- Email: [email protected]
Suicide Intent Scale in the prediction of suicide
Journal of Affective Disorders 136 (2012) 167–171
Contents lists available at SciVerse ScienceDirect
Journal of Affective Disorders j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a d
Brief report
Suicide Intent Scale in the prediction of suicide J. Stefansson, P. Nordström, J. Jokinen ⁎ Department of Clinical Neuroscience, Psychiatry Section, Karolinska Institutet, Karolinska University Hospital, Solna, SE-171 76 Stockholm, Sweden
a r t i c l e
i n f o
Article history: Received 13 October 2010 Received in revised form 15 November 2010 Accepted 15 November 2010 Available online 8 December 2010
Keywords: Suicide intent Prediction Suicide attempt Suicide Scales ROC Karolinska Institutet
a b s t r a c t Objective: To assess the predictive value of the Suicide Intent Scale in patients with high suicide risk. The secondary aim was to assess if the use of the factors of the Suicide Intent Scale may offer a better predictive value in suicide risk detection. Finally a shorter version of the scale was created after an item analysis. Method: Eighty-one suicide attempters were assessed with the Beck's Suicide Intent Scale (SIS). All patients were followed up for cause of death. Receiver-operating characteristic (ROC) curves and tables were created to establish the optimal cut-off values for SIS and SIS factors to predict suicide. Results: Seven patients committed suicide during a mean follow up of 9.5 years. The major finding was that mean SIS scores distinguished between suicides and survivors. The positive predictive value was 16.7% and the Area Under Curve (AUC) was 0.74. Only the planning subscale reached statistical significance. Four items were used to test a shorter version of the SIS in the suicide prediction. The positive predictive value was 19% and the AUC was 0.82. Conclusions: The Suicide Intent Scale is a valuable tool in clinical suicide risk assessment, a shorter version of the scale may offer a better predictive value. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Over the past 30 years, Beck's Suicide Intent Scale (SIS) has been the prevailing psychometric scale for assessing suicide intent in suicide attempters (Freedenthal, 2008). In a recent review article, five out of 13 studies showed a positive relationship between SIS scores and suicide over a follow-up period ranging from 10 months to 20 years (Freedenthal, 2008). Only two earlier clinical studies have used receiver operating characteristics (ROC) to assess the optimal threshold of the SIS in suicide prediction (Niméus et al., 2002; Harriss and Hawton, 2005). Earlier studies of the factorial structure of the SIS have identified between two and four factors (Antretter et al., 2008). Mieczkowski found two factors: planning subscale and lethal
⁎ Corresponding author. Department of Clinical Neuroscience/Psychiatry, Karolinska Institutet, R5, Karolinska University Hospital, Solna, SE-171 76 Stockholm, Sweden. Tel.: + 46 8 51776759; fax: + 46 8 303706. E-mail address: [email protected] (J. Jokinen). 0165-0327/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jad.2010.11.016
intent, and recently Misson et al. presented a four factor solution of the SIS in suicide attempters. In a recent study by Antretter et al. only one factor: “subjective part” of the SIS consisting of items 9 to 14 (the same items as in factor lethal intent by Mieczkowski) was strongly supported, whereas an acceptable model fit for the ‘objective part’ was not found in eleven clinical samples (Antretter et al., 2008). They concluded that possible future revisions of ‘objective’ SIS items may be worth consideration. There is a need for validated clinical tools for suicide risk assessment that can be easily administered. To the best of our knowledge, only two studies have compared underlying factors of the SIS in suicide prediction (Niméus et al., 2002; Harriss and Hawton, 2005) and only one study has assessed the suicide predictive validity of individual items of the scale (Niméus et al., 2002). We hypothesized that high scores with Suicide Intent Scale may predict future suicide after attempted suicide. The aim of the present study was to assess the predictive value of the Suicide Intent Scale in patients with high suicide risk, i.e. patients admitted to a psychiatric clinic after a suicide attempt. The secondary aim was to assess the predictive value
168
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171
of the factors of the Suicide Intent Scale to detect future suicide in suicide attempters. Finally a shorter version of the scale was created after an item analysis. 2. Methods
Åsberg, 1979). The mean score of MADRS was 16 (median 17, S.D. 9, range 0–37). The Beck Hopelessness Scale is a 20-item true/false instrument with statements of pessimistic beliefs about oneself and the future (Beck et al., 1974b). Mean level of hopelessness was 10.4, (median 11, S.D. 6.2, range 0–20).
2.1. Study setting 2.4. Outcomes Patients having their clinical follow-up after a suicide attempt at the Karolinska University Hospital were asked to participate in a study of biological and psychological risk factors for suicidal behaviour. The Regional Ethical Review Board in Stockholm approved the study protocols (Dnr 93-211) and the participants gave their written informed consent to the study. 2.2. Subjects This is a cohort study involving 81 suicide attempters (35 men, mean age 39 years, S.D.=11.8, range 20–69 and 46 women, mean age 35 years, S.D.=12.1, range 18–68). Patients were included to the study between 1993 and 1998. Inclusion criteria were a recent suicide attempt (a time limit of one month), fair capacity to communicate verbally and in writing in the Swedish language and an age of 18 years or older. Exclusion criteria were schizophrenia spectrum psychosis, dementia, mental retardation and intravenous drug abuse. Suicide attempt was defined as any nonfatal, self-injurious behaviour with some intent to cause death. The participants were interviewed by a trained psychiatrist using the SCID I research version interview to establish diagnosis according to DSM-III (American Psychiatric Association). Axis II diagnoses were established with SCID II interview. Ninety-four percent of participants had at least one current Axis I psychiatric diagnosis; 80% of patients fulfilled criteria for mood disorder, 5% for adjustment disorder and 4% for anxiety disorders, one patient had substance related disorder, one patient had anorexia nervosa and one an unspecified psychiatric disorder (not psychotic). Twenty-one percent of the patients had a co morbid substance related disorder (mostly alcohol dependence). Among Axis II diagnoses, 39% of the patients fulfilled criteria for a personality disorder. Fourteen patients (17%) had used a violent suicide attempt method. 2.3. Assessments Beck's Suicide Intent Scale (SIS), is an instrument using 15-items designed to examine the factual aspects of the suicide attempt; such as the patients thoughts and feelings and the circumstances at the time of the suicide attempt (Beck et al., 1974a). One patients SIS rating was incomplete survivor and was not used in the statistical analysis. The mean value of SIS was 16, (median 16, S.D. 5.7, range 2–27, n =80). One two-factor model of the SIS (Factor 1: Lethal intent and Factor 2: Planning) (Mieczkowski et al., 1993) and one four-factor model of the SIS (Conception, Preparation, Precautions and Communication) (Misson et al., 2010) were composed and tested separately. The factor Lethal Intent is identical to the factor Conception according to Misson consisting of items 9–14 of the SIS also named as the “subjective part” according to Antretter as well as. To evaluate severity of depression, the Montgomery–Åsberg Depression Rating Scale (MADRS) was used (Montgomery and
By use of the unique personal identification number patients were linked to the Cause of Death register, maintained by the National Board of Health and Welfare in Sweden (http:// www.socialstyrelsen.se). Seven patients had committed suicide before January 2009; suicides were ascertained from the death certificates. Five patients committed suicide within 6 years, two patients died of suicide after 11 years from entering to the study (time to suicide: median 4 years, mean 6 years, range between 1.7 and 12.8 years). The follow up time ranged between 10 and 15 years. There was no age difference between suicides and survivors. 2.5. Data analysis Characteristics of the population were described by using the mean, the median and the range for quantitative variables. Shapiro–Wilk test was used to test if data was normally distributed. Parametric statistics, t-test one tailed was applied for between-group comparisons, suicide victims vs. survivors if data was normally distributed. If skewed, nonparametric statistics (Kruskal–Wallis' test) in continuous variables was applied for between-group comparisons. An ad hoc ROC analysis was used to find optimal thresholds for SIS and SIS factors to predict suicide. Receiver-operating characteristic (ROC) curves and tables were created for scales to establish the optimal cut-off values. ROC areas under the curves (AUCs) were calculated as a measure of the diagnostic performance, and differences were calculated and tested according to the methods of Hanley and McNeil. The cut-off point that optimized sensitivity (proportion of suicides correctly identified) and specificity (proportion of survivors correctly identified) was used. Pearson Chi-square and Fisher's exact test were used for cross tabulations of categorical variables. Statistical analyses were performed using JMP VI software, SAS Institute inc., Cary, NC, USA. The p value was set at b0.05.
25 20 15
SIS ratings 10 5 0
Suicide victims
Survivors
Fig. 1. Suicide Intent scores in suicide victims (mean + SE) (20.1 ± 1.2) and in survivors (15.7 ± 0.7) (n = 80, p = 0.026).
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171
169
1,00 0,90 0,80 0,70 0,60
Sensitivity
0,50 0,40 0,30 0,20 0,10 0,00 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00
1-Specificity Fig. 2. ROC curve for Suicide Intent Scale in suicide prediction, AUC = 0.74.
shows positive predictive values of SIS, SIS planning subscale and SIS shorter version in suicide prediction. The Beck Hopelessness Scale did not distinguish between suicides and survivors (p b 0.69). The correlations between SIS and BHS as well as SIS and MADRS were non significant. There were no differences of SIS ratings in suicide attempters with or without co morbid substance abuse or personality disorder. There were no differences of SIS ratings between suicide attempters who had used a violent suicide attempt method vs. non violent method.
3. Results The SIS scores were normally distributed but not the scores of the factors. Seven suicides (8.6%) occurred during the followup time: 3 women (6.5%) and 4 men (11.4%). The major finding was that mean SIS scores distinguished between suicides (mean+SD) (20.1±3.2) and survivors (15.7±5.8) (n=80, t ratio=1.98, p=0.026, t-test, one-tailed), Fig. 1. To estimate which cut-off level of SIS scores optimally predicts suicide, we analyzed the ROC curves and the ROC tables. The cut-off 16 gave a specificity of 52% and a sensitivity of 100%. The positive predictive value was 16.7% and the AUC was 0.74, Fig. 2. The ratings of SIS factors in suicide victims and survivors are presented in Table 1. Only the planning subscale reached statistical significance. From an analysis of the separate scores of each item in the SIS we found that four items of the scale were different in suicide victims compared with survivors under the significance level p b 0.1. We constructed a new scale with these items. Items 4, 7, 12 and 13 were used to test a shorter version of the SIS in suicide prediction. ROC analysis revealed an optimal cut-off of 6 which gave specificity of 59% and sensitivity of 100%. The positive predictive value was 19% and the AUC was 0.82. Table 2
4. Discussion In this follow up study of 81 suicide attempters, 8.6% committed suicide during a mean follow-up time of almost ten years. This can be compared with clinical studies reporting suicide mortality between 4 and 12% in suicide attempters (Beck and Steer, 1989; Lindqvist et al., 2007; Niméus et al., 2002; Nordström et al., 1995; Skogman et al., 2004; Suominen et al., 2004; Runeson et al., 2010). We found that suicide intent assessed shortly after a suicide attempt predicted subsequent suicide. This is in line with five earlier studies in different clinical populations showing the Suicide Intent Scale as a significant predictor of suicide (Harriss et al., 2005; Hawton and Harriss,
Table 1 Suicide victims (N = 7)
Survivors (N = 74)
Rating
Items
Mean
Median
SD
Range
Mean
Median
SD
Range
Statistic
SIS planning
1–7, 15
10.1
10
2.5
7–15
7.3
7
3.8
0–15
SIS lethal intent/Conception/“subjective part”
9–14
9.7
10
1.1
8–11
8.3
8.5
2.9
1–12
Preparation
5, 6, 7, 15
3.9
3
2.1
2–8
0.98
2.6
2.4
0–8
Precautions
1–3
4.4
5
1.5
2–6
3.5
4
1.9
0–6
Communication
4, 8
2.1
2
0.4
2–3
1.5
2
0.9
0–4
SIS shorter version
4, 7, 12, 13
7
7
1
6–8
5
5
1.9
0–8
Z = 2.0 p b 0.045 Z = 1.2 p b 0.22 Z = 1.3 p b 0.19 Z = 1.3 p b 0.21 Z = 1.9 p b 0.058 Z = 2.8 p b 0.0046
170
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171
Table 2 Positive predictive values of SIS, SIS planning subscale and SIS shorter version in suicide prediction. Test
Cut off
Suicide+
Suicide−
Sensitivity
Specificity
Positive predictive value
AUC
Test ⁎
SIS
N16
7
35
100% (7/7)
52% (38/73)
16.7% (7/42)
0.74
p = 0.012
SIS planning
b16 N7
0 7
38 39
100% (7/7)
46% (39/72)
15.2% (7/46)
0.73
p = 0.012
SIS shorter version
b7 N6
0 7
33 30
100% (7/7)
59% (43/73)
19% (7/37)
0.82
p = 0.003
b6
0
43
⁎ Fisher exact (2-sided).
2006; Niméus et al., 2002; Pierce, 1987; Suominen et al., 2004). One study found a positive relationship for women only (Skogman et al., 2004). Two of the earlier positive reports studied elderly patients (Pierce, 1987; Hawton and Harriss, 2006). However there are several large negative studies. So far seven studies, with follow-up periods from 113 days to 10 years, did not find a statistically significant association between the SIS scores and later suicide (see review by Freedenthal, 2008). One of the negative studies showed that the total SIS score did not predict suicide whereas items 1, 2 and 3 measuring precaution did so (Beck and Steer, 1989). Only two studies have compared the underlying factors of the Beck's Suicide Intent Scale in suicide prediction (Niméus et al., 2002; Harriss and Hawton, 2005). In the study of Niméus and his coworkers all the subscales/factors were predictive for future suicide. We found that only higher scores in the Planning subscale were a significant predictor of future suicide. The Planning subscale showed a very similar predictive value and the AUC compared with the SIS total score, whereas Lethal intent factor scores did not predict suicide. The seven attempters who later killed themselves had reported more planning at the time of their index attempt than the 74 patients who did not commit suicide. This is partly in line with the results of Harris and Hawton who found a stronger association between the circumstances section of the SIS (items 1–8) and suicide especially in female deliberate self harm patients. Interestingly the planning subscale was associated with lower levels of CSF 5-HIAA, a replicated biomarker of suicide risk (Mann et al., 1996). We have earlier reported that CSF 5-HIAA was a short term predictor of suicide compared to suicide intent and hopelessness assessed after a suicide attempt in male mood disorder inpatients (Samuelsson et al., 2006). In this study, we found that the cut-off point of 16 was optimal and the positive predictive value was 16.7%, which is higher than the PPV in the study of Niméus et al. who reported a PPV of 9.7% in the whole sample. They also found that Suicide Intent Scale may offer a better prediction if targeted in elderly suicide attempters with PPV of 22.5% for those 55 years or older. Harris and Hawton reported a low PPV of 4.0% in a large group of deliberate self harm patients. They concluded that the SIS cannot predict which individual patients will ultimately die by suicide due to a large number of false positives. Our optimal cut-off of 16 was somewhat lower than in the study of Niméus et al. but higher than in deliberate self harm patients (Niméus et al., 2002; Harriss and Hawton, 2005). This could partly be explained by suicidal intent being lower in deliberate self harm patients.
We constructed a shorter version of the SIS with items 4, 7, 12 and 13. When we used this shorter version of the SIS in suicide prediction we found that an optimal cut-off of 6 gave specificity of 59% and sensitivity of 100%. The positive predictive value was 19% and the AUC was 0.82 which gave a better prediction by reducing the number of false positives. In the study of Niméus et al., shorter versions of the SIS or the factors were not superior to the original SIS scale in predicting suicide. Having a shorter test may increase the clinical utility of the scale. The utility and predictive value of the shorter version should be replicated in a larger cohort of suicide attempters. Limitation of this study is a small number of patients; furthermore suicidal intentions are very difficult to measure per se (Freedenthal, 2007). A patient who tried to die may deny suicidal intent to avoid hospitalization. Shame, ambivalence, confusion or intoxication can contribute to recall bias. In summary, our findings support the use of information about suicidal intent as part of a clinical suicide risk assessment. Further work needs to be done to test the utility of a shorter version of the SIS combined with other clinical rating scales measuring other types of suicide risk factors such as violence. Role of funding source Funding for this study was provided by the Swedish Research Council (Project number K2009-61P-21304-04-4) by Söderström-Königska Foundation and by the Thurings Foundation. The Swedish Research Council, Söderström-Königska Foundation and the Thurings Foundation had no further role in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication. Conflict of interest No conflicts of interests to declare for any of the co authors.
Acknowledgements We want to acknowledge Professor Marie Åsberg for inspiring us with studies in suicidology Dr Kaj Forslund for careful clinical assessments and Dr. Large who made helpful comments on the electronic version of the paper resulting in an improved final paper version. References Antretter, E., Dunkel, D., Haring, C., Corcoran, P., De Leo, D., Fekete, S., Hawton, K., Kerkhof, A.J., Lonnqvist, J., Renberg, E.S., Schmidtke, A., Van Heeringen, K., Wasserman, D., 2008. The factorial structure of the Suicide Intent Scale: a comparative study in clinical samples from 11 European regions. Int. J. Meth. Psychiatr. Res. 17, 63–79.
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171 Beck, A.T., Steer, R.A., 1989. Clinical predictors of eventual suicide: a 5- to 10-year prospective study of suicide attempters. J. Affect. Disord. 17, 203–209. Beck, A.T., Schuyler, D., Herman, I., 1974a. Development of suicidal intent scales. In: Beck, A.T., Resnik, H.L.P., Lettieri, D.J. (Eds.), The Prediction of Suicide. Charles Press, Philadelphia, PA, pp. 45–56. Beck, A.T., Weissman, A., Lester, D., Trexler, L., 1974b. The measurement of pessimism: the hopelessness scale. J. Consult. Clin. Psychol. 42 (6), 861–865. Freedenthal, S., 2007. Challenges in assessing intent to die: can suicide attempters be trusted? Omega J. Death Dying 55, 57–70. Freedenthal, S., 2008. Assessing the wish to die: a 30-year review of the Suicide Intent Scale. Arch. Suicide Res. 12 (4), 277–298. Harriss, L., Hawton, K., 2005. Suicidal intent in deliberate self-harm and the risk of suicide: the predictive power of the Suicide Intent Scale. J. Affect. Disord. 86, 225–233. Harriss, L., Hawton, K., Zahl, D., 2005. Value of measuring suicidal intent in the assessment of people attending hospital following self-poisoning or self-injury. Br. J. Psychiatry 186, 60–66. Hawton, K., Harriss, L., 2006. Deliberate self-harm in people aged 60 years and over: characteristics and outcome of a 20-year cohort. Int. J. Geriatr. Psychiatry 21, 572–581. Lindqvist, D., Niméus, A., Träskman-Bendz, L., 2007. Suicidal intent and psychiatric symptoms among inpatient suicide attempters. Nord. J. Psychiatry 61, 27–32. Mann, J.J., Malone, K.M., Psych, M.R., Sweeney, J.A., Brown, R.P., Linnoila, M., Stanley, B., Stanley, M., 1996. Attempted suicide characteristics and cerebrospinal fluid amine metabolites in depressed inpatients. Neuropsychopharmacology 15 (6), 576–586. Mieczkowski, T.A., Sweeney, J.A., Haas, G.L., Junker, B.W., Brown, R.P., Mann, J.J., 1993. Factor composition of the Suicide Intent Scale. Suicide Life-Threat. Behav. 23, 37–45.
171
Misson, H., Mathieu, F., Jollant, F., Yon, L., Guillaume, S., Parmentier, C., Raust, A., Jaussent, I., Slama, F., Leboyer, M., Bellivier, F., Courtet, P., 2010. Factor analyses of the Suicidal Intent Scale (SIS) and the Risk-Rescue Rating Scale (RRRS): toward the identification of homogeneous subgroups of behaviors. J. Affect. Disord. 121 (1–2), 80–87. Montgomery, S.A., Åsberg, M., 1979. A new depression scale designed to be sensitive to change. Br. J. Psychiatry 134, 382–389. Niméus, A., Alsén, M., Träskman-Bendz, L., 2002. High suicidal intent scores indicate future suicide. Arch. Suicide Res. 6, 211–219. Nordström, P., Samuelsson, M., Åsberg, M., 1995. Survival analysis of suicide risk after attempted suicide. Acta Psychiatr. Scand. 91, 336–340. Pierce, D., 1987. Deliberate self-harm in the elderly. Int. J. Geriatr. Psychiatry 2, 105–110. Runeson, B., Tidemalm, D., Dahlin, M., Lichtenstein, P., Långström, N., 2010. Method of attempted suicide as predictor of subsequent successful suicide: national long term cohort study. BMJ 13 (341), c3222. Samuelsson, M., Jokinen, J., Nordstrom, A.L., Nordstrom, P., 2006. CSF-HIAA, suicide intent and hopelessness in the prediction of early suicide in male high-risk suicide attempters. Acta Psychiatr. Scand. 113, 44–47. Skogman, K., Alsen, M., Öjehagen, A., 2004. Sex differences in risk factors for suicide after attempted suicide—a follow-up study of 1052 suicide attempters. Soc. Psychiatry Psychiatr. Epidemiol. 39, 113–120. Suominen, K., Isometsa, E., Ostamo, A., Lönnqvist, J., 2004. Level of suicidal intent predicts overall mortality and suicide after attempted suicide: a 12-year follow-up study. BMC Psychiatry 4, 11.
Contents lists available at SciVerse ScienceDirect
Journal of Affective Disorders j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a d
Brief report
Suicide Intent Scale in the prediction of suicide J. Stefansson, P. Nordström, J. Jokinen ⁎ Department of Clinical Neuroscience, Psychiatry Section, Karolinska Institutet, Karolinska University Hospital, Solna, SE-171 76 Stockholm, Sweden
a r t i c l e
i n f o
Article history: Received 13 October 2010 Received in revised form 15 November 2010 Accepted 15 November 2010 Available online 8 December 2010
Keywords: Suicide intent Prediction Suicide attempt Suicide Scales ROC Karolinska Institutet
a b s t r a c t Objective: To assess the predictive value of the Suicide Intent Scale in patients with high suicide risk. The secondary aim was to assess if the use of the factors of the Suicide Intent Scale may offer a better predictive value in suicide risk detection. Finally a shorter version of the scale was created after an item analysis. Method: Eighty-one suicide attempters were assessed with the Beck's Suicide Intent Scale (SIS). All patients were followed up for cause of death. Receiver-operating characteristic (ROC) curves and tables were created to establish the optimal cut-off values for SIS and SIS factors to predict suicide. Results: Seven patients committed suicide during a mean follow up of 9.5 years. The major finding was that mean SIS scores distinguished between suicides and survivors. The positive predictive value was 16.7% and the Area Under Curve (AUC) was 0.74. Only the planning subscale reached statistical significance. Four items were used to test a shorter version of the SIS in the suicide prediction. The positive predictive value was 19% and the AUC was 0.82. Conclusions: The Suicide Intent Scale is a valuable tool in clinical suicide risk assessment, a shorter version of the scale may offer a better predictive value. © 2010 Elsevier B.V. All rights reserved.
1. Introduction Over the past 30 years, Beck's Suicide Intent Scale (SIS) has been the prevailing psychometric scale for assessing suicide intent in suicide attempters (Freedenthal, 2008). In a recent review article, five out of 13 studies showed a positive relationship between SIS scores and suicide over a follow-up period ranging from 10 months to 20 years (Freedenthal, 2008). Only two earlier clinical studies have used receiver operating characteristics (ROC) to assess the optimal threshold of the SIS in suicide prediction (Niméus et al., 2002; Harriss and Hawton, 2005). Earlier studies of the factorial structure of the SIS have identified between two and four factors (Antretter et al., 2008). Mieczkowski found two factors: planning subscale and lethal
⁎ Corresponding author. Department of Clinical Neuroscience/Psychiatry, Karolinska Institutet, R5, Karolinska University Hospital, Solna, SE-171 76 Stockholm, Sweden. Tel.: + 46 8 51776759; fax: + 46 8 303706. E-mail address: [email protected] (J. Jokinen). 0165-0327/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jad.2010.11.016
intent, and recently Misson et al. presented a four factor solution of the SIS in suicide attempters. In a recent study by Antretter et al. only one factor: “subjective part” of the SIS consisting of items 9 to 14 (the same items as in factor lethal intent by Mieczkowski) was strongly supported, whereas an acceptable model fit for the ‘objective part’ was not found in eleven clinical samples (Antretter et al., 2008). They concluded that possible future revisions of ‘objective’ SIS items may be worth consideration. There is a need for validated clinical tools for suicide risk assessment that can be easily administered. To the best of our knowledge, only two studies have compared underlying factors of the SIS in suicide prediction (Niméus et al., 2002; Harriss and Hawton, 2005) and only one study has assessed the suicide predictive validity of individual items of the scale (Niméus et al., 2002). We hypothesized that high scores with Suicide Intent Scale may predict future suicide after attempted suicide. The aim of the present study was to assess the predictive value of the Suicide Intent Scale in patients with high suicide risk, i.e. patients admitted to a psychiatric clinic after a suicide attempt. The secondary aim was to assess the predictive value
168
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171
of the factors of the Suicide Intent Scale to detect future suicide in suicide attempters. Finally a shorter version of the scale was created after an item analysis. 2. Methods
Åsberg, 1979). The mean score of MADRS was 16 (median 17, S.D. 9, range 0–37). The Beck Hopelessness Scale is a 20-item true/false instrument with statements of pessimistic beliefs about oneself and the future (Beck et al., 1974b). Mean level of hopelessness was 10.4, (median 11, S.D. 6.2, range 0–20).
2.1. Study setting 2.4. Outcomes Patients having their clinical follow-up after a suicide attempt at the Karolinska University Hospital were asked to participate in a study of biological and psychological risk factors for suicidal behaviour. The Regional Ethical Review Board in Stockholm approved the study protocols (Dnr 93-211) and the participants gave their written informed consent to the study. 2.2. Subjects This is a cohort study involving 81 suicide attempters (35 men, mean age 39 years, S.D.=11.8, range 20–69 and 46 women, mean age 35 years, S.D.=12.1, range 18–68). Patients were included to the study between 1993 and 1998. Inclusion criteria were a recent suicide attempt (a time limit of one month), fair capacity to communicate verbally and in writing in the Swedish language and an age of 18 years or older. Exclusion criteria were schizophrenia spectrum psychosis, dementia, mental retardation and intravenous drug abuse. Suicide attempt was defined as any nonfatal, self-injurious behaviour with some intent to cause death. The participants were interviewed by a trained psychiatrist using the SCID I research version interview to establish diagnosis according to DSM-III (American Psychiatric Association). Axis II diagnoses were established with SCID II interview. Ninety-four percent of participants had at least one current Axis I psychiatric diagnosis; 80% of patients fulfilled criteria for mood disorder, 5% for adjustment disorder and 4% for anxiety disorders, one patient had substance related disorder, one patient had anorexia nervosa and one an unspecified psychiatric disorder (not psychotic). Twenty-one percent of the patients had a co morbid substance related disorder (mostly alcohol dependence). Among Axis II diagnoses, 39% of the patients fulfilled criteria for a personality disorder. Fourteen patients (17%) had used a violent suicide attempt method. 2.3. Assessments Beck's Suicide Intent Scale (SIS), is an instrument using 15-items designed to examine the factual aspects of the suicide attempt; such as the patients thoughts and feelings and the circumstances at the time of the suicide attempt (Beck et al., 1974a). One patients SIS rating was incomplete survivor and was not used in the statistical analysis. The mean value of SIS was 16, (median 16, S.D. 5.7, range 2–27, n =80). One two-factor model of the SIS (Factor 1: Lethal intent and Factor 2: Planning) (Mieczkowski et al., 1993) and one four-factor model of the SIS (Conception, Preparation, Precautions and Communication) (Misson et al., 2010) were composed and tested separately. The factor Lethal Intent is identical to the factor Conception according to Misson consisting of items 9–14 of the SIS also named as the “subjective part” according to Antretter as well as. To evaluate severity of depression, the Montgomery–Åsberg Depression Rating Scale (MADRS) was used (Montgomery and
By use of the unique personal identification number patients were linked to the Cause of Death register, maintained by the National Board of Health and Welfare in Sweden (http:// www.socialstyrelsen.se). Seven patients had committed suicide before January 2009; suicides were ascertained from the death certificates. Five patients committed suicide within 6 years, two patients died of suicide after 11 years from entering to the study (time to suicide: median 4 years, mean 6 years, range between 1.7 and 12.8 years). The follow up time ranged between 10 and 15 years. There was no age difference between suicides and survivors. 2.5. Data analysis Characteristics of the population were described by using the mean, the median and the range for quantitative variables. Shapiro–Wilk test was used to test if data was normally distributed. Parametric statistics, t-test one tailed was applied for between-group comparisons, suicide victims vs. survivors if data was normally distributed. If skewed, nonparametric statistics (Kruskal–Wallis' test) in continuous variables was applied for between-group comparisons. An ad hoc ROC analysis was used to find optimal thresholds for SIS and SIS factors to predict suicide. Receiver-operating characteristic (ROC) curves and tables were created for scales to establish the optimal cut-off values. ROC areas under the curves (AUCs) were calculated as a measure of the diagnostic performance, and differences were calculated and tested according to the methods of Hanley and McNeil. The cut-off point that optimized sensitivity (proportion of suicides correctly identified) and specificity (proportion of survivors correctly identified) was used. Pearson Chi-square and Fisher's exact test were used for cross tabulations of categorical variables. Statistical analyses were performed using JMP VI software, SAS Institute inc., Cary, NC, USA. The p value was set at b0.05.
25 20 15
SIS ratings 10 5 0
Suicide victims
Survivors
Fig. 1. Suicide Intent scores in suicide victims (mean + SE) (20.1 ± 1.2) and in survivors (15.7 ± 0.7) (n = 80, p = 0.026).
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171
169
1,00 0,90 0,80 0,70 0,60
Sensitivity
0,50 0,40 0,30 0,20 0,10 0,00 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00
1-Specificity Fig. 2. ROC curve for Suicide Intent Scale in suicide prediction, AUC = 0.74.
shows positive predictive values of SIS, SIS planning subscale and SIS shorter version in suicide prediction. The Beck Hopelessness Scale did not distinguish between suicides and survivors (p b 0.69). The correlations between SIS and BHS as well as SIS and MADRS were non significant. There were no differences of SIS ratings in suicide attempters with or without co morbid substance abuse or personality disorder. There were no differences of SIS ratings between suicide attempters who had used a violent suicide attempt method vs. non violent method.
3. Results The SIS scores were normally distributed but not the scores of the factors. Seven suicides (8.6%) occurred during the followup time: 3 women (6.5%) and 4 men (11.4%). The major finding was that mean SIS scores distinguished between suicides (mean+SD) (20.1±3.2) and survivors (15.7±5.8) (n=80, t ratio=1.98, p=0.026, t-test, one-tailed), Fig. 1. To estimate which cut-off level of SIS scores optimally predicts suicide, we analyzed the ROC curves and the ROC tables. The cut-off 16 gave a specificity of 52% and a sensitivity of 100%. The positive predictive value was 16.7% and the AUC was 0.74, Fig. 2. The ratings of SIS factors in suicide victims and survivors are presented in Table 1. Only the planning subscale reached statistical significance. From an analysis of the separate scores of each item in the SIS we found that four items of the scale were different in suicide victims compared with survivors under the significance level p b 0.1. We constructed a new scale with these items. Items 4, 7, 12 and 13 were used to test a shorter version of the SIS in suicide prediction. ROC analysis revealed an optimal cut-off of 6 which gave specificity of 59% and sensitivity of 100%. The positive predictive value was 19% and the AUC was 0.82. Table 2
4. Discussion In this follow up study of 81 suicide attempters, 8.6% committed suicide during a mean follow-up time of almost ten years. This can be compared with clinical studies reporting suicide mortality between 4 and 12% in suicide attempters (Beck and Steer, 1989; Lindqvist et al., 2007; Niméus et al., 2002; Nordström et al., 1995; Skogman et al., 2004; Suominen et al., 2004; Runeson et al., 2010). We found that suicide intent assessed shortly after a suicide attempt predicted subsequent suicide. This is in line with five earlier studies in different clinical populations showing the Suicide Intent Scale as a significant predictor of suicide (Harriss et al., 2005; Hawton and Harriss,
Table 1 Suicide victims (N = 7)
Survivors (N = 74)
Rating
Items
Mean
Median
SD
Range
Mean
Median
SD
Range
Statistic
SIS planning
1–7, 15
10.1
10
2.5
7–15
7.3
7
3.8
0–15
SIS lethal intent/Conception/“subjective part”
9–14
9.7
10
1.1
8–11
8.3
8.5
2.9
1–12
Preparation
5, 6, 7, 15
3.9
3
2.1
2–8
0.98
2.6
2.4
0–8
Precautions
1–3
4.4
5
1.5
2–6
3.5
4
1.9
0–6
Communication
4, 8
2.1
2
0.4
2–3
1.5
2
0.9
0–4
SIS shorter version
4, 7, 12, 13
7
7
1
6–8
5
5
1.9
0–8
Z = 2.0 p b 0.045 Z = 1.2 p b 0.22 Z = 1.3 p b 0.19 Z = 1.3 p b 0.21 Z = 1.9 p b 0.058 Z = 2.8 p b 0.0046
170
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171
Table 2 Positive predictive values of SIS, SIS planning subscale and SIS shorter version in suicide prediction. Test
Cut off
Suicide+
Suicide−
Sensitivity
Specificity
Positive predictive value
AUC
Test ⁎
SIS
N16
7
35
100% (7/7)
52% (38/73)
16.7% (7/42)
0.74
p = 0.012
SIS planning
b16 N7
0 7
38 39
100% (7/7)
46% (39/72)
15.2% (7/46)
0.73
p = 0.012
SIS shorter version
b7 N6
0 7
33 30
100% (7/7)
59% (43/73)
19% (7/37)
0.82
p = 0.003
b6
0
43
⁎ Fisher exact (2-sided).
2006; Niméus et al., 2002; Pierce, 1987; Suominen et al., 2004). One study found a positive relationship for women only (Skogman et al., 2004). Two of the earlier positive reports studied elderly patients (Pierce, 1987; Hawton and Harriss, 2006). However there are several large negative studies. So far seven studies, with follow-up periods from 113 days to 10 years, did not find a statistically significant association between the SIS scores and later suicide (see review by Freedenthal, 2008). One of the negative studies showed that the total SIS score did not predict suicide whereas items 1, 2 and 3 measuring precaution did so (Beck and Steer, 1989). Only two studies have compared the underlying factors of the Beck's Suicide Intent Scale in suicide prediction (Niméus et al., 2002; Harriss and Hawton, 2005). In the study of Niméus and his coworkers all the subscales/factors were predictive for future suicide. We found that only higher scores in the Planning subscale were a significant predictor of future suicide. The Planning subscale showed a very similar predictive value and the AUC compared with the SIS total score, whereas Lethal intent factor scores did not predict suicide. The seven attempters who later killed themselves had reported more planning at the time of their index attempt than the 74 patients who did not commit suicide. This is partly in line with the results of Harris and Hawton who found a stronger association between the circumstances section of the SIS (items 1–8) and suicide especially in female deliberate self harm patients. Interestingly the planning subscale was associated with lower levels of CSF 5-HIAA, a replicated biomarker of suicide risk (Mann et al., 1996). We have earlier reported that CSF 5-HIAA was a short term predictor of suicide compared to suicide intent and hopelessness assessed after a suicide attempt in male mood disorder inpatients (Samuelsson et al., 2006). In this study, we found that the cut-off point of 16 was optimal and the positive predictive value was 16.7%, which is higher than the PPV in the study of Niméus et al. who reported a PPV of 9.7% in the whole sample. They also found that Suicide Intent Scale may offer a better prediction if targeted in elderly suicide attempters with PPV of 22.5% for those 55 years or older. Harris and Hawton reported a low PPV of 4.0% in a large group of deliberate self harm patients. They concluded that the SIS cannot predict which individual patients will ultimately die by suicide due to a large number of false positives. Our optimal cut-off of 16 was somewhat lower than in the study of Niméus et al. but higher than in deliberate self harm patients (Niméus et al., 2002; Harriss and Hawton, 2005). This could partly be explained by suicidal intent being lower in deliberate self harm patients.
We constructed a shorter version of the SIS with items 4, 7, 12 and 13. When we used this shorter version of the SIS in suicide prediction we found that an optimal cut-off of 6 gave specificity of 59% and sensitivity of 100%. The positive predictive value was 19% and the AUC was 0.82 which gave a better prediction by reducing the number of false positives. In the study of Niméus et al., shorter versions of the SIS or the factors were not superior to the original SIS scale in predicting suicide. Having a shorter test may increase the clinical utility of the scale. The utility and predictive value of the shorter version should be replicated in a larger cohort of suicide attempters. Limitation of this study is a small number of patients; furthermore suicidal intentions are very difficult to measure per se (Freedenthal, 2007). A patient who tried to die may deny suicidal intent to avoid hospitalization. Shame, ambivalence, confusion or intoxication can contribute to recall bias. In summary, our findings support the use of information about suicidal intent as part of a clinical suicide risk assessment. Further work needs to be done to test the utility of a shorter version of the SIS combined with other clinical rating scales measuring other types of suicide risk factors such as violence. Role of funding source Funding for this study was provided by the Swedish Research Council (Project number K2009-61P-21304-04-4) by Söderström-Königska Foundation and by the Thurings Foundation. The Swedish Research Council, Söderström-Königska Foundation and the Thurings Foundation had no further role in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication. Conflict of interest No conflicts of interests to declare for any of the co authors.
Acknowledgements We want to acknowledge Professor Marie Åsberg for inspiring us with studies in suicidology Dr Kaj Forslund for careful clinical assessments and Dr. Large who made helpful comments on the electronic version of the paper resulting in an improved final paper version. References Antretter, E., Dunkel, D., Haring, C., Corcoran, P., De Leo, D., Fekete, S., Hawton, K., Kerkhof, A.J., Lonnqvist, J., Renberg, E.S., Schmidtke, A., Van Heeringen, K., Wasserman, D., 2008. The factorial structure of the Suicide Intent Scale: a comparative study in clinical samples from 11 European regions. Int. J. Meth. Psychiatr. Res. 17, 63–79.
J. Stefansson et al. / Journal of Affective Disorders 136 (2012) 167–171 Beck, A.T., Steer, R.A., 1989. Clinical predictors of eventual suicide: a 5- to 10-year prospective study of suicide attempters. J. Affect. Disord. 17, 203–209. Beck, A.T., Schuyler, D., Herman, I., 1974a. Development of suicidal intent scales. In: Beck, A.T., Resnik, H.L.P., Lettieri, D.J. (Eds.), The Prediction of Suicide. Charles Press, Philadelphia, PA, pp. 45–56. Beck, A.T., Weissman, A., Lester, D., Trexler, L., 1974b. The measurement of pessimism: the hopelessness scale. J. Consult. Clin. Psychol. 42 (6), 861–865. Freedenthal, S., 2007. Challenges in assessing intent to die: can suicide attempters be trusted? Omega J. Death Dying 55, 57–70. Freedenthal, S., 2008. Assessing the wish to die: a 30-year review of the Suicide Intent Scale. Arch. Suicide Res. 12 (4), 277–298. Harriss, L., Hawton, K., 2005. Suicidal intent in deliberate self-harm and the risk of suicide: the predictive power of the Suicide Intent Scale. J. Affect. Disord. 86, 225–233. Harriss, L., Hawton, K., Zahl, D., 2005. Value of measuring suicidal intent in the assessment of people attending hospital following self-poisoning or self-injury. Br. J. Psychiatry 186, 60–66. Hawton, K., Harriss, L., 2006. Deliberate self-harm in people aged 60 years and over: characteristics and outcome of a 20-year cohort. Int. J. Geriatr. Psychiatry 21, 572–581. Lindqvist, D., Niméus, A., Träskman-Bendz, L., 2007. Suicidal intent and psychiatric symptoms among inpatient suicide attempters. Nord. J. Psychiatry 61, 27–32. Mann, J.J., Malone, K.M., Psych, M.R., Sweeney, J.A., Brown, R.P., Linnoila, M., Stanley, B., Stanley, M., 1996. Attempted suicide characteristics and cerebrospinal fluid amine metabolites in depressed inpatients. Neuropsychopharmacology 15 (6), 576–586. Mieczkowski, T.A., Sweeney, J.A., Haas, G.L., Junker, B.W., Brown, R.P., Mann, J.J., 1993. Factor composition of the Suicide Intent Scale. Suicide Life-Threat. Behav. 23, 37–45.
171
Misson, H., Mathieu, F., Jollant, F., Yon, L., Guillaume, S., Parmentier, C., Raust, A., Jaussent, I., Slama, F., Leboyer, M., Bellivier, F., Courtet, P., 2010. Factor analyses of the Suicidal Intent Scale (SIS) and the Risk-Rescue Rating Scale (RRRS): toward the identification of homogeneous subgroups of behaviors. J. Affect. Disord. 121 (1–2), 80–87. Montgomery, S.A., Åsberg, M., 1979. A new depression scale designed to be sensitive to change. Br. J. Psychiatry 134, 382–389. Niméus, A., Alsén, M., Träskman-Bendz, L., 2002. High suicidal intent scores indicate future suicide. Arch. Suicide Res. 6, 211–219. Nordström, P., Samuelsson, M., Åsberg, M., 1995. Survival analysis of suicide risk after attempted suicide. Acta Psychiatr. Scand. 91, 336–340. Pierce, D., 1987. Deliberate self-harm in the elderly. Int. J. Geriatr. Psychiatry 2, 105–110. Runeson, B., Tidemalm, D., Dahlin, M., Lichtenstein, P., Långström, N., 2010. Method of attempted suicide as predictor of subsequent successful suicide: national long term cohort study. BMJ 13 (341), c3222. Samuelsson, M., Jokinen, J., Nordstrom, A.L., Nordstrom, P., 2006. CSF-HIAA, suicide intent and hopelessness in the prediction of early suicide in male high-risk suicide attempters. Acta Psychiatr. Scand. 113, 44–47. Skogman, K., Alsen, M., Öjehagen, A., 2004. Sex differences in risk factors for suicide after attempted suicide—a follow-up study of 1052 suicide attempters. Soc. Psychiatry Psychiatr. Epidemiol. 39, 113–120. Suominen, K., Isometsa, E., Ostamo, A., Lönnqvist, J., 2004. Level of suicidal intent predicts overall mortality and suicide after attempted suicide: a 12-year follow-up study. BMC Psychiatry 4, 11.