Self-harm in a prison mental health unit: a pilot trial of a structured professional judgement approach to risk assessment and safety planning Open Access

Internationally, rates of self-harm (SH) and suicide in prison are known to be significantly higher than in the general population (Fazel et al., 2016; Hawton et al., 2014). Demographic groups with elevated suicide risk (e.g. males, minority ethnic groups including First Nations people in Australia, and the socioeconomically disadvantaged), are disproportionately represented within the prison population (Australian Institute of Health and Welfare, 2023b). Those in prison also have elevated rates of factors known to be associated with SH/suicide risk, including mental illness, substance use problems and trauma (Gottfried and Christopher, 2017).

Despite the well-established increase in risk, evidence to support approaches to reducing SH and suicide in prison settings is lacking. A recent systematic review noted a limited ability to draw conclusions on the efficacy of interventions intended to reduce suicidal thoughts and SH in criminal justice settings due to a lack of high-quality evidence (Carter et al., 2022). However, in the absence of a strong evidence base, there are calls for a multi-factored approach to risk reduction, including universal means restriction, risk screening and assessment, improved and increased staff training, increased social support, continued monitoring of people at high risk and the provision of mental health treatment (Barker et al., 2014; Stijelja and Mishara, 2022). Screening to identify those at increased risk has long been considered an important first step in reducing the risk of SH and suicide in prison settings (Gallagher and Dobrin, 2005) but systematic reviews have so far concluded that no “gold standard” approach exists (Gould et al., 2018; Perry et al., 2010). Furthermore, there has been limited focus on post-screening assessment and management of identified risk of SH/suicide (Bruno et al., 2024; Carter et al., 2022).

Outside of prison settings, structured professional judgement (SPJ) approaches, more often used in the context of clinical violence risk assessment, have been increasingly recommended for SH/suicide risk assessment and management (Gray et al., 2021). The SPJ approach provides a structured evidence-based framework to guide clinical formulation, management plans and judgements in relation to risk, as well as increasing the transparency of clinical decision-making. The focus is on identification and exploration of evidence-based risk factors as they guide understanding of risk of SH/suicide for an individual patient to develop an effective management plan to reduce, rather than attempt to predict, such outcomes. SPJ approaches also allow for individual, case specific risk markets to be considered. However, existing guidelines for taking an SPJ approach to suicide risk assessment and management focus on risk factors relevant to the general population or inpatient psychiatric settings (Bouch and Marshall, 2005; Fagan et al., 2009; Gray et al., 2021). These guidelines may not be relevant to the prison environment, particularly as it is known that there are risk factors for SH or suicide that are unique to the prison population (Fazel et al., 2008). These include factors such as having a serious (e.g. murder/manslaughter charge), remand/unsentenced status, long/life sentence, violent offense or single cell occupation.

The current study describes a pilot trial of a health service-improvement initiative intended to provide a structured framework to support mental health clinicians in assessing and managing risk of SH/suicide in a prison setting. The Suicide/Self-Harm, Legal, Individual, Psychiatric, Safety Plan (SLIPS) framework was developed by a team of clinician-researchers (psychiatrists and psychologists) with experience in the clinical assessment and management of suicide/SH risk in prison and forensic mental health settings. Specifically, the team used an evidence-based expert consensus approach, whereby published literature on risk factors for SH /suicide in prison settings was reviewed and integrated with input from clinical and academic experts. SLIPS thus takes a SPJ approach, incorporating assessment of risk factors identified within the literature as being associated with suicide and SH in prison settings, providing structure but not restricting clinical judgement.

In addition, the final stage of the SLIPS framework explicitly links the structured risk assessment to a risk management approach in the form of collaborative safety planning, an evidence-based intervention for reducing risk of suicidal/SH behaviours (Stanley and Brown, 2012). One recent meta-analysis by Nuij et al. (2021) examined safety-planning type interventions and found evidence for preventing suicidal behaviour. However, only six studies were included in the review, and these studies were conducted with military veterans, active-duty soldiers, or adults reporting to ED/s services. As far we are aware, there is no existing research on the efficacy of using collaborative safety planning in prison settings. This need for further research in this area was highlighted by Doupnik et al. (2020) as a in their systematic review and meta-analysis on suicide prevention interventions in acute care settings.

In the current pilot trial, the incidence of recorded SH behaviours among people detained in a prison mental health unit is examined over a three-year period, including one year following the introduction of the SLIPS intervention.

The mental health screening unit (MHSU) is a 43-bed mental health unit for male patients situated in a prison setting in the Metropolitan Remand & Reception Centre (MRRC) in Sydney, New South Wales (NSW), Australia. Although operationally run by Corrective Services New South Wales, health services in NSW prisons are provided by a specialist health network, the Justice Health and Forensic Mental Health Network (Justice Health NSW).

Mental health care and treatment in the MHSU is provided by a team of Justice Health NSW psychiatrists and mental health nurses. Sentenced and remand individuals in prison of any classification or protection status can be accepted to the MHSU if they present with acute mental illness, have a suspected mental illness and are at risk of harm to self or others (determined on screening), require assessment and commencement of treatment for major mental illness or have established mental illness and are awaiting placement in a declared mental health facility. Demand for beds in the unit is consistently high and referrals to the unit can occur upon initial reception to the MRRC or from health centres in adult male correctional centres around the state of NSW. Referrals are prioritised at a weekly bed demand meeting, focusing on the clinical needs and risks of the patients on the waitlist (Inspector of Custodial Services, 2021).

For the current study, we included all individuals who were admitted to the MHSU at any point during the three-year study period from 1 August 2020, through 31 July 2023. The only exception to this were two individuals who arrived in the unit on the last day of data collection (i.e. days in unit recorded as 0).

The SLIPS framework uses an SPJ approach to support mental health clinicians to undertake the assessment and management of risk of SH/suicide in prison (Supplementary File 1). The acronym SLIP refers to the four categories of risk factors for SH/suicide – Suicide/Self-Harm, Legal, Individual and Psychiatric, with the final “S” representing Safety Plan, the framework for managing SH/suicide risk:

• Suicide or SH factors, including a history of self-injurious behaviours and suicidal ideation, have a well-established link to future SH and suicide in prison settings (Favril et al., 2020; Zhong et al., 2021).

• Legal factors such as being on remand (Zhong et al., 2021) or receiving a long sentence (Fazel et al., 2008) have also been found to be associated with suicide and SH, as has a previous history of incarceration (Rivlin et al., 2013; Webb et al., 2011). This category also considers the type and nature of the offence for which a person is incarcerated; considering the elevated risk of SH found amongst those with sexual offences (Webb et al., 2012) as well as those with charges of serious violence (Fazel et al., 2008; Zhong et al., 2021). Also considered under this category are factors relating to an individual’s experience of custody. Prison misconduct and disciplinary infractions have been found to be associated with SH in custody, as has experiencing bullying, physical or sexual victimisation in the prison context (Favril et al., 2020).

• Individual risk factors include those known to increase an individual’s risk, acknowledging that an individual’s experience of external stressors (e.g. family, relationship, financial and health) are unique. Included in this category are cultural considerations – for example, in Australia, individuals of First Nations background are around two and half times more likely than non-Indigenous Australians to die by suicide (AIHW, 2023b). Studies from prison and community settings further suggest that individual risk factors can also include a history of trauma (Ford et al., 2020; Marzano et al., 2011) and a lack of social support in custody (Rivlin et al., 2013; Zhong et al., 2021).

• Psychiatric factors have clear links with SH/suicide in custody. Having any psychiatric diagnosis has been found to be associated with SH (Favril et al., 2020). Considered within this category is a patient’s history of mental illness as well as their current mental state, expressed feelings of hopelessness and coping style, whether they are experiencing intoxication or withdrawal from substances and their compliance and response to biopsychosocial treatments.

• The Safety Plan, adapted from Stanley and Brown (2012), and to be completed collaboratively with the patient. The safety plan considers identifying warning signs, supporting internal coping strategies, identifying helpful people and social settings, highlighting personal and professional supports, and considering other individual-specific strategies that may assist the patient to navigate a crisis without engaging in SH behaviour.

All Justice Health NSW staff working in the MHSU were trained in the SLIPS framework prior to implementation. The training package was developed and delivered by members of the research team including those who worked clinically in the unit (CB, CM, KS, TM). The training included an overview of the published research related to understanding and identifying risk of SH and suicide in custodial settings, background to the development of SLIPS, implementation guidelines, an example case study modelling the use of clinical notes and safety plans and some sample scenarios. Training for all staff began in March 2022 and was completed by 1 August 2022, when SLIPS was implemented in the MHSU. New staff who began working in the MHSU after this date were trained as part of their orientation/onboarding. All staff had access to copies of the SLIPS manual and guidelines and posters were placed in clinic rooms for quick reference.

The SLIPS framework was implemented in the MHSU on 1 August 2022. The initial plan was to attempt to use the SLIPS framework with all patients of the MHSU, commencing as soon as clinically feasible following admission, and to thus incorporate SLIPS into the standard admission process for all new admissions to the unit.

Two months following implementation, the procedure was updated in response to feedback from unit staff with regard to resourcing and operational barriers to completing SLIPS for all patients. The updated procedure specified that psychiatry staff would complete the “SLIP” assessment component only when clinically indicated. High-risk patients (i.e. those under “camera” observation or those required to share a cell for risk management purposes – “two-outs”) were prioritised for assessment. Following assessment, the safety planning component was completed by nursing staff in collaboration with the patient. All documentation relating to SLIPS was recorded in the patient’s electronic case notes and a copy of the safety plan provided to the patient.

Data were collected for a three-year period: two years retrospectively for the period pre-implementation of SLIPS and one year following implementation. Justice Health NSW electronic recording systems were used for data collection. Through the Patient Administration System, the researchers obtained information on which patients were in the MHSU at any point during each month of the study period and their admission and discharge dates. Some individuals had multiple distinct stays in the MHSU during the study period. Monthly bed occupancy rates at the total unit level were also extracted.

From the Justice Health Electronic Health System (JHeHS), clinical and demographic information was collected (date of birth, First Nations background, primary mental health diagnosis and history of SH). Clinical notes for each patient for the period that they were admitted to the MHSU within the data collection period were manually examined and incidents relating to SH extracted.

For each calendar month during the data collection period the primary outcome of interest – recorded incidents of SH – was extracted. These incidents were varied, and included cutting, scratching, wound opening, strangling, hitting/punching, ingestion of items and self-starvation. Thoughts of SH or suicide reported by patients on the unit to staff and recorded were also examined. For each month, the number of incidents were recorded for each individual, as well as overall for the unit. For the 12-month period post-implementation, the number of SLIP assessments and safety plans completed were also recorded on an individual and unit level.

Poisson generalised linear models were used to estimate the difference in the number of individuals with recorded SH events between the SLIPS intervention period and the period prior. To look at the difference between the pre-SLIPS and SLIPS intervention period on the number of SH incidents per month per patient, a Poisson generalised linear mixed model was used, with a random effect for individual accounting for repeated observations on patients across multiple months. Here, a within-between model approach was used to partition the within patient differences and between patient differences (“group level differences”). The within patient effect is the average change in the incident rate in individuals whose incident rates were observed both before and after SLIPS was introduced (across n = 79 individuals), whereas the group level difference is the difference between periods in patients’ average incident rates. This was achieved by including as fixed effects the mean of the dummy variable for treatment (0 = pre, 1 = post) per patient (contextual effects), which equates to the proportion of observations taken during the SLIPS intervention and the monthly deviation from this mean per patient (within patient effects). Where the random effect variance for individual was estimated to be 0, we removed the random effect from the model. We also included bed days (log variable) as an offset variable to account for variation in incident rates due to the length of stay. The pattern of reported thoughts of SH/suicide during the pre- and post-intervention periods was similarly analysed with within-between Poisson generalised linear mixed models, with the same terms as above.

This study was approved by the Justice Health and Forensic Mental Health Network Human Research Ethics Committee (G946/20); the Aboriginal Health and Medical Research Council Ethics Committee (1903 / 22); and the Corrective Services NSW Ethics Committee (D2021/1452316).

Table 1 provides an overview of demographic characteristics of individuals admitted to the unit across the three years of data collection. Duration of stay varied greatly among individuals, though the median across years was around 40 days. Between 20% and 30% of people admitted to the unit had a First Nations background. The majority had a known history of SH and a primary diagnosis of a psychotic disorder.

Fifty-three individuals (18.9%) received SLIPS intervention contact during the one-year implementation period. Twenty-seven individuals (9.6%) had both a SLIP assessment and safety plan created. Twenty further individuals (7.1%) had a SLIP assessment only. This occurred in cases where clinicians felt the person did not need a safety plan, the person was not well enough to engage with creation of the safety plan or the person was transferred or released before a safety plan was created. The remaining six individuals (2.1%) only had a record of a completed safety plan with no SLIP assessment; this process did not follow the intended implementation plan. Of note, some individuals had multiple safety plans completed over time to allow for updates to be made, though the numbers presented here only count each individual once.

Table 2 and Figure 1 provide descriptive and visual overviews of the SH outcomes across the three years. The proportion of individuals engaging in acts of SH was 8.7% in Year 1, 6.2% in Year 2 and 6.1% in Year 3. However, the overall number of SH incidents increased over the same period from 34 in Year 1 to 40 in Year 2 to 53 in Year 3.

Table 4 provides an overview of the effect of SLIPS on the number of individuals with recorded SH and the number of SH incidents during the study period. We observed no statistically significant effect of SLIPS on the number of individuals experiencing SH. In addition, no statistically significant effects of time or SLIPS were observed for the number of SH incidents.

As seen in Table 3 and Figure 2, the proportion of individuals reporting thoughts of SH (TOSH) increased from 14.9% in Year 1 to 15.8% in Year 2 to 20.6% in Year 3 with the number of TOSH events also increasing over time from 186 in Year 1 to 207 in Year 2 to 279 in Year 3.

Table 5 provides an overview of the effect of SLIPS on the number of individuals with recorded TOSH during the study period. We observed no statistically significant effect of SLIPS on the number of individuals experiencing TOSH. For number of TOSH incidents, we found evidence for a group-level difference in TOSH incidences across the periods (i.e. between effect) [F (1,2186) = 8.61, p = 0.003], such that TOSH incidents increased post-SLIPS implementation (pre: M = 0.27, SD = 1.43; post: M = 0.34, SD = 1.46). However, the intervention did not have a detectable effect on individual-level TOSH incident rates (i.e. within effects; p = 0.073).

This study establishes the incidence of SH in a prison mental health unit over a three-year period and evaluates the impact of the pilot introduction of an SPJ approach to SH/suicide risk assessment and management (the SLIPS framework). Although the proportion of people on the unit with an SH event appeared to reduce over the three-year period of observation, the introduction of the SLIPS framework in the final year was not significantly associated with any further reduction. Use of the SLIPS framework by clinical staff was at a lower level than expected but there was evidence of an increase in the reporting of SH/suicide thoughts by patients engaging with staff following SLIPS implementation.

After the prison MHSU in Sydney first opened in 2006, data collected over a 12-month period indicated that there were 604 admissions, the average patient age was 33.2 years (18–65 years), 18.6% of patients were of a First Nations background and the average length of stay was 12.4 days (0–78 days) (Adams et al., 2009). Between 60% and 70% of people were diagnosed with a schizophrenia-related disorder. In the current study, nearly two decades later, some shifts in the patient profile were apparent. There were fewer admissions to the unit (fewer than 300 annually) and a longer duration of stay (median stays ranged from 38 to 44 days). The current study also revealed an increase in the proportion of patients who were of a First Nations background in the MHSU, in line with the rising rate of over-representation of First Nations people in custody in Australia (e.g. 26.8% December 2021 and 28.4% December 2022; NSW Bureau of Crime Statistics and Research, 2022, 2023).

Some international reviews have examined different types of prison mental health services (e.g. Forrester et al., 2013; McKenna et al., 2021), but few studies have reported on the rates of mental illness and other demographics of people admitted to specific mental health units in prison. Those that do show lower rates of psychotic disorder than found in the current study. For example, in the UK, Forrester et al. (2010) found approximately 60% of men admitted to a prison health-care wing had a psychotic disorder, and in Ireland, Giblin et al. (2012) found that 35% of those admitted to a High Support Unit for mentally ill and vulnerable prisoners reported psychotic symptoms. However, direct comparisons are difficult given the great variations in legislation, policy and systems of care in different jurisdictions.

In the current study, a total of 34 SH incidents were recorded during the first year of observation, 40 in the second and 53 in the third (at least one SH incident from 8.7% of patients in the first year, 6.2% in the second and 6.1% in the third). Although there are few comparable studies, surveys of people released from prison in Australia indicate that approximately 1 in 20 men released from prison report intentional SH on at least one occasion during their time in prison (Australian Institute of Health and Welfare, 2023a). The annual rates of SH found in the MHSU are driven by the vulnerable nature of the population, with complex mental health problems and high rates of prior SH. The extent to which even higher rates of SH might be seen in such groups in the absence of high levels of surveillance, high staff to patient ratios and highly restricted access to means for SH is unknown. The negative impact of imprisonment itself as well as the additionally restricted conditions typically imposed on those deemed to present a risk of harm need to be better understood. Overall, international meta-analytic work suggests a lifetime pooled prevalence of non-suicidal self-injury (NSSI) of 5.5% among non-clinical samples of adults, with higher rates in younger people (Swannell et al., 2014).

Although the rate of SH events in the unit decreased over the three-year study period, there was no evidence of a significant effect of the SLIPS framework on the number of individuals self-harming, nor the number of incidents of SH in the unit. This lack of evidence for a reduction in the rates of SH aligns with other studies examining interventions aimed at reducing SH, where inconclusive or non-significant results have been found (Carter et al., 2022), highlighting the challenges of developing and implementing effective harm reduction strategies in this context. Implementation challenges were apparent in the current study where only one-fifth of those admitted to the unit received the SLIPS intervention due to staffing/resource constraints, and only one in 10 had both a SLIP assessment and safety plan created. These low rates of utilisation of the SLIPS framework likely limited the effectiveness of the intervention.

As a secondary or potentially mediating outcome, we also examined rates of reported thoughts of self-harm (TOSH) on the basis that the SLIPS approach is designed to encourage more engagement between patients and clinicians with regard to SH ideation and that patients might feel more able to report thoughts of SH as a result. We did find that the number of reported TOSH increased post-SLIPS implementation. Given that thoughts of SH are likely to be under-reported and thus unable to be addressed, even outside prison settings (Hallford et al., 2023), interventions that facilitate collaborative care planning with patients to reduce SH behaviour and suicide may expect to generate a rise in reported thoughts of SH. Research focused on understanding the under-reporting of thoughts of SH in the community highlight stigma-related concerns (Hom et al., 2017), as well as fears of overreaction and loss of autonomy (Richards et al., 2019). These concerns are likely heightened in prison settings, where disclosing thoughts of SH can result in aversive outcomes including increased restrictions and monitoring. Reliance on routinely recorded clinical data in the current study makes it difficult to explore the reasons for the change in rates of TOSH seen and the lack of significant impact on actual SH events undermines the proposition that SLIPS led to an increase in TOSH reporting, which in turn enabled the identified risk to be better managed. It is also possible that clinician knowledge and awareness of the SLIPS initiative could have led to staff asking patients more regularly or responsively about thoughts and feelings regarding SH, even without a recorded SLIP assessment or safety plan. Some research suggests that, in fact, asking about suicidality can lead to improved mental health among individuals who are suicidal (Dazzi et al., 2014).

This study had several strengths, including the use of an SPJ approach to developing a framework for supporting clinicians to undertake the difficult job of assessing and managing SH and suicide risk in prison. The SLIPS framework was developed by a multi-disciplinary group of researchers and clinicians, incorporated evidence-based risk factors specifically relevant to the prison setting and explicitly linked the structure assessment process to a management tool. Our use of routinely recorded data from clinical notes ensured that the study was not biased by selection, attrition or interview or recall biases, which are issues that would likely arise with other methods of data collection in this setting. We were also able to investigate both SH events as well as reported thoughts of SH, something not possible in studies using incident reports (Adams et al., 2009).

The main limitations of the current study were the problems with implementation of the SLIPS intervention, problems commonly encountered in studies attempting to trial new interventions in such difficult and low resource clinical settings (Biddle et al., 2018). The SLIPS framework was designed to be used with every patient in the prison mental health unit to identify risk factors to inform decision-making and management plans. However, limited staffing and access to patients meant that only people deemed very high risk for SH /suicide were prioritised for SLIPS assessment and management. Many clinicians may have only felt the need to use the SLIPS framework when they already perceived an obvious high risk, perhaps suggesting a failure to fully use the standardised approach to risk assessment (Hawgood and De Leo, 2016). This selective focus on a smaller sample of the most severe cases may have limited the ability of the intervention to demonstrate an effect on SH events on the unit as a whole, including as a result of the anticipated statistical power being reduced. The reduced sample also prevented any subgroup analyses from being pursued.

In addition, the framework was intended to be completed as a whole, with the safety plan based on the information arising from the SLIP assessment. Separating completion of the two elements of SLIPS as occurred in practice may have contributed to nurses feeling underequipped to properly develop a safety plan given the SLIP assessment component was completed by the assessing psychiatrist. Other issues included COVID-19-related delays to study commencement resulting in research and clinical team turnover, which may have resulted in reduced enthusiasm by newer staff in the MHSU to trial a novel intervention without additional resources. There may be limited generalisability of the implementation needs and the study findings to other settings, particularly those with different populations and policies. For instance, the MHSU is an all-male unit and past research suggests that rates of SH in prison are higher for women than men (Favril et al., 2020). These types of interventions should also be tested in non-mental health areas, including in broader prison settings.

A further limitation of the study is the pre–post approach to the trial, which was particularly problematic given the descriptive evidence that the rate of SH changed over time, even prior to SLIPS implementation. Where possible, future research should aim to use controlled trials to more effectively examine the impact of interventions. Finally, reliance on routinely collected clinical data to examine SH and thoughts of SH restricted the ability to explore in more detail the circumstances relevant to these events as well as the links between them. Emerging research has suggested that NSSI differs from suicidal behaviour, and people engaging with NSSI may not have suicidal ideation or be attempting to end their life (Klonsky et al., 2014; Muehlenkamp, 2014), and future research should aim to examine these types of incidents distinctly where possible.

Given the high rates of SH/suicide in prison and the need to prioritise risk-reduction strategies specific to this setting, the SLIPS framework (Suicide/Self-Harm, Legal, Individual, Psychiatric + Safety Plan) was developed according to SPJ principles that present a gold-standard approach in other mental health contexts. In a pilot trial of SLIPS in a prison mental health unit in New South Wales, Australia, challenges in the feasibility of implementation in such a complex context were apparent. No significant reduction in the number of SH incidents was seen as a direct result of the implementation but an increase in post-SLIPS engagement of patients with staff to communicate thoughts of SH to staff was noted. Future research needs to better understand the barriers to implementation and enable rigorous evaluations of SH risk assessment and management interventions in complex prison settings.

The authors gratefully acknowledge the Justice Health and Forensic Mental Health Network (JHFMHN) for their support for this project, particularly those at the Mental Health Screening Unit (MHSU).

Funding: This research project is funded by an NHMRC Emerging Leadership Investigator Grant (GNT1175408) awarded to Professor Kimberlie Dean and a Suicide Prevention Research Fund Innovation Grant from Suicide Prevention Australia.

Credit statement: CM: Conceptualisation, Methodology, Investigation, Formal analysis, Writing – Original draft preparation; CB: Conceptualisation, Methodology, Investigation, Formal analysis, Writing – Original draft preparation; MR: Investigation, Formal analysis, Writing – Original draft preparation; KS: Conceptualisation, Methodology, Writing – Reviewing and Editing; TM: Conceptualisation, Methodology, Writing – Reviewing and Editing; SS: Conceptualisation, Methodology; ES: Visualisation, Formal analysis; Writing – Reviewing and editing; VN: Conceptualisation, Methodology, Writing – Reviewing and editing; KD: Conceptualisation, Methodology, Writing – Reviewing and editing, Supervision, Funding acquisition.

Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committees (see “Ethical Approval” section in manuscript for more details) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Conflicts of interest: The authors have no conflicts of interest to disclose.

The supplementary material for this article can be found online.