Combining scales to assess suicide risk
Introduction
Suicide is a major health issue. One suicide is completed every 40 seconds, leading to approximately one million deaths every year worldwide (WHO, 2002). Moreover, suicide is the third most important cause of death worldwide among people aged 15–44 (Holmes et al., 2007). Notwithstanding human costs, the economic burden of suicidal behavior has been estimated annually in $33 billion in the United States (Coreil et al., 2001). Fortunately, suicidal behavior might be prevented to a great extent (Jamison, 2000). Treating subjects at risk with the appropriate preventive measures, such as cognitive behavior therapies (Brown et al., 2005) can reduce suicide rates up to 25% (Isaacson, 2000). More recently, a 75% reduction of suicide rates has been reported in a large depression care program (Hampton, 2010). In order to detect subjects at risk, researchers have investigated the factors underlying suicidal behavior. The most relevant risk factors are major depression (Mann et al., 1999b), high impulsiveness (Patton et al., 1995), aggressiveness (Mann et al., 1999b), personality disorders (Mann et al., 1999a), life events (Kolves et al., 2006), and social-demographic factors (Smith et al., 1988).
Unfortunately, most of these studies did not measure the effectiveness of the risk factors to identify subjects at risk. They just tested if there was a statistically significant relationship between the studied variable (e.g. high impulsiveness) and suicidal behavior. Therefore, the clinical usefulness of these studies is limited. One notable exception is the seminal Pokorny's article (Pokorny, 1983). Pokorny applied discriminant analysis to several features including, among others, socio-demographic variables, the 24 items of the brief psychiatric rating scale, and the items included in the nurses' observation scale for inpatient evaluation. Although it was an innovative approach, his results showed a weak performance, with accuracy, sensitivity and specificity levels below 70%. More recently, Hendin (Hendin et al., 2010) slightly improved these results achieving an accuracy of 71.67% with a specificity of 74% and a sensitivity of 60%. The improvement was basically due to the use of a different set of predictive variables, as they used a simple classifier consisting on the sum of the individual scores associated to each variable. This study suggested that the use of a more suitable set of predictive variables together with the use of more sophisticated classifiers might improve the classification accuracy of people at risk of suicidal behavior. This intuition was confirmed by us (Delgado-Gomez et al., 2011) in a study aimed at discriminating between suicide attempters (SA) and non-SA. In this study, we used two personality scales as predictors, and a collection of modern classification techniques such as linear discriminant analysis, Fisher linear discriminant analysis, boosting, and support vector machines (SVM). The best results were obtained with SVM, which achieved a classification accuracy of 80.3%, with a specificity of 86.8% and a sensitivity of 76.1%. Recently, Stefansson et al. (Stefansson et al., 2010) have shown that the prediction of suicide can be improved by means of an appropriate selection of the items. However, their results were obtained ad hoc. Therefore, we have applied the Lars-en algorithm in order to automatically select the most discriminative items (Delgado-Gomez et al., 2012). Using the selected items of scales measuring life events and personality disorders, it was possible to achieve a classification accuracy of 83%. A question that remains open is the accuracy that could be obtained if the Lars-en algorithm were applied to a set of items assessing the most relevant risk factors for suicidal behavior.
The present study extends our previous findings and is conceived as a further step toward the development of more precise and reliable measures of suicide risk (García-Nieto et al., 2012). The major goal is to maximize the classification accuracy applying simultaneously Lars-en to sociodemographic factors and items from four scales measuring some of the most relevant risk factors for suicidal behavior (impulsive aggression, life events, and personality disorders). As a by-product, we developed a small set of items to classify subjects as SA or non-SA. This set of items might help to develop a tool to support clinical decisions with regard to suicide risk in settings where time is scarce.
Section snippets
Participants
To accomplish our objectives, data from 687 subjects were used. Participants were 18 years or older and provided written informed consent before participating in the study. Subjects that showed incapability to provide informed consent were excluded (e.g. presence of a life-threatening medical condition, significant organic brain disease). The cases included 249 first-time SA (157 women and 92 men) admitted to two university hospitals in Madrid, Spain, between 1999 and 2003. Non-SA (n = 438)
Results
Classification results are displayed in Table 3. The items selected by the Lars-en algorithm attained an average accuracy of 85.3% in classifying SA.
For a better understanding of the predictive capacity of these scales, Fig. 1 shows the respective average receiver operating characteristic (ROC) curves together with the average area under the curves.
Once we observed that a suitable accuracy could be obtained, it was necessary to decide which items should be part of an accurate scale for
Discussion
In this article, the Lars-en algorithm analyzes the accuracy that can be obtained in the classification of individuals as SA or non-SA, by selecting the most discriminating items of well-known psychiatric scales. Our results indicate that SA can be accurately classified using a set of 27 items. This set of items showed an average accuracy of 86.4%, a specificity of 89.6%, and a sensitivity of 80.8% in classifying SA.
According to their weights, the items most closely associated to SA status came
Conclusion
The reduced number of items selected by the Lars-en algorithm in this study suggests that this set of items could be used as a quick, feasible but accurate instrument to assist clinicians in the evaluation of suicide risk. For instance, it might help primary care physicians in deciding which patients are at risk of suicide and should be referred to a psychiatrist or even hospitalized. It might also assist psychiatrists in evaluating short-term suicide risk in the emergency departments.
Contributors
Dr. Baca-Garcia designed the study and wrote the protocol. Dr. Blasco-Fontecilla and Dr. Lopez-Castroman managed the literature searches. Dr. Delgado-Gomez, Dr. Aguado and Dr. Ruiz-Hernandez undertook the statistical analysis. Dr. Blasco-Fontecilla and Dr. Delgado-Gomez wrote the first draft of the manuscript. All authors contributed to and have approved the final manuscript.
Conflict of interest
None author report any conflict of interest.
Acknowledgments
This article was supported by the National Alliance for Research on Schizophrenia and Affective Disorders (NARSAD), Fondo de Investigacion Sanitaria (FIS) PI060092, Fondo de Investigacion Sanitaria FIS RD06/0011/0016, ETES (PI07/90207), the Conchita Rabago Foundation, and the Spanish Ministry of Health, Instituto de Salud Carlos III, CIBERSAM (Intramural 521 Project, P91B; SCO/3410/2004). Dr. Blasco-Fontecilla acknowledges the Spanish Ministry of Health (Rio Hortega CM08/00170), Alicia
References (41)
- et al.
Aggression in humans correlates with cerebrospinal fluid amine metabolites
Psychiatry Research
(1979) - et al.
Improving the accuracy of suicide attempter classification
Artificial Intelligence in Medicine
(2011) - et al.
Suicide attempters classification: toward predictive models of suicidal behavior
Neurocomputing
(2012) - et al.
Brief suicide questionnaire. Inter-rater reliability
Revista de Psiquiatria y Salud Mental
(2012) - et al.
Personality and attempted suicide. Analysis of anger, aggression and impulsivity
Journal of Psychiatric Research
(2009) - et al.
Imagery about suicide in depression–“flash-forwards”?
Journal of Behavior Therapy and Experimental Psychiatry
(2007) - et al.
The social readjustment rating scale
Journal of Psychosomatic Research
(1967) - et al.
Recent life events and suicide: a case-control study in Tallinn and Frankfurt
Social Science & Medicine
(2006) - et al.
Impulsivity, aggression and suicidal behavior in unipolar and bipolar disorders
Journal of Affective Disorders
(2011) - et al.
Admixture analysis of age at first suicide attempt
Journal of Psychiatric Research
(2009)
Marital status and the risk of suicide
American Journal of Public Health
Suicide attempts and impulsivity
European Archives of Psychiatry and Clinical Neurosciences
Differences in aggression in New York and Madrid: impact on suicidal behavior
Journal of Clinical Psychiatry
Cross-validation of the Suicidal Intent Scale
Psychological Reports
An exploratory study of the relationship between diverse life events and personality disorders in a sample of suicide attempters
Journal of Personality Disorders
Suicide rates, life satisfaction and happiness as markers for population mental health
Social Psychiatry and Psychiatric Epidemiology
Cognitive therapy for the prevention of suicide attempts
Journal of the American Medical Association
The gender paradox in suicide
Suicide and Life-Threatening Behavior
Risk of repetition of suicide attempt, suicide or all deaths after an episode of attempted suicide: a register-based survival analysis
The Australian and New Zealand Journal of Psychiatry
Social and behavioral foundations of public health
Cited by (24)
Serum β-endorphin levels are associated with addiction to suicidal behavior: A pilot study
2020, European NeuropsychopharmacologyCitation Excerpt :The protocol also included the S-PLE, a 6-item instrument created to improve the identification of individuals at risk of SB (Artieda-Urrutia et al., 2015). The 27-item version of the PLE was constructed using some of the most discriminative variables extracted from different scales frequently used to evaluate SB and sociodemographic factors (age, gender) (Blasco-Fontecilla et al., 2012a). The PLE had an AUC of 0.92, an average accuracy of 86.4%, a specificity of 89.6% and a sensitivity of 80.8% in classifying suicide attempters.
A qualitative study of the chronic feeling of emptiness
2020, Annales Medico-PsychologiquesComputerized adaptive test and decision trees: A unifying approach
2019, Expert Systems with ApplicationsCitation Excerpt :For example, Gibbons, Weiss, Frank, and Kupfer (2016) argued that CATs are ideal for measuring severity but not for diagnosis screening, distinguishing between CATs and Computerized Adaptive Diagnosis (CADs) and developed a DT based CAD for detecting major depression disorder. Recently, Delgado-Gomez, Baca-Garcia, Aguado, Courtet, and Lopez-Castroman (2016) compared the performance of a DT and a CAT for identifying suicidal behaviour using the personality and life events scale (Blasco-Fontecilla et al., 2012). Their results showed that a DT required fewer items than a CAT for obtaining a similar classification rate.
Serotonin transporter gene expression predicts the worsening of suicidal ideation and suicide attempts along a long-term follow-up of a Major Depressive Episode
2018, European NeuropsychopharmacologyCitation Excerpt :While this result needs to be considered with caution due to the small number of events (5 patients attempted suicide between the 8th and 30th weeks) that could have dimly emphasize its predictive value (Button et al., 2013; Seeley et al., 1984), it is worth noting that the proportion of suicide attempters in our cohort of depressed individuals was concordant with the expected rates in a context of clinical care (Nicoli et al., 2012; Weissman et al., 1999). It is also of interest to note that Blasco-Fontecilla et al. reported an AUC value of 0.92 for the prediction of suicide attempts based on selected items from clinical scales in an heterogeneous psychiatric population (Blasco-Fontecilla et al., 2012), that is relatively close to our predictive value. Secondly, it is worth noting that generalizability of our findings should be limited due to our relative small sample size, specifically regarding the occurrence of suicide attempts.
Usefulness of the Short Personality and Life Event Scale (S-PLE) for detection of suicide attempters
2017, Revista de Psiquiatria y Salud MentalComputerized Adaptive Test vs. decision trees: Development of a support decision system to identify suicidal behavior
2016, Journal of Affective DisordersCitation Excerpt :Second, a feature selection algorithm can select directly the most appropriate items. For instance, Blasco-Fontecilla et al. built a small but performing scale to identify suicide attempters, the Personality and Life Event scale (PLE scale), using the Lars-en algorithm (Blasco-Fontecilla et al., 2012). Third, De Beurs et al. have recently proposed the use of a CAT created from Beck's Scale for Suicide Ideation to facilitate the assessment (De Beurs et al., 2014).