Predicting the risk of repetition after self harm: cohort study
BMJ 2005; 330 doi: https://doi.org/10.1136/bmj.38337.584225.82 (Published 17 February 2005) Cite this as: BMJ 2005;330:394All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Sir,
We read with interest the paper by Kapur et al about the ability of
psychiatrists and emergency physicians to predict the likelihood of repeat
episodes of self harm (1). As stated in the article the MASSH forms used
in the study gave the assessing physicians three options; high, moderate
or low risk. Despite these potential three outcomes the authors have
analysed the data in a dichotomous manner as either “high risk” or “not
high risk”. This may represent a difference in perspective between the
emergency physicians and the psychiatrists. From the Emergency Department
(ED) perspective we are most interested in identifying low risk patients
as these are a group who are perceived as not requiring immediate referral
to specialist services. A more logical analysis from the ED perspective
would be to classify patients as “low risk” vs “not low risk” i.e. the
high/moderate risk stratification would be combined as shown below.
Table 1: Predictive value of assessments (% (95% confidence interval)) ED assessments vs MH assessments Sensitivity *: 82.5 (79.5 to 85.2) vs 69.9 (66.2 to 73.4) Specificity #: 33.3 (32.9 to 33.6)vs 45 (44.4 to 45.5) PPV ~: 14.1 (13.6 to 14.5)vs 15.4 (14.6 to 16.2) NPV ^: 93.5 (92.4 to 94.5)vs 91.3 (90.2 to 92.3) NNM £: 43 vs 23
* Sensitivity – if someone repeats self harm within 12 months, how
likely is he/she to have been identified as moderate or high risk at the
initial assessment.
# Specificity – If someone does not repeat self harm within 12
months, how likely is he/she to have been identified as at low risk at the
initial assessment.
~ Positive predictive value – what proportion of those identified as
moderate or high risk at then initial assessment actually go on to harm
themselves again within the next 12 months.
^ Negative predictive value – what proportion of those identified as
low risk would not go on to harm themselves again within the next 12
months.
£ Number needed to miss – how many patients need to be assessed before
one patient is labelled as low risk who is not.
Kapur et al did not report the negative predictive value, yet
arguably, it is clinically more relevant to the initial assessment than
the positive predictive value. Screening assessments should aspire to a
high sensitivity and high negative predictive values, not, as Kapur et al
seem to suggest, to a high positive predictive value.
These results suggest that there are indeed differences between
emergency department and mental health staff. Neither group performs well
at this level though on these figures for every 100 patients assessed; ED
staff would miss 2 patients who repeat whereas mental health staff would
miss 4. In our experience, ED assessments are usually rapid (usually
minutes) and integrated with clinical assessment of the severity of the
self harm. In contrast, assessment by mental health staff typically takes
much longer. It is therefore interesting to note that the increased time
does not improve the predictive nature of the assessment.
Simon Carley
Consultant in Emergency Medicine
Bernard Foëx
Consultant in Emergency Medicine and Critical Care
John Butler
Consultant in Emergency Medicine and Critical Care
On behalf of the Manchester Royal Infirmary emergency department
journal club.
1. Kapur N, Cooper J, Rodway C, Kelly J, Guthrie E, Mackway-Jones K.
Predicting the risk of repetition after self harm: cohort study. BMJ
2005;330:394-395
Competing interests:
None declared
Competing interests: Table 1: Predictive value of assessments (% (95% confidence interval)) ED assessments vs MH assessmentsSensitivity *: 82.5 (79.5 to 85.2) vs 69.9 (66.2 to 73.4)Specificity #: 33.3 (32.9 to 33.6)vs 45 (44.4 to 45.5)PPV ~: 14.1 (13.6 to 14.5)vs 15.4 (14.6 to 16.2)NPV ^: 93.5 (92.4 to 94.5)vs 91.3 (90.2 to 92.3)NNM £: 43 vs 23
This article attracted my attention because we have just completed a
study in which subjects (mental health professionals (MHPs)) assigned
formal probabilities to the semantic lables 'high', 'medium' and 'low'
(HML). Our results indicated wide dipersal of the probabilities associated
with these terms (for example: some subjects denoted 'H' as being a
probility of a given outcome of >0.2 while others considered 'H' to be
> 0.001: a two-hundred fold difference). We concluded that expression
of probability in HML terms is 'at best highly degraded' and 'at worst,
information free'. In the study by Kapur it is not so surprising that the
predictive validity was low as the fact that any predictive validity was
demostrated at all.
It would be of some interest to see if predictive ability were
improved if professionals assigned probability to the given outcome rather
than expressing this as a semantic label. However, some of our previous
work has shown the great difficulty that MHPs in manipulating even the
basic principles of probability theory such that there would be limited
scope for optimisim on this point.
But I would like to congratulate Kapur et al for their simple but
informative study.
Competing interests:
None declared
Competing interests: No competing interests
Prognosis of behaviour is much more difficult than a clinical
prognosis.
Now and then studies and opinions on self- harm and suicide give rise to
endless and inconclusive discussions. In the context of prison inmates,
both waiting for the trial and convicted, the rate of self- harm and
suicide is much higher (about ten times) than in general population.
Reasons are often of a demonstrative nature, or the inmate wants to get
some benefit (for instance, to have a short holiday in a hospital, or to
obtain the detention at home). However, the prison population is not
representative of the general population. Risk assessment is a part of
clinical documents of the inmate, is classified as low, moderate and high,
and is based on clinical (psychological) conditions, anamnesis for self-
harm and drug abuse, familiar support and length of the detention. Self-
harm and suicide are more frequent among those waiting for the trial than
among convicted. In my opinion, this could be due to anxiety for a
verdict, capable of altering severely the future of the person. In any
case, a prognosis is practically impossible, but often the psychiatrist of
the prison, guilty of non- predicting a suicide, will face a trial for
culpable homicide.
Competing interests:
None declared
Competing interests: No competing interests
Dear sir,
This article was very interesting and amusing as well. The reason why I am
using the word "amusing" is that i personally believe that any person
who has harmed himself once has the potential of repeating it and there is
no place of prediction scoring system in this. An individual is not a
machine which works in a set format. God has made all of us differntly.
Each individual has his own thought process, his likes and dislikes, his
hopes and ambitions and his way of reacting to different conditions so any
individual who has tried to do self harm once should be assumed to be at
risk of repeating it again and looked after with great care.
I personally believe that we should not use predicting systems for these
sort of patients as if someone scores low on the predicting system he will
recieve less care. All individuals who have tried to do self harm are
emotionally labile and and there emotions can fluctuate widely.
Another point to note is that even Consultant Psychatrists are sometimes
surprised by the behaviour of there patients so it would be unsafe to try
to develop a predicting system for A&E doctors or G.P.
Kind Regards
Girish Chawla
Competing interests:
None declared
Competing interests: No competing interests
EDITOR- Back in 2002 the BMJ published a trial of a general practice
based intervention to prevent repetition of ‘deliberate self-harm’[1].
The article was illustrated by two pictures, one on the front cover of the
paper version of the journal, which depicted self-cutting. Readers
pointed out that this was misleading because self-poisoning (rather than
self-injury) made up the great majority of self-harm episodes[2]. Times
have moved on. The terminology has changed - the prefix 'deliberate' has
been dropped from 'self-harm' in response to the heterogeneous nature of
the phenomenon and the concerns of service users [3][4]- but the choice of
illustrative material seems to have remained the same.
In ‘This week in
the BMJ’ (19th February) the synopsis of our article on risk assessment
following self-harm[5]was accompanied by a picture of an individual about
to cut themselves with a blade. The terms used to describe aspects of
suicidal behaviour have evolved (attempted suicide, parasuicide,
deliberate self-harm). Self-harm is likely to be with us for the
foreseeable future. What needs to be emphasised is that it refers to both
self-poisoning and self-injury.
Nav Kapur, Senior Lecturer in Psychiatry
Jayne Cooper, Research Fellow
Centre for Suicide Prevention, Department of Psychiatry and Behavioural
Sciences, University of Manchester, Manchester M13 9PL
nav.kapur@manchester.ac.uk
References
[1]Bennewith O, Stocks N, Gunnell D, Peters TJ, Evans MO, Sharp DJ.
General practice based interventions to prevent repeat episodes of
deliberate self harm: cluster randomised controlled trial. BMJ 2002; 324:
1254-1257
[2]Horrocks, J., Owens, D., House, A. Pictures of self-injury misrepresent
trial. BMJ 2002; 325:28.
[3] National Collaborating Centre for Mental Health. Self-harm: the short
term physical and psychological management and secondary prevention of
self-harm in primary and secondary care (full guideline). National
Clinical Practice Guideline 16. Leicester and London: British
Psychological Society and Royal College of Psychiatrists, 2004.
[4] Royal College of Psychiatrists Assessment following self-harm in
adults. London: Royal College of Psychiatrists 2004.
[5] Kapur N, Cooper J, Rodway C, Kelly J, Guthrie E, Mackway-Jones K.
Predicting the risk of repetition after self harm: cohort study. BMJ
2005;330:394-395
Competing interests:
We were the authors of the article which is the subject of this letter
Competing interests: No competing interests
Re: Screening assessments should aim for high sensitivity, not high specificity.
Sir,
We would like to thank all those who commented on our paper [1] but
we would particularly like to respond to the comments of Carley et al. We
agree that Emergency Department and psychiatric perspectives are sometimes
different but we suspect that Carley et al may have misunderstood the main
aim of our study. Our aim was to investigate how well clinicians
predicted the risk of repetition following an episode of self-harm.
The paper was not an investigation of how accurately clinicians
identified non-repeaters. Indeed there may be problems with the ‘low
risk’ ‘not low risk’ dichotomy that Carley et al propose. The reality is
that in busy acute medical settings ‘low risk’ individuals from a
challenging patient group may get very little in the way of treatment[2].
Neither was the paper an evaluation of a screening tool to assist in
the Emergency Department management of patients who had harmed themselves.
Of course, if our aim had been to develop such a screening tool one
approach would be to maximise sensitivity and negative predictive value by
setting the threshold for ‘high risk’ at a fairly low level. We have taken
just such an approach in the development of a simple four-item scale to be
used in the Emergency Department and we hope to publish the findings soon.
We agree that assessments carried out by staff in the Emergency
Department are generally of shorter duration than those carried out by
mental health professionals. We also agree that the predictive value of
mental health assessments with respect to repetition is no better. Does
this mean there is no role for specialist assessment following self-harm?
We do not think so and neither do others [2][3]. The clinical evaluation
of patients following self-harm has been referred to as one of the most
complex assessments in psychiatry [4]. Assessments have a much broader
purpose than simply predicting the risk of repetition (for example,
exploring social circumstances and personal relationships, examining the
role of recent life events, identifying psychological symptoms). This is
reflected in the growing emphasis on 'needs assessment' [2] in recent
national guidance.
Nav Kapur
Jayne Cooper
Kevin Mackway-Jones
Else Guthrie
References:
[1] Kapur N, Cooper J, Rodway C, Kelly J, Guthrie E, Mackway-Jones K.
Predicting the risk of. repetition after self harm: cohort study. BMJ
2005;330:394-395
[2] National Collaborating Centre for Mental Health. Self-harm: the
short term physical and psychological management and secondary prevention
of self-harm in primary and secondary care (full guideline). National
Clinical Practice Guideline 16. Leicester and London: British
Psychological Society and Royal College of Psychiatrists, 2004.
[3] Royal College of Psychiatrists Assessment following self-harm in
adults. London: Royal College of Psychiatrists 2004
[4] Isacsson, G, Rich C. Management of patients who deliberately
harm themselves. BMJ 2001;322:213-215.
Competing interests:
None declared
Competing interests: No competing interests