There has been increasing concern about patients with personality disorders in terms of risk to others and risk to themselves, as reflected in a number of official inquiries. Recently there has been an inquiry that hinged upon the lack of clarity in psychiatric risk assessment of borderline personality disorder (BPD) (Reference Brown, Fraser and MorleyBrown et al, 1999). It is well known that patients with BPD pose difficulties with regard to clinical management and we are aware of much anecdotal evidence of difficulties of risk assessment for these patients. However, we have not been able to locate any current research into the efficacy of risk assessment of BPD and the associated problem of assigning correct Care Programme Approach (CPA) status based upon risk.

American long-term studies of BPD over 15 years indicate an overall risk of suicide of 9% (Reference Paris, Nowlis and BrownParis et al, 1987; Reference StoneStone, 1990). This has been confirmed by a more recent review (Reference PerryPerry, 1993). Paris et al and Stone found that suicide rates for BPD are highest for patients in their 20s, peak at 30 and rapidly diminish thereafter. A number of studies of young male suicides have found that 30% of such cases rate a retrospective diagnosis of BPD (Reference Rich, Fowler and FogartyRich et al, 1988; Reference Runeson and BeskowRuneson & Beskow, 1991). It is possible that the suicide rate for BPD is increasing, though it may also be that the incidence of BPD itself is increasing (Reference ParisParis, 1991).

A comprehensive review by Fine and Sansone (Reference Fine and Sansone1990) points out that there are two sorts of suicidality in patients with BPD, and that this problem is unique to the disorder. Acute suicidality represents a sudden surge of risk that may occur, for instance, as a result of an interpersonal crisis and is managed by hospitalisation. Chronic suicidality represents the chronic despair of the patient with BPD and is best managed by containment by the therapist or professional team involved in the patient's care, with the patient remaining in the community. Distinguishing between acute and chronic suicide risk is a tricky assessment problem (Reference ParisParis, 1991). Patients with BPD transfer anxiety to those working with them by their characteristic psychic defence of projective identification; with this influx of subjective anxiety, objective assessment is made more difficult.

The Borderline Team at the Regional Department of Psychotherapy, Newcastle, run a unique regional outpatient psychotherapy service for patients with BPD. The service has instituted Care Programme Approach (CPA) management of patients with BPD at two levels, minimal and complex, which relate to the therapist's perception of significant risk. Thus, if the patient is not judged to be at significant risk, the CPA is instituted at a minimal level with normal clinical communication when necessary to others involved in the care of the patient. However, if the patient is assessed to be at significant risk, the CPA is set at complex, with regular reviews of treatment with all professionals involved in the care of the patient at 3- or 6-month intervals. In order for this care management plan to be effective the therapist must be effective, in assessing significant risk in order to assign an appropriate CPA level. Additionally, for patients at significant risk, the therapist needs to distinguish acute from chronic suicide risk, as the management of these two categories will be different, with hospitalisation appropriate only for acute risk.

One way of assisting the therapist to make crucial clinical decisions would be to be able to augment clinical assessment of risk with a valid self-report measure of risk that could be used sequentially to measure surges in risk. The Clinical Outcomes in Routine Evaluation System (CORE) psychotherapy evaluation questionnaire (Core System Group, 1998) is a recently published outcome measure that contains a risk sub-scale and can be used to measure risk over the week preceding completion.

Despite uptake by the Borderline Team of over 100 cases of BPD per year, the majority of patients with BPD in the immediate region are managed in community mental health teams (CMHTs). A valid and easy to use questionnaire would also be very helpful for CMHT professionals in assessing and managing risk in their patients with BPD.

We wanted to know whether the CORE risk subscale was clinically valid for our BPD patients.

The study

An explanatory letter and a CORE evaluation questionnaire were sent out to all patients under treatment by the Borderline Team in mid-December 1998, asking the patient to complete the questionnaire and return it to the department before their next therapy session. The CORE questionnaire contains six risk questions, which together comprise the risk sub-scale. The six questions are divided into two questions for each category: risk to others; risk of suicide; and risk of self-harm.

These are:

1. (a) Risk to others

   1. (i) I have been physically violent to others;

   2. (ii) I have threatened or intimidated another person.

2. Risk of suicide

   1. (i) I have made plans to end my life;

   2. (ii) I would be better off dead.

3. Risk of self-harm

   1. (i) I have thought of harming myself;

   2. (ii) I have hurt myself physically or taken dangerous risks with my health.

Each question is answered on a five-point scale that varies from zero (not at all) to four (most or all of the time) and relates to the past week only. This gives a possible score range of 0-8 for each of the three risk categories, in the past week.

For the same week in December the therapist treating each patient was asked to rate the clinical risk for each of the three categories (risk to others, suicide and self-harm) as significant or not significant. They were also asked to rate significant risk as stable or fluctuating. This assessment was carried out before the patient's therapy session, with the therapist using his or her impression of the patient gleaned from normal clinical interaction with the patient over the preceding weeks.

Findings

The therapists gave information on 72 patients in treatment, giving current estimation of risk and information about whether the patient was on minimal or complex CPA.

From the 72 patients in treatment we received 47 completed CORE questionnaires, a response rate of 65.3%. In terms of age and gender there were no differences between responders and non-responders. We were concerned to know whether the responders differed markedly in severity of risk or in frequency of therapy from non-responders. Of the respondents 29 patients (61.7%) had been placed on minimal CPA compared with 18 patients (38.3%) who had been placed on complex CPA. Of the non-respondents, 17 patients (68.8%) had been placed on minimal CPA and only eight (31.8%) had been placed on complex CPA. Patients are seen for psychotherapy in the department either weekly or fortnightly; of the respondents 26 (76.6%) were being seen weekly and 11 (23.4%) fortnightly, while of the non-responders 19 (76%) were being seen weekly and six (24%) fortnightly. These results are shown in Table 1.

Table 1. A comparison of CPA levels and frequency of therapy for responders and non-responders

	CPA level		Frequency of therapy
Total patients in therapy (n=72)	Complex	Minimal	Weekly	Fortnightly
Respondents (n=47)	18 (38.3%)	29 (61.7%)	36 (76.6%)	11 (23.4%)
Non-respondents (n=25)	8 (31.2%)	17 (68.8%)	19 (76.6%)	6 (24.0%)

We were able to conclude that although responders were slightly more likely to be at risk than non-responders, frequency of therapy did not correlate with response/non-response. Of those on complex CPA, representing significant risk in one, two or three of the risk areas (including responders and non-responders), eight were rated as at stable chronic risk while 18 were rated as at fluctuating risk.

Risk of self-harm

For risk of self-harm, therapists estimated 22 to be at significant risk. Of the at risk patients, 21 were on complex CPA with one on minimal CPA representing a current change of risk for one patient.

The therapists' estimation of significant self-harm v. no risk was compared with the CORE sub-scale of self-harm behaviour. There were significant statistical differences between therapists' estimation of risk of self-harm behaviour and the CORE mean score (t=-3.90, P<0.001). Patients rated at risk of self-harm scored on average significantly more (4.87) than those not at risk of self-harm (2.13).

A summary of findings with regard to therapists' estimation of risk of self-harm, suicide and risk to others and CORE sub-scale values is shown in Table 2.

Table 2. Therapist assessment of risk/no risk compared to means and standard deviations of CORE risk sub-scores and suggested cutting score for significant risk

	CORE risk sub-score
Therapists' assessment	Mean	(s.d.)	Suggested cut-off score for significant risk
Risk to others (n=8)	3.33*	2.07	3 or above=
No risk to others (n=39)	1.07*	1.66	significant risk to others
Total (n=47)	1.36	1.86
Risk of suicide (n=13)	7.12**	1.45	5 or above=
No risk of suicide (n=34)	3.10**	2.46	significant risk of suicide
Total (n=47)	3.79	2.77
Risk of self-harm (n=22)	4.87***	2.25	3 or above=
No risk of self-harm (n=25)	2.13***	2.31	significant risk of self-harm
Total (n=47)	3.06	2.62

Discussion

In all three categories, self-harm, suicide and risk to others, there was very clear separation of the therapist estimation of risk v. no risk for each relevant CORE risk sub-score. For no risk to others as estimated by therapists, the mean CORE risk sub-score was very low at 1.07, while for risk, the risk sub-score was 3.33. The two CORE risk sub-score questions both relate to behaviours (physical violence and threatening another person), so it seems that sporadic threatening behaviour is not seen by therapists as a significant risk in patients with BPD. For no risk of suicide, as estimated by therapists, the main CORE risk was 3.10, while for those at risk, the risk sub-score was 7.12, which represents a wide separation. The two CORE risk sub-scale questions for suicide represent both thoughts of suicide and the possible translation of thoughts into action (‘I have made plans to end my life’). It seems that therapists were able to tolerate patients' frequent thoughts of suicide as not representing significant risk, while they felt that significant risk for suicide was indicated by active planning and frequent thoughts of death together. This indicates that the therapists were able to carry the chronic despair of their patients without assigning significant risk, unless this was accompanied by active planning.

In relation to self-harm, the therapists' rating of significant risk gave a CORE sub-scale mean score of 4.87 compared with 2.13 for those not rated as at risk. The two self-harm CORE sub-scale questions refer both to thoughts of self-harm (‘I have thought of harming myself’) and to actual self-harm (‘I have hurt myself physically or taken dangerous risks with my health’). It seems that therapists were able to distinguish adequately between thoughts of self-harm and actual acts of self-harm in their patients.

The psychotherapists treating patients with BPD in the Regional Department of Psychotherapy in Newcastle are all analytically trained and experienced in working with patients with BPD. This research indicates that the therapists are able to carry a degree of anxiety about chronic feelings of suicide and self-harm before assigning significant risk. It does seem that the significant correlation of CORE risk sub-scales and therapists' estimation of risk indicates that the CORE risk scales themselves are clinically valid. We suggest cut-off points for the risk sub-scales indicating significant risk in each category for our patients with BPD.

We think that the CORE risk sub-scales can be used in psychotherapy to augment clinicians' risk assessment at entry to the service and during treatment to quantify risk, and to highlight changes or surges of risk. A sudden shift in risk might, for instance, provides important back up empirical data when hospitalisation for increased risk is contemplated. Furthermore, we think CORE risk sub-scales with our suggested cut-offs could be used in a general psychiatric service to delineate significant risk and therefore to assign appropriate CPA status. This would represent a significant advance in CPA assignment and risk assessment for in-patient or CMHTs managing patients with BPD.