Risk Assessment: New Developments to Think About

Risk Assessment: New Developments to Think About Texas Council on Sex Offender Treatment 26 th Annual Conference Galveston Texas March 7, 2018 Andrew J. R. Harris andrew@offenderrisk.com Three Generations of Risk Assessment Bonta (1996) First Generation = Clinical Judgement Unstructured, Non-replicable, Personal Discretion Based on experience and level of knowledge of the literature Non-standard (even within same institution) Level of prediction little better than chance Second Generation = Actuarial Assessment Static, Actuarial, Structured, Replicable, Less open to Interpretation Based on factors empirically related to recidivism Standardized assessment, Static - Cannot measure change Moderate Levels of prediction, ROC s upper 60 s to lower 70 s Third Generation = Dynamic Assessment Based on factors empirically related to recidivism Standardized assessment, Measures change Actuarial measure with dynamic factors 2 Prediction of sexual recidivism Measures Designed for Sexual Recidivism d (95% CI) N (k) Empirical Actuarial.67 (.63-.72) 24,089 (81) Structured Judgement.46 (.29-.62) 1,131 (6) Unstructured.42 (.32-.51) 6,456 (11) Hanson & Morton-Bourgon, 2007 3 1

Why Assess Risk? Promoting public safety Routine interventions Targeting scarce resources Officer time Treatment Exceptional measures 4 Three Linked Research Projects The First: Meta-analytic Reviews R. Karl Hanson and Colleagues Public Safety Canada Hanson & Bussière, 1996, 1998 Static risk factors Hanson & Morton-Bourgon, 2004, 2005 Promising Stable risk factors Hanson & Morton-Bourgon, 2007 Risk assessments 5 Three Linked Research Projects The Second: Dynamic Predictors 1998, 2000 File review study Interviewed the supervising officer n = 208 community sexual recidivists n = 201 community sexual non-recidivists Canada-wide study Federal parole and Provincial probation Provided a strong list of likely STABLE predictors 6 2

Three Linked Research Projects The Third: Dynamic Supervision 2000-2007 Follow 1000+ in-community sex offenders - for a 43 month period - prospective design Multiple jurisdictions Continuous intake - consecutive new cases Trained officers submitting data American participants = Alaska & Iowa 7 A Directed Program of Research Hanson & Bussière Meta-analysis 1996 Hanson & Harris Dynamic Predictors 1998 Hanson & Harris Dynamic Supervision 2001-2007 RRASOR, 1997 SONAR, 2000 STABLE- 2000 & ACUTE- 2000 STABLE- 2007 & ACUTE- 2007 Dynamic Supervision of Sexual Offenders STATIC-99(R) STATIC-2002(R) 8 Static, Stable, and Acute Risk Factors Definitions Static Non-changeable life factors that relate to risk for sexual recidivism, generally historical in nature Stable Personality characteristics, skill deficits, and learned behaviours that relate to risk for sexual recidivism that may be changed through intervention Acute Risk factors of short or unstable temporal duration that can change rapidly, generally as a result of environmental or intra-personal conditions 9 3

Implied Time Lines for Each Level of Assessment STATIC Now till forever STABLE One year either side of today ACUTE Since I saw him last STATIC Development Team Kelly Babchishin R. Karl Hanson Andrew J.R. Yolanda Fernandez Harris L. Maaike Helmus Amy Phenix 11 David Thornton Risk Assessment Pyramid ACUTE STABLE Static 4

Risk Assessment Pyramid ACUTE STABLE Actuarial Risk Factors Static Risk Assessment Pyramid ACUTE STABLE Actuarial Risk Factors Static Redemption Assessment NOTE: New Scoring Manual 2016 Go to: static99.org 5

Common Language for Risk Assessment R. Karl Hanson & Andrew Brankley With Thanks to Maaike Helmus! Why do we need a common language metric? Get away from categorical levels of measurement (High, Moderate, Low) that mean something different to everyone Base ideas about risk in the data Allow for better communication and more precision A standardized metric can be replicated for all tests! Common for similar scales to place offenders in different nominal categories Barbaree, Langton, & Peackock, 2006; Jung, Pham, & Ennis, 2013; Mills & Kroner, 2006 Developing a Common Language for Risk Assessment Justice Center. (2014). A common language for risk assessment: Experts convene in Washington. Retrieved from http://csgjusticecenter.org/reentry/posts/a-common-languagefor- riskassessments-experts-convene-in-washington/ 6

Risk Categories That Work! Describes people (not risk scales) Psychologically meaningful characteristics Tell us what to do Linked to realistic options for action Evidence-based, scientifically credible Universal applicable to all risk scales Simple Common Professional Language Easy to implement Across jurisdictions/scales/offenders Meaningful (Perceptible) Differences Between Categories Descriptive Summary Needs: Criminogenic & Non-criminogenic Personal and social resources Correctional Strategies & Responses Interventions, supervision, custody Outcomes Base rate of re-offending Expected outcomes with appropriate services Council of State Government Justice Center Standardized Risk Levels Level I Level II Level III Level IV Level V 7

Level I Descriptive Summary: Needs: None or transitory Resources: Clear personal and social resources Reoffending Base Rate: Same as non-offender community at large (e.g., <5%) Sex offenders: Similar to non-sex offenders committing a sex offense (< 2%) Correctional Strategies: Human services: Largely unnecessary Community Supervision: Minimal (not necessary for compliance) Custody: Counterproductive Outcomes: Short-term: No change in probability of re-offending Level II Descriptive Summary: Needs: Few, if any, more acute than chronic. Resources: Clear personal and social resources Reoffending Base Rate: Higher than Level I but lower than typical offender Correctional Strategies: Human services: Brief interventions Community Supervision: Simple case management Custody: May be productive for crisis management but unnecessary long-term Outcomes: Short-term: Probability of re-offending reduces quickly to non-offender levels (Level I) Level III Descriptive Summary: Needs: Typical for offenders; some severe or multiple less chronic needs Resources: Some, sporadic use Reoffending Base Rate: Average base rate Correctional Strategies: Human services: Structured services target criminogenic needs over months (e.g. ~ 100-200 hours of service) Community Supervision: Change-focused supervision Custody: May support short-term risk management Outcomes Short-term: Risk can significantly reduce with intervention BUT still higher than community at large (Level II) 8

Level IV Descriptive Summary: Needs: Multiple, moderate to severe, more chronic. Resources: Few, sporadic prosocial use Reoffending Base Rate: Higher than typical offender Correctional Strategies: Human services: Structured comprehensive services with community follow-ups and supports (e.g. ~ 300+ hours) Community Supervision: Intensive and change-focused Custody: May be productive for short-term risk management and beginning treatment Outcomes: Short-term: Risk can with appropriate strategies BUT still average for offender population at large. Level V Descriptive Summary: Needs: Multiple, chronic, severe, and entrenched Resources: Few; used for procriminal pursuits Reoffending Base Rate: 85% (Top 5% of offenders) Not currently possible to empirically identify this group with sex offenders Correctional Strategies: Human services: Structured, comprehensive (e.g. ~ 300+ hours) Community Supervision: Long-term supervision with risk management focus and gradual movement to change focus Custody: Incapacitation for community safety Outcomes Short-term: Risk still significantly higher than average even with appropriate long-term correctional strategies Calculating Risk Categories Recidivism 1.00 Risk Scores Distribution Base Rate ~.40 Median 0.00 9

Upper Boundary ~5% Recidivism Calculating Risk Categories Lower Boundary ~85% Recidivism 1.00 Cat I ~0.85 Cat V ~0.40 ~0.05 0.00 Category I: Risk Score Cutoff Category V: Risk Score Cutoff Calculating Risk Categories Boundaries = Appropriate Treatment Effect Size Odds Ratio: ±1.44 Cat III 1.00 ~0.85 Cat II Cat IV ~0.40 Cat I Cat V ~0.05 0.00 Category III: Risk Score Cutoffs When is a Sex Offender, No Longer a Sex Offender? Hanson, R. K., Harris, A. J. R., Letourneau, E. Helmus, L. M., & Thornton, D. (2017). Reductions in risk based on time offense free in the community: Once a sex offender, not always a sex offender. In Press: Psychology, Public Policy, and Law See also Hanson, R. K., Harris, A. J. R., Helmus, L. M., & Thornton, D. (2014). High-risk sex offenders may not be high risk forever. Journal of Interpersonal Violence, doi: 10.1177/0886260514526062 Harris, A. J. R., & Hanson, R. K. (2004). Sexual offender recidivism: A simple question. [Corrections User Report No 2004-01]. Ottawa, Canada: Public Safety and Emergency Preparedness Canada. Retrieved from http://www.publicsafety.gc.ca/res/cor/rep/2004-03-se-off-eng.aspx. 10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 All three Groups (Full Sample: Number of guys followed up by Years) 4500 4000 3500 3000 2500 Routine Tx Need HR/HN 2000 1500 1000 500 0 Sexual Recidivism by Time Interval (68% CI) No recidivism after 20.5 Years How much community supervision is enough? What is a sufficiently low risk (Probability of reoffence) that the guy can be in the community without further surveillance? Zero Percent risk (Is this ever possible or desirable?) Same risk as men in the general non-offender population who spontaneously emit a sexual offence? Same risk as men in the non-sexual-offender offender group who spontaneously emit a sexual offence? The idea of an efficiency trade-off How many low risk guys are we willing to pay to follow to catch the one guy who will re-offend? 11

Alfred Blumstein & Kiminori Nakamura Blumstein Born 1930 Leading Criminologist Professor of Urban Systems and Operations Research at Carnegie Mellon University Idea of Redemption Research Nakamura Grad Student Blumstein, A. & Nakamura, K. (2009). Redemption in the presence of widespread criminal background checks. Criminology, 47, 327-359. doi: 10.1111/j.1745-9125.2009.00155.x 0.045 0.040 0.035 0.030 0.025 0.020 0.015 0.010 0.005 High Group Logistic Regression With Standard Error Bars (Percent Recidivism by Time Interval [6 Month gates ]) 0.000 5 Years 10 Years 15 Years 20 Years Moderate Group Logistic Regression With Standard Error Bars (Percent Recidivism by Time Interval [6 Month gates ]) 0.045 0.040 0.035 0.030 0.025 0.020 0.015 0.010 0.005 0.000 5 Years 10 Years 15 Years 20 Years 12

0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5 21.5 22.5 23.5 24.5 25.5 Low Group Logistic Regression With Standard Error Bars (Percent Recidivism by Time Interval [6 Month gates ]) Logistic Regression was not Significant for the low guys (p =.12) 0.045 0.040 0.035 0.030 0.025 0.020 0.015 0.010 0.005 0.000 5 Years 10 Years 15 Years 20 Years STATIC-99R Categories Logistic Regression (Percent Recidivism by Time Interval (Years) [6 Month gates ]) 0.035 0.030 0.025 0.020 0.015 High Moderate Low 0.010 0.005 0.000 Conclusions Like regular offenders, SO s are less likely to re-offend the longer they remain offencefree in the community Eventually, they are less likely to reoffend sexually than a non-sexual offender is to commit an out of the blue sexual offence These findings should be considered in the re-assessment of men who have spent protracted periods in the community Jurisdictions should take these findings into account with regard to the Risk Principle and the utilization of public resources 13

Hanson, Harris, Letourneau, Helmus, & Thornton (2017) Reductions in Risk Based on Time Offense Free in the Community: Once a Sexual Offender, Not Always a Sexual Offender Journal of Psychology, Public Policy, and the Law (In Press) Basically the same samples More sophisticated analysis Recidivism whole sample (sexual): 9.1% @ 5 years; 13.3% @ 10 years; 16.2% @ 15 years; 18.2% @ 20 years; 18.5% @ 25 years Five Year Hazard Rates: 9.1% @ 5 years; 4.1% Btwn years 5 to 10; 2.9% Btwn years 10 to 15; 2.0% Btwn years 15 to 20; 0.3% Btwn years 20 to 25 Note: each year offence free was associated with a 12% decrease in the odds of recidivism No interaction between time free and Static- 99R scores hence, relative risk reductions were constant across risk levels. Desistance Rates set a (0.0019 / 6 Mth interval) = 2% out of the blue sexual offences Recidivism By Risk Level: (Unadjusted for follow-up time or sample type) Level 1 = 1.9% Level 2 = 3.6% Level 3 = 7.6% Level 4a = 14.7 Level 4b = 27.5% What happens when the guy has new nonsexual offences nonsexual recidivism resets the individual s relative risk to what it would have been 3.3 years previously 14

What causes this decline? People make better life choices over time Anti-social choices sometimes have very negative consequences (Pun+) vs. Pro-social choices more often have positive outcomes (R+) {Skinner??} Age one of the big themes Crime is a young man s game Treatment (Tx) and Supervision frailty - those with the greatest propensity for recidivism do it early in the follow-up and effectively remove themselves from follow-up STABLE-2007 STABLE-2007 1 Significant Social Influences 2 Capacity for Relationship Stability 3 Emotional ID with Children 4 Hostility Toward Women 5 General Social Rejection 6 Lack of Concern for Others 7 Impulsive 8 Poor Problem Solving Skills 9 Negative Emotionality 10 Sex Drive Sexual Preoccupation 11 Sex as Coping 12 Deviant Sexual Preference 13 Co-operation with Supervision 44 Think of Your Normal Interview Significant Social Influences* Who do you hang around with? What do you do with Joe? Intimacy Deficits* Anybody special in your life? How do you get along with women? Children? Would you say you are a loner? Who do you care most about in the world? General Self-regulation* Ever play sports? How often in Emerg? Money problems? Housing problems? Quit jobs a lot? What sort of things cause you problems in your life? What do you do about them? What do you think of (the guy who caused the problem)? And do you meet people like that a 45 lot? 15

Think of Your Normal Interview Sexual Self-regulation Sexual outlets? Habits? How often? How often do you think about sex? Feel pressured by sexual needs? Ever use sexual outlets to change your mood or make yourself feel better? You mentioned a child in your list of friends. You seem to have a history of people making allegations against you. Co-operation with Supervision* (Your call) You can re-order the STABLE interview to 46 suit your situation. NEW DYNAMIC META!!! Van der Berg, Smid, Schepers et al., (2017) Psychological Assessment Dynamic measures have small to moderate predictive ability Sexual = 41 studies, 22 unique samples, N = 10,368 Adult Males Sexual Offenders Follow-up = 67 Months Samples from Australia, Canada, United States, New Zealand, United Kingdom published between 2000 and 2014 New Dynamic Meta - Answers Dynamic tools in terms of Andrews & Bonta s (2010) risk and need principles, to be a useful tool for improving sex offender treatment. (p. 1) Dynamic tools significantly contribute to the prediction of sexual, violent, and other recidivism Incremental validity of dynamic over static risk assessment was established for all outcome measures. (sex, violence, other) Change scores significantly predicted all three types of recidivism Dynamic tools should be used to guide offenders toward appropriate treatment approaches and levels, focus treatment on criminogenic needs, and to evaluate the success of treatment programs. 16

NOTE: New Scoring Manual 2016 Go to: static99.org Risk Assessment Pyramid ACUTE Criminogenic Risk/Needs Actuarial Risk Factors STABLE Static Redemption Assessment Risk Assessment Pyramid ACUTE Criminogenic Risk/Needs Actuarial Risk Factors STABLE Static Treatment Needs Redemption Assessment 17

Who Can I Use This Stuff With? Population STATIC-99 STABLE-2007 ACUTE-2007 Adult Male Sexual Offenders Adult male offenders offencefree in the community for many years Juvenile offenders aged 16 & 17 With caveats With Caution With Caution With Caution Juvenile offenders less than 16 years Adult female offenders Research use only Research use only Research use only 52 PROTECTIVE FACTORS Structured Assessment of Protective Factors Sex Offense version (SAPROF-SO) Willis, Kelley, & Thornton (2017) Structured Professional Judgement Rates the presence and strength of positive protective factors Helpful for Tx and release Info: Gwenda Willis g.willis@auckland.ac.nz Sex & Violence Risk and General Recidivism Risk Sex/Violence Score (Sum of four factors) Score General Recidivism Score (Sum of all seven factors) Score Victim Access Copy Score Over Hostility Copy Score Over Sexual Pre-occupation Rejection of Supervision Copy Score Over Copy Score Over Emotional Collapse Change in Social Supports Substance Abuse Sex/Violence Total General Recidivism Total 18

NOTE: New Scoring Manual 2017 Go to: static99.org Risk Assessment Pyramid Short Acting Risk Factors ACUTE Criminogenic Risk/Needs Actuarial Risk Factors STABLE Static Treatment Needs Redemption Assessment Risk Assessment Pyramid Short Acting Risk Factors ACUTE Crisis/Relapse Avoidance Criminogenic Risk/Needs Actuarial Risk Factors STABLE Static Treatment Needs Redemption Assessment 19

Kelly Babchishin s Thesis (2013) Sex Offenders do Change on Risk-relevant Propensities Evident from a Longitudinal Study of The ACUTE-2007 Study Number One To better understand the measurement properties and construct validity of the ACUTE-2007 317 adult male SO s on community supervision At least three assessments Results Study Number One Explored the factor structure and construct validity Risk assessment instruments sample several risk-relevant items but include few because more doesn t improve predictive validity (more items, sampling theory each adds a slice of error) ACUTE-2007 correlates well with other risk measures Multiple assessments predict better Here is the Kicker!!!!! The construct being measured does not change across time and the items perform the same across time Hence, observed changes can be assumed to reflect true change on risk-relevant propensities (p. 108) 20

Study Number Two 1. To examine the average rate of change for SO s on the ACUTE-2007 2. To examine the extent to which individual SO s vary in their rates of change 3. To examine the relationship between initial scores on the ACUTE-2007 and the rate of change Study 2 Method & Findings Used Hierarchical Linear Modeling (HLM) 1. Rate of Change SO s found to decrease on risk-relevant propensities over time (p. 179) 2. Extrafamilial offenders had lower rates of change than Intrafamilial (Extrafamilial higher risk this opens gates to further research) 3. Rate of change variance Those with high initial scores showed greater decline than those with lower initial scores (regression to the mean change due to natural variation as opposed to true change) + Total scores predict recidivism but, reassessment scores predicted sexual recidivism and initial scores predicted violent recidivism 4. Need larger samples for better findings Future Work Need to understand better why individual items work relate items to constructs (antisociality, impulsivity) While most offenders decreased their risk over time a substantial proportion did not (and some actually increased) Why?? And how do we effect these guys? What is the relationship between observed change and treatment? (dosage, targets, quality of, therapist engagement) While there was change high risk guys remained problematic Data indicates supervision officers may need more training on helping with collapse issues (poor coping skills) Officers trained in effective change promotion may well effect change more readily than those who act as case managers How often do you have to re-assess to reliably measure change How long are these risk assessments good for (shelf life) 21

For those who like to read Harris, A. J. R., & Hanson, R. K. (2010). Clinical, actuarial, and dynamic risk assessment of sexual offenders: Why do they keep changing things? Journal of Sexual Aggression, 16(3), 296-310. Q s?? 22