Self-harm in prisons in England and Wales has reached record levels in recent years (Ministry of Justice, 2020). Self-harm does not necessarily always reflect suicide intention; for some it is used as a way to cope with emotional distress, however, it is considered an important risk factor for suicide.
Indeed, research has shown that for many people in prison, there is a direct link between their self-harm behaviour and subsequent suicide (Hawton et al., 2014). The Prisons and Probation Ombudsman (2014) found that over a third of self-inflicted deaths in prison had a history of self-harm and we know that both men and women in prison are more likely to die by suicide than their counterparts in the community (Fazel et al., 2005; 2009).
Incidents of self-harm also, understandably, put a lot of pressure on prisons and staff who are under a great deal of scrutiny regarding increasing rates. Self-harm incidents can also increase tensions between staff groups and affect the interaction between staff and the people under their care (Taylor et al., 2009; Marzano et al., 2015). Being able to efficiently identify risk of self-harm is therefore crucial.
A recent systematic review of screening tools used in prisons identified a need to develop more robust models, specific to the prison setting (Gould et al., 2018). Ryland et al.’s (2020) recent paper reports on the development and evaluation of such a screening tool used to identify male prisoners at high risk of self-harm.
From a review of relevant prison suicide research, the authors developed an 11-item screening tool (called OxSHIP) covering a range of factors:
- Historical (e.g. have you ever attempted suicide or self-harmed outside of prison?),
- Clinical (e.g. have you ever had psychiatric treatment by a medical professional?),
- Social (e.g. do you have friends or family you feel close to?); and
- Prison-related (e.g. have you been in prison before?).
Two adult male prisons in England (Wormwood Scrubs in London and Woodhill in Milton Keynes) took part in the study. The screening tool was administered to new receptions into the prisons during their first week in custody. Participants were followed up for six months to identify those who subsequently self-harmed. Time from screening to first episode of self-harm, transfer, release or six months post- screening (i.e. end of the follow up period) was considered the ‘at-risk’ period. Self-harm reported by the individuals, their peers or prison officers, was recorded by the prison Safer Custody Team for inclusion. Details of the harm and any treatment were recorded on a standardised form.
Analysis was conducted using Cox proportional hazard regression adjusted for the score at screening. The pre-specified cut-off was a score of 5 (low risk was categorised as a score of <5 and high risk as ≥ 5) and this was used to calculate the sensitivity, specificity, and positive and negative predictive values of the model.
The screening tool was administered to 542 male prisoners. Of those, 17 (3.1%) self- harmed in the subsequent six months. Multivariate analyses revealed hazard ratios for the screening items ranged between 0.3 and 9.3. The strongest associations with self-harm were:
- Previous self-harm in prison (adjusted hazard ratio [aHR] = 9.3 [95% CI: 3.3 to 26.6])
- Current thoughts of self-harm (aHR = 7.6 [CI: 2.1 to 27.4])
Of those with a score of 4 and lower (n= 412; 76.0%), 11 (2.7%) later self-harmed. Of those with a score of 5 and higher (n= 130; 34.0%), 6 (4.6%) subsequently harmed themselves. The full 11-item screening tool resulted in an area under the curve (AUC) of 0.66 (0.53 to 0.79).
Multivariate modelling revealed two items from the screening tool were not associated with the outcome (i.e. the first episode of self-harm by an individual). These were:
- ‘Have you ever attempted suicide or self-harmed outside prison?’ and
- ‘Do you have friends or family that you feel close to?’
These items were removed from the model and weighting the remaining nine predictors, according to effect size, resulted in an increased AUC from 0.66 (0.53 to 0.79) to 0.84 (0.76 to 0.92).
The full 11-item screening tool showed moderate discrimination (i.e. ability to predict those who would subsequently self-harm versus those who would not) and poor calibration (i.e. how closely the expected probability compared with the observed risk). The model demonstrated low sensitivity, moderate specificity, very low positive predictive value and a high negative predictive value. The authors concluded that this version of the tool should not be used to identify those at high risk of self-harm.
The authors explored how the model could be improved and made post-hoc adjustments. Two items were removed and the new nine-item model showed improved predictive power (shown by an increased AUC), demonstrating good to excellent discrimination. However, the secondary analyses did not follow a protocol and therefore the authors recognise the model requires testing with new large samples.
Strengths and limitations
The authors of this paper have published widely in the area of suicide and self-harm in prison. The screening tool was short with only 11 items, an important consideration in a busy prison environment. In addition, the items included in the tool included a range of factors recognised by the literature as important risk factors for self-harm in the prison population. The paper also transparently reported the analysis and essential information as part of that (i.e. discrimination and calibration). Nevertheless, Ryland et al., (2020) recognise a number of limitations:
- Self-harm incidents were self-reported and were not confirmed via medical records
- The original model was not adequately powered and the low number of self-harm events created model instability
- As a consequence of the low event rate, there was limited information to determine which items should be removed from the model
- Although the revised model showed increased predictive power, which was promising, analyses did not follow a pre-specified protocol and so the model is therefore likely to perform differently with new samples.
Implications for practice
The risk factors tested by the model in this study show potential for future work to explore further development of screening tools. However, there are a number of challenges with developing an efficient and effective screening tool, including the importance of testing individual items and the need to test newly derived models with different samples. The authors recommend that without current valid screening tools in prison: “safety planning for all prisoners should be considered.” Such safety planning includes: the restriction of access to means for self-harm and suicide, assessment and treatment for mental health problems and ensuring prisoners are aware of the various avenues of support available to them (e.g. clinical services, Listener peers, Samaritans etc.); things that should be routine practice for prisons.
Given the high rates of self-harm and the elevated risk of people in prison dying by suicide, research that aims to further our understanding of the identification and management of these phenomena remains an important area of focus. Something the authors point out in their paper is that there has been debate as to whether screening tools are any more effective than clinical judgement. Further work to explore the association between risk factors and self-harm behaviour in prison is needed to add to this debate. As well as being able to predict who are at high risk of self-harm in prison, there is also perhaps a need to understand why the majority do not self-harm and, why of those that do, only some go on to die by suicide (Forrester & Slade, 2014). Such work has great potential to inform service delivery for the better.
Statement of interests
Verity is a researcher at the Offender Health Research Network with a particular interest in suicide and self-harm research in the criminal justice system. She has previously worked with one of the authors of this paper on an NIHR funded study regarding the identification and management of self-harm and suicide across the criminal justice pathway.
Ryland H, Gould C, McGeorge T, Hawton K, Fazel S (2020). Predicting self harm in prisoners: Risk factors and a prognostic model in a cohort of 542 prison entrants. European Psychiatry, 63(1), e42, 1–8 https://doi.org/10.1192/j.eurpsy.2020.40
Fazel S, Benning R, Danesh J. (2005) Suicides in male prisoners in England and Wales, 1978-2003. Lancet 366, 1301-1302.
Fazel S, Benning R. (2009) Suicides in female prisoners in England and Wales, 1978-2004. British Journal of Psychiatry 194(2), 183-184.
Forrester A, Slade K. (2014) Preventing self-harm and suicide in prisoners: job half done. Lancet 383, 1109-1111.
Gould C, McGeorge T, Slade K. (2018) Suicide screening tools for use in incarcerated offenders: a systematic review. Archives of Suicide Research 22, 345-364.
Hawton K, Linsell L, Adeniji T, et al (2014) Self-harm in prisons in England and Wales: an epidemiological study of prevalence, risk factors, clustering, and subsequent suicide. Lancet 383, 1147-1154.
Marzano L, Adler J R, Ciclitira K. (2015) Responding to repetitive, non-suicidal self-harm in an English male prison: Staff experiences, reactions, and concerns. Legal and Criminological Psychology 20, 241-254.
Ministry of Justice. (2020). Safety in Custody Statistics, England and Wales: Deaths in Prison Custody to December 2019 Assaults and Self-harm to September 2019. London: Author.
Prisons and Probation Ombudsman. (2014). Learning from PPO investigations: Self-inflicted deaths of prisoners on ACCT . London: Author.
Taylor T L, Hawton K, Fortune S, et al (2009) Attitudes towards clinical services among people who self-harm: Systematic review. British Journal of Psychiatry 194, 104-110.