The suicide of a patient is an occupational hazard for psychiatrists and is identified as one of the most stressful adversities of psychiatric training (Reference Kozlowska, Nunn and CousensKozlowska et al, 1997). Trainees are vulnerable because of their close involvement with patients, relative inexperience, inadequate support and limited understanding of normal reaction to suicide. There are limited data on the impact of patient suicide on trainees.

Brown (Reference Brown1987) reported that 25% of psychiatric trainees had experienced patient suicide, of which 77% had a ‘severe’ or ‘strong’ impact. A study of psychotherapists' responses to patient suicide reported that all the subjects had reacted initially with guilt, anger, disbelief and shock, followed by grief, shame, despair and loss of self-esteem (Reference Goldstein and BuongiornoGoldstein & Buongiorno, 1984).

Chemtob et al (Reference Chemtob, Hamada and Bauer1988) undertook a national survey of psychiatrists in North America using a well-validated instrument — the Impact of Events Scale (IES; Reference Horowitz, Wilner and AlvarezHorowitz et al, 1979) — to assess the impact of patient suicide on their personal and professional lives. When compared with a group defined by Horowitz et al (Reference Horowitz, Weiss and Kltreider1984), 53% of their subjects initially were ‘clinically stressed’. In an Irish study, 36% of consultant psychiatrists had experienced an impact in the ‘clinical’ range of the intrusion and 17.5% in the avoidance sub-scale of the above instrument (Reference Cryan, Kelly and MccaffreyCryan et al, 1995). Dewar et al (Reference Dewar, Eagles and Klien2000) found that, in Scotland, 47% of the psychiatric trainees had experienced a patient suicide, of which 31% reported it to have a deleterious effect on their personal and professional lives.

We report on the impact of patient suicide on psychiatric trainees immediately after the event as well as at the time of study.

The study

All 89 psychiatric trainees (senior house officers and registrars) in the St George's Hospital Regional Training Scheme in London were sent the following questionnaires:

1. (a) General information about the doctor, information about the suicide victim, outcome of the coroner's inquest, staff/coroner/family attitudes and support available to the trainee after suicide.

2. (b) Impact of suicide on trainee's personal and professional life on a seven-point scale consisting of 19 questions (Reference Chemtob, Hamada and BauerChemtob et al, 1988).

3. (c) Impact on cognitive and emotional state immediately after suicide and currently as rated by the IES (Reference Horowitz, Wilner and AlvarezHorowitz et al, 1979). This consists of 15 questions about avoidance and intrusive symptoms of posttraumatic stress disorder with a total score in the range 0-75; a score of 20 or above is considered to be clinically significant.

The differences between the immediate impact and that at the time of study on both intrusion and avoidance subscales were tested statistically using a two-tailed Wilcoxon matched-pair signed-rank test. The t-test for two independent samples was used to test differences between groups in the variables rating the impact of suicide.

For trainees who experienced more than one suicide, only the first suicide was investigated.

Findings

Impact of suicide on trainees

The immediate reactions reported by the trainees after the suicide are shown in Table 1.

Table 1. Initial reactions reported by doctors (n=23)

Reaction	Number (%)
Shock	19 (83)
Self-blame/guilt	14 (70)
Grief	9 (39)
Fear of negligence	7 (30)
Anger	3 (13)
‘Difficult’ experience	14 (70)
Useful learning experience	12 (52)
Did not want to talk about it	2 (9)
Earlier education would have been useful	1 (4)

The effect on personal and professional life was rated using the scale devised by Chemtob et al (Reference Chemtob, Hamada and Bauer1988). The score for each item was divided, as in the original study, into three levels of severity. Eight items yielded mean responses in the middle third, one in the upper third and 10 in the lowest third. On comparing the results of the two studies, our group was found to have less overall impact (Table 2).

Table 2. Ratings of the impact of a patient suicide on personal and professional life

	Impact variable	Present study	n	Chemtob et al (Reference Chemtob, Hamada and Bauer1988)	n	P value
1	Increased attention to legal aspects of practice	3.7 (2.00)	23	3.5 (1.9)	130	0.651
21	Increased tendency to hospitalise	3.6 (1.87)	23	2.6 (1.6)	130	0.009
3	More conservative patient selection	2.3 (1.46)	23	2.2 (1.6)	130	0.784
4	Increased focus on suicide cues	4.7 (2.01)	23	4.4 (1.7)	128	0.458
5	Increased concern with death issues	3.2 (1.90)	23	3.1 (1.8)	128	0.812
6	Increased use of collegial consultation	3.4 (1.92)	23	3.0 (1.8)	128	0.342
7	More conservative record-keeping	3.6 (1.76)	23	3.3 (1.9)	128	0.491
8	Increased use of peer consultation	3.0 (1.70)	23	2.8 (1.7)	129	0.612
9	Disturbed relationship with colleagues	1.5 (1.14)	23	1.5 (1.1)	130	1.000
10	Disturbed relationship with friends	1.3 (0.63)	23	1.4 (0.9)	130	0.618
11	Loss of self-esteem	2.8 (1.38)	23	2.9 (1.6)	128	0.783
12	Dreams related to suicide	1.2 (1.07)	23	1.5 (1.0)	130	0.199
131	More intense dreaming	1.0 (0)	23	2.6 (1.7)	34	0.00
141	Disturbed relationship with family	1.1 (0.29)	23	1.8 (1.3)	130	0.00
151	Intrusive thoughts of suicide	1.3 (0.72)	23	2.9 (1.6)	129	0.00
16	Guilt	3.0 (1.50)	23	3.1 (1.7)	129	0.796
171	Anger	1.8 (1.05)	23	3.4 (1.8)	129	0.00
18	Emotional numbness	2.2 (1.31)	23	2.4 (1.6)	129	0.580
19	Social withdrawal	1.2 (0.75)	23	1.5 (1.1)	129	0.221

Twenty-one trainees completed the IES twice, so as to measure the response in the 2 weeks after the suicide and that currently. In the immediate aftermath, they had a mean intrusion score of 11.9 (s.d.=8.5) and a mean avoidance score of 9.7 (s.d.=9.21). Chemtob et al (Reference Chemtob, Hamada and Bauer1988) and Horowitz et al (Reference Horowitz, Weiss and Kltreider1984) defined clinically significant on the IES as scores greater than 12 on the intrusion subscale and greater than 10 on the avoidance sub-scale. On this basis, clinical intrusion was experienced in the immediate aftermath by seven (33%) and clinical avoidance by six (29%) trainees of our study.

On adding the sub-scales together, 11 trainees (52%) were found to have a clinically significant total score, yielding them clinically stressed according to Chemtob et al (Reference Chemtob, Hamada and Bauer1988). There was no significant change in intrusion between the immediate aftermath and at the time of the study (median change=0, lower quartile=-7, upper quartile=+2; P=0.118). Although there was a tendency for the avoidance score to decrease over time (median=-2, lower quartile=-3, upper quartile=0), it was not significant at the 5% level (P=0.076). On combining the two sub-scales, six trainees (29%) were found to be clinically stressed despite the lapse of time.

Comment

The study was confined to psychiatrists in one regional training scheme in the UK and the sample size was small. The training scheme was well established at the time of the survey and may not fully represent trainees elsewhere. The reasons for the 40% non-response rate were not determined, although it is possible that no suicide was experienced. Alternatively, it may reflect the distress and avoidance consequent to the event. Reliance on recall of the impact is a limitation of this and earlier studies. An active risk was identified in almost half of the cases before the suicide in our study, contrasting with the National Confidential Inquiry's report in which 85% of the cases contacting the services in the week before suicide were assessed as having low or absent immediate risk (Reference Appleby, Shaw and AmosAppleby et al, 1999). Low recognition rate may be due to poor information. In our study, information from previous medical notes or GPs had been available in less than half of the cases. The finding that less than half of the patients were seen recently by a senior colleague may suggest poor supervision, which in turn may be owing to time limitation. One of the reasons for little contact with the families after the event could be a fear of accusation, although it was not explored any further. There appeared to be dissonance of the perception by the trainee of the appropriateness of the treatment with the degree of self-blame and guilt that was held. This may be amenable to simple cognitive strategies or informal support from their senior colleagues.

The experience of suicide is potentially adaptive because almost half of the trainees found it to be a useful learning experience and many reported increased focus on suicide cues, although the consequent increased tendency to hospitalise may not always be appropriate with adverse clinical implications. The intensity of the impact of a patient suicide was highlighted by Chemtob et al (Reference Chemtob, Hamada and Bauer1988), who suggested that on average it was as great as that of losing a parent. Surprisingly, we found no significant change over time in the IES. This contrasts with the results of Chemtob et al (Reference Chemtob, Hamada and Bauer1988) and Cryan et al (Reference Cryan, Kelly and Mccaffrey1995) and may be owing to small sample size, the trainee status of our group and the shorter time interval since the suicide. Brown (Reference Brown1987) proposed that training programmes should collaborate to establish policies for the preparation of trainees should they experience patient suicide. Furthermore, he recommended “psychological first aid” for the trainee in the acute phase, facilitated by the supervisor and the training director, followed by “psychological autopsy” 2 to 6 months later. Schneidman (Reference Schneidman1981) recommended “psychological autopsy” after every death to investigate the psychological aspects of the death as well as it being a therapeutic tool to the survivors. Dewar et al (Reference Dewar, Eagles and Klien2000) reported that the trainees who had received training for the potential consequences of suicide found it of value. Our trainees did not deem preparation in the form of earlier discussion or didactic teaching to be useful, but we believe that an interactive training approach using a modelling technique could be beneficial. In addition, both formal and informal support should be made available routinely.

Our main recommendations for future research on this subject are the use of prospective methodology, a larger sample with a control group and an investigation of the benefits of risk assessment, preparation and supervision on the impact of patient suicide on psychiatrists.