Occupational health and safety (OHS) is a part of everyday life, with risk assessments (RA) playing a pivotal role in creating safer work, study, and public spaces. RA instruments’ effectiveness in acute mental health units (AMHU) has provoked controversy and contradictory research regarding their utility.
Research literature reflects a lack of standard RAs to address persistent OHS issues in AMHUs, ranging across (1) somewhat useful when combined with training and early interventions (Maguire, 2019), (2) moderately effective but not fit for purpose (Yang et al., 2010), and (3) purposeless fear-based instruments, reinforcing stigmatisation that contributes to the ongoing OHS hardships for patients and staff (Szmukler & Rose, 2013; Slemon et al., 2017).
Tonso et al. (2016) concluded that 83% of staff working in Victorian AMHUs reported experiencing workplace violence in a twelve month period. A not unnatural accompaniment to this dangerous work is a seemingly endless array of assessments to identify and manage the different types of risks in AMHUs (Anderson & Jenson, 2019). Patients in AMHUs are some of the most vulnerable people in society and are more likely to self-harm, suicide, commit acts of violence and aggression, abscond, and suffer from co-occurring disorders like substance abuse (Briner & Manser, 2013).
Health workers use three different approaches to accurately measure the risk associated with patients. The original ‘unstructured clinical approach’ is a basic informal interview with a patient along with a review of the patient’s files and is considered the least reliable of the three approaches to assessing risk (Yang et al., 2010; Maguire, 2019). This approach relies heavily on the assessor and has resulted in devastating consequences, eventually leading to the more structured evidence-based algorithmic approach of the ‘actuarial risk assessment’ (ARA) instruments (Yang et al., 2010).
Most ARAs are observational checklists designed from empirical and theoretical concepts to predict violence and anti-social behaviour. Patients are then graded low, medium or high risk based on the results of the observations, and the interventions will respond accordingly. ARAs like the Brøset Violence Checklist (BVC), the Dynamic Appraisal of Situational Aggression (DASA) instrument and the Suicide Assessment Five-step Evaluation and Triage (SAFE-T) are commonly used ARA instruments in AMHUs (Wand, 2011; Yang et al., 2010; Maguire, 2019). Studies suggest ARAs provide more accuracy than the unstructured clinical approach; however, they are criticised for being non-therapeutic, identifying only a limited number of risk factors, considered monocultural and lacking robust empirical validation.
Most ARAs are derived from forensic patients and they tend not to consider protective factors that might mitigate the potential risk (Wand, 2011; Yang et al., 2010; Maguire, 2019). The data-driven, static nature of the ARAs once again leads back to how they are being utilised and observed by staff in AMHUs. The ‘structured clinical judgement’ (SCJ) is a far more collaborative and integrative approach combining the data of the ARAs, history, knowledge and protective factors of the patient, subsequently assessed with the experience and structured professional judgement of staff (Maguire, 2019; Wand, 2011).
Some ARA instruments, like the Short-term Assessment of Risk and Treatability (START) and the Historical Clinical Risk-20 Version 3 (HCR-20 V3) were designed to be used in conjunction with the SCJ approach in order to give staff the freedom to interpret data, mitigate risk alongside protective factors and provide appropriate treatment and interventions to manage risk. (Kivisto, 2016; Szmukler & Rose, 2013)
The use of ARAs and SJCs has reduced more restrictive interventions like seclusion and physical, mechanical and chemical restraints whilst creating safer work environments (Kivisto, 2016; Anderson & Jenson, 2019).
Research indicates no gold standard for assessing risk in AMHUs but a variety of different RAs and screening instruments, either generic, modified, or adapted to suit certain population groups, different countries, and metropolitan and regional areas (Wand, 2010). This review will analyse RA tools and strategies within AMHU’s, adaptations of screening tools to address risk challenges for services and discuss the ethical considerations of using ARAs on people experiencing mental illness.
"Violent acts are only committed by a very small portion of people and only under specific circumstances or environments".
Yang et al. (2010) assert that violent acts are only committed by a very small portion of people and only under “specific circumstances or environments”. Researchers have tried to identify specific character traits, behaviours and predisposing factors that may influence potential violent and aggressive patients in AMHUs’. (Kivisto, 2016) The BVC was developed in a secure AMHU in Norway, examining daily reports over five years by identifying fifty-six different types of behaviour that occurred daily before a violent incident.
The most common behaviours identified were confusion, irritability, boisterousness, verbal threats, physical threats and attacking objects. The results became a short-term, six-item violence ARA tool identifying the presence of certain behaviours in a twenty-four hour period. A score of one is given if the behaviour is present and zero if the behaviour is absent, totalling a possible tally of six.
The checklist states if the patient continually displays a specific type of behaviour, but if this is normal for that patient, then a zero score is given. If the overall score is zero, then the risk of violence is small; if the score is one to two, then the risk of violence is moderate, and preventative measures should be taken. If the score is two plus, then the risk of violence is high, and specific plans should be developed to manage the potential violence. (Woods & Almvik, 2002) Almvik et al. (2000) reported the BVC had an accuracy rating of 63% for predicting violence in the ensuing twenty-four hour period and a 92% rate for predicting non-violent behaviour from patients in the same time frame. Very similar results were found in a study of forensic patients. (Hvidhjelm et al. 2014) In a critical review, Anderson & Jenson (2019) found the DASA outperformed several other ARA instruments in MHU’s. In a more recent study of schizophrenia patients, the BVC was outperformed by the DASA, which predicted violent and non-violent behaviour more accurately in a twenty-four hour period (Maguire, 2019). Kivisto (2016) argues that both ARAS and SPJ might be equally capable of identifying risks; only SPJ can provide appropriate management and intervention for risks identified.
The DASA was created based on the BVC with some slight alterations of listed behaviours and emphasised being used as an SJC rather than a straight ARA. Recently, Teresa Maguire conducted two research studies as part of her doctoral thesis to test the DASA’s validity and an electronic version of the DASA called the E-DASA in a forensic MHU.
The original DASA manual highlighted the need for interventions as part of using the instrument, which should also be accompanied by documentation to validate its use. The study developed ‘aggression prevention protocols’ which reflected the risk level of the DASA to mitigate exacerbating behaviours. The study reported a reduction of harsh restrictive interventions in the AMHU and encouraged staff to respond to different risk levels according to a structured protocol. Many staff were ambivalent to the use of the ARA tool and most resisted combining less restrictive intervention like ‘Safewards‘ in response to increased level of risk. Verbal aggression rates were reduced amongst patients.
However, the use of seclusion as an intervention for patients remained the same and, in some cases, increased. (Maguire, 2019) It is unclear from all ARA clinical trials how much the outcome is being influenced by the trial itself, young and inexperienced staff, unconscious bias of the research and the high rates of ‘false positives’ (overly predicting risk) because of rapid early interventions. Evidence suggests that ARAs have very low accuracy, producing too many errors and that trying to predict human behaviour is impossible. (Szmukler & Rose, 2013; Yang et al. 2010; Slemon et al. 2017; Wand, 2011; Buchanan et al. 2015) While ARAs might be a useful communication tool for staff and encourage added attentiveness for their patients, it is unclear whether they are used as intended, preventing and predicting potential risks.
The prevailing medical view that 90% of people who die from suicide have a mental disorder is challenged by newer modelling, suggesting the number could be much lower (Pridmore, 2015). However, a large percentage of people who die from suicide had contact with an AMHU in the preceding twelve months (Stene-Larsen & Reneflot, 2019) and suicide remains the leading cause of death of Australians aged between fifteen and forty. (Beyond Blue, 2020) Suicide RA should be “individualised, ongoing, incorporating past and present suicidality” (Alonzo, 2018, p. 295).
The Suicide Assessment Five-step Evaluation and Triage (SAFE-T) card is a basic five-step ARA instrument designed to help MHU staff identify both protective and potential risk factors. According to Fowler (2012), accurately predicting the genuine risk of suicide is extremely difficult but the SAFE-T is recommended if used in conjunction with highly trained and skilled staff who consider an extensive range of personal and external factors relating to the patient. Discussion of suicide and self-harm in the community and treatment services is polarising, often leading to mixed messages and misunderstood policies. Providing AMHU staff with appropriate resources, support, and training can have a significant impact on preventing self-harm and suicide.
Dual diagnosis substance use disorder and mental illness increase the likelihood of violent behaviour in AMHUs. (Kivisto, 2016; Szmukler & Rose, 2013) In 2009 a Western Australian AMHU reported upwards of forty-eight percent of patients presented with a dual diagnosis substance use disorder and mental illness. An in-vivo education program was conducted, instructing nurses on using an amended version of the most commonly used drug and alcohol screening instrument, the ASSIST, and provide brief interventions for patients experiencing dual diagnosis (World Health Organisation, 2020). Nurses reported becoming more confident in identifying casual substance use, abuse, and clinical dependence and, in turn, how to provide the appropriate interventions and referrals. The program improved accurate identification of drug and alcohol challenges for patients, improved assessments and overall treatment plans (Heslop et al., 2013). The in-vivo program highlighted the importance of cost-effective targeted ongoing training, education, and collaboration within the acute AMHU sector and the potential benefits to patients and staff.
The cross-over ARA research reflects a lack of clear differentiation between forensic MH patients and patients in an AMHUs. Studies on patients lived experience or input into AMHUs are nonexistent. The ethos of ‘nothing about me without me’ contradicts the utility of an ARA. Adapting drug and alcohol screening tools for AMHU addresses the high levels of dual diagnosis presentations in AMHUs. Assessments are only as good as the accuracy, relevance, context of the information, and the assessor conducting the assessment.
Evidence suggests senior/junior health professionals rarely rely on ARAs (Solely for risk determination), and the latter believe them to be defensive instruments designed to protect an organisation, disperse accountability amongst protocols, and ease collective societal consciousness (Kivisto, 2016; Szmukler & Rose, 2013).
Staff in AMHUs are constantly assessing risk as part of their job and sometimes find themselves working as ‘risk managers’ and seeing their patients as the embodiment of risk instead of a human being experiencing a nightmarish hell (Al-Azzawi, 2016 p.25). Patients want clarity and explanations so they can understand what is happening and why (Learmonth & Sethi 2016). While SPJ is a more collaborative approach, ARAs are officious, ethically questionable, discriminatory, trauma-inducing, and erode the most important health care asset, a trusting therapeutic relationship between patients and staff.