremental predictive validity above other criminal history variables (both sexual and nonsexual) to the prediction of both sexual and nonsexual rearrest (Freeman & Sandler, 2010). Offenders in the present study averaged having committed 1.66 (SD = 1.86) of these types of crime in their past.

Dependent Variables

Sexual disciplinary ticket

The main outcome measure in the present study was whether an offender was issued an institutional disciplinary ticket for a sexual violation during his index incarceration. This variable was coded dichotomously (0 = no, 1 = yes). Of the final sample, 7.6% (n = 283) of the offenders received a sexual disciplinary ticket during their index incarceration.

Nonconsensual sexual disciplinary ticket

Not all institutional sexual violations are for nonconsensual acts. That is, two inmates caught engaging in a consensual sexual activity in violation of institutional rules will still receive a sexual disciplinary ticket, even if no force or coercion was used. Therefore, each offender who received a sexual disciplinary ticket was coded for whether his sexual violation(s) was nonconsensual (0 = all consensual, 1 = at least one nonconsensual). Due to missing data, however, consensual/nonconsensual information was only available for 65.7% (n = 186) of offenders with sexual violations. The reasons for missing data are not known, but likely include situations where the investigation could not determine whether the act was consensual or nonconsensual but did determine that it was sexual in nature, and situations where record-keeping was not sufficiently detailed. Of those 186 offenders in the sample who received sexual disciplinary tickets and had complete information, 73.7% (n = 137) had committed nonconsensual acts, suggesting that a majority of institutional sexual misconducts involve coercion or force and thus have some similarities with sexual offenses committed in the community.

Violent (nonsexual) disciplinary ticket

To compare the Static-99R’s ability to predict sexual misconduct versus its ability to predict general misconduct, four nonsexual types of disciplinary tickets were examined. All four of these variables were coded the same way as the sexual disciplinary ticket variable: a dichotomous (0 = no, 1 = yes) coding of whether or not an offender was issued an institutional disciplinary ticket for each type of violation during his index incarceration. The first of these comparison ticket types was any violent (nonsexual) ticket (e.g., fighting). Of the final sample, 29.3% (n = 1,084) of the offenders received a violent (nonsexual) disciplinary ticket during their index incarceration.

Drug-related disciplinary ticket

The second of the comparison ticket types was any nonviolent ticket related to drugs (e.g., possession of contraband narcotics). Of the final sample, 12.1% (n = 447) of the offenders received a nonviolent, drug-related disciplinary ticket during their index incarceration.

Nonviolent disciplinary ticket

The third of the comparison ticket types was any nonviolent ticket not related to drugs (e.g., refusal of a direct order, attempted bribery). Of the final sample, 77.0% (n = 2,852) of the offenders received a nonviolent disciplinary ticket unrelated to drugs during their index incarceration.

Any nonsexual disciplinary ticket

The fourth and last of these comparison ticket types was any nonsexual ticket. This category included any of the other three nonsexual types of tickets listed above: (a) violent (nonsexual), (b) drug, and (c) nonviolent. Of the final sample, 79.6% (n = 2,946) of the offenders received at least one nonsexual disciplinary ticket during their index incarceration.

Analyses

First, group differences between those offenders who received a sexual disciplinary ticket and those who received no sexual disciplinary ticket, as well as between those offenders who received a sexual disciplinary ticket for a nonconsensual act and those who received no sexual disciplinary ticket, were assessed using one-way analyses of variance (ANOVAs; for continuous variables) and chi-square analyses (for categorical variables). Relationships between the individual independent variables and the dependent variables were then assessed through ROC area under the curve (AUC) analyses. AUC analyses judge the ability of a variable to predict an outcome (e.g., receiving a sexual disciplinary ticket) by reporting the chance that someone positive on the outcome randomly selected from the file will be higher on the variable than someone not positive on the outcome randomly selected from the file. An AUC = .50, therefore, indicates predictive accuracy no better than chance, while an AUC = 1.00 indicates perfect predictive accuracy.

To then assess the ability of the Static-99R to predict the likelihood of an offender receiving a sexual disciplinary ticket after controlling for other possible predictors, two Cox regressions were conducted. The dependent variable for the first regression was receiving any sexual disciplinary ticket, and the dependent variable for the second regression was receiving a nonconsensual sexual disciplinary ticket. Cox regression was preferred to binary logistic regression because Cox regression takes time at risk for the event to occur into account. The varying lengths of incarceration for offenders in the present analysis, therefore, made Cox regression more appropriate. For comparison to the Cox regression results, binary logistic regressions were also estimated that included all of the same predictor variables and had a fixed follow-up period of 3 years. As results of the binary logistic regressions were almost identical to those for the Cox regressions (i.e., the pattern of the results was the same), only the Cox regression results are discussed below.

Results

Group Differences

Results of the one-way ANOVAs and chi-square analyses revealed several significant group differences between those offenders who did not receive a sexual disciplinary ticket, and (a) those offenders who received a sexual disciplinary ticket of any kind and (b) those offenders who received a sexual disciplinary ticket for a nonconsensual act. As can be seen in Table 1, offenders who were issued sexual disciplinary tickets of any kind and offenders who were issued sexual disciplinary tickets for nonconsensual acts were younger, had higher Static-99R scores (including having a higher proportion of offenders with unrelated, stranger, and male victims), and had more extensive criminal histories on every measure (other than prior number of sexual arrests) than those offenders who did not receive a sexual disciplinary ticket (all ps ≤ .001). Furthermore, the institutional misbehavior of offenders who received sexual disciplinary tickets was not limited to sexual misconduct. Offenders who received sexual disciplinary tickets for any reason and those who received sexual disciplinary tickets specifically for nonconsensual acts also had a significantly higher proportion of offenders who also received violent (nonsexual) disciplinary tickets, drug-related disciplinary tickets, nonviolent disciplinary tickets, and any disciplinary tickets (all ps ≤ .001). No significant differences were found between any of the groups with regard to type of index sexual conviction.

Univariate Predictive Validity

Sexual Ticket Outcomes

AUC analyses were then used to test the ability of study variables to predict the two sexual study outcome measures. As can be seen in Table 2, all but one of the variables were significant predictors of an offender receiving a sexual disciplinary ticket; the exception was number of prior sexual arrests (AUC = .54; p = ns). Of the significant predictors, the strongest was the Static-99R (AUC = .69; p ≤ .001), followed by age at conviction (reversed so that younger age equals more likely to have a misconduct: AUC = .64; p ≤ .001) and prior number of sexual tickets (AUC = .64; p ≤ .001). The AUC analyses predicting a nonconsensual sexual act produced similar results, with all of the variables being significant predictors, other than number of prior sexual arrests (AUC = .55; p = ns) and number of prior criminal sexual act arrests (AUC = .54; p = ns). Of the significant predictors, the strongest was the Static-99R (AUC = .71; p ≤ .001), followed by prior number of sexual tickets (AUC = .67; p ≤ .001).

Table 2:

Receiver Operative Characteristic AUC Values for Sexual Tickets.

	Sexual Ticket Outcome
	Any		Nonconsensual
	vs. No Ticket		vs. No Ticket		vs. Consensual
Predictor Variables	AUC	95% CI	AUC	95% CI	AUC	95% CI
Age at conviction^a	.64***	[.60, .67]	.59***	[.54, .64]	.39*	[.30, .49]
Prior disciplinary ticket count	.55**	[.51, .59]	.57**	[.51, .62]	.55	[.45, .64]
Prior sexual ticket count	.64***	[.60, .68]	.67***	[.61, .73]	.60***	[.51, .69]
Sexual arrests	.54	[.50, .57]	.55	[.49, .60]	.54	[.45, .64]
Criminal sexual act arrests^b	.55**	[.52, .59]	.54	[.49, .60]	.47	[.37, .57]
Violent felony arrests	.60***	[.56, .63]	.65***	[.60, .70]	.65**	[.56, .74]
Variety of offending types^c	.60***	[.57, .64]	.65***	[.60, .69]	.65**	[.56, .74]
Static-99R^d	.69***	[.66, .72]	.71***	[.67, .75]	.58	[.48, .68]

Note. AUC = area under the curve; CI = confidence interval.

AUC values can be said to differ significantly if their 95% confidence intervals do not overlap.

In years and reversed for the analysis due to its negative correlation with the outcome. ^bAccording to New York State penal law, criminal sexual act offenses are nonconsensual oral sex and sodomy offenses. ^cVariety is a count variable of eight types of arrest: assault, burglary, theft, public order, criminal mischief, custody, marijuana, and robbery. ^dIn the same analyses, the Static-99 yielded AUCs = .69 (p ≤ .001), .73 (p ≤ .001), and .61 (p ≤ .05), respectively.

p ≤ .05. **p ≤ .01. ***p ≤ .001.

As a supplemental analysis, a set of AUC analyses was then conducted to test whether any of the predictor variables could discriminate between those offenders who received sexual disciplinary tickets for consensual acts and those who received sexual disciplinary tickets for nonconsensual acts. The analytic power for these analyses was significantly lower than for the previous two sets of AUC analyses, as the sample for this third set of analyses was reduced to the 186 offenders who received a sexual disciplinary ticket and for whom information was available on whether or not the act was consensual. The results of these analyses should, therefore, be interpreted with caution. As can be seen in the last AUC column in Table 2, only four of the variables significantly discriminated between the two groups. The strongest two discriminators were number of prior violent felony arrests (AUC = .65; p ≤ .01) and variety of offending types (AUC = .65; p ≤ .01), while the Static-99R was not found to be significantly discriminative (AUC = .58; p = ns).

Nonsexual Ticket Outcomes

Although the results of the previous analyses showed the Static-99R to be a significant predictor of an offender receiving a sexual disciplinary ticket (in general, or a nonconsensual sexual ticket specifically), the question remained as to whether the Static-99R was specifically predicting sexual misconduct, or simply predicting general misconduct. To address this question, four other series of AUC analyses were conducted to examine the four nonsexual disciplinary ticket outcomes: (a) violent (nonsexual) ticket, (b) drug ticket, (c) nonviolent ticket, and (d) any nonsexual ticket (i.e., violent, drug, or nonviolent). The results of the analyses are displayed in Table 3. As can be seen in the bottom row, the Static-99R was a significant predictor of all types of institutional misconduct, although the AUC values for all four nonsexual ticket outcomes were lower than the AUC values for the two main sexual ticket outcomes reported above (i.e., any sexual ticket [AUC = .69] and nonconsensual sexual ticket [AUC = .71]). Furthermore, the Static-99R predicted a nonconsensual sexual ticket (vs. no sexual ticket) significantly better than all the nonsexual outcome measures other than violent (nonsexual) misconduct, as evidenced by the 95% confidence intervals for the AUC values not overlapping. Thus, it appears that the Static-99R does predict general institutional misconduct, but the instrument is still more focused on the prediction of sexual misconduct.

Table 3:

Receiver Operative Characteristic AUC Values for Nonsexual Tickets.

	Nonsexual Ticket Outcome
	Violent	Drug Nonviolent	Nondrug Nonviolent	Any
Predictor Variables	AUC (95% CI)	AUC (95% CI)	AUC (95% CI)	AUC (95% CI)
Age at conviction^a	.64 [.62, .66]	.63 [.61, .66]	.63 [.61, .66]	.64 [.62, .66]
Prior disciplinary ticket count	.54 [.52, .56]	.55 [.52, .58]	.53 [.51, .55]	.53 [.51, .55]
Prior sexual ticket count	.52 [.51, .55]	.53 [.50, .56]	.51 [.49, .53]	.51 [.49, .54]
Sexual arrests	.50 [.48, .52]	.49 [.46, .52]	.51 [.49, .53]	.52 [.50, .54]
Criminal sexual act arrests^b	.52 [.50, .54]	.53 [.50, .56]	.52 [.50, .54]	.52 [.50, .55]
Violent felony arrests	.60 [.58, .62]	.62 [.60, .65]	.57 [.55, .60]	.58 [.56, .60]
Variety of offending types^c	.63 [.61, .65]	.64 [.62, .67]	.61 [.59, .63]	.63 [.60, .65]
Static-99R^d	.66 [.64, .68]	.63 [.61, .66]	.63 [.61, .65]	.64 [.62, .66]

Note. AUC = area under the curve; CI = confidence interval. Variables can be said to be a significant predictor of an outcome if the 95% confidence interval for that variable’s AUC does not include the value of .50. Furthermore, AUC values can be said to differ significantly if their 95% confidence intervals do not overlap.

Incremental Predictive Validity

Finally, Cox regression was used to investigate the ability of the Static-99R to predict both an offender receiving any sexual disciplinary ticket and an offender receiving a nonconsensual sexual disciplinary ticket while controlling for time at risk (calculated as the time from conviction to the time when institutional records were examined and coded) and for other predictors. Before the Cox regressions were estimated, however, the model variables were checked for possible collinearity and multicollinearity. The former was assessed through correlations, while the latter was assessed through auxiliary regression analyses (in which each independent variable was regressed on the others). Results of the analyses showed no evidence of either collinearity or multicollinearity, so for each of the two main sexual ticket outcomes (any and nonconsensual), two models were estimated: (a) one model including all predictor variables other than the Static-99R and (b) one model including all predictor variables and the Static-99R. These second models, therefore, tested the ability of the Static-99R to bring incremental validity to the prediction of receiving a sexual disciplinary ticket (first any and then only nonconsensual) above and beyond other predictors already in the model. Results of the analyses are presented in Table 4 (any sexual disciplinary ticket) and Table 5 (nonconsensual sexual disciplinary ticket).

Table 4:

Cox Survival Analysis Predicting Likelihood of Receiving a Sexual Disciplinary Ticket (N = 3,703).

	Model Without Static-99R		Model With Static-99R
Predictor Variables	β	Exp(β) (95% CI)	β	Exp(β) (95% CI)
Age at conviction (in years)	−0.04***	0.96 [0.95, 0.97]	−0.02**	0.98 [0.96, 0.99]
Prior disciplinary tickets	0.03***	1.03 [1.02, 1.04]	0.02***	1.02 [1.01, 1.03]
Prior sexual ticket count	0.07***	1.07 [1.05, 1.10]	0.06***	1.06 [1.04, 1.09]
Criminal sexual act arrests^a	0.18*	1.20 [1.03, 1.40]	0.11	1.11 [0.95, 1.30]
Violent felony arrests	0.00	1.00 [0.91, 1.10]	−0.03	0.97 [0.88, 1.07]
Variety of offending types^b	0.12**	1.12 [1.04, 1.21]	0.05	1.01 [0.98, 1.14]
Race/ethnicity (Black)	0.11	1.12 [0.85, 1.48]	0.15	1.12 [0.88, 1.53]
Race/ethnicity (Hispanic)	−0.12	0.89 [0.60, 1.31]	−0.06	0.95 [0.64, 1.40]
Static-99R^c			0.21***	1.23 [1.16, 1.31]
Overall model χ²	201.11***		240.33***

Note. CI = confidence interval.

According to New York State penal law, criminal sexual act offenses are nonconsensual oral sex and sodomy offenses. ^bVariety is a count variable of eight types of arrest: assault, burglary, theft, public order, criminal mischief, custody, marijuana, and robbery. ^cIn the same model, the Static-99 yielded an Exp(β) = 1.20 (1.13, 1.28; p ≤ .001).

p ≤ .05. **p ≤ .01. ***p ≤ .001.

Table 5:

Cox Survival Analysis Predicting Likelihood of Receiving a Nonconsensual Sexual Disciplinary Ticket (N = 3,606).

	Model Without Static-99R		Model With Static-99R
Predictor Variables	β	Exp(β) (95% CI)	β	Exp(β) (95% CI)
Age at conviction (in years)	−0.03**	0.97 [0.95, 0.99]	−0.01	1.00 [0.97, 1.02]
Prior disciplinary tickets	0.03***	1.03 [1.02, 1.04]	0.02***	1.02 [1.01, 1.04]
Prior sexual ticket count	0.08***	1.09 [1.06, 1.12]	0.08***	1.08 [1.05, 1.11]
Criminal sexual act arrests^a	0.09	1.09 [0.87, 1.38]	−0.02	0.98 [0.77, 1.25]
Violent felony arrests	0.11	1.11 [0.98, 1.26]	0.07	1.07 [0.94, 1.22]
Variety of offending types^b	0.13*	1.14 [1.02, 1.27]	0.05	1.05 [0.94, 1.18]
Race/ethnicity (Black)	0.17	1.12 [0.78, 1.83]	0.22	1.25 [0.81, 1.91]
Race/ethnicity (Hispanic)	−0.01	0.91 [0.50, 1.65]	−0.01	0.99 [0.55, 1.80]
Static-99R^c			0.27***	1.30 [1.19, 1.43]
Overall model χ²	191.31***		218.60***

Note. CI = confidence interval.

p ≤ .05. **p ≤ .01. ***p ≤ .001.

Any Sexual Disciplinary Ticket

As can be seen in Table 4, the model excluding the Static-99R was significantly predictive of an offender receiving any sexual disciplinary ticket (Model χ² = 201.11; p ≤ .001). Five variables were found to significantly contribute to this model: (a) age at conviction (each year older decreased the hazard of receiving a ticket by 4%), (b) prior number of disciplinary tickets (each additional prior ticket increased the hazard by 3%), (c) prior number of sexual disciplinary tickets (each additional prior ticket increased the hazard by 7%), (d) prior number of criminal sexual act arrests (each additional criminal sexual act arrest increased the hazard by 20%), and (e) variety of offending types (each additional type of offending increased the hazard by 12%). When offender Static-99R scores were entered into the model, the overall model fit chi-square rose to 240.33 (p ≤ .001), a significant increase of 39.22 (p ≤ .001). Thus, including the Static-99R brought a significant increase in the accuracy of predicting whether an offender would receive a sexual disciplinary ticket above and beyond the other variables already in the model, with each additional point on the Static-99R increasing the hazard of an offender receiving a sexual disciplinary ticket by 23%. Furthermore, once the Static-99R was included in the model, the only other three variables that significantly contributed to the prediction of receiving a sexual disciplinary ticket were age at conviction (each year older lowered the hazard of receiving a ticket by 2%), prior number of disciplinary tickets (each additional prior ticket increased the hazard by 2%), and prior number of sexual disciplinary tickets (each additional prior ticket increased the hazard by 6%).

Nonconsensual Sexual Disciplinary Ticket

As can be seen in Table 5, the model excluding the Static-99R was significantly predictive of an offender receiving a nonconsensual sexual disciplinary ticket (Model χ² = 191.31; p ≤ .001). Four variables were found to significantly contribute to this model: (a) age at conviction (each year older decreased the hazard of receiving a nonconsensual ticket by 3%), (b) prior number of disciplinary tickets (each additional prior ticket increased the hazard by 3%), (c) prior number of sexual disciplinary tickets (each additional prior ticket increased the hazard by 9%), and (d) variety of offending types (each additional type of offending increased the hazard by 14%). When offender Static-99R scores were then entered into the model, the overall model fit chi-square rose to 218.60 (p ≤ .001), a significant increase of 27.29 (p ≤ .001). Thus, including the Static-99R brought a significant increase in the accuracy of predicting whether an offender would receive a nonconsensual sexual disciplinary ticket above and beyond the other variables already in the model, with each additional point on the Static-99R increasing the hazard of an offender receiving a sexual disciplinary ticket by 30%. Furthermore, once the Static-99R was included in the model, the only other two variables that significantly contributed to the prediction of receiving a sexual disciplinary ticket were prior number of disciplinary tickets (each additional prior ticket increased the hazard by 2%) and prior number of sexual disciplinary tickets (each additional prior ticket increased the hazard by 8%). Figure 1 displays the logistic rate of offenders receiving sexual misconduct tickets (any and specifically nonconsensual) by the Static-99R score.

Figure 1:

Sexual Misconduct Ticket Rate by Static-99R Score.

Discussion

This study sought to examine the predictive accuracy of the Static-99R for institutional misconduct in a sample of incarcerated adult male sex offenders. The accurate identification of those incarcerated offenders who are at greatest risk of engaging in institutional sexual misconduct is of great value to correctional institutions as they strive to address the sexual victimization of prisoners (and staff), and maintain a safe and secure institutional environment. As expected, the Static-99R was able to predict institutional sexual misconducts overall, as well as institutional sexual misconducts where it was clear the sexual activity was nonconsensual. In multivariate analyses, the Static-99R continued to be a significant predictor of sexual misconducts, even after taking prior misconducts and other relevant historical variables into account. Higher-risk sex offenders are not only of greater concern in terms of treatment, release decisions, and supervision once back in the community, but they are also of concern while incarcerated because they pose a greater risk of institutional sexual misconducts resulting in sexual victimization of other prisoners and the subsequent strains on institutional security and atmosphere.

The institutional misbehavior of offenders who received sexual disciplinary tickets was not limited to just sexual misconduct. Offenders who received sexual disciplinary tickets for any reason and those who received sexual disciplinary tickets for nonconsensual acts both had a significantly higher proportion of offenders who also received violent (nonsexual) disciplinary tickets, drug-related disciplinary tickets, and nonviolent tickets. These results indicate sex offenders engaging in sexual misconduct may resemble general offenders also participating in nonsexual misconducts. That is, the institutional sexual misbehavior of these offenders may also be an indicator of general rule-breaking and antisociality, rather than specifically sexual deviance.

Limitations

The biggest limitations of this study were those of missing data. For example, we did not have information available to classify all the sexual misconducts as either consensual or nonconsensual, and there could be something systematic rather than random as to why approximately one third of the misconducts could not be classified. We also did not have more detailed information available for misconducts to learn more about these incidents. For example, we did not know what proportion of misconducts with nonconsensual acts involved violence or physical injury to the victim, and we do not know about victim characteristics (e.g., age, stature, relationship to the perpetrator). Prospective research where all of the details of institutional misconducts are carefully recorded would shed light on these questions.

Furthermore, as in all follow-up research relying on official records, we do not know about all institutional misconducts. Some consensual sexual behavior undoubtedly took place without being detected by or reported to prison authorities, and some nonconsensual sexual behavior may also have been missed because the victim did not report the incident or there was insufficient evidence to draw a conclusion.

Last, while we had sufficient statistical power to detect a significant association between Static-99/99R scores and institutional sexual misconducts (including misconducts that involved nonconsensual acts), we did not have a sufficiently large number of misconduct cases to examine the probabilistic estimates associated with specific Static-99/99R scores. Aggregation of large-sample results (ideally from geographically and demographically diverse offender populations) would be needed to estimate the likelihood of institutional sexual misconduct (any or specifically nonconsensual) for different scores. Until such analyses can be conducted, practitioners hoping to use the Static-99/99R as an indicator of institutional sexual misconduct likelihood should use the instrument strictly to rank offenders in terms of risk (i.e., the higher an offender’s Static-99/99R score, the more likely the offender is to commit institutional sexual misconduct) and refrain from drawing any conclusions about an offender’s probabilistic likelihood of committing an institutional sexual misconduct. The Static-99/99R norms for likelihood of sexual recidivism in the community should not be used.

Implications and Future Directions

Even with these limitations in mind, however, we believe the findings have important implications for correctional administration. Aggression and sexually inappropriate behavior present safety threats to staff and inmates in correctional facilities and degrade institutional milieu. The need for increased identification, assessment, and prevention of institutional misconduct is critical, as sexual victimization within prisons not only affects the inmates’ mental and physical well-being while incarcerated, but may also affect their ability to adjust to community life (Wolff, Shi, Blitz, & Seigel, 2007). As noted by Wolff et al. (2007)

If the goal is to reduce sexual victimization inside prisons (as suggested by the Prison Rape Elimination Act), action is required by prison officials and researchers to identify those at elevated risk, to develop effective placement strategies that minimize the proximity of inmates who have predatory tendencies to those at risk of victimization, to accurately and reliably measure the prevalence of sexual victimization, and to train officers and inmates on the meaning and practice of “zero tolerance.” (p. 554)

Although a wealth of research has been dedicated to assessing risk of recidivism for sex offenders in the community, few studies have examined the ability of risk-assessment instruments to assess risk for institutional misconduct, especially sexually behaviors. Results of the current study suggest that sexually inappropriate behavior in prison as well as institutional aggression may not be all that different from that demonstrated in the community. In fact, it appears the Static-99/99R may be of use to correctional officials looking to predict institutional misconduct, particularly sexual misconduct, and to develop effective prevention and risk-management strategies to address this security risk. Given that numerous correctional facilities already complete the Static-99/99R for programmatic purposes (i.e., to assign sex offender inmates into appropriate treatment programs), results of the current study suggest that this assessment may also be beneficial and cost-effective for security classification of sex offenders. It may also assist correctional facilities as they implement proactive measures to prevent sexually abusive inmate-on-inmate behaviors, as suggested by the Prison Rape Elimination Act (2003).

Overall, enhanced knowledge regarding the risk factors for inmates who may engage in sexual misconduct as well as tools to identify this risky behavior is critical. Identifying which inmates are most likely to engage in sexually aggressive behavior while incarcerated is important for correctional staff and treatment providers so that risk-management and security plans can be implemented effectively. Results of this study suggest that the Static-99/99R may be effective tools to achieve this purpose.

Footnotes

Authors’ Note:

Data for this project were furnished to the researchers by the New York State Office of Mental Health (NYS OMH). However, the NYS OMH was not responsible for the methods of statistical analysis or the conclusions reached. Any opinions and suggestions within the article are those of the authors alone, and not representative of the views of the NYS OMH.

Jeffrey C. Sandler does research for the New York State Office of Mental Health through the Research Foundation for Mental Hygiene. His research has addressed the impact of public policies designed to manage sex offenders, sex offender risk assessment, and issues surrounding female sex offenders.

Naomi J. Freeman received a PhD in criminal justice from the University at Albany and an MA in forensic psychology from Castleton State College. She is an adjunct professor at the School of Criminal Justice, University at Albany, and she currently serves as the deputy director for the Division of Forensic Services at the New York State Office of Mental Health, which oversees and implements New York State’s civil management initiative as well as State mental health services to individuals involved with the criminal justice system.

Paul Farrell currently works for the New York State Division of Criminal Justice Services. He received his MA and BA from the University at Albany.

Michael C. Seto is the director of forensic rehabilitation research at the Royal Ottawa Health Care Group and cross-appointed to the University of Toronto, Ryerson University, Carleton University, and University of Ottawa. He has published and presented extensively on paraphilias, sexual offending, and risk assessment. He is the author of two books, Pedophilia and Sexual Offending Against Children: Theory, Assessment, and Intervention and Internet Sex Offenders, both published by the American Psychological Association.

References

Beck

A. J.

Johnson

(2012). Sexual victimization reported by former state prisoners, 2008 (Research Report No. NCJ 237363). Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

Buffington-Vollum

Edens

J. F.

Johnson

D. W.

Johnson

J. K.

(2002). Psychopathy as a predictor of institutional misbehavior among sex offenders: A prospective replication. Criminal Justice and Behavior, 29, 497-511. doi:10.1177/009385402236730

Caperton

J. D.

Edens

J. F.

Johnson

J. K.

(2004). Predicting sex offender institutional adjustment and treatment compliance using the Personality Assessment Inventory. Psychological Assessment, 16, 187-191. doi:10.1037/1040-3590.16.2.187

Cunningham

M. D.

Sorensen

J. R.

Vigen

M. P.

Woods

S. O.

(2011). Correlates and actuarial models of assaultive prison misconduct among violence-predicted capital offenders. Criminal Justice and Behavior, 38, 5-25. doi:10.1177/0093854810384830

Edens

J. F.

Buffington-Vollum

Colwell

K. W.

Johnson

D. W.

Johnson

J. K.

(2002). Psychopathy and institutional misbehavior among incarcerated sex offenders: A comparison of the Psychopathy Checklist-Revised and the Personality Assessment Inventory. International Journal of Forensic Mental Health, 1, 49-58. doi:10.1080/14999013.2002.10471160

Freeman

N. J.

Sandler

J. C.

(2010). The Adam Walsh Act: A false sense of security or an effective public policy initiative? Criminal Justice Policy Review, 21, 31-49. doi:10.1177/0887403409338565

Hanson

R. K.

Bussière

M. T.

(1998). Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology, 66, 348-362. doi:10.1037/0022-006X.66.2.348

Hanson

R. K.

Helmus

Thornton

(2010). Predicting recidivism amongst sexual offenders: A multi-site study of the Static-2002. Law and Human Behavior, 34, 198-211. doi:10.1007/s10979-009-9180-1

Hanson

R. K.

Morton-Bourgon

(2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73, 1154-1163. doi:10.1037/0022-006X.73.6.1154

10.

Hanson

R. K.

Morton-Bourgon

K. E.

(2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment, 21, 1-21. doi:10.1037/a0014421

11.

Hanson

R. K.

Thornton

(1999). Improving risk assessments for sex offenders (User Report 99-02). Ottawa, Ontario: Department of the Solicitor General of Canada.

12.

Hare

R. D.

(1991). The Hare Psychopathy Checklist-Revised. Toronto, Ontario, Canada: Multi-Health Systems.

13.

Harris

Phenix

Hanson

R. K.

Thornton

(2003). STATIC-99 coding rules revised 2003. Ottawa, Ontario: Department of the Solicitor General of Canada.

14.

Hastings

M. E.

Krishnan

Tangney

J. P.

Stuewig

(2011). Predictive and incremental validity of the violence risk appraisal guide scores for male and female jail inmates. Psychological Assessment, 23, 174-183. doi:10.1037/a0021290

15.

McDermott

B. E.

Dualan

I. V.

Scott

C. L.

(2011). The predictive ability of the Classification of Violence Risk (COVR) in a forensic psychiatric hospital. Psychiatric Services, 62, 430-433. doi:10.1176/appi.ps.62.4.430

16.

Monahan

Steadman

H. J.

Appelbaum

P. S.

Grisso

Mulvey

E. P.

Roth

L. H.

Silver

(2005). Classification of violence risk (COVR). Lutz, FL: Psychological Assessment Resources.

17.

Morey

L. C.

(1991). The Personality Assessment Inventory: Professional manual. Odessa, FL: Psychological Assessment Resources.

18.

Morris

R. G.

Longmire

D. R.

Buffington-Vollum

Vollum

(2010). Institutional misconduct and differential parole eligibility among capital inmates. Criminal Justice and Behavior, 37, 417-438. doi:10.1177/0093854810361672

19.

Newbury

Shuker

(2012). Personality Assessment Inventory (PAI) profiles of offenders and their relationship to institutional misconduct and risk of reconviction. Journal of Personality Assessment, 94, 586-592. doi:10.1080/00223891.2012.669220

20.

Prison Rape Elimination Act, 42 U.S.C. §15601 (2003).

21.

Quinsey

V. L.

Harris

G. T.

Rice

M. E.

Cormier

C. A.

(2006). Violent offenders: Appraising and managing risk (2nd ed.). Washington, DC: American Psychological Association.

22.

Sex Offender Management and Treatment Act, Mental Hygiene Law, §2, Article 10 (2007).

23.

Steiner

Wooldredge

(2009). Implications of different outcome measures for an understanding of inmate misconduct. Crime & Delinquency. Advance online publication. doi:10.1177/0011128709335151

24.

Vitacco

M. J.

Gonsalves

Tomony

Smith

B. E. R.

Lishner

D. A.

(2012). Can standardized measures of risk predict inpatient violence? Combining static and dynamic variables to improve accuracy. Criminal Justice and Behavior, 39, 589-606. doi:10.1177/0093854812436786

25.

Wolff

Shi

Blitz

Seigel

(2007). Understanding sexual victimization inside prisons: Factors that predict risk. Criminology & Public Policy, 6, 535-564. doi:10.1111/j.1745-9133.2007.00452.x