Suicide is a major public health issue, and reduction or prevention depends on accurate identification of at-risk patients. Use of a precise evaluation method of lifetime suicidal behaviour and ideation may provide more accurate determination of risk in clinical and research settings.1 However, assessment of suicide risk is a complex task. One of the most common problems that arises in risk assessment is the failure to adequately document clinical judgments and observations.2 Furthermore, use of psychometric scales to assess suicidal behaviour has yet to be incorporated into everyday clinical practice.3 Indeed, a survey of 400 psychiatrists showed that only 10% habitually use structured scales and questionnaires to assess suicide risk.4

A comprehensive mental status examination targeting not only the present condition, but also including questions about past suicidal behaviour is valuable from a preventive point of view, especially since a history of suicidal behaviour is the most reliably replicated risk factor for future suicide attempt or completion.5

To address both the inconsistencies in nomenclature and their impact on accurate identification of suicide risk, as well as the need for a measure to assess the severity of suicidal ideation and behaviour and to track changes in them, a team of investigators developed the Columbia-Suicide Severity Rating Scale (C-SSRS).6 The scale was designed to6: “(i) provide definitions of suicidal ideation and behaviour and nonsuicidal self-injurious behaviour and corresponding probes; (ii) quantify the full spectrum of suicidal ideation and suicidal behaviour and gauge their severity over specified periods; (iii) distinguish suicidal behaviour from nonsuicidal self-injurious behaviour; and (iv) employ a user-friendly format that allows integration of information from multiple sources”. These criteria are considered essential for judging the utility of scales assessing suicide, and the C-SSRS meets all of these criteria.6,7

Additionally, the C-SSRS has been used to compare the full range of suicidal ideation and suicidal behaviour across clinical trial visits. It also has criteria for mental health referral that have shown adequate predictive power.6 The C-SSRS is the prospective counterpart to the FDA-commissioned Columbia Classification Algorithm for Suicide Assessment (C-CASA)8 adverse event system. C-CASA suicide attempt classification is required in central nervous system and obesity trials.6 The neuropharmacology division of the FDA has specifically requested that any instrument used prospectively to monitor trial visits should link to C-CASA suicide attempt definitions. The C-SSRS is the only scale currently deemed to do so7 and thus is approved by the FDA for prospective monitoring.9

The C-SSRS has not been validated in Spanish, the third most commonly spoken language in the world. The aim of this study was to translate and validate a Spanish version of the C-SSRS for broad use, and to examine its psychometric properties (reliability, construct and discriminant validity, and sensitivity to change) in a sample of psychiatric outpatients.

Suicidal behaviour is an important public health concern, and the development of valid methods for identifying patients at greater risk for engaging in suicidal behaviour has important implications for research and suicide prevention.1 However, the role of empathy is crucial when evaluating and treating suicidal individuals.18

We found adequate Sp-C-SSRS convergent and divergent validity, and sensitivity to change, as well as internal consistency of the severity and intensity of ideation subscales, which support the Sp-C-SSRS as a reliable and valid instrument for assessing suicidal ideation and behaviour in daily clinical practice and research settings.

There are many scales and measures that quantitatively and qualitatively measure the presence of suicidal ideation, intent, and motivation as well as the intensity, duration, frequency, and consequences of suicide attempts.19 Even when using the same measures across studies and across populations–as in this study–researchers often assume that all respondents are basing their responses on the same conceptualizations and definitions of suicide-related terms. The refinement of investigative tools and techniques can only improve their efficiency, effectiveness, sensitivity and specificity by clearly describing the studied populations and behaviours.19 In addition, a standardized nomenclature will lead to a standardized set of questions for determining the presence or absence of suicidal cognitions, motivations, emotions, and behaviours. Indeed, Silverman et al.20 provided a nomenclature built with an eye toward application in clinical research.

Posner et al.6 found good internal consistency for the C-SSRS intensity of ideation subscale with Cronbach's alpha values ranging between 0.73 and 0.93. However, in our sample the Cronbach coefficient value for the Sp-C-SSRS intensity of ideation subscale was 0.53, above what would be considered unacceptable. This rating reflects low internal reliability, showing that perhaps intensity of suicidal ideation is the most complex or heterogeneous aspect of suicidal behaviour. In this sense, it should be noted that three of the intensity subscale items assess “quantitative” aspects of the intensity of ideation such as frequency, duration, and controllability. Meanwhile, the other two items are more related to “qualitative” issues such as deterrents and reasons for ideation.

Regarding to this, the principal component analysis performed in the total sample confirms the existence of two components. The inconsistencies found in the internal reliability may be due to this new two-component solution. To the best of our knowledge, no previous studies have been done about this issue. We were able to obtain a one-component solution, even though the two-component showed a better fit, as shown in “Results” section. When examining the two-component solution, we observed that the first component, consisted of frequency, duration, and controllability, is more related to the characteristics of suicidal ideation, while the second component was represented by cognitive aspects (container vs content). These results suggest the need for further research on the structure of the C-CCSR intensity subscale.

We found lower internal validity for the intensity of ideation subscale. Because the samples in the original study6 and in this one are not similar or comparable, it was expected that the results would differ somewhat. The samples differed in at least 3 aspects: (i) Age: the original study6 was conducted with adolescents, while the mean age in our sample is nearly 47 years. It is known that scores tend to be less dispersed in adolescent samples than in others.21 Furthermore, educational level and personality characteristics of adolescents vs adults may also have led to differences. (ii) Cultural issues: the constructs of individualism and collectivism have been extensively researched.22 Empirical evidence supports the notion that Americans differ in terms of individualism and collectivism from those from Mediterranean countries (Spain). The US occupies the individualistic end of the spectrum while Spain would be in an intermediate position. However, prior data suggest a possible role of individualism in suicidality in Western societies.23 (iii) The same cultural influences may have been at work with respect to the clinicians conducting the evaluation, which may also have resulted in lower internal validity of the Sp-C-SSRS.

In fact, severity and intensity of ideation scores do not show correlation coefficients as high as expected (from 0.44 to 0.19). Nevertheless, in the initial study,6 C-SRRS severity and intensity of ideation subscales were moderately correlated (from 0.56 to 0.52). Pearson correlations for the Sp-C-SSRS scores were especially low in patients without any previous SA (0.19). Perhaps this reflects the complex relationship between suicidal ideation severity and intensity, which may generate different kinds of patient profiles (i.e. patients with a high score on the intensity of ideation subscale but no specific plan or intent to act, or patients with severe active suicidal ideation but with low frequency, duration, or controllability of those ideas).

In terms of construct validity, we found a moderate relationship between Sp-C-SSRS severity of ideation scores and the suicide-related item 3 of the HDRS, showing good convergent validity for the total sample. However, this correlation was lower for patients with SA (0.20). This finding is likely due to the fact that the C-SSRS severity of ideation subscale provides a retrospective lifetime assessment of the severity of suicidal behaviour, whereas the suicide-related item 3 of the HDRS rates only the past 2 weeks. Supporting this hypothesis, in our study, the best correlation was found in the subsample of patients without SA.

The Sp-C-SSRS severity of ideation subscale moderately but significantly correlated with the SIS (this analysis was conducted only for patients with SA). Posner et al.6 also found only moderate convergent validity between these two measures. Thus, this finding was expected and was likely due to non-overlap in items and to different operationalization of construct.6 Namely, the C-SSRS severity of ideation subscale was designed to reflect five different types of ideation of increasing severity whereas the SIS was designed to measure the seriousness of the intent to commit suicide among patients who have actually attempted suicide.

On the other hand, divergent validity with measures of other constructs is essential.6 We found that the Sp-C-SSRS severity of ideation subscale discriminates among the five levels of the HDRS suicide-related item 3 scores in the expected direction. Patients scoring higher on HDRS item 3 (“suicide ideas or gesture” and “attempts of suicide”) obtained significantly higher scores on the Sp-C-SSRS severity of ideation subscale, while patients with low scores on HDRS item 3 (“absent”, “life not worth living”, and “wish to be dead”) obtained significantly lower scores on this Sp-C-SSRS subscale than the other two groups. Of note, the Duncan post hoc test established a significant difference precisely between ideation (from “absent” to “wish to be dead”) and behaviour (“suicide gesture” and “attempts of suicide”) only. Posner et al.6 also demonstrated strong divergent validity of the C-SSRS with some items of scales for depression assessment (Beck Depression Inventory – BDI, and Montgomery-Asberg Depression Rating Scale – MADRS).

Likewise, the Sp-C-SSRS suicidal behaviour lethality subscale demonstrated good discriminant validity with the cut-off point of the MDS, similar to findings by Posner et al.,6 who found a robust correlation between these two measures.

Regarding sensitivity to change, Posner et al.6 found that a decrease in the severity of ideation (from “active suicidal thoughts” to “wish to die” or “no ideation”) was accompanied by a corresponding decrease in the Scale for Suicide Ideation (SSI) score, suggesting that the C-SSRS severity of ideation subscale is sensitive to clinical change. Our study found that decreases in Sp-C-SSRS severity and intensity of ideation subscale scores were associated with a decrease in HDRS suicide-related item 3. However, statistical significance was reached only with the intensity of ideation subscale. These data suggest that both Sp-C-SSRS suicide ideation subscales (i.e. severity and intensity subscales) are sensitive to clinical change. The lack of concordance between linear regression models and interaction plots may possibly be due to lack of statistical power associated with the small sample size used in the 3-month follow-up study.

Some limitations of this analysis should be noted. First of all, the study design precluded examination of other psychometric properties: (i) no independent evaluators assessed suicidal behaviour, so the sensitivity to change of the Sp-C-SSRS suicidal behaviour subscale could not be analyzed and (ii) interrater reliability was not evaluated (the data about the clinicians’ identity were not available), although previous data have demonstrated good interrater reliability of the original C-SSRS scale.24 Secondly, only patients with a current diagnosis of mood disorders were included as the non-suicidal control group, and only mid-term follow-up was conducted. For this 3-month follow-up, only Sp-C-SSRS and HDRS were used as assessment measures, limiting our ability to conduct a more robust validation study. Also, the small sample size used in the follow-up study may limit statistical power to detect concordance between regression analyses and interaction plots. Likewise, the comparison groups (patients with and without SA) are not completely similar in some socio-demographic and clinical features, and only mood disorders are included in the non-suicide attempter group. However, the demographic differences between suicide attempters and non-attempters reflect typical characteristics of populations displaying suicidal behaviour. Thirdly, the HDRS suicide item (item 3) may not be the best comparator for the suicidal ideation question, as the top anchor point is occurrence of a suicide attempt. The Montgomery-Asberg Depression Rating Scale (MADRS) or other single question suicide risk scales would have more data points to assess suicidal ideation. Fourthly, intent is not truly assessed by the C-SSRS scale, so using the SIS to determine construct validity for the SA sub-sample of patients may not be the best choice. Nonetheless, the medical lethality of a suicide attempt is commonly, but not always, associated with severity of ideation. Additionally, we used validated Spanish versions of all the measures, except the Medical Damage Scale, which has not been validated in Spanish.

In conclusion, our data indicate that the Sp-C-SSRS is a suitable instrument for assessing and monitoring suicidal ideation and behaviour in psychiatric patients, both in clinical and research settings. The impact of our findings is twofold if we take into account the FDA requirement for all central nervous system clinical drug trials to include a Columbia Classification Algorithm of Suicide Assessment (C-CASA)-compatible screening instrument for assessing and documenting the occurrence of treatment-emergent suicidal ideation and behaviour.7

This work was partly supported by the Spanish Ministry of Economy and Competitivity (grants PI10/01632, PI080247, and PI1200906). P Buron was funded by the Fundación para el Fomento en Asturias de la Investigación Científica Aplicada y la Tecnología, FICYT, through a Severo Ochoa predoctoral grant. E Jimenez is funded by the Institut d’investigacions Biomèdiques August Pi i Sunyer, IDIBAPS, through a predoctoral grant.

Dr Benabarre has received research grants and served as a speaker for the following companies: Bristol-Myers Squibb, Eli Lilly, Glaxo-Smith-Kline, Janssen-Cilag, Pfizer, and the Spanish Ministry of Economy and Competitivity.

Dr Oquendo receives royalties for use of the Columbia Suicide Severity Rating Scale and received financial compensation from Pfizer for the safety evaluation of a clinical facility, unrelated to this study. She has received unrestricted educational grants and/or lecture fees from Astra-Zeneca, Bristol-Myers Squibb, Eli Lilly, Janssen, Otsuka, Pfizer, Sanofi-Aventis, and Shire. Her family owns stock in Bristol-Myers Squibb.

Dr Posner has received salary support from the Research Foundation for Mental Hygiene (RFMH). She has served as director of the Center for Suicide Risk Assessment.

The rest of the authors declare no conflict of interest.