Self-injury incidents in correctional service of canada institutions over a thirty-month period (Full report)

Number: R-233 - Summary

Date: December 2010

Note: Arthur Gordon, Ph.D. Health Services, Correctional Service Canada

Alternative format:

Table of Contents

• Acknowledgements
• Executive Summary
• List of Tables
• List of Appendices
• Introduction
• Method
• Discussion
  • Is OMS Sufficient to Identify Self-Injury?
  • Location of Incidents
  • Degree of Harm
  • Self-Injury and Suicide
  • Impact of Self-Injury Incidents on Staff
  • Future Directions
• References

List of Tables

• Table 1 Coded Variables from OMS Including Definitions and Coding Descriptions
• Table 2 Self-Injury Incidents As Reported in OMS-SITREP
• Table 3 Distribution of Self-Injury Incidents by Region and Gender
• Table 4 Institutional Self-Injury Incidents by Race
• Table 5 Self-Injury Incidents by Institution Type Over a 30-Month Period
• Table 6 Change in Number of Self-injury Incidents Over Time by Region (OMS-SITREP data)
• Table 7 Method of Self-Injury Incidents As Reported by SITREP and OMS-SITREP
• Table 8 Methods of Self-injury in Men and Women
• Table 9 How the Self-injury Incident Was Reported
• Table 10 Degree of Harm Caused by Self-injury Acts
• Table 11 Distribution of Incidents - Data from OMS-SITREP

Acknowledgements

The author would like to express a sincere thanks to Debra Coradazzo and Linda Fulton for their help writing the OMS queries. Also thank you to Sarah Wootten and Nicholas McAlister for help reviewing the incidents and to Lynn Stewart for her incredible editorial and analytical input.

Executive Summary

Between April 1, 2006, and September 30, 2008, 1,230 self-injury incidents were reported in the Offender Management System (OMS), a national database that contains information on all federal offenders, and/or in the Situation Reports (SITREP), which are daily reports designed to keep senior managers abreast of significant incidents across CSC. Descriptions of the incidents based on these sources were examined. The following are the main findings:

• The number of self-injury incidents across Correctional Service of Canada (CSC) that were reported increased by 73% during the 30-month study period from 197 incidents (between April and September, 2006) to 341 incidents (April - September 2008). The pattern of change is similar whether the data are based on SITREP alone or the more complete dataset (OMS-SITREP) used in this study. The data do not, however, determine what factors drove this increase. The number of self harming incidents may be due to an actual increase in self injurious behaviour related possibly related to a changing offender profile, or to greater staff awareness and changes to reporting standards that occurred within CSC during this time period, or, indeed, other factors.
• Women offenders were proportionately more likely than male offenders to engage in self-injurious behaviours. Women were more likely than men to self-injure more than once.
• The number of self-injury incidents among Aboriginal offenders was greater than would be expected based on their population in the institutions.
• The five treatment centres and the maximum/multi-security institutions had the most self-injury incidents, the most offenders who self-injured, and the most offenders who self-injured repeatedly.
• Male offenders tended to self-injure by slashing or overdosing, while women offenders showed higher use of ligatures or head-banging.
• Ninety percent of self-injury incidents resulted in no or minor injury to the offender.
• Sixty-seven percent of offenders self-injured only once during the study period. While a similar percentage of men and women offenders engaged in self-injury more than once, women who repeatedly self-injured had more such incidents than did men.
• Twenty-six of the 1,230 self-injury incidents resulted in death. However, 77% of the inmates who died had no previous documented incidents of self-injury during their incarceration suggesting that self-injury appears to be a distinct phenomenon that should be studied independently of attempted suicide.
• Based on the results of this report, it can be concluded that data extraction through OMS queries is not sufficient to provide an accurate measure of self-injurious incidents due largely to variability in how such incidents are coded in OMS. Creating a field in OMS that would indicate whether an incident involved self-injury would provide a more efficient and reliable measure of self-injury. In addition, future research should focus on the psychological and behavioural characteristics of self-injuring offenders to better inform prevention, management and treatment options.

Introduction

Improving the Correctional Service of Canada's (CSC's) capacity to address the mental health needs of offenders has been a priority for the Service for several years. Addressing the mental health needs of offenders creates safer institutions and enhances successful reintegration, which contributes to public safety. CSC is required to provide health care services directly to its offenders, and is legislatively mandated to do so through the Corrections and Conditional Release Act (CCRA 1992, c.20 (Revised 1996)) which states that:

• 86 (1) The Service shall provide every inmate with
• (a) essential health care, and
• (b) reasonable access to non-essential mental health care that will contribute to the inmate's rehabilitation and successful reintegration into the community.

Recently, significant gains have been made in strengthening the continuum of mental health care for offenders through the implementation of a more comprehensive mental health screening process, building the capacity in regular institutions to respond to mental health needs, and improving mental health training for mental health and correctional staff working in both the institutions and the community. Despite these gains, improving the management and treatment of offenders who engage in self-injurious behaviour (SIB) continues to be a challenge that CSC is committed to addressing.

The safety and well-being of offenders under CSC jurisdiction is paramount. Moreover, dealing with self-injuring individuals is extremely stressful for staff and can equally have psychological effects on other offenders who witness or hear of such actions. The impact is significantly greater when individual offenders engage in repeated acts of self-injury, as staff must remain vigilant, working toward ensuring that the next act will be interrupted in time for a safe resolution.

An earlier study (Gordon & Laishes, 2009) analyzed incidents of self-injury reported in the Situation Reports (SITREP) over a 30-month period. That study concluded that self-injury appeared to be a growing problem across CSC, that women were relatively more likely to self-injure than men, that men and women engage in different types of self-injurious acts, and that acts of self-injury, even among those offenders who self-injured repeatedly, did not seem related to suicide attempts. A significant limitation to this study was that only those incidents identified in SITREPs were considered. These reports are issued daily and are designed to keep senior managers abreast of significant incidents across CSC. To that end, National Headquarters staff review all reports submitted by regional duty officers on a daily basis and prepare and disseminate a summary of these incidents. Given that there are over 17,000 reportable incidents in CSC each year, the SITREP report cannot hope to capture more than a sampling of the most significant events. Therefore, it was recognized that using data drawn solely from SITREP concerning the incidence and nature of self-injurious events may not provide completely accurate information about self-injury in CSC.

The Offender Management System (OMS), a database that captures all relevant information on offenders throughout their sentence, may be a better source of data. All reportable incidents lead to a situation report being entered in OMS. These situation reports provide more detail about the incidents than can be captured in a SITREP summary. Moreover, OMS is a searchable database so that relevant incidents can be identified more easily through queries. In short, the OMS is a more robust and complete source of information about individual offenders and incidents.

Consistent with a growing consensus in the literature, in this study SIB will refer to acts that cause or have the potential to cause, immediate physical harm. Thus, behaviours such as slashing that cause immediate damage are included in this definition, as well as behaviours such as the use of a ligature that is apprehended quickly that does not cause any immediate harm but has the clear potential to do so. Behaviours that can cause harm but only over extended periods (e.g., history of substance abuse) are not considered to be SIB. This definition of SIB does not assume that the self-injurious acts were carried out in order to end life (i.e., suicide), nor does it exclude the possibility that a self-injurious act was suicidal in intent (for a more detailed discussion, see Power and Brown (2010), which provides an excellent review of the literature, discussing definitional issues).

One common form of SIB, namely drug overdose, deserves additional comment. Despite CSC's concerted efforts to eliminate drugs from its prisons, some inmates obtain and use street drugs. Other inmates may obtain quantities of prescription drugs, either from other offenders or by accumulating their own prescriptions. While overdosing on such drugs may well lead to immediate harm, it is clear that many offenders ingest drugs in order to "get high" rather than to cause injury, and that any harm caused may be accidental. Some may argue that acts motivated by such hedonistic interest should not be equated with acts clearly designed to injure (e.g., slashing). Unfortunately, establishing the motivation for drug-related incidents is beyond the scope of this study. Moreover, our definition of self-injury relates to potential harm (which a drug overdose implies) rather than intent. To that end, the present study treated all overdose situations as self-injurious, while recognizing that the motivation behind many of these acts may have been to create pleasure rather than damage.

The present study sought to replicate and extend the review by Gordon and Laishes (2009) by drawing data on self harming incidents from the OMS. Retaining the same 30-month study period as used previously allows for direct comparisons between the results provided by the two data sources. This report reviews OMS files on self injury incidents, providing details about the self-injurious actions, degree of harm caused, and staff response.

Method

Data extraction for the present study involved a number of steps after it became clear that simply conducting a query for self-injury incidents in the Offender Management System (OMS) would not accurately identify all relevant incidents. A full description and discussion of these steps is included in Appendix A.

All incidents in CSC are documented in OMS and each is labelled with a category according to the nature of the incident. As a first step, an OMS query was executed for the period April 1, 2006 through September 30, 2008 using the categories most likely to be relevant to self-injury. These categories included self-inflicted injury, overdose interrupted, suicide, death (overdose, unknown causes, other ^{Footnote 1}), and medical emergency. In addition, two word strings (self harm*, auto-mutilation) were included in the query to identify relevant incidents through their description in the incident synopsis. The query identified 1184 incidents.

Every query incident was reviewed manually to ensure that each described a valid self-injurious event. As is detailed in Appendix A, 113 incidents were eliminated from the data base because: a) they were duplicates of the same incident (n=37), b) they described medical emergencies that were unrelated to self-injury (n=32), c) they were categorized as "death" but were clearly not related to self-injury; some incidents were not even related to death^{Footnote 2} (n=15), and d) the incident category was relevant (e.g., self-inflicted injury), but the situation report did not provide any evidence that self-injury had occurred (n=29). Degree of harm was assessed based on the statements of the staff completing the reports. Those incidents where staff did not provide a clear statement that fit one of the categories of degree of harm were coded "not stated".

In order to capture incident-specific information from the earlier analysis of SITREP data, an attempt was made to merge cases from the two sources (i.e., SITREP and OMS). As a result, 159 incidents that had been described in SITREP but were not identified through the OMS query were added to the final database. The most common discrepancy between OMS- and SITREP-identified incidents was that the latter identified 139 cases that clearly described self-injury but had been categorized using labels not included in the OMS query. For example, OMS labelled these incidents as a disciplinary incident (n=30), a cell extraction (n=21), a staff or inmate assault (n=14) or an "other" incident (n=62). Other OMS categories included minor disturbance, possess contraband or under the influence. Still other SITREP incidents that were added to the database did not have a corresponding OMS entry but were clearly relevant self-injury events.

Taken together, these steps resulted in a final database of 1,230 self-injurious incidents reported between April 1, 2006 and September 30, 2008. Because the final database was a combination of two data sources, the present paper will refer to the result as the OMS-SITREP database to distinguish it from data derived solely from SITREP.

Each incident was then coded on a number of variables. Some of this information was available directly from the OMS query (e.g., date and location of incident, race) while other information was extracted by reviewing the full OMS file (e.g., self-injuring acts, degree of injury). Table 1 identifies the variables coded including a definition and coding description.

Discussion

The present study sought to replicate and extend the results of Gordon and Laishes' (2009) analysis of self-injury incidents by focusing on incidents as reported in OMS. For the most part, the results of the two studies are highly similar. Both data sources demonstrate that the reported incidence of self-injurious incidents increased markedly over the 30-month study period. Notably, incidence data from SITREP correlated very significantly with data from OMS-SITREP suggesting that both data sources are identifying and describing similar trends.

For the most part, data from SITREP and OMS-SITREP led to similar conclusions about the nature of self-injurious actions. Analysis of both data sets showed that women offenders were involved in proportionately more incidents than men, and that men and women offenders differed in terms of the types of self-injurious actions they were likely to be involved in with men more frequently slashing and women using of a ligature. Also, both analyses described a small group of offenders, both men and women, who were responsible for a disproportionately high number of self-injury incidents.

The two data sources did yield somewhat different results on some dimensions. By considering a much broader range of incidents, the OMS-SITREP data suggested that self-injurious incidents were increasing more rapidly in some CSC regions (i.e., Quebec and Pacific), that slashing is a more common method of self-injury than had been assumed, and that relatively more incidents were reported to staff than by the self-injuring offender. OMS also proved a much richer data source to assess issues around discovery of incidents and degree of harm. In short, while SITREP can provide a reliable source of information about overall trends, the richness of OMS can provide more detail about the self-injury incidents themselves. However, it must be noted that some of the most relevant data came from manually reviewing each OMS file rather than from the OMS query.

Is OMS Sufficient to Identify Self-Injury?

While OMS-SITREP clearly identified many more relevant self-injury incidents than did SITREP alone, it appears that OMS queries do not provide totally reliable, accurate and complete data. Almost 10% of the cases identified as relevant by the OMS query had to be eliminated, largely because an in-depth review of each incident revealed that the incidents were not related to self-injury. An additional 159 legitimate incidents were identified through SITREP but were not identified through the OMS query. As is detailed in Appendix A, the major drawback to relying on OMS queries is that many self-injury incidents are labelled in categories that would not normally be associated with self-injury (e.g., disciplinary problem, cell extraction, assault on staff). Even though such categories are usually understandable and correct in the context of the incident, these categories would not likely be included in a search query for self-injurious incidents. One could include these apparently unrelated categories in a query and then manually delete those cases which were not relevant, but such a process would be extremely inefficient. A better solution would be to add a self-injury data field to OMS whereby each incident would be identified as involving some form of self-injury (or not). This additional variable would increase the likelihood that all relevant incidents could be identified through a single query, regardless of how the incident was categorized. While it is understood that adding fields to a database as complex as OMS is a major undertaking, being able to accurately track incidents as important as self-injury may well be worth the effort.

Location of Incidents

While almost all major men's and women's institutions experienced some self-injury incidents, such events were more likely to occur in treatment centres or maximum/multi-security institutions. Even at the same security level, some institutions experienced more incidents than others. Thus Kingston Penitentiary (93 incidents) and Kent Institution (83 incidents) had twice the self-injury incidents as the average for maximum security facilities (M = 42.6). Similarly, Shepody Healing Centre (82 incidents) and RPC - Prairies (72 incidents) had the most incidents of the treatment centres (overall M = 55.4 incidents). Only one medium security institution (Dorchester Institution) stood out as having many more incidents than its counterparts (73 incidents versus a mean of 18.2 for all mediums). Moreover, these same institutions tended to house a greater number of offenders who self-injured repeatedly (see Appendix B for details on all institutions).

It is not clear why some institutions experience relatively more self-injury incidents. It seems most likely that each region evolves a practice of housing self-injuring offenders in the one or two institutions that may be in a better position to manage them, if only because of past experience dealing with the behaviour. Such seems to be the case in the Atlantic region where most offenders who self-injure tend to be sent to either Shepody Healing Centre or its co-located partner, Dorchester Institution. Minimum security institutions, possibly because of the vetting process for transfer to these facilities, are much less likely to house inmates with histories of self injury. In any case, the present data do identify those institutions that could currently benefit most from the resources necessary to manage this challenging population.

It is also not clear why so many self-injuring offenders are classified as maximum security. Certainly, repeated self-injury suggests the need for more stringent controls offered in maximum security institutions. It remains to be seen whether other aspects of these offenders' behaviour (e.g., violence toward others, high risk and high needs profiles) also warrant that classification. Again, the Atlantic region stands apart as many of the self-injuring offenders seem to be located at Dorchester Institution, a medium-security institution, rather than at the maximum security Atlantic Institution.

One could argue that the increased confinement and reduced social contact of a maximum security environment is not well suited to offenders who are displaying self-injurious behaviour (Konrad et al., 2007). In fact, the close confinement (including frequent segregation status) that many self-injuring offenders experience may actually increase the likelihood of further self-injury (Dear, 2006). Further research to address the impact of such environmental factors on incidence of self-injury would be useful and important in developing strategies about how self-injuring offenders should best be managed.

Degree of Harm

The present data suggest that the vast majority (90%) of self-injurious incidents result in at worst, minor injury. While this finding is consistent with the conclusion that most self-injurious acts are not intended to end life, it would be a mistake to conclude that one could or should reduce vigilance. Some incidents (e.g., ligature) may result in no physical harm if apprehended quickly, but could be fatal if intervention occurs only moments later. For example, of the 172 apprehended ligature incidents, 159 (85%) resulted in minor or no harm. On the other hand, 54% of those self-injurious acts that resulted in death involved a ligature. Thus, relative lack of injury from use of ligature and other forms of self-injury (e.g., overdose, slashing) may be more due to the diligence and quick intervention of staff rather than the inherent risks in the act itself. Unfortunately, while 34% of incidents are reported to staff by the offender him/herself thus ensuring rapid staff intervention, offenders who used a ligature self-reported only seven percent of the time, thus putting even more pressure on staff vigilance.

Self-Injury and Suicide

The present results support and extend Gordon and Laishes' (2009) conclusion that most self-injurious acts do not seem to be designed to end life. A similar conclusion has been reached for self injurious behaviour in the broader community (Suyemoto, 1998). In the present case, 77% of the offenders whose incidents resulted in death had no previous self-injury incidents throughout their current incarceration. Of the six who had previously self-injured, two had only one previous incident, one had two incidents, and the most recent previous incident for five of the six was at least a year prior to the fatal incident. Again, these data should not be taken to minimize the seriousness of the problem of self-injury. Rather, the data suggest that self-injury may involve a very different set of dynamics, predictors and potential interventions than do suicidal behaviours.

Impact of Self-Injury Incidents on Staff

As the number of self-injury incidents increases, so might the impact on staff who deal with these offenders, particularly those who self-injure repeatedly. There appears to be little research on the impact on staff of dealing with self-injury, but a recent study suggests the psychological impact may be significant (Marzano, & Adler, 2007). As was noted earlier, some forms of self-injury can be particularly lethal if not apprehended quickly. In the absence of effective tools to prevent such incidents, front-line staff must be constantly vigilant waiting for the next incident, hoping their intervention will be timely.

CSC has a well-established Critical Incident Stress Management (CISM) program that helps staff deal with a wide variety of stressful incidents. During the present study, we reviewed 593 self-injury incidents to determine if CISM had been offered to staff. In only 4% of cases was there a clear affirmative attestation. This result should not be taken to imply that CISM was not offered in 96% of cases. Rather, the situation report may have simply not made reference to such activity even if it had occurred. However, it is possible that self-injury is sufficiently common that it may not be seen as a critical incident requiring CISM intervention. Moreover, one might question whether CISM is always the most relevant intervention, or whether other approaches may better help staff deal with these highly stressful situations. CSC's Mental Health Branch is currently developing strategies to ensure that interdisciplinary treatment and management plans are in place for higher-risk self-injuring offenders. Having all staff use similar approaches in managing these offenders, along with the support of dealing with such offenders as part of a team, may help ameliorate the impact of these incidents on staff. Given the crucial role that front-line staff play in limiting damage from self-injury, it is essential that they be given the support and skills necessary to optimize their performance. This is also an area were additional research would be helpful in determining approaches to improve CSC's response to this important discussion.

Future Directions

The present study replicates Gordon and Laishes (2009) in demonstrating an increased number of self-injurious incidents across CSC over a recent 30-month time period. These studies have not addressed the possible reasons underlying this increase. It may be as simple as increased or improved reporting of such incidents over time. Alternatively, it is recognized that the nature of the CSC population has shifted over time to younger offenders serving shorter sentences. It is possible that some feature of the changing population correlates with increased likelihood of self-injury. In fact, self-injury is typically associated with younger perpetrators (Cutter, Jaffe & Segal, 2008). It is also possible that there may be more tension in institutions and more vulnerable offenders may be self-injuring to either cope with the stress or seek to escape it. Such a dynamic has been described in the literature (Dear, 2006), but it is not clear what the source of such systemic tension might be. Unfortunately, the present study does not have data that speak to these or other possible reasons for the increased rates of self-injury. It is important for research to address the issue as the results may suggest systemic approaches to ameliorate the problem.

There are also many questions that research could address to understand the self-injuring offender and the dynamics of such acts in prison. Determining what motivates self-injury should help define appropriate intervention strategies. It is far from clear that the offenders who self-injured during the present study period represent a homogeneous group. Do those who self-injure infrequently do so for different reasons then those offenders who repeatedly self-injure? Do CSC's current interventions (e.g., psychological services) with infrequent self-injurers contribute to their low rate of self-injury? Research should determine also the offender's mental health and cognitive status. Clinical experience suggests that many repetitive male self-injurers tend to be cognitively impaired, suggesting that treatment may have to address specific learning impediments. Unfortunately, there is relatively little research on self-injury adult prison populations (Power & Brown, 2010) but CSC researchers have several relevant studies underway.

The Correctional Service of Canada, like many correctional jurisdictions, is facing a significant challenge to address self-injurious behaviour in its prisons. While research is needed to better inform management and treatment strategies, the reality of prisons is that one cannot always wait for research before intervening. To that end, the Mental Health Branch of CSC is developing a multifaceted strategy that will include 1) developing a systematic approach to respond to each case of self-injury which may include developing clinical management plans for those deemed to be at a higher risk for continued self-injury, 2) developing regional resources to assist institutions in developing treatment plans and management strategies, 3) piloting a formal treatment program for the highest frequency self-injurers, 4) studying the impact of intervening with self-injury on both operational and clinical staff and developing appropriate strategies to minimize impact, and 5) continuing to promote and provide relevant research activities to help inform the preceding steps.

References

Cutter, D., Jaffe, J., & Segal, J. (2008). Self-Injury: Types, Causes and Treatment. Retrieved from http://www.helpguide.org/mental/self_injury.html.

Corrections and Conditional Release Act, C.C.R.A. 1992, c.20 (Revised 1996).

Dear, G. E. (Ed)(2006). Preventing Suicide and Other Self-Harm in Prison. New York: Palgrave Macmillan.

Gordon, A., & Laishes, J. (2009). A Study of Reported Self-Harm Incidents in CSC. Retrieved from Correctional Service Canada website: http://infonet/Corporate/National/Resources/publications1117.htm?lang=en#R

Konrad, N., Daigle, M., Daniel, A., Dear, G., Frottier, P., Hayes, L., Kerkhof, A., Liebling, A., & Sarchiapone, M. (2007). Preventing suicide in prisons, Part I: Recommendations from the International Association for Suicide Prevention Task Force on Suicide in Prisons. Crisis: The Journal of Crisis Intervention and Suicide Prevention, 28(3), 113-121.

Marzano, L., & Adler, J.R. (2007). Supporting staff working with prisoners who self-harm: A survey of support services for staff dealing with self-harm in prisons in England and Wales. International Journal of Prisoner Health, 3(4), 268-282.

Minister of Public Works and Government Services Canada (2007). Annual Report of the Office of the Correctional Investigator 2006-07. Ottawa, ON.

Power, J. & Brown, S. L. (2010). Self-Injurious Behaviour: A Review of the Literature and Implications for Corrections. Ottawa, ON: Correctional Service of Canada.

Suyemoto, K.L. (1998). The functions of self-mutilation.. Clinical Psychology Review, 18(5), 531-554.

Appendices

Appendix A: Methodology for Tracking Self-injury Incidents in CSC

One important component of a more comprehensive strategy to manage self harm in institutions is an efficient and reliable means to track such incidents over time. These data are necessary to describe the problem, suggest possible interventions, as well as measure the impact of such interventions over time. Indeed, in his 2007-08 annual report, the Correctional Investigator recommended that CSC "(e)stablish a consistent framework for recording and reporting attempted suicides, self-inflicted injuries and overdoses"^{Footnote 10}. A self-injury tracking system must provide quick, reliable and accurate data. The popular consensus seems to be that appropriately crafted OMS queries will meet these conditions. The present study evaluated this assumption by trying to identify all self-injury incidents in CSC institutions between April 1, 2006 and September 30, 2008.

OMS query by incident type

An OMS query spanning the period April 1, 2006 through September 30, 2008 captured incidents that would most likely involve self-injury. The query included the following incident categories^{Footnote 11}:

• Self-inflicted injury
• Overdose interrupted
• Suicide
• Death (overdose, unknown causes, other)
• Medical emergency

This query yielded 1127 incidents. However, from our experience with SITREP data, we recognized that some self-injury incidents are assigned to categories that one would not immediately identify as being related to self-injury (e.g., cell extraction, other incidents, etc). To attempt to capture such incidents, we included the word strings "self harm*" and "automutilation" as a search variable in the OMS query. This addition should identify those incidents that refer to self-injury in the incident synopsis regardless of the categorization of the incident.

This strategy did indeed identify an additional 57 self-injury incidents^{Footnote 12}. These additional incidents had been categorized in OMS as "disciplinary problem" (n=26) "other incident" (n=20), "cell extraction" (n=4), "assault staff" (n=3), "possession of contraband" (n=2), and "threaten staff" and "fire" (n= 1 each). In total then, the OMS query identified 1184 self-injury incidents.

"Cleaning" the OMS data

The next step was to review each of the 1184 identified incidents to ensure that they were valid (i.e., each involved acts of self-injury). To that end, each incident synopsis and, where necessary, the OMS incident summary document, was reviewed to ensure that the incident did indeed involve an act of self-injury. As a result, 113 incidents (9.5% of the total) were eliminated from the data set for the following reasons.

a) Duplicate incidents: When an incident involves more than one offender, the same incident information is logged into the OMS record of all involved offenders. Thus, if one offender self-injured, but two others were in some way involved (e.g., as bystanders), the self-injury incident would be recorded in OMS for all three offenders^{Footnote 13}. Moreover, the query would capture the same incident three times. We deleted 37 incidents because they were duplicates that did not involve an offender who self-injured.

b) Medical Emergency: Although some incidents labelled as medical emergencies clearly involve self-injury, other medical emergencies may be unrelated. We eliminated 32 incidents that were unrelated to self-injury while retaining eight relevant incidents.

c) Death: Although some self-injury incidents may well result in death, some OMS incidents categorized as "death" may not have involved self-injury. We eliminated 4 incidents because they were duplicate entries, 6 incidents because death was clearly related to an existing medical condition or natural causes, and 5 incidents because the events had nothing to do with self-injury or death^{Footnote 14}.

d) Mis-categorized Incidents: In reviewing the incidents, it became clear that even if the incident category was very relevant (e.g., self-inflicted injury, suicide), the incident itself might be unrelated to self-injury. For example, the synopsis for an incident labelled as O.D. Interrupted read: "ION Scanner low reading for cocaine for Inmate XX's visitor. Open visit was permitted." Another "self-inflicted injury" incident reported that the "Inmate picked up a plastic chair and threw in the general direction of another inmate in the yard." In total, 29 such incidents were eliminated as being invalid reports of self-injury.

Based on this review, the self-injury database was reduced by 113 to 1071 incidents. That is, OMS-reported incidents proved invalid 9.5% of the time. Moreover, given the time required to review all 1184 incidents, the query approach did not prove an efficient means of identifying only legitimate self-injury incidents.

Matching OMS and SITREP data

Gordon and Laishes (2009) identified 510 institutional self-injury incidents reported in SITREP and captured considerable data on these events. As the same data was required for the analysis of cases extracted through the OMS query, we attempted to merge the SITREP cases with the 1071 OMS incidents, fully expecting that the OMS query would have captured all of the cases reported by SITREP. In fact, only 328 SITREP cases had an OMS-query match. Fully 182 self-injury incidents described in SITREP were not identified through the OMS query.

To understand this discrepancy, each unmatched SITREP incident was reviewed to determine why the OMS query had not captured it. This review revealed the following:

a) Inconsistent dates: In 12 cases, the incident date reported by OMS and SITREP were different, but the incident itself appeared to be the same.^{Footnote 15} We adjusted the final database to ensure that each of these incidents was counted only once.

b) No incident in OMS: In 13 cases, OMS showed no record of an incident of any sort at or around the date reported by SITREP. The reasons for these discrepancies are not immediately clear. In three cases, the offender was housed at Pinel Institute and it may be that these incidents did not get reported in OMS. Additionally, it is possible that the Regional Duty Officer reported these incidents to those creating SITREP, but OMS reports were never completed by the institution.

c) Categorization issues: The OMS query did not identify 139 cases listed in SITREP because the OMS report categorized the incident using a label that was not included in the query. Most commonly, these incidents were categorized in OMS as "other incident" (n=62), a "disciplinary incident" (n=30), a "cell extraction" (n=21), or a "staff" or "inmate assault" (n=14). Other categories included "minor disturbance", "possess contraband" or "under the influence".

Notwithstanding the OMS categories used, the behaviours described in these 139 incidents were all were legitimately self-injurious, involving acts such as slashing, head banging, use of ligature and overdose. Obviously, it would be very inefficient to write an OMS query to include categories such as "disciplinary" or "other" incident if one then had to eliminate manually the resulting irrelevant cases.

As a result of the SITREP/OMS reconciliation, 159 SITREP cases were added to those extracted through the OMS query to yield a final database of 1230 self-injury incidents between April 1, 2006 and September 30, 2008.

A major question remains: can we assume that the 1230 incidents identified by combining valid OMS-derived incidents with those identified through SITREP represents an accurate and complete count of self-injury incidents? While combining the data from OMS and SITREP clearly provides a more accurate measure than either source alone, we cannot assume that combining data from both sources successfully identified all relevant incidents. Clearly, SITREP is better at identifying those incidents that OMS does not categorize as being self-injury related (e.g., "other" incidents, "disciplinary" incidents). However, SITREP is also much more selective in reporting incidents. In the current situation, SITREP reported on only 41% of the total 1230 self-injury incidents. It remains possible that we have not captured those self-injury incidents that a) were not considered appropriate for reporting in SITREP but b) were categorized in such a way that the OMS query did not identify them. Short of manually reviewing every incident in all OMS categories (over 17,000 in 2008-09), the approach taken in the present study may represent a conservative, best estimate of self-injury incidents.

We began this exercise by attempting to identify all self-injury incidents in CSC institutions over a 30-month period. Table A1 summarizes the steps taken and the impact on the number of incidents identified. Depending on which steps one adopted, the number of self-injury incidents in our target period could range between 1071 and 1230. Moreover, we added and/or deleted 329 incidents at various stages. While each step in this process seems necessary to derive an accurate and valid estimate of self-injurious incidents, these steps clearly do not meet the criterion of providing a quick and easily obtained result. Unfortunately, the data also suggest that, while an OMS query provides an efficient means of extracting data, one cannot assume the data are completely accurate or valid. An ideal solution would combine the efficiency of an OMS query with the accuracy afforded by reviewing individual incidents, regardless of how they are categorized.

The main impediment to developing an appropriate self-injury tracking tool would seem to be the categories applied to such incidents, particularly in OMS. This problem has at least three aspects. First, staff may select an appropriate category to reflect self-injury (e.g., self-inflicted injury), but the incident itself may not be self-injury related. This problem may be resolved through training. But to the extent that it does occur (and it did over 30 times in the present project), one cannot rely on extracting OMS data based solely on incident type without reviewing each incident to ensure that it does in fact involve acts of self-injury.

A second concern is that the OMS category chosen may be appropriate to describe the self-injury incident, but the category is too broad and captures incidents unrelated to self-injury. A prime example of this is the "medical emergency" category defined as "An event which requires medical intervention, which is not attributable to assaultive behaviour and which normally requires outside treatment with or without hospitalization"^{Footnote 16}. By this definition, one could legitimately classify many self-injury incidents as "medical emergencies". However, many other events totally unrelated to self-injury (e.g., heart attack, seizure) might equally be classified as a medical emergency. Thus, one would need to review manually each incident identified by the OMS query to determine those actually related to self-injury.

The most common problem and impediment to an effective tracking tool is the use of incident categories that appear to be unrelated to self-injury. The present study found numerous self-injury incidents categorized as "disciplinary", "cell extraction", "assault" or "other^{Footnote 17}". In most cases, the category chosen was appropriate to the incident. For example, the response to an offender's self-injuring behaviour may have required a cell extraction, or the offender may have assaulted staff as they tried to intervene. However, one would not likely include such categories in an OMS query to identify self-injury incidents. Moreover, self-injury is likely involved in very few cell extractions or staff assaults.

The recent (July 2008) reissuing of CD 568-1 including definitions of incident categories is not likely to address the concerns raised here and may, in fact, exacerbate the problem. Thus, the CD defines "self-inflicted injuries" as "The deliberate harm of oneself without the intent to commit suicide as determined by a mental health professional (i.e. Psychologist or Psychiatrist)". Similarly, the category "suicide attempt" also requires a mental health professional to make a determination of intent. The requirement that a mental health professional determine motivation and intent may discourage staff from using this category, particularly for offenders who engage in self-injuring behaviours frequently. If these categories are not readily available, staff may increasingly choose other categories that may or may not be related to self-injury (e.g., non-serious bodily injury). Thus, it appears that issues around categorizing incidents may continue to impede development of a quick and reliable method to track self-injury incidents.

The concerns described above could be addressed if OMS contained a searchable field that asked the question "Did this incident involve some form of self injury?" A 'yes" in this field would identify a self-injury incident independently of the category chosen to describe the incident. For example, this field could readily distinguish between relevant and irrelevant "medical emergencies" or "cell extractions". It would require very little additional work from staff inputting the data (but staff would have to check "no" for most incidents). Above all, this field would allow for determining the total number of self-injury incidents based on a one-variable query. That is, we would have a tracking system that is efficient, reliable and valid. It remains to be seen whether, from an OMS perspective, this solution is also feasible. However, the currently data network based on OMS clearly does not meet CSC's need for an efficient, reliable and effective self-injury tracking system.

Appendix B: Self Injury Incidents by Institution