Research Reports

Federal Offender Population Movement: A Study of Minimum-security Placements

---

An operational study prepared for the
Executive Committee of the
Correctional Service of Canada

Laurence L. Motiuk,
Ralph Serin,
Fred Luciani,
Ben Vuong

Research Branch
Correctional Service Canada

March 2001

This report is also available in French. Ce rapport est également disponible en français. Veuillez vous adresser à la direction de la recherche, Service Correctionnel du Canada, 340 avenue Laurier ouest, Ottawa (Ontario) K1A 0P9. Should additional copies be required they can be obtained from the Research Branch, Correctional Service of Canada, 340 Laurier Ave., West, Ottawa, Ontario, K1A 0P9

---

EXECUTIVE SUMMARY

Background

The Federal Offender Population Movement study was designed to improve our understanding of the movement of federally sentenced men to and from minimum-security in terms of classification policies, procedures and practices. Classification data were derived from the Offender Management System (OMS), Criminal Justice Information Library (CJIL), specialized research bases, field surveys, and exercises involving operational representatives. Operational case file information was drawn over an 18-month review period extending from April 1, 1999 to September 30, 2000. Among the more salient findings are the following:

Movement to Minimum-security

• Custody Rating Scale (CRS, initial security level placement) and Security Reclassification Scale (SRS, subsequent security level placement) completion rates for the review period were approximately 90% of eligible files. The results suggest additional attention should be paid to CRS completions for all re-admissions and generally all security reviews requiring an SRS completion.
• The CRS national concordance (classification derived from objective instrument and decision converge) rate was 76.5% and the SRS rate was 81% (or 84.4% when the discretionary override rule was applied).
• Variations in regional concordance rates were evident suggesting other factors unique to each region having an impact on Offender Security Level (OSL) decision-making.
• Approximately three-fifths of the minimum-security stock population were direct placements and two-fifths were indirect transfers from higher security levels.
• Twenty-five percent of direct placements and one-fifth of indirect transfers to minimum-security were overrides of the CRS or SRS, respectively.

Classification Tools

• Modifications to the SRS were identified that would limit but not totally exclude offenders with an escape history from receiving a scale rating of minimum-security.
• There was moderate inter-rater reliability using CRS.

Overrides

• A review of case files concluded that it was difficult to identify the reasons for overriding the CRS or SRS.
• 30% of override reasons found in a randomly selected sample of CRS files were deemed inappropriate.
• A survey of field opinion about the CRS indicated a wide spread desire for additional or refresher training.
• The CRS was re-applied to a random sample of current files by three operational representatives resulting in modest inter-rater reliability outcomes.

Movement from Minimum-Security

• The national escape rate was 4 per 100 offenders; directly placed offenders were less likely to escape than those indirectly transferred (from higher security) to minimum-security. Offenders whose CRS and SRS scale ratings were concordant with Offender Security Level (OSL) decisions were found to have escaped less often than those over-ridden to minimum-security.
• Clear variations in escape rates exist between the regions that reflect differential transfer source and override practices.
• The national rate of return to higher security was 17.3 per 100 offenders. Relative to concordant decisions, override decisions resulted in twice the return rate to increased security.
• Total CRS, SIR scores and a number of individual scale items, including SRS items, were identified that distinguish offenders at minimum-security who were successfully released, those who escaped, and those who were returned to higher security.

Discussion

CRS and SRS completion rates are good, but additional gains are possible. The CRS completion rate, particularly for re-admissions, could be improved while the SRS rate is close to optimal. The policy, procedural and training suggestions contained in this report could improve scale completion and concordance rates and lead to a reduction in rates of escape and return to higher security.

Override use continues to be frequent and seen to require development of policy and practice guidelines. The apparent need for training (either initial or refresher), as reflected by a strong consensus for additional training among surveyed staff, and quality assurance (monitoring completion and accuracy of the scales) would optimize classification processes and outcomes.

Many offenders who were admitted to minimum-security were not rated minimum on either the CRS or the SRS. On the other hand, even more offenders rated minimum were overridden to higher levels of security. In either case, there would appear to be substantial benefits in identifying factors that distinguish between successful and unsuccessful overrides and in re-examining the many offenders rated minimum-security who were overridden to higher security levels.

Analyses demonstrate that the CRS and SRS scales can be adjusted to target offenders with previous escape history in order to limit their admission to minimum-security. Also, other factors were identified that are related to escapes from minimum-security.

As a general rule, regions that rely more heavily on indirect transfers from higher security and override the classification scales had higher escape rates. Relative to concordant cases, overrides of the scales were more likely to be returned to higher security, by a factor of two to one.

This comprehensive and integrated investigation of minimum-security population movement provided a number of insights. In particular, information regarding security classification practices, offender movement decision-making, the use of overrides, the functioning of objective scales, regional and national offender profiles, and staff opinions on classification issues. Further consolidation of research in these areas should yield additional gains in both our understanding about security placements and outcomes. Direction in the areas of policy and training should address concerns regarding the use of overrides, improve the reliability of the classification scales and yield improvements in terms of classification decisions and subsequent outcomes.

TABLE OF CONTENTS

• Executive Summary
• Table of Contents
• Introduction
• Format of the Report
• Review Period
• Samples
• Data Sources
• 1. POPULATION MOVEMENT TO MINIMUM-SECURITY
• Introduction
  1. Custody Rating Scale and Security Reclassification Scale Completion Rates
  2. National and Regional Custody Rating Scale and Security Reclassification Scale Offender Profiles
  3. Custody Rating Scale, Security Reclassification Scale and Offender Security Level Concordance, Override and Scale Yields
  4. Override Impacts on Minimum-security Profiles
• 2. ASSESSING SECURITY CLASSIFICATION TOOLS AND SCORING OPTIONS
• Introduction
  1. Re-Application of the Custody Rating Scale
  2. Custody Rating Scale User Survey Findings
  3. Factors Related to Escape, Return to Higher Security and Successful Release from Minimum-Security
• 3. EXAMINING OVERRIDES
• Introduction
  1. Operational Team Review of a Sample of Override Reasons
  2. Reasons for Overriding the Custody Rating Scale and Security Reclassification Scale
• 4. MOVEMENT OUT OF MINIMUM-SECURITY
• Introduction
  1. Release from Minimum-security
  2. Adjusting Custody Rating Scale and Security Reclassification Scale Security Level Cut-off Values: Reducing the Number of Escape Offenders in the Minimum-Security Yield
  3. Effects of the Security Reclassification Scale Scoring Adjustments on Offenders with Serious Escape History
  4. National and Regional Rates of Escape by Admission Source and Concordant Status
  5. National and Regional Rates of Return to Higher Security
• CONCLUSIONS AND RECOMMENDATIONS
• APPENDICES

LIST OF TABLES

• Table 1-A: National and Regional Custody Rating Scale Completion Rates by Admission Type
• Table 1-B: National and Regional Security Reclassification Scale Completion Rates
• Table 1-C: National and Regional Custody Rating Scale and Security Reclassification Scale Total Mean Scores
• Table 1-D: Custody Rating Scale Concordance and Override Rate
• Table 1-E: Security Reclassification Scale Concordance and Override Rate
• Table 1-F: Combined Custody Rating Scale and Security Reclassification Scale Yields and Offender Security Level Distribution
• Table 1-G: Regional Distribution of Offenders Rated and Assigned to Minimum-Security
• Table 2-A: Re-application of the Custody Rating Scale
• Table 2-B: Factors Relating to Escape, Returns to Higher Security and Successful Releases from Minimum-Security
• Table 2-C: Relationship between Classification Scales and Outcome (point biserial correlation)
• Table 3-A: Reasons for Overriding the Custody Rating Scale
• Table 3-B: Reasons for Overriding the Security Reclassification Scale
• Table 4-A: National and Regional Releases from Minimum-Security
• Table 4-B: Impact of Different Cut-off Scores for the Custody Rating Scale and the Security Reclassification Scale
• Table 4-C: Security Reclassification Scale Yield by Security Rating: Security Reclassification Scale Models
• Table 4-D: Security Reclassification Scale and Offender Security Level Concordance and Overrides by Security Reclassification Models
• Table 4-E: Rates of Escape Per 100 Offenders by Admission Type and Concordant Status
• Table 4-F: Rates of Return Per 100 Offenders from Minimum-security to Increased Security

Introduction

Minimum-security facilities play an integral role in advancing a number of Correctional Service of Canada (CSC) objectives including the assignment of offenders to least restrictive measures of confinement, minimizing the negative effects of incarceration while preparing offenders for release, and providing them opportunities for safe release. In the last decade, the minimum-security population has doubled as the result of steady growth in custodial populations, policy initiatives, improved selection and infrastructure, and expanded program opportunities at minimum-security. Concomitant with the growing emphasis on minimum-security facilities is the need to continue to explore and improve objective assessment methods, to monitor and track offender movement and to identify conditions that relate to successful institutional adjustment and safe release.

This report addresses operational issues relating to both the initial penitentiary placement and reclassification transfer of federally sentenced offenders to minimum-security, the general use of classification tools in the selection of these offenders and a review of the profile of offenders who are assigned and discharged from minimum facilities. The report provides an analytical framework to examine security placement issues. Case file data was drawn from the Offender Management System (OMS) and other sources, representing both general custody population and specifically the minimum-security male offender population incarcerated between April 1, 1999 to September 30, 2000. The focus of the review is on the flow of offenders to and from minimum-security, the use of overrides, and the impact of these decisions.

Format of the Report

Although detailed, the report is not intended to be an exhaustive examination of the intricate nature of security classification procedures, minimum-security assessment or population movement. Rather it targets specific issues or concerns identified in the Terms of Reference ("Study of Population Movements at Minimum-security"). In this manner, it represents an opportunity to integrate research into a discussion paper and to identify a work plan for subsequent investigations.

Data sources are referenced and computation exercises are explained. Unless otherwise noted, data analysis is performed by inspection. Where deemed appropriate statistical methods or graphic illustration are employed to examine data but, in such instances, an explanation or working interpretation of these methods is provided. Data highlights are summarized and where the evidence is persuasive, conclusions or recommendations are provided. Appendices provide fuller detail or finer break down of the data summaries explored in the body of the report.

In order to provide clarity to the results, the report has been organized into four sections: movement to minimum-security; a review of instruments; overrides; and outcomes.

Review Period

Offender samples used in the analyses are from a review period beginning April 1, 1999 and extending to September 30, 2000. These 18 months capture a period in which amendments to the Custody Rating Scale (CRS) cut-off values were fully implemented. Further, the Security Reclassification Scale (SRS) had been implemented for sufficient time to ensure reliable data.

Samples

For the review period, a number of samples were collected to address the specific issues raised in the work plan. The samples represent the following populations:

• A General Admission Flow Population - defined as all offenders admitted to a CSC facility during the review period.
• A General Release Population - defined as all releases, regardless of release type that occurred during the review period. Within the General Release Population we have also distinguished offenders who were successfully released from a minimum-security facility (Day Parole, Full Parole, Statutory Release, Warrant Expiry) from those who escape or are returned to higher security for adjustment related issues.
• A Minimum Stock Population - defined as all offenders reported incarcerated at a minimum-security facility on April 1, 1999.
• Average Minimum Population - The stock populations reported to OMS on the first day of each month of the review period were used to develop an "average" population profile of minimum-security facilities.
• A Minimum-security Admission Population - defined as all offenders transferred to a minimum facility either by way of initial penitentiary placement procedures or reclassification and transfer procedures from a higher security level.

Data Sources

Data sources included the Offender Management System (OMS), the Criminal Justice Information Library (CJIL), Research bases, field surveys, and from exercises involving operational representatives. Security and Reintegration Divisions maintain information on the number and profiles of offenders who escape. This information was used to ensure that the definition of escape and computation of escape rates were consistent and comparable with currently reported results.

1. POPULATION MOVEMENT TO MINIMUM-SECURITY

Introduction

The two sources from which most offenders are placed at a minimum-security facility are via Initial Placement admission directly from the community or Reclassification, and transfer from a higher security level institution. Initial placement admissions consist of warrant of committal admissions, revocation re-admissions, and an assortment of other admissions types including foreign transfer, international exchanges, conditional release termination or interruption. The majority of initial placement admissions undergoes assessment at either a regional reception unit or while in a remand centre where the local CSC community office initiate the penitentiary placement procedures.

Offenders reclassified to minimum-security typically transfer from medium security institutions or occasionally from multilevel institutions. It is infrequent that an offender classified as a maximum-security risk is reclassified directly to minimum without spending some time in medium-security institutions as a medium-security risk.

The CRS or the SRS are administered to all offenders undergoing initial placement assessment (direct placement) or a security review for reclassification transfer (indirect). Not all offenders receive an OSL decision consistent with the CRS or SRS rating, presumably for case-specific reasons. Offenders can therefore be identified according to whether their classification decision is consistent (concordant) with their scale ratings or whether their classification decision override the scale ratings. Importantly, overrides of scale ratings can be to either higher or lower security. The distinction between a direct and indirect placement, and a concordant and override decision of the scales are central to this report.

1-1 Custody Rating Scale and Security Reclassification Scale Completion Rates

CSC policy requires that all offenders admitted to federal custody are administered the CRS as part of the initial assessment procedure. Policy also requires that all incarcerated offenders be administered the SRS as part of the security review procedure. Security reviews for reclassification consideration are required to take place at least annually. National and regional CRS completion rates by admission type are presented in Table 1-A. SRS completion rates for offenders for the same 18-month period are presented in Table 1-B.

There were 11,093 offenders admitted during the review period for which a CRS was expected. This includes warrant of committal and re-admissions, exchange of service and transfers from foreign countries and the provinces, but excludes provincial offenders admitted to federal custody and offenders held under authority of a suspension warrant.

Table 1-A: National and Regional Custody Rating Scale Completion Rates by Admission Type

Region	Admissions	Warrant of Committal (%)	Revocation With Offence (%)	Revocation Without Offence (%)	Other* (%)	Total(%)
Atlantic	1,288	100	76.1	61.6	62.5	85.3
Quebec	2,610	100	89.7	82.8	91.7	93.3
Ontario	2,701	99.9	80.3	64.4	89.5	87.2
Prairie	3,408	99.9	90.1	78.4	56.76	92.3
Pacific	1,086	100	39.3	33.8	41.2	76.4
National	11,093	99.9	81.0	70.6	79.5	88.9

(Other includes Exchange of Services admissions, Transfer from Foreign Country, Conditional Release Inoperative, Termination of Conditional Release without and with Offence, Other)

A complete CRS was administered on approximately 89% of all eligible federal admissions. Regional completion rates ranged from 93% in Quebec to 76% in the Pacific region. The completion rate for warrant of committal admissions was essentially 100% while revocation re-admissions with or without a new offence, and "other" admissions lacked a CRS in 20-30% of cases, except in the Pacific region where completion rates were noticeably lower.

Incarcerated offenders are expected to undergo a security review in the event of program reviews, adjustment difficulties, serious incidents indicative of changes in security risk, or at least annually. This makes the establishment of an SRS completion rate for the 18-month review period problematic. As a result it was decided to examine SRS completions rates from the perspective of a one-day snapshot of the population within the review period. The on-register population was identified on September 1, 2000 for all offenders residing at a minimum-security. Those who were admitted with less than one year from this date, as well as women offenders were removed. This is because the SRS is not completed on minimum-security offenders or women offenders. Also, an SRS would not be anticipated unless an offender had least one-year incarceration from the admission date.

Table 1-B: National and Regional Security Reclassification Scale Completion Rates

Region	Incarcerated Offenders	Completion Rate(%)
Atlantic	470	89
Quebec	1,407	89
Ontario	1,734	90
Prairie	1,052	93
Pacific	1,015	95
National	5,678	91

A complete SRS was administered on approximately 91% of eligible offender files. The completion rate ranged from 89% in the Atlantic and Quebec regions to 95% in the Pacific region.

1-2 National and Regional Custody Rating Scale and Security Reclassification Scale Offender Profiles

The scales provide an objective format and standardized criteria in order to profile the offender security classification risk. Total mean scores for the two CRS sub-scales and all SRS totals reported to OMS are presented in Table 1-C. The mean scores for all scale items is available in Appendix 1-2-1.

CRS total and item scores were available on 9,866 files for offenders admitted during the review period. The national mean score on the Institutional Adjustment scale (50) places the admission population comfortably within minimum-security risk range (0-85.5), while the mean Security Risk score (76) is within the medium-security risk range (65-133). Each subscale of the CRS produces a security classification rating and where there is disagreement the offender is assigned the higher security classification rating. Given the consistently higher security risk scores, the CRS outcome is often determined by the Security Risk scale.

In four of the five regions there is little or no distinction among the regions on their mean Institutional Adjustment scores and only slight differences in their Security Risk scores. All four regions are within approximately 2 points of the national average on both sub-scales. The results suggest that Atlantic, Quebec, Ontario and Prairie regions share a similar admission profile. The admission profile in the Pacific region, on the other hand, reported substantially higher mean scores on both the Institutional Adjustment and Security Risk scales. In fact, the Pacific region reported higher mean item scores on all 12 items and substantially higher scores on Institutional Incidents, Escape History, and Sentence Length items. This would normally suggest that the region's admission profile is a substantially higher security classification risk. However, contrary results on the SRS (i.e., comparable mean score relative to other regions) suggest inconsistent scale application may be occurring.

Table 1-C: National and Regional Custody Rating Scale and Security Reclassification Scale Total Mean Scores

Region	CRS Sample	CRS: Institutional Adjustment Mean Score	CRS: Security Risk Scale Mean Score	SRS Sample	SRS Mean Score
Atlantic	1,098	51.2	73.2	1,382	20.0
Quebec	2,435	48.6	73.8	4,116	19.5
Ontario	2,356	48.1	75.2	3,930	18.7
Prairie	3,147	48.4	76.3	4,891	19.6
Pacific	830	64.2	82.1	2,653	19.1
National	9,866	50.0	75.6	16,972	19.3

There were over 19,000 SRS reports entered on the CJIL during the 18-month review period. In many cases, multiple SRS were reported on the same day or within days of each other, and these were deemed either training exercises or errors in application that led to re-entry of information. As a result, duplicate SRS were removed if reported on the same day or within 30 days of each other. Also, a handful of SRS reports that could not be attributed to a particular region were removed, resulting in a total sample of 16,972 SRS reports for subsequent analyses.

The national SRS total mean score was 19.3 which falls comfortably within the medium-security range (> 15.0 and < 26.0). Regional total mean scores varied slightly from the national mean ranging from a high of 20.0 in the Atlantic to 18.7 in the Ontario regions. Individual mean item scores also did not appear to vary across regions suggesting consistency in the application of the SRS. The automated download of many items may contribute to the consistency in the application of the SRS.

A review of the individual items also showed only slight variability in item mean scores among the regions. It is interesting to note that the Pacific region's SRS total and item mean scores were among the lowest of the five regions in contrast to their CRS scores that were substantially higher than other regions. The region reported substantially higher CRS Institutional Incident History mean scores but reported the lowest item score on the SRS Serious Disciplinary Offence item. It is possible that the higher CRS incident scores may be the result of incidents incurred in provincial rather than federal custody.

1-3 Custody Rating Scale, Security Rating Scale, and Offender Security Level Concordance, Override and Scale Yields

CRS/SRS and OSL concordance, override and scale yield analyses were conducted on all initial and reclassification decisions reported during the review period. The national CRS and SRS concordance rates, override and scale yield trends are presented in Tables 1-D and 1-E. Similar rates and yields for each region are found in Appendix 1-3-1 and 1-3-2. Before reviewing tables, it may be helpful to point out that the sum of the percentages on the diagonal represents the total concordance rate. The sum of the percentages above the diagonal represents the total override rate to higher security. The sum of the percentages below the diagonal represents the total override rate to lower security. The CRS and SRS yields by security classification rating are found on the row margins while the OSL distribution by security classification decision is found on the column margins.

As noted earlier, CRS ratings were matched with an OSL decision for direct placement in 9,658 files of offenders admitted during the review period. (For 204 files, matches could not be made because of missing OSL decisions - likely involving offenders who were re-released or returned to provincial custody before the penitentiary placement process was completed). An SRS rating was matched with an OSL decision for reclassification in 16,982 files of offenders undergoing security review during the review period.

The national CRS/OSL concordance rate was 76.5%. The SRS/OSL concordance rate was 84.4% when discordant cases falling within the 5% discretionary rule are removed from the SRS analysis. (The SRS protocol recognizes that the classification of offenders whose SRS scores fall close to the cut-off values are often the most contentious and as a result overrides within a 5% range of the cut-off are not considered overrides. The SRS/OSL concordance rate was 80.7% when no adjustment is made for discordant cases falling within the discretionary range).. The CRS was overridden in 23.6% of applications and the SRS in 15.6%. In both cases overrides to higher security substantially exceeded those to lower security.

Table 1-D: Custody Rating Scale Concordance and Override Rate

		OSL Decision
		MIN % (#)	MED % (#)	MAX % (#)	CRS Yield % (#)
CRS Rating	MIN	22.1 (2,134)	8.9 (855)	0.1 (10)	31.0 (2,999)
	MED	7.2 (694)	49.5 (4,775)	2.4 (228)	59.0 (5,697)
	MAX	0.2 (20)	4.8 (462)	5.0 (480)	10.0 (962)
	OSL Distribution	29.5 (2,848)	63.1 (6,092)	7,3 (718)	100 (9,658)

Table 1-E: Security Reclassification Scale Concordance and Override Rate

		OSL Decision
		MIN % (#)	MED % (#)	MAX % (#)	SRS Yield % (#)
SRS Rating	MIN	14.2 (2,413)	6.9 (1,164)	0.09 (15)	21.2 (3,592)
	MED	3.9 (667)	57.3 (9,737)	6.7 (1,145)	68.0 (11,549)
	MAX	0.01 (1)	1.7 (283)	9.2 (1,557)	10.8 (1,841)
	OSL Dist.	18.1 (3,081)	65.9 (11,184)	16.0 (2,717)	100 (16,982)

Regional CRS/OSL concordance rates ranged from 84.4% in the Atlantic region to 69.7% in the Ontario region (see Appendices 1-3-1 and 1-3-2). Within the regions there appear to be distinctions or preferences in the direction of override decisions. For example, overrides to higher security far exceed overrides to lower security in the Atlantic and the Quebec regions. Overrides to lower security far exceed those to higher security in the Prairie and the Pacific regions, and in the Ontario region the percentage of overrides to both higher and lower security are approximately equal at 15%.

SRS/OSL concordance ranged from 84% in the Atlantic and Prairie regions to 77% in the Ontario region. Overrides of the SRS to higher security levels once again predominate in all regions, and in particular, in the Pacific region where overrides to higher security levels (15.7%) exceed those to lower security (4.5%) by a ratio of more than three to one.

Relative to the scale ratings, OSL decisions tend to be conservative as evidenced by the substantially higher rate of overrides to higher security levels at the national level and in four of the five regions. The disparities in regional concordance rates and in override trends suggest that operational factors that are unique to the regions or disparities in the application of the scales may influence OSL decision-making.

The CRS cut-off values were derived from a 1987 offender population. They were re-designed in 1998 to provide a 35% minimum-security yield. The minimum-security yield in the current review was only 31%. This suggests that offenders admitted during the review period presented a higher security risk as evidenced by the decrease in minimum-security yield. The SRS minimum-security cut-off value was designed to produce a 24% yield, compared to 21% of the offenders in the current review who received a scale rating of minimum-security. Recent policies to deter low risk offenders from federal custody, a reduction in crime rates (particularly violent crime rates), and the proportional growth of offenders serving indeterminate sentences may all contribute to these differences in rates. The OSL distribution to minimum-security at admission was 29.5%, and at reclassification, 18.1%, deviating from the scale yields by less than 3%, although (as discussed below) not all offenders rated minimum-security were transferred to minimum and not all transferred offenders were rated minimum-security.

In the Atlantic, Ontario, and the Prairie regions the CRS minimum yields and OSL distribution were within two percent of the national average (see Appendices 1-3-1 and 1-3-2). The Quebec minimum CRS yield was somewhat higher (35.2%) while in the Pacific region the minimum yield was substantially lower (19.3%). The CRS yield exceeded the OSL distribution to minimum-security in all regions except the Prairies. It is also worthy of note that in the Pacific region both the CRS yield and OSL distribution (14.0%) to minimum-security were far below the national average. (The substantially higher CRS scores in the Pacific region resulted in a CRS maximum yield of over 24% or double the national average, although only 10.4% of admissions received a maximum OSL decisions).

SRS minimum-security yields and distributions were somewhat more consistent in that SRS yields and OSL distributions in four regions were within two percent of the national average. The Ontario SRS minimum yield was 25% of admissions, substantially higher then the national average, while the OSL distribution to minimum-security in the Pacific was 14%, substantially lower than the average.

Table 1-F summarizes that the differences between the combined CRS and SRS yields and OSL distribution in terms of the total number of cases affected and potential implication for offender classification practice.

Table 1-F: Combined Custody Rating Scale and Security Reclassification Scale Yields and Offender Security Level Distribution

	OSL Decision	Scale Placement (CRS and SRS)
Minimum- security	5,974 (22.4%)	6,587 (24.7%) + 600 cases (9.1%)
Medium- security	16,983 (63.8%)	17,504 (65.7%) + 521 cases (3.0%)
Maximum- security	3,683 (13.8%)	2,549 (9.6%) - 1,134 (30.8%)

Relative to OSL decisions, the scales tend to be more liberal in terms of offender security classification. That is, they rate more offenders to be minimum and fewer to be maximum security than OSL decisions. Improving concordance with the scales could result in up to 600 additional minimum-security decisions and 1,134 fewer maximum-security decisions. While Parole Officer discretion remains a vital principle of offender classification, the objective of assigning offenders to the least restrictive level of confinement may well be served by a better understanding of override practices. In this manner a greater potential from the use of classification scales may be realized.

1-4 Override Impacts on Minimum-Security Profiles

Differences in scale yields and OSL distribution that impact on the movement of offenders to minimum-security are discussed below. Table 1-F illustrates how the use of overrides impact on the profile of offenders admitted to minimum-security.

Many offenders who receive OSL decisions of minimum-security are not rated minimum by the scales. Twenty-five percent of direct placed offenders to minimum-security (714) were overrides of the CRS. Twenty-two percent of indirect placed offenders (668) to minimum-security were overrides of the SRS. In both cases the overrides to minimum- security consist largely of medium-rated offenders. In effect, 1,382 offenders were transferred to minimum-security during the review period who were not rated minimum by the classification scales.

Also, many offenders who are rated minimum-security by the scales do not receive OSL decisions of minimum-security. Twenty-eight percent of offenders rated minimum-security by the CRS received OSL decisions of either medium- (855) or maximum-security (10). An even higher percent, 33%, of offenders who were rated minimum-security by the SRS received either medium- (855) or maximum-security (15) OSL decisions. In fact, 1,735 offenders rated minimum-security by the classification scales were not transferred to a minimum-security institution.

The objective is to select and assign the most suitable offenders to minimum-security in a manner that can be understood and replicated. To do so requires a better understanding of those offenders who succeed at, but who were not rated minimum-security. Equally important is an understanding of offenders rated but not assigned to minimum-security. The volume of offenders in these categories supports an examination of these groups in subsequent research initiatives.

Table 1-G: Regional Distribution of Offenders Rated and Assigned to Minimum-Security

OSL Decision - CRS Rating	Min Med %	Med Min(%)	CRS/OSL Concordance(%)	OSL - SRS	Min Med %	Med Min(%)	SRS/OSL Concordance(%)
Atlantic	11.8	23.5	84.4		21.0	22.2	84.3
Quebec	20.9	29.1	76.6		23.5	27.4	79.2
Ontario	30.5	35.6	69.7		16.9	40.7	77.3
Prairie	26.5	20.4	80.0		24.1	23.9	84.2
Pacific	24.1	46.3	72.0		21.1	43.9	79.9
National	24.4	28.5	76.5		21.7	32.4	80.7

(Note: Minimum rated - Maximum placed or Maximum rated - Minimum placed offenders are not presented in Table 1a-3-3 because of their infrequent occurrence).

Regional variation in the use of overrides, their effects on scale yields and OSL distribution, and population movement to minimum-security are also apparent in Table 1-F. For example, overrides of the CRS of medium or maximum rated offenders to minimum-security were only 11.8% in the Atlantic region compared to 26.5% in the Prairie region. Overrides of CRS minimum rated offenders to higher security, on the other hand, were sharply higher in the Pacific region (46.3%).

With respect to the use of overrides of the SRS, the Ontario region (16.9%) reported the lowest rate of medium or maximum-rated offenders overridden to minimum-security. Override rates in the other regions ranged from 21% to 24%. The use of overrides with the SRS of minimum-rated offenders to higher security was again high in the Pacific (43.9%).

In summary, there are regional differences with respect to completion and concordance rates for classification scales. In addition, the use of overrides varies across regions and by the type of security review - direct versus indirect, but these impact minimum-security yields. The nature of overrides and their impact on escape, transfer to increased security, and successful release will be considered in a later section of this report.

2. ASSESSING SECURITY CLASSIFICATION TOOLS AND SCORING OPTIONS

Introduction

The CRS and SRS scales are reflected in security classification policy in order to standardize procedures and advance corporate objectives. Many of the analyses undertaken in the preparation of this report rely on assessment tools that have been developed and evaluated over many years. The application of these tools has been integral to our understanding of security classification practice and offender movement issues. This section examines the reliability of the CRS, the results of a user survey, the impacts of CRS and SRS scoring options on population movement, and the identification of factors related to adjustment at minimum-security.

2-1 Re-Application of the Custody Rating Scale

A total of 40 CRS records prepared during the review period were randomly selected and represented both concordant and override outcomes. Three operational representatives (Quebec, Ontario, and Prairies) were then asked to re-apply the CRS, using information available from OMS sources that existed at the time of the original CRS decision date. The group worked independently on the same first 20 files and jointly on the remaining files. Their CRS ratings were then compared with the original CRS results. In 14 of 20 re-applications of the scale, the 3 proxy CRS ratings agreed with the original rating. In 14 of 20 cases there was also agreement between their consensus ratings and the original CRS rating. Finally, in 17 of 20 cases there was agreement on the CRS rating among the team. This latter finding suggests a slight gain in reliability of ratings when the conditions under which the ratings are completed are similar.

Next, the group reviewed and evaluated 20 files for override decisions and rationales reported in the original correctional plan and provided their judgements as to the appropriateness of the override.

Table 2-A: Re-application of the Custody Rating Scale

	Agreement/Disagreement
Rater Comparisons	Number Agreements	Number Disagreements
Original vs Proxy	14	6
Original vs Consensus	14	6
Within Proxy	17	3

Analyses between the original and proxy CRS ratings identified sources of scoring inconsistency. Those items that require the user to apply their judgement in accordance with prescribed guidelines yield lower reliability. For example, inter-rater correlation was lowest (r< 0.5) for Street Stability. Recent revision to the Street Stability coding guidelines may yield improvements in reliability in future reviews. On many other items, correlation results approached commonly accepted criteria for inter-rater reliability. In comparisons with the original scores, the items that appear to be consistently scored include Age at Admission (r = 0.85), Escape History (r = 0.9), and Number of Prior Convictions (r = 0.85), items that rely on file information and that is readily accessible at admission. Other items, Sentence Length (r = 0.88) and Prior Statutory Release (r = 0.88), also appear to be scored consistently. These results suggest that a revitalized curriculum and a consistent approach to training may improve the reliability of the CRS.

2-2 Custody Rating Scale User Survey Findings

Our knowledge of the views and experiences of the users of standardized assessment tools is largely anecdotal. The opinion of the users of assessment tools, their degree of training and their access to support materials, however, can impact security classification practices. This in turn can influence population movement. Therefore, a brief user survey of a number of factors that may impact on the functioning of the CRS was undertaken.

In order to determine consumer familiarity and confidence with the CRS, a survey was sent to all Case Management staff involved in its use (actual scoring or quality control). As a group, they reported an average of 105 months experience working in the Service, and an average of 65 months experience in Case Management. There were 75 respondents, 54% of whom work in the Prairies region.

Of the respondents, 80% use the CRS at least weekly. Overall, the level of confidence in the CRS was good; however, there was some room for improvement. Importantly, only 3% lacked confidence in the scale. In terms of staff's familiarity with the CRS, there is also room for improvement. For instance, staff reported being somewhat unfamiliar with several aspects of the CRS - 12% regarding the Service's policy on the CRS; 11% regarding procedures; 8% regarding its application; 82% regarding its development and supporting research; 69% regarding performance and accountability reports. Importantly, only 8% of staff reported being unfamiliar with where to find information to complete the CRS. Disconcertingly, 31% reported the scoring guidelines to be barely clear. Equally important, 73% of the respondents felt there was sufficient information in order to apply the CRS. Finally, with respect to overrides, staff reported often needing to use them, at least sometimes, both to increased security (73%) and decreased security (76%).

With respect to training, there are also important findings. For instance, 72% of respondents described their training as being the reading of training materials they found themselves; 69% reviewed materials from their supervisors; 40% completed cases with supervision; and 23% received in-class instruction during orientation. (Note that the same respondent can report multiple training sources). Specific CRS training was received at institutional and regional meetings by 34% and 14% of the respondents respectively. Also, 10% reported that they received specific CRS training at the staff college. Importantly, the respondents indicated an interest in receiving additional training. Their needs varied from completion and scoring (52%), policy (41%), accountability (38%), procedures (35%), and supporting research (35%). Only 12% of the respondents reported not needing additional training regarding the CRS.

Even for staff who use the scale regularly and who report confidence in the scale, 20% reported the scoring guidelines to be unclear and that there was insufficient information to apply the CRS. While staff reports confidence in the CRS, there appears to be considerable room for improvement. Key issues to be addressed include revisions to the scoring descriptions and training. From the survey results, it appears that only after these have been addressed will the nature and frequency of override usage likely change.

2-3 Factors Related to Escape, Return to Higher Security and Successful Release from Minimum-Security

Identifying factors that discriminate offenders who are likely to escape, be returned to higher security or who are successfully released plays an important role in improving assessment and selection of minimum-security offenders. In the following review three groups were compared - a sample of 142 offenders involuntarily transferred to increased security, an equal number of randomly selected offenders successfully released from minimum-security (on Day Parole, Full Parole or Statutory Release) and the escape sample (n = 142). Included among the factors analyzed were items from CRS, the SIR-R1, OIA, and a selection of specific variables known to distinguish offenders according to criminal risk. Factors that differentiated among the three groups are presented in Table 2-B.

Table 2-B: Factors Relating to Escape, Returns to Higher Security and Successful Releases from Minimum-security

	Escape	Return to Higher Security	Successfully Released	F-Ratio or χ2
CRS Item Scores
Age at Sentence	6.9	3.9	1.6	19.9**
Security Risk - Total	74.8	70.9	62.7	7.6**
Street Stability - Adjustment Scale	24.2	20.8	19.2	7.5**
Street Stability - Risk Scale	7.6	6.6	6.1	6.7**
Age at Admission	12.5	12.5	4.9	36.2**
Incident History Score	10.7	13.4	6.9	5.0**
Institutional Adjustment - Total	48.1	44.6	33.2	14.5**
SIR-R1 Variables
SIR-R1 - Total	-6.8	-5.5	-1.2	14.5**
Age at First Adult Conviction	0.9	0.2	0.9	9.4**
Interval at Risk Since Last Conviction	0.1	0.1	0.6	6.7**
Number of Dependents at Most Recent Admission	0.09	0.17	0.31	4.0*
Employment Status at Arrest	0.15	0.32	0.36	6.8**
SRS Item Scores
SRS - Total	14.6	15.2	13.9	3.8*
Age at Review	2.3	1.7	1.8	18.6**
Pay Grade	-1.2	-1.3	-1.1	4.2*
Recorded Incidents	1.0	1.0	0.9	3.3*
Miscellaneous Variables
Direct/ Indirect Population Split(%)	43/58	63/37	67/33	χ2 = 20.7**
Concordant/ Override Split(%)	74/26	58/42	76/24	χ2 = 12.3*

---

A number of significant mean and distribution differences were identified to suggest that many variables may be useful in discriminating escape, return and successful release from minimum-security. For example, total mean score differences on Security Risk, Institutional Adjustment and SIR-R scales are apparent. Offenders who escape were significantly higher risk on these scales than the offenders returned to higher security, who in turn were higher risk than those successfully released from minimum-security.

A number of static, individual scale items, such as age at sentence or SRS review, incident history, street stability, and number of dependants appear to distinguish offender groups. Further dynamic variables, such as employment status and pay grade, also showed potential to distinguish among these groups.

The data also suggest that an offender overridden to minimum-security is more likely to be returned to higher security than to escape. This conclusion may be spurious, however, as the base rate for return to increased security is four times that of escapes. Further, disciplinary transfers may serve to pre-empt escape risk, and therefore, attenuate the actual escape rate. Compared to offenders who are successfully released, offenders who escape are significantly higher risk scores on all reported variables.

In general, offenders who escaped or were returned to higher security were more likely to be indirectly transferred to minimum-security. At admission they tended to be younger, unemployed, lacking dependents at the time of conviction, come from unstable community conditions, and be incident-prone during previous incarcerations.

With respect to SRS variables that assess current in-custody behaviour, those offenders who escape were more likely to have incurred serious disciplinary offences and incident reports, be at lower inmate pay levels, and be younger than successfully released offenders. Offenders returned to higher security share some, but not all, of the characteristics of the escape group, and still remain distinguishable from the successfully released group.

The relationship between the total scores of the CRS, SRS, and SIR-R, and minimum-security outcomes was also explored to determine the predictive validity of the risk instruments currently available to classification decision-makers. A point biserial correlation procedure was employed. This procedures provides an indication of how effective a scale score is in distinguishing between dichotomous groups - all offenders successfully released versus all offenders who escaped; all offenders successfully released versus all offenders returned to higher security, etc. The results of this analysis are presented in following table.

Table 2-C: Relationship between Classification Scales and Outcome (point biserial correlation)

	CRS Institutional Adjustment	CRS Security Risk	SRS	SIR-R1 Score
Escape	0.17**	0.15**	0.00	0.20**
Return to Higher Security	0.10*	0.05	0.17*	0.10
Successful Release	-0.27***	-0.20**	-0.18*	0.29***

---

A number of statistically reliable relationships were found. Institutional Adjustment and Security Risk scores were significantly related to escape and successful release; SRS scores to returns to higher security and successful release; and SIR-R1 scores to escape and successful release.

The results suggest that discernible differences can be identified between offenders who escape, those who are returned to higher security, and those who are successfully released. Many of the factors that discriminate offender groups are embedded within the risk instruments currently employed in security classification practice.

3. EXAMINING OVERRIDES

Introduction

As a rule-of-thumb, an overall 15% override rate is generally accepted as indicative of the appropriate application of a security classification scale. Higher concordance may limit the case specificity of the classification decision, but lower concordance may suggest imprecise use of the classification scale. That is, the scale results should be an anchor to the final OSL decision. The overall override rate as discussed above was 23.5% for the CRS and 15.7% for the SRS. The override results related specifically to minimum-security were somewhat higher (29% and 32%, respectively). To better understand the nature and use of overrides, a number of methods were employed. These included an evaluation by operational representatives of randomly selected CRS reports of override, a review and categorization of reasons for overrides, a re-application of randomly selected CRS reports, and the survey of staff views of the CRS discussed previously.

3-1 Operational Team Review of a Sample of Override Reasons

A total of 60 CRS records prepared during the review period and representing both concordant and override applications were randomly selected from all five regions. Three operational representatives (from Quebec, Ontario, and Prairie regions) were asked to independently re-apply the CRS to 40 cases for comparison with the original CRS rating. They also assessed rationales reported in the original correctional plan to determine the "appropriateness" of the explanation for a sample of 20 override decisions.

Overrides were considered to be "inappropriate" if there was no reason given, if the reason given was not related to security risk, or if the reason was not sufficiently supported by file evidence. Generally, the review team was left to their collective experience and judgement in making the distinction between appropriate and inappropriate usage of override discretion.

The sample consisted of two groups. Case files rated medium-security with OSL override decisions to minimum-security (n = 15) and case files rated minimum-security (n = 5) overridden to medium-security. In 14 of the 20 files reviewed, the operational representatives judged the override to be appropriate. That is, they agreed there was sufficient evidence relating to security classification risk to warrant overriding the scale rating. In most (but not all) of the override cases judged as appropriate, it was also concluded that the rationale was clearly supported, documented and well-argued in the assessment for decision commentary.

The remaining 6 cases (30%) of CRS overrides were judged to be inappropriate. According to the reviewers, the evidence to support the use of an override was not given or was insufficient. In two of the cases judged inappropriate, no rationale could be clearly discerned from the commentary, while the remaining rationales were described as "weak" or "poor" by the review team.

It should be cautioned that these results are based on a small, exploratory sample and under somewhat specific conditions. Accordingly, they may not fully generalize to broader case management practice. Nonetheless, the results are consistent with the view that the practice of overrides warrants systematic study and that additional training is required.

3-2 Reasons for Overriding the Custody Rating Scale and the Security Reclassification Scale

A random sample of 329 CRS overrides, prepared during the review period was selected. For each case, the correctional plan and the assessment for decision narratives were scanned for the specific override reason. The narratives were then given to five operational staff, experienced in initial placement, who were asked to categorize the override reason. A number of reason categories were provided to the team. However, the team was not restricted to these assigned categories. The reviewers were instructed to prioritize the override reasons in the cases of multiple reasons. Two of the team members reviewed cases in both languages while three reviewed English-only cases. It should also be noted that the reviewers were inadvertently provided files that were not override cases and where limited information made identifying a reason for override problematic. This, however, reflects the state of securing OMS-based cases for review, whether for research or quality assurance. The most frequently reported primary and secondary reasons aggregated for all five reviewers are presented in Table 3-A.

It may be instructive to review some of the reasons that were not often used to override the CRS. For example, a deportation order where public risk was a factor was reported only once as a primary or secondary reason. Incompatibility with offenders was reported in only two cases. Protective custody was reported as a primary reason in three cases and a secondary reason in four cases. Cell availability, medical factors and public notoriety played no role as a prime reason for overriding the CRS. A sex offence history where public concerns were an issue, deportation where escape was a factor, and facilitating program participation at lower security were reported in 10 to 20 cases as primary override reasons. In total, these represent 2.0% of the reasons provided by staff in making override decisions.

In a large number of cases, a primary or secondary reason for overriding the CRS could not be found among the 26 available categories provided to the reviewers. The "Other" category was selected as a primary reason for overriding scale in 29% of the reports and included factors such as Schedule 1 issues, concerns for untreated sex offenders, and good or poor institutional adjustment. In a surprising number of reviews (19%) either vague or no reasons were found by the reviewers. However, it was not always clear whether this resulted from insufficient commentary on the files or difficulty in determining the actual override status.

Some of the reasons for overriding such as "Risk can be managed at lower security" or "Prior successful adjustment at lower security", suggest a fundamental disagreement with the classification scale outcome. Further, offender "motivation" was employed as both a positive and negative factor. This assessment of dynamic factors likely resulted from interview contact with the offender. Although previous breach of trust is well represented in the CRS, it was frequently reported as a reason to override the scale, (i.e., "History of breaches", "History of success in trust situations" and "Prior successful adjustment at lower security"). This suggests a lack of confidence in the weight assigned to breaches of trust in the scale or perhaps a misunderstanding of how to interrupt the scale.

Table 3-A: Reasons for Overriding the Custody Rating Scale

Reason	Primary # (%)	Secondary # (%)
No history of violence	49 (4)	5 (0)
Recent, uncontrolled substance use	34 (3)	49 (8)
Motivation to change as a negative factor	34 (3)	26 (4)
Motivation to change as a positive factor	52 (4)	62 (10)
History of Breaches of Trust	50 (4)	49 (8)
History of success in situations of trust	20 (2)	29 (5)
Prior successful adjustment at lower security	82 (7)	13 (2)
Program - requires program to reduce security risk.	37 (3)	44 (7)
Risk can be managed at lower security.	172 (14)	96 (15)
Important factors not reflected in the CRS	38 (3)	7 (1)
Vague or unclear reasons given	166 (13)	31 (5)
No reasons given	72 (6)	4 (0)
Other - specify	364 (29)	121 (19)
All non-listed categories	77 (6)	95(15)
Total	1,247	631

The reasons for SRS overrides were derived from the CJIL database from entries submitted directly with the application of the scale. Unlike the CRS, the SRS application allows for one override reason to be identified. The frequency for SRS overrides are presented in Table 3-B.

Escape played a role in 2.4% of the overrides. In the original scoring of the SRS, escape was not a scored item but it was added to the current version in November 2000. Accordingly, escape should play a minor role in future reviews of SRS overrides. Escape, assault causing serious harm, and instigating disruption/damage are often sole reasons to conduct a security review and sufficient to determine security reclassification outcome. The SRS was designed to assess a 12-month period of incarceration behaviour. When applied to event-driven security reviews such as those noted, overrides of the scale have been known to increase. Twenty-three percent of overrides occurred with SRS scores that were within 5% of the designated cut-off values or within the range where officer discretion is recognized and encouraged. Discordant outcomes falling within the authorized discretionary range are not treated as overrides and were not included in these analyses.

Table 3-B: Reasons for Overriding the Security Reclassification Scale

Reason	Frequency # (%)
Escape/attempt with violence	9 (0.3)
Escape/attempt without violence	67 (2)
Serious escape history or current escape	34 (0.1)
Assault with serious harm	107 (3)
Instigator in disruption without damage	80 (3)
Major source in distribution of contraband	170 (5)
Major force in disruptive activity	106 (3)
SRS score within 5% of cut-off values	749 (23)
Other	1,953 (60)
Total	3,275

The most frequently reported reason for overriding the SRS was for "Other" (60%) reasons. From the information provided in the override comment field "Other", overrides were often based on program-related factors (the offender needed to complete a specific program before reclassification could be considered; the offender was meeting correctional plan expectations, etc).. Sexual offending and protection issues were also reported for overrides where the reason was noted as "Other". In many cases, it was difficult to interpret the use of the "Other" category.

Often overrides appear to be used to reconcile the classification decision with the security rating of the assigned institution. In many of these situations, assignment was determined by a custodial factor (such as protection or medical needs), rather than risk factor. Occasionally, override results from the application of the scale to event-driven security reviews - applications for which it is not optimally suited. The likelihood of reducing the use of scale overrides by providing better instructions and improving training appears likely.

The current research regarding the nature of override use should be considered exploratory. We have investigated overrides from a number of perspective - an analysis of override trends, a case audit of rationales, and a review and description of override reasons - that has advanced our understanding of override use. Override discretion is widely authorized but only supported by limited policy and procedural guidelines. For the most part, the decision to override the scales seems locally based and case-specific. The operational reviewers who had access to either the full OMS report or to limited segments of the override argument, report many cases of inappropriate or poorly articulated rationales. Often, there was no clearly defined protocol for presenting or defending the use of overrides.

While additional research is indicated, it is important to first establish policy that better defines authority and procedural obligations associated with override discretion. A co-ordinated approach to training of both users and first level supervisors is also clearly indicated. A specific override field in the CRS and improved instructions in the application of the SRS is necessary to facilitate ongoing monitoring and analysis of override discretion.

4. MOVEMENT OUT OF MINIMUM-SECURITY

Introduction

Minimum-security facilities are the primary releasing facilities within the Service. Day or full parole release, and preparing offenders for statutory release are important considerations in the assessment and selection of minimum-security candidates. It is also important to limit the number of offenders in minimum-security who might not benefit from the experience. That is, they are unlikely to adjust to a minimum program or are prone to escape because of the absence of perimeter restraints. This section examines the volume and nature of releases from minimum-security, the rate of return of offenders to higher security, and the rates of escape from minimum-security.

4-1 Release from Minimum-Security

Table 4-A summarizes the volume and nature of lawful release from minimum-security by regions for the review period. OMS indicates that a total of 3,353 offenders were released during the review period, the majority (68%) of whom were released on a discretionary release, primarily on day parole (63%). Thirty-one percent of offenders released from minimum-security were statutory releases and the remaining releases consisted of warrant expiry releases, court ordered returns to provincial custody, and foreign transfers.

There were regional differences in terms of release from minimum-security. For instance, the Pacific (80%) and the Atlantic (78%) regions had a much higher proportion of their minimum population on discretionary release than the Prairie (63%) region. Quebec (67%) and Ontario (68%) regions had about two-thirds of their minimum population released on a discretionary release.

Table 4-A: National and Regional Releases from Minimum-Security

	Day Parole # (%)	Full Parole # (%)	Statutory Release # (%)	Other # (%)	Total
Atlantic	222(73)	14(5)	60(20)	5(2)	301
Quebec	664(63)	42(4)	348(33)	7(1)	1,061
Ontario	538(61)	58(7)	278(32)	3(0.3)	877
Prairie	524(59)	34(4)	318(36)	8(1)	884
Pacific	173(75)	12(5)	45(20)	0(0)	230
National	2,121(63)	160(5)	1,049(31)	23(1)	3,353

Explanations for these differences in discretionary release rates are not readily apparent. The Atlantic region's minimum population is drawn largely from their admission or direct placement pool which may contribute to both the higher discretionary release rate and the lower escape rate (see below). This explanation, however, does not hold in the Pacific region where, despite a minimum population drawn largely from the indirect placement process, the region maintains a very high discretionary release rate. Additional research into the antecedents of discretionary release would be helpful in better understanding the release potential of minimum-security facilities.

4-2 Adjusting Custody Rating Scale and Security Reclassification Scale Security Level Cut-off Values: Reducing the Number of Escape Offenders in the Minimum-Security Yield

The question was raised as to whether the scales could be better used to identify and prevent the transfer of potential escapees to minimum facilities. Examining various security level cut-off options and applying them retrospectively to a group of offenders who actually escaped determines the utility of adjusting the scales.

There were 142 escapes during the review period, many of them were rated minimum-security by either the CRS or SRS. The minimum-security cut-off values of the scales were lowered to determine what effect this would have on the number of false negatives (escapees incorrectly rated as minimum-security by the scales). CRS or SRS scores were available on 99 of the 142 offenders who escaped during the review period. Seventy-one of the 99 escapees or about 72% were rated minimum-security by one of the classification scales. Table 4-B illustrates the effects of lowering the minimum-security cut-off values of both the CRS and SRS. Each successive lowering of the cut-off reduces the number of offenders rated as minimum-security. This is also referred to as the yield of both the CRS and SRS. The more conservative the cut-off, the fewer false negatives (escapees).

Table 4-B: Impact of Different Cut-off Scores for the Custody Rating Scale and the Security Reclassification Scale

Minimum-security Cut-off Options	Escapees Rated Minimum % (#) (n = 71)	Concordance Rates(%) CRS SRS	Min. Yield(%) CRS SRS	Med. Yield(%) CRS SRS	Max. Yield(%) CRS SRS
Current Values: CRS: Inst. Adj: 85.5 Sec. Risk: 63.5 SRS: 15.0	72 (71)	76.5 84.4	31.0 21.2	59.0 68.0	10.0 10.8
Option 1: CRS: Inst. Adj: 82.5 Sec. Risk: 60.5 SRS: 14.0	41 (41)	74.9 82.4	26.3 13.8	63.8 75.3	10.0 10.8
Option 2: CRS: Inst. Adj: 79.5 Sec. Risk: 58.5 SRS: 12.0	30 (30)	74.2 79.5	24.0 5.1	66.0 84.1	10 10.8

Option 1 lowered the CRS minimum-security cut-off values to 82.5 on the Institutional Adjustment scale and 60.5 on the Security Risk scale, and to 14.0 on the SRS. This resulted in a 46% reduction of the number of escapees (31) rated as minimum-security. Importantly, CRS concordance drops by only 1.6% and on the SRS by 2%. This Option would reduce the CRS yield to minimum-security to 26.3%, which is a reduction of about 4.7% from the current yield. The projected SRS minimum-security yield is 13.8%, or a reduction of 6.2% from the current yield.

Option 2 further reduces the number of escapees who received a minimum-security rating to 30 with a minimal reduction in concordance. The minimum-security yields, however is markedly reduced, to 5.1%.

These analyses illustrate that adjustments to the scales that lower cut-off values can be effective in limiting the number of escapees who are rated minimum. There are associated cost in terms of concordance and minimum yields, however. Perhaps these can be minimized and offset by improving assessment methods. The results invite further discussion and exploration of cut-off values that provide a balance between the need to minimize escapes and ensure a broad candidate pool of minimum-security offenders. An alternative strategy to changing the cut-offs is to adjust scoring by increasing weighting of key items. These are the next set of analyses presented.

4-3 Effects of the Security Reclassification Scale Scoring Adjustments on Offenders with Serious Escape History

Table 4-C presents the comparative yield for the original SRS rating, the yield from the revised (current) scoring model, and an enhanced model weighted to address serious escape history. In the latter case, additional weighting was assigned to any offender with a CRS Escape History of escape or attempt escape:

• from minimum in the last 2 years;
• with violence from maximum, medium or minimum within the last 5 years or;
• two or more escapes from any level within the last 5 years.

Table 4-C: Security Reclassification Scale Yield by Security Rating: Security Reclassification Scale Models

SRS Model	Minimum (%)	Medium (%)	Maximum (%)
Original SRS	21.3	68.0	10.8
Revised SRS	20.8	69.2	10.1
Enhanced SRS (Weighted for Escape)	19.7	70.2	10.1
OSL Decisions	18.2	65.9	16.0

---

There was little or no effect on the SRS yield to security placement as a result of the revised scoring or the enhanced model. Relative to the original scale, both the revised and the enhanced models tended to be slightly more conservative in assignment to minimum and maximum security. Relative to actual OSL decisions, all models tend to be more liberal, that is, assigning more offenders to lower security levels.

Table 4-D presents the SRS rating, OSL decision concordance and override results from the original, revised scales and the enhanced scale.

Table 4-D: Security Reclassification Scale/Offender Security Level Concordance and Overrides by Security Reclassification Scale Models

SRS Scores	Concordance (%)	Overrides to Higher Security (%)	Overrides to Lower Security (%)
Original SRS	80.7	13.4	5.6
Revised SRS	79.8	14.3	5.9
Enhanced SRS Weighted for Escape	79.3	14.0	6.6

The effect on concordance levels and override use of the revised and enhanced models is minimal. It should be noted that the revised model does increase the number of overrides to lower security - consisting largely of medium-rated offenders who received minimum-security OSL decisions.

While there appears to be little difference in the overall yields across SRS scoring models, the effect on those offenders with serious escape history was more substantial. The total number of offenders with serious escape histories in the sample was 1,898. Of these, 220 (11.6%) were rated minimum-security by the original SRS scoring model. The revised model reduces the number of minimum-rated offenders to 193 (10.2%). The enhanced model that was weighted for serious escape history reduced the number of offenders falling in the minimum category to 14 (0.7%). Therefore, an enhanced SRS model weighted against offenders with serious escape history would reduce the number of these offenders receiving a minimum rating by over 93% with only minimal loss in yield and concordance rates. Given the relationship between escape history and subsequent escape, this model should contribute to a reduced escape rate from minimum-security.

4-4 National and Regional Rates of Escape by Admission Source and Concordant Status

For the purpose of this report, escape is defined as "all escapes from a minimum-security facility reported to the Sensational Incident Reporting system, between April 1, 1999 to September 30, 2000". This definition is consistent with the one used by both the Performance Assurance and Correctional Programs and Operations sectors. The rate of escape is calculated as the number of escapes per 100 offenders, based on the minimum-security average population during the review period. This definition excludes escape from escorted temporary absence (ETA), unlawfully at large (UAL) from unescorted temporary absence (UTA), or work release, and absconding while under Section 81 status, a total of 15 incidents. National and regional escape rates, converted to represent annual rates, are presented next.

Table 4-E: Rates of Escape Per 100 Offenders by Admission Type and Concordant Status

	Atlantic	Quebec	Ontario	Prairie	Pacific	National
Overall	2.7	4.7	2.9	3.7	5.6	4.0
Pen Placed	3.8	4.3	1.7	2.8	1.1	3.0
Reclassified	1.0	5.2	4.9	5.0	7.4	5.2
Concordant	3.3	4.0	3.4	3.6	5.0	3.8
Override	0.0	7.1	1.4	4.0	7.5	4.4

The National escape rate from minimum-security was 4.0 per 100 average offenders. Escape rates per 100 offenders vary across regions by admission type and concordant status. The escape rate for directly placed offenders (3.0) was substantially lower than that for indirectly transferred offenders (5.2). The rate for offenders whose OSL decisions were concordant with their CRS or SRS ratings was 3.8 per 100, which was lower than that for offenders where overrides were used (4.4 per 100).

Regional escape rates varied from 2.7 per 100 offenders in the Atlantic to 5.6 in the Pacific regions. All regions, except for the Atlantic, followed the national trend and reported higher escape rates for indirectly transferred offenders. Also, all but one of the Atlantic region's escapes was directly placed to minimum-security. The opposite was the case in the Pacific region where all but one of their escapes were indirect transfers.

In three regions, overriding the CRS or the SRS contributed to higher escape rates. The override escape rate was substantially higher (7.1) than the concordant rate (4.0) in the Quebec region. Nineteen of the region's escapes (33.3%) during the review period involved overrides of the scales and these were equally divided between direct and indirect placements.

For the Atlantic and the Ontario regions, the escape rate for offenders whose CRS or SRS ratings were overridden was lower than that of offenders whose ratings were concordant with OSL decisions. The frequency of indirect transfers of medium-rated offenders overridden to minimum-security in the Atlantic (12%) and Ontario (17%) regions was much lower than the national figure and may explain why their concordant escape rate was higher relative to the override rate.

The override rate (30%) of direct placements in Ontario, however, was much higher than the national rate. Nonetheless, their override escape rate (1.4) was much lower than the national average (4.4). These results suggests that the Ontario region's low overall escape rate is the result of a direct placement process that is careful in selecting minimum-security offenders and effectively applying override discretion. It should be noted, that Ontario has a large number of minimum-rated offenders placed in medium security (false positives), and the accuracy of these override decisions is difficult to evaluate.

There were 142 escapes from minimum-security during the review period. Of these, 82 were indirect transfers from higher security and 60 were direct placements from admission units. OSL decisions were concordant with the scales in 105 of cases of escape. Overrides occurred in 37 cases. Relative to the proportion of offenders by admission source, the results suggest that offenders who escape from minimum-security were more likely to be indirect transfers. For the majority of the regions, overrides contributed to higher escape rates.

4-5 National and Regional Rates of Return to Higher Security

Minimum-security is often a testing ground for release potential. Offenders who fail to adjust or whose security risk increases can be quickly re-assessed and transferred to higher security.

A total of 721 offenders were involuntarily transferred to higher security during the review period. Approximately 17 for every 100 average offenders incarcerated at minimum-security facilities were returned to higher security on an involuntary transfer warrant.

There was little difference in the rate of return to higher security by type of placement, direct (18) or indirect (17). Rates of return to higher security, varied sharply among the regions, ranging from a low return rate in the Ontario (10) to unusually high rates in the Atlantic (42) and Pacific (35) regions. Minimum facilities in the Pacific region rely heavily on indirect transfers for offenders and this may have contributed to their higher return rate. This is not the case in the Atlantic region where direct placements constitute the majority of the minimum population and no explanation for their high return rate is apparent.

While the escape rate of indirect transfers was substantially higher than direct placements, there was little or no difference in their rates of return to higher security. At this point, the reasons for these differences in return rates among regions cannot be determined. It is clear, however, that returns to higher security is an outcome that should be considered in evaluating the effectiveness of classification decisions.

Table 4-F: Rates of Return Per 100 Offenders from Minimum-Security to Increased Security

	Atlantic	Quebec	Ontario	Prairies	Pacific	National
Overall	41.8	16.0	10.3	13.1	35.5	17.3
Direct Placement (Concordant)	33.9	14.6	8.4	10.8	24.6	14.5
Direct Placement (Override)	37.2	29.5	12.5	40.3	58.8	28.1
Direct Placement	34.3	17.7	9.6	18.8	32.8	17.7
Indirect Transfer (Concordant)	49.6	7.8	6.8	2.8	21.6	10.4
Indirect Transfer (Override)	20.5	25.6	26.8	9.0	71.3	29.0
Indirect Transfer	53.9	14.0	11.4	5.0	36.5	16.8
Concordant	44.2	12.6	8.3	7.8	26.5	14.0
Override	28.6	27.6	16.2	28.3	67.5	28.5

CONCLUSIONS AND RECOMMENDATIONS

The purpose of this project was to provide a consolidated investigation into population movement to and from minimum-security. For an 18-month period, over 26,000 classification decisions were reviewed in terms of type of placement (direct or indirect). Completion rates for classification scales (CRS and SRS), concordance between scale results and OSL decision, the use of overrides in decision-making, and outcomes (escape, return to increased security, successful release) were examined.

CRS and SRS completion rates for the period under review were approximately 90%. The national concordance rate for the CRS was 76.5% and for the SRS, 80.7% (reaching 84.4% when the 5%, discretionary override rule was applied). Regional concordance rates for both scales varied among the regions. CRS concordance rates ranged from 84% in the Atlantic region to 70% in the Ontario region, while SRS concordance ranged from 84% in the Atlantic and Prairie regions to 77% in the Ontario region.

Actual and potential movement to minimum-security is affected by override decisions involving both minimum and medium rated offenders. Twenty-five percent of direct placements and 22% of indirect placements to minimum facilities were overrides of offenders rated medium (and to a minor extent, maximum) security by either the CRS or SRS. The number of overrides to medium-security of CRS minimum-rated (28%) and the SRS minimum-rated offenders (33%) suggests that a substantial number of minimum-rated offenders are being denied access to minimum facilities. The potential for increasing the initial placement or transfer of these offenders to minimum-security should be explored.

Further, the effect of scale modifications designed to limit offenders with serious escape histories from receiving a minimum-security rating on the CRS or SRS was investigated. Adjusting the item weight of the escape variable on the SRS scale would reduce the number of offenders with serious escape history currently rated minimum by 93%, without effecting the yield or concordance rate of the SRS.

Reasons for overriding the scales were examined using SRS information drawn from the CJIL base, a random sample review of selected paragraphs of OSL decisions, and a full OMS audit of 20 CRS files conducted by operational representatives. This review concluded that the reasons for overriding were not easily categorized. Also, 30% of the reasons for overriding the CRS were listed as "Other" and 60% of SRS were similarly listed. The full audit of the OSM file noted that 30% of the overrides were judged to be "inappropriate" either because the reasons given were not based on criminogenic factors, the reason was not substantiated or the override was poorly argued. Importantly, many reasons given for overriding the classification scales included factors that were already well represented by the classification scale items.

A survey of CRS operational users (n = 75) was conducted and responses were received from all regions. While there was moderate confidence in the scale, knowledge of how to apply the scales warrants improvement. The majority of respondents (88%) requested some form of training while a re-application of the CRS by field staff found only 70% agreement with original scale outcome. It is reasonable to conclude that supplemental training on all aspect of scale application would reduce the use of overrides.

The majority of offenders (68%) released from minimum-security was released on some form of discretionary release (DP - 63%, FP - 5%), although there was substantial regional variation in rates of discretionary release. An explanation for the differences in rates was not apparent, however, there may be benefits in reviewing the practices of regions with high rates of discretionary release.

Relative to the national escape rate (4.0/100 offenders), direct placed offenders had a 25% lower rate of escape and indirect transferred offenders had a 23% higher rate of escape. In fact, there is a 42.3% relative increase in the escape rate when comparing direct placed offenders and indirect transfers (3.0/100 direct, 5.2/100 indirect). Also, there is a 13.6% relative increase in escape rate when comparing concordant and override placements (concordant escape rate is 3.8/100 offenders and 4.4/100 for overrides). Indirect, discordant placements resulted in the highest escape rate.

The rate of return to increased security was 17.3 per average 100 offenders, although rates ranged from 10.4 to 29.0 depending on type of placement and concordance status. Relative to concordant decisions, override decisions resulted in twice the return rate to increased security by offenders.

Efforts were undertaken to distinguish offenders who succeed at minimum-security from those who fail. Offenders who escaped during the review period were compared with offenders who were returned involuntarily to higher security, and also with offenders who were successfully released from minimum-security. These groups were compared using the CRS, SIR-R1, and OIA case needs domains. Offenders who escaped had significantly higher scores on the two sub-scales of the CRS and on total SIR-R1 scores relative to offenders returned to higher security, who in turn had higher scores than the successfully released group. In addition, groups were distinguishable on a number of individual items from each scale including age, incident history, street stability, and number of dependents. Only the employment domain of the OIA appears to distinguish the groups, in that the successful release group had more stable histories of employment than the other groups. An analysis of dynamic risk factors and pre-indicators using a small subsample yielded inconclusive findings.

APPENDICES

Appendix A:

Custody Rating Scale and Security Reclassification Scale, Total and Item Means

CRS Item Scores	Atlantic	Quebec	Ontario	Prairie	Pacific	National
Ins. Incident History	14.3	17.5	17.7	13.4	26.9	16.7
Escape History	2.8	1.8	1.5	2.1	4.1	2.1
Street Stability (Institutional Adjustment)	22.3	20.7	20.3	21.5	23.2	21.3
Alcohol/Drug Abuse	4.1	3.5	3.2	3.9	4.4	3.7
Age (At Sentencing)	7.6	5.0	5.3	7.5	5.6	6.2
Total Institutional Adjustment	51.2	48.6	48.1	48.4	64.2	50.0
Number Prior Convictions	10.0	8.7	9.8	10.1	10.1	9.7
Most Severe Outstanding Charge	3.9	3.0	3.2	4.7	3.8	3.4
Severity of Current Offence	23.1	25.3	26.5	26.1	28.2	25.9
Sentence Length	9.2	12.1	13.2	9.5	14.3	11.4
Street Stability (Sec. Risk)	7.0	6.5	6.3	6.7	7.2	6.6
Prior Parole &/or Stat. Rel.	1.3	1.5	1.0	0.7	1.2	1.1
Age at First Admission	16.6	13.2	13.7	17.2	15.7	15.2
Total Security Risk	73.2	73.8	75.2	76.3	82.1	75.6

Security Reclassification Scale Total and Item Means - Original Model

SRS Item Scores	Atlantic	Quebec	Ontario	Prairie	Pacific	National
Serious Disciplinary Offence	1.7	1.6	1.3	1.4	1.3	1.4
Recorded Incidents	1.2	1.1	1.2	1.2	1.2	1.2
Pay Grade	-0.8	-0.8	-1.0	-0.8	-1.0	-0.9
Segregation Period	2.3	2.9	2.2	2.2	2.0	2.3
Detention Referral	1.3	1.1	1.2	1.3	1.4	1.2
Correctional Plan Progress	2.5	2.7	2.5	2.5	2.5	2.5
Correctional Plan Motivation	2.9	3.0	2.8	2.9	2.9	2.9
Drug & Alcohol Rating	2.8	2.7	2.6	2.8	2.8	2.7
Successful ETA Releases	-0.2	-0.3	-0.1	-0.1	-0.5	-0.2
Successful UTA Releases	0.8	0.8	0.8	0.8	0.8	0.8
Age At Review	2.2	2.1	2.1	2.3	2.1	2.2
Psychological Concerns	1.3	1.2	1.3	1.3	1.4	1.3
CRS Ins. Incident History	2.0	1.7	1.9	1.8	2.1	1.9
Total Security Reclassification	20.0	19.5	18.7	19.6	19.1	19.3

---

Appendix B:

Regional Custody Rating Scale Concordance and Override Rates, Custody Rating Scale-Offender Security Level Concordance

Atlantic Region						Quebec Region
		OSL Decision				OSL Decision
		MIN %(#)	MED %(#)	MAX %(#)	CRS Yield %(#)	MIN %(#)	MED %(#)	MAX %(#)	CRS Yield %(#)
CRS Rating	MIN	24 (253)	8 (79)	0.39 (4)	33 (336)	25 (606)	10 (249)	0.04 (1)	35 (856)
	MED	3 (34)	54 (560)	2 (23)	60 (617)	7 (162)	47 (1,153)	3 (70)	57 (1,385)
	MAX	0.1 (1)	2 (20)	6 (60)	8 (81)	0.33 (8)	3 (78)	4 (106)	8 (192)
	OSL Dist.	29 (288)	64 (659)	8 (87)	100 (1,034)	32 (776)	61 (1,480)	7 (17)	100 (2,433)

		Ontario Region				Prairie Region
OSL Decision						OSL Decision
		MIN %(#)	MED %(#)	MAX %(#)	CRS Yield %(#)	MIN %(#)	MED %(#)	MAX %(#)	CRS Yield %(#)
CRS Rating	MIN	21 (481)	12 (271)	0.2 (4)	33 (756)	23 (708)	6 (182)	0.03 (1)	29 (891)
	MED	9 (214)	44 (999)	3 (65)	56 (1,278)	8 (256)	53 (1,632)	2 (63)	63 (1,951)
	MAX	0.3 (6)	6 (134)	5 (116)	11 (256)	0.1 (3)	4 (109)	4 (119)	6 (231)
	OSL Dist.	31 (701)	61 (1,404)	8 (185)	100 (2,290)	31 (976)	63 (1,923)	6 (183)	100 (3,073)

		Pacific Region
		OSL Decision
		MIN %(#)	MED %(#)	MAX %(#)	CRS Yield %(#)
CRS Rating	MIN	10 (86)	9 (74)	0 (0)	19 (160)
	MED	3 (28)	52 (431)	0.9 (7)	56 (466)
	MAX	0.2 (2)	15 (121)	10 (79)	24 (202)
	OSL Dist.	14 (116)	76 (626)	10 (86)	100 (828)

Appendix C:

Regional Security Reclassification Scale Concordance and Override Rates Security Reclassification Scale, Offender Security Level Concordance

		Atlantic Region				Quebec Region
		OSL Decision				OSL Decision
		MIN %(#)	MED %(#)	MAX %(#)	SRS Yield %(#)	MIN %(#)	MED %(#)	MAX %(#)	SRS Yield %(#)
SRS Rating	MIN	14 (199)	4 (57)	0.07 (1)	19 (257)	15 (620)	6 (235)	0.07 (3)	21 (858)
	MED	4 (53)	55 (762)	6 (89)	65 (904)	5 (191)	56 (2,293)	9 (358)	69 (2,842)
	MAX	0 (0)	1 (17)	15 (204)	16 (221)	0.02 (1)	2 (69)	8 (346)	10 (416)
	OSL Dist.	18 (252)	60 (836)	21 (294)	100 (1,382)	20 (812)	63 (2,597)	17 (707)	100 (4,116)

		Ontario Region				Prairie Region
		OSL Decision				OSL Decision
		MIN %(#)	MED %(#)	MAX %(#)	SRS Yield %(#)	MIN %(#)	MED %(#)	MAX %(#)	SRS Yield %(#)
SRS Rating	MIN	15 (580)	10 (403)	0.2 (8)	25 (991)	14 (696)	4 (218)	0 (0)	19 (914)
	MED	3 (118)	55 (2,177)	7 (299)	66 (2,594)	5 (221)	59 (2,908)	5 (234)	69 (3,363)
	MAX	0 (0)	2 (63)	7 (282)	9 (345)	0 (0)	2 (100)	11 (514)	13 (614)
	OSL Dist.	18 (698)	76 (2,643)	15 (589)	100 (3,930)	19 (917)	66 (3,226)	15 (748)	100 (4,891)

		Pacific Region
		OSL Decision
		MIN %(#)	MED %(#)	MAX %(#)	SRS Yield %(#)
SRS Rating	MIN	12 (315)	9 (249)	0.11 (3)	21 (567)
	MED	3 (84)	60 (1,593)	6 (164)	69 (1,841)
	MAX	0 (0)	1 (340)	8 (245)	9 (245)
	OSL Dist.	14 (399)	72 (1,876)	14 (378)	100 (2,653)