Risk assessment in psychiatry is a subject that has received much attention in recent years, both in government policy and from the media. With the increasing emphasis on the role of psychiatry in protecting public safety, risk assessment and risk management have come to be seen as core skills in mental health workers. The 1999 policy booklet from the Department of Health, Modernising the Care Programme Approach, states that ‘risk assessment and risk management is at the heart of effective mental health practice and needs to be central to any training developed around the Care Programme Approach’ (Department of Health, 1999a : p. 22) and the National Service Framework for Mental Health (Department of Health 1999b ) contains the expectation that local health and social care communities ‘should focus on ensuring that staff are competent to assess the risk of violence or self-harm, to manage individuals who may become violent, and to know how to assess and manage risk and ensure safety’ (p. 12, Executive Summary).

The emergency clinic at the Maudsley Hospital is a service that provides emergency mental health care for the local catchment area. A central part of its work is the assessment of patients for admission to hospital or for community follow up and as such is ideal for an audit of risk assessment. The emergency clinic accepts referrals from local general practitioners, self-referrals and is also classified as a place of safety for the purpose of Section 136 of the Mental Health Act 1983. It is staffed by one staff grade psychiatrist, one junior doctor and three nurses from 9 a.m. to 5 p.m., and between the hours of 5 p.m. and 9 a.m. by two nurses and one junior doctor on-call.

This study aimed to assess the quality and type of risk assessment that was recorded in the notes for patients who were seen out of hours by doctors working in the emergency clinic and to evaluate the effect of a multi-component intervention on this.

Method

Case notes from the emergency clinic were identified for all patients attending in July 1999. The emergency clinic notes of the first 35 separate new cases who were seen by doctors out of hours and who fulfilled our criteria for needing a risk assessment (Table 1) were selected. The entries recorded by the doctors for the initial assessment were analysed and the numbers of well-established risk factors that were recorded (either as present or absent) in each entry noted. It was noted whether an explicit statement about the level of risk was recorded and also whether there was a statement reflecting an attempt to balance factors increasing risk against those reducing it. Lastly, the management plan for each patient was analysed: the communication between the doctor and the patient's care provider and family was recorded as well as whether or not admission was considered and if there was an emergency plan and a plan explicitly to address the risk.

Table 1. Criteria for risk assessment

Criteria for risk assessment for harm	Cases (%) 1999 (n=35)	Cases (%) 2000 (n=35)
to self
Past self-harm attempt	11 (31)	12 (34)
Current depression	14 (40)	7 (20)
Currently expressing suicidal ideation	17 (49)	15 (43)
Current drug or alcohol misuse	7 (20)	11 (31)
Current psychoses	15 (43)	16 (46)
to others
History of violence	1 (3)	6 (17)
Thoughts of violence	3 (9)	4 (11)
Current psychotic illness	15 (43)	16 (46)
Current drug or alcohol misuse	7 (20)	11 (31)

Results

Thirty-nine sets of notes in 1999 and 37 sets in 2000 had to be analysed in order to obtain 35 patients for each year that fulfilled the criteria for warranting assessment. These 35 assessments were made by 18 different junior doctors in 1999 and by 16 in 2000. For both sets of patients, all 35 who fulfilled the criteria for a risk assessment required a risk assessment for harm to self. In addition, 21 in 1999 and 27 in 2000 (Fisher's exact test=0.06) also required a risk assessment for risk of harm to others according to our criteria. The patient groups did not differ significantly in their risk factors between 1999 and 2000 (Table 1). Of all the cases that required a risk assessment, 54%, both in 1999 and in 2000, had been seen previously in the trust.

Table 2 shows the risk factors recorded in 1999 and 2000. The only statistically significant difference between the risk assessments recorded in 1999 and in 2000 was the evidence in the notes for weighing up the risk of self-harm that increased from 14% in 1999 to 34% in the notes for weighing up the risk of self-harm that increased from 14% in 1999 to 34% in 2000 (Fisher's exact test=0.034). Comparable statements concerning risk of violence were practically non-existent in both years.

Table 2. Risk factors and management recorded in notes 1999 and 2000

	1999 (%)	2000 (%)
Risk factors for harm to self recorded in case notes	n=35	n=35
Previous self-harm attempt	17 (49)	16 (46)
Violent or perceived lethal method	10 (29)	8 (23)
Expressing current suicidal ideation	28 (80)	30 (86)
Plan to end life	23 (66)	25 (71)
Feelings of hopelessness	8 (23)	3 (9)
High level of subjective distress	5 (14)	9 (26)
Feelings of no control in life	1 (3)	2 (6)
Misuse of drugs or alcohol	25 (71)	23 (66)
Displaying impulsivity	2 (6)	4 (11)
Living alone	21 (60)	26 (74)
Poor physical health	19 (54)	21 (60)
Recent significant loss or threatened loss	2 (6)	5 (14)
Recent disengagement from services/stopped medicines	18 (51)	19 (54)
Recently discharged from hospital	20 (57)	20 (57)
Family history of suicide	5 (14)	1 (3)
Risk factors explicitly considered in notes	5 (14)*	12 (34)*
Low-/medium-/high-risk stated in records	4 (11)	7 (20)
Risk factors for harm to others recorded in case notes	n=21	n=27
Previous history of violence	2 (10)	5 (19)
Misuse of drugs or alcohol	18 (86)	17 (63)
Delusions of being persecuted or controlled	9 (43)	12 (44)
Thoughts of harming others	8 (38)	4 (15)
History of antisocial behavior	6 (29)	8 (30)
Impulsive or showing emotional lability	1 (5)	0 (0)
Rootlessness or social restlessness	0 (0)	0 (0)
Problems with stability in relationships or work	5 (24)	6 (22)
History of non-compliance or disengaging	3 (14)	1 (4)
History of childhood adversity	5 (24)	3 (11)
Significant recent stress	3 (14)	2 (7)
Minimisation of previous violent incidents	1 (5)	0 (0)
Violence within social network	1 (5)	0 (0)
Others expressed concern about violence risk	1 (5)	1 (4)
Risk factors explicitly considered in notes	0 (0)	2 (7)
Low/medium/high risk stated in records	0 (0)	0 (0)
Management	n=35	n=35
Arrangement for follow up	33 (94)	35 (100)
Direct communication with care provider	1 (3)	3 (9)
Letter to care provider	25 (71)	28 (80)
Direct communication with family/friends	10 (29)	9 (26)
Admitted	12 (34)	16 (46)
Consideration of admission	19 (54)	21 (60)
Explicit plan to address risk	2 (6)	0 (0)
Emergency plan	21 (60)	24 (69)

Table 2 also shows the recorded components of management plans. These did not differ significantly between 1999 and 2000. While crisis plans were commonly formulated, specific plans to manage risk were rare.

Discussion

This audit assessed only the documentation associated with risk assessment and management. The clinical encounter in respect of risk may have been more comprehensive and relevant information may have been communicated verbally rather than in writing. We found that despite recent trends towards greater emphasis on risk assessment in psychiatry and despite local specifically targeted opportunities, the standard of documentation of risk of harm to others remained unchanged between 1999 and 2000. It is possible that the intervention was too weak, the numbers analysed were too small, or that there were differences between the doctors in 1999 and 2000 that have not been accounted for. The documentation of risk of suicide did improve significantly between the 2 years however, suggesting that the intervention did have some impact.

Unlike the assessment of the risk of self-harm, the assessment of the risk of harm to others is a new task. Few psychiatrists have received any relevant instruction while medical students. The standard textbooks of general psychiatry usually fail to mention it. Risk assessment has not been regarded, until very recently, as a standard psychiatric skill. We know that changing the behaviour of clinicians is very difficult (NHS Centre for Reviews and Dissemination, 1999). It is thus not surprising that there was little evidence of change in our study. It is further confusing that it seems only in the UK is risk assessment regarded as being ‘at the heart of effective mental health practice’ (Department of Health, 1999a : p. 22).

A number of senior house officers emphasised that in the context of working with patients in a crisis, their first concern is the wellbeing of the patient. This is what they as doctors have been trained to do. To think about the risk to others does not seem to come naturally, especially when dealing with a very ill patient. Some resistance to considering risk may also come from the belief that doctors are there to treat patients, not to protect the public. The poor predictive value of risk assessment may also be a factor. What is the practical value of a risk assessment when individual risk factors are common (as in an inner-city emergency clinic) yet serious incidents rare (Reference SzmuklerSzmukler, 2001)?

We believe there is a need for us to be clear about the value and limitations of risk assessment. The assessment of the risk of violence to others is inherently imprecise (as it is for the risk of suicide). Practice aimed at avoiding later blame (managing the ‘risk to oneself’) is not a good reason to advocate risk assessment. We suggest that the focus of risk assessment should not be primarily to prevent violent acts, but to alert the clinician that a particular patient may present a higher risk than others, and that the consequences of an inadequate treatment plan could prove damaging.