Predictive Validity and Cutoff Scores of the Revised Suicide Crisis Inventory in Korean Adults: A One-Year Follow-Up Study
Article information
Abstract
Objective
This study aimed to examine the predictive validity of the revised Suicide Crisis Inventory (SCI-2) and determine its optimal cutoff score.
Methods
Data from 662 community adults participating in a one-year follow-up study were analyzed. Receiver operating characteristic analysis was conducted to examine whether the SCI-2 could predict suicide attempts and ideation with intent at the follow-up and to determine the optimal cutoff score for identifying individuals at high risk for suicide.
Results
The SCI-2 demonstrated adequate predictive validity for suicide attempts and ideation with intent at the one-year follow-up. Based on Youden’s index and Runeson et al.’s criteria, a cutoff score of 102 was proposed as the threshold for high-risk groups.
Conclusion
The SCI-2, a measure of Suicide Crisis Syndrome, demonstrated predictive validity using longitudinal data. It is effective in identifying high-risk individuals in a community population. These findings highlight the SCI-2 as a valuable tool for early suicide risk detection and prevention.
INTRODUCTION
Common methods for assessing suicide risk rely on the evaluation of recent suicidal ideation and past attempt history. However, these approaches have limitations in identifying high-risk individuals who conceal their suicidal thoughts or intention for various reasons [1-3]. Moreover, the predictive value of a past history of suicidal behavior for future attempts is limited [4-6]. To address these limitations, Galynker et al. [7] proposed assessing Suicide Crisis Syndrome (SCS), which includes cognitive and emotional symptoms that emerge during an acute suicide crisis and does not rely on current suicidal thoughts or past behaviors. Galynker [8] proposed diagnostic criteria for the Diagnostic and Statistical Manual of Mental Disorders (DSM) system, including five symptom groups: frantic hopelessness/entrapment, affective disturbances, loss of cognitive control, hyperarousal, and social withdrawal. Among these, frantic hopelessness/entrapment is considered the core symptom, describing a cognitive state in which individuals feel unable to resolve their problems or endure their pain any longer [9,10].
To assess SCS, Galynker et al. [7] developed and validated the 48-item Suicide Crisis Inventory (SCI) in a study involving psychiatric inpatients in the United States. Subsequently, Bloch-Elkouby et al. [11] developed and validated the 61-item Revised Suicide Crisis Inventory (SCI-2) in a sample of inpatient and outpatient psychiatric patients. Initially, the SCI-2 was validated with psychiatric patients in the US [7,11-16]. More recently, the SCI-2 has also been validated in community samples in Eastern and Western countries, including South Korea [17], Russia [18], India [19], Taiwan [20], and Brazil [21].
Previous studies have shown that SCS predicts suicidal behavior in the near future [7,11-16,22]. SCS, as measured by the SCI, significantly predicted suicide attempt within 4 to 8 weeks following discharge [7] and within the next 1 month in psychiatric inpatients [13,14]. Barzilay et al. [12] reported that the SCI predicted suicide attempts in inpatient and outpatient psychiatric patients at one-month follow-up, while it did not predict suicidal thoughts. Similarly, the SCI-2 predicted suicide attempt and preparatory behaviors in psychiatric patients one month later, but not suicidal thoughts [11]. Taken together, SCS consistently predicted near-term suicide attempts in psychiatric patients, although its ability to predict future suicidal thoughts is unclear.
Park et al. [17] validated the SCI-2 in a Korean adult sample using cross-sectional data and confirmed the one- and five-factor structures of the SCI-2. Additionally, they identified an alternative four-factor structure that is also suitable for Korean: 1) hopelessness and overwhelming distress; 2) affective, cognitive, and physical disturbances; 3) extreme anxiety; and 4) social withdrawal. In this four-factor structure, the affective disturbances factor was integrated with the loss of cognitive control and hyperarousal factors. Interestingly, the extreme anxiety subscale was separated as an independent factor from the affective disturbances factor. Park et al. [17] reported that both five- and four-factor structures were valid and applicable to Korean adults. Yet, they did not examine predictive validity and cutoff scores to identify high-risk individuals. So far, one study examined cutoff scores in predicting prospective suicide attempts or preparatory acts among psychiatry patients at one-month follow-up [11]. However, it is unclear whether these cutoffs are applicable to community adults or individuals in other cultural contexts.
This study aimed to examine predictive validity of the SCI-2 and cutoff scores in identifying high-risk individuals in a community sample in South Korea. To address this goal, we used longitudinal data in which community adults were followed for one year. Based on previous studies, we hypothesized that the predictive validity of the SCI-2 would be good in predicting suicide attempts. We also examined the predictive value for suicidal ideation with intent of the SCI-2. To compare the predictive value of the SCI-2 with other common suicide risk indexes, we examined the predictive value of past history of suicidal ideation and attempts.
METHODS
Participants
This study used the data derived from a community-based longitudinal study in predicting suicide crisis in South Korea. Data was collected through an online survey targeting community adults aged 19 years and older from 2021 to 2023. This study utilized the 2nd and 3rd wave data collected in 2022 and 2023, in which the SCI-2 was measured. The research was approved by the Institutional Review Board (IRB) of Chungbuk National University (CBNU202010-HR-0164).
A total of 837 participants were followed up for a year. Among them, 662 (79.09%) who continued to participate at one-year follow-up were included for the analysis. Demographic characteristics of the baseline sample are presented in Table 1. Group comparisons between participants who were followed up and dropped out indicated that gender (χ2=22.05, p<0.001) and marital status (χ2=15.36, p<0.001) significantly differ between the groups, in which men and married individuals are more likely to discontinue at the follow-up. No group differences were found for age, education level, living status, employment status. No significant differences in suicidal behaviors were found between participants who were followed up and those who dropped out (Table 1).
Demographic and clinical characteristics of the study sample
At the one-year follow-up, 9.97% (n=66) endorsed suicidal ideation with intent (Columbia Suicide Severity Rating Scale [C-SSRS] ≥4) and 2.42% (n=16) endorsed suicide attempts in the past year.
Measures
The SCI-2
The SCI-2 is a 61-item self-report measure of acute suicidal crisis syndrome [11]. The Korean version of the SCI-2 [17] was used for this study. The SCI-2 is composed of five factors [11,17]: entrapment (10 items), affective disturbance (18 items), loss of cognitive control (15 items), hyperarousal (13 items), and social withdrawal (5 items). In addition, an alternative four-factor structure of the SCI-2 identified in Korea comprises the following dimensions [17]: hopelessness and overwhelming distress (9 items), affective, cognitive, and physical disturbances (32 items), extreme anxiety (5 items), and social withdrawal (5 items). The SCI-2 assesses symptoms of the SCS using a 5-point Likert scale (0: not at all to 4: very much) with the score range of 0 to 244. In this study, we assessed the SCS within the past year, and SCI-2 scores at baseline (Wave 2) were used to predict prospective suicidal behaviors occurring between Wave 2 and 3. The internal consistency coefficients (Cronbach’s α) of the SCI-2 were reported as 0.98 for the total scale, ranging from 0.86 to 0.95 for the five-factor subscales and from 0.89 to 0.93 for the four-factor subscales [17]. In this study, the Cronbach’s α values at baseline were 0.98 for the total SCI-2 scale and 0.96, 0.94, 0.88, 0.95, and 0.92 for its five original subscales. For the alternative four-factor structure of the SCI-2, the Cronbach’s α values were 0.94, 0.98, 0.86, and 0.92, respectively.
The C-SSRS screen version
The C-SSRS screen version was used to assess the severity of suicidal ideation and behavior [23]. The C-SSRS screen version consists of five items that measure the severity of suicidal ideation and one item that assesses suicide attempts or preparatory behaviors. We used the C-SSRS screen version to measure the severity of suicidal ideation over the past year and assessed the occurrence of suicide attempts within the past year by asking, “Have you attempted suicide in the past year?” In this study, participants with a C-SSRS suicidal ideation severity score of 4 or higher were classified as the suicidal ideation with intent group [24].
Statistical analysis
We conducted receiver operating characteristic (ROC) curve analyses with the area under the curve (AUC) to examine the predictive validity of the SCI-2 and determine the cutoff score. Using ROC analysis, we examined the predictive validity of the SCI-2 and its factors, as well as lifetime or past-year suicide ideation and attempts, in predicting prospective suicidal ideation with intent or attempts at one-year follow-up. According to Muller et al. [25], an AUC below 0.7 indicates poor predictive accuracy; an AUC between 0.7 and 0.8 indicates fair accuracy; an AUC between 0.8 and 0.9 indicates good accuracy; and an AUC of 0.9 or above indicates excellent predictive accuracy. If the AUC was 0.7 or higher, the cutoff score for the SCI-2 was determined by considering predictive indices, including sensitivity, specificity, positive predictive value, and negative predictive value, and the Youden index (J=sensitivity+ specificity-1) [26]. We also considered guidelines for detecting suicide risk when determining the cutoff [27]. The guidelines suggest the following priorities: 1) a score with a sensitivity exceeding 80% and a specificity exceeding 50% [28]; 2) among scores with a specificity of 50% or higher, the one with the highest sensitivity; and 3) the score with the highest sensitivity. All analyses were conducted using SPSS version 25.0 (IBM Corp.).
RESULTS
Predictive validity of the SCI-2
As presented in Table 2 and Figure 1, the predictive validity of the SCI-2 for prospective suicidal ideation with intent and attempts at one-year follow-up was good, with AUCs of 0.800 and 0.803, respectively. Among five factors, affective disturbance had the highest predictive value for suicide attempts, with an AUC of 0.809, while the other factors demonstrated fair predictive validity. All five factors of the SCI-2 predicted prospective suicidal ideation with intent at a fair level. The alternative four factors of the SCI-2 also demonstrated fair to good predictive validity for prospective suicidal ideation with intent and suicide attempts, with the “affective, cognitive, and physical disturbances” factor showing the highest predictive value. Among suicidal indices based on past suicidal behaviors, only past-year suicide attempts demonstrated fair predictive validity (AUC=0.709) for prospective suicide attempts. The other indices predicted neither ideation nor attempts.
Predicting suicidal ideation with intent and suicide attempts at one-year follow-up using the SCI-2 and past-year suicidal behaviors
ROC curves for the prediction of SI and SA at one-year follow-up using the SCI-2 and suicidal behaviors in the past year. A: ROC curves for SI at one-year follow-up. B: ROC curves for SA at one-year follow-up. SI, suicidal ideation with intent; SA, suicide attempts; ROC, receiver operating characteristic; SCI-2, revised Suicide Crisis Inventory; C-SSRS, Columbia Suicide Severity Rating Scale.
Optimal cutoff scores of the SCI-2
Using the results from ROC analysis, we examined optimal cutoff scores of the SCI-2 (61 items) for identifying individuals at high risk for suicidal ideation with intent and attempts (Supplementary Table 1). In predicting prospective suicide attempts, a score of 117 had the highest Youden index (0.558). Based on the criteria established by Runeson et al. [28], scores ranging from 80 to 120 met the criteria, exceeding a sensitivity of 0.80 and a specificity of 0.50. In predicting prospective suicidal ideation with intent, the highest Youden index (0.466) was observed for SCI-2 scores of 102, 106, and 109. Also, scores ranging from 75 to 106 met the Runeson et al.’s criteria, exceeding a sensitivity of 0.80 and a specificity of 0.50.
Among these potential cutoffs, scores between 80 and 102 had the highest sensitivity of 0.94 in predicting suicide attempts, with 102 having the highest Youden index for attempt and ideation with intent. Thus, a cutoff score of 102 on the SCI-2 was proposed as the optimal cutoff for identifying individuals at high risk of suicide.
DISCUSSION
Using one-year follow-up data, this study examined predictive validity of the SCI-2 and identified an optimal cutoff score for assessing future suicide risk. Overall, the SCI-2 proved to be an effective tool for predicting prospective suicide attempts among community-dwelling adults in Korea. A cutoff score of 102 was proposed as the threshold for identifying high-risk groups.
The SCI-2 demonstrated good predictive value in this study, with an AUC of 0.80 for both prospective suicide attempts and ideation with intent. These results were comparable to, or slightly higher than, those from previous studies, where the predictive power of the SCI for future suicide attempts ranged from fair to moderate, with AUCs of 0.73 to 0.78 [7,12]. While the predictive validity of the SCI-2 for suicidal ideation with intent was good, with an AUC of 0.80 in this study, previous research has reported inconsistent findings. This discrepancy may stem from differences in how suicidal ideation groups are classified. Most prior studies [11,12] used a cutoff score of 1 (wish to be dead) or 2 (non-specific active suicidal ideation) on the C-SSRS suicidal ideation severity scale, whereas we used a more stringent cutoff of 4 (active suicidal ideation with some intent to act) [24]. A C-SSRS score of 4 or higher indicates not only suicidal ideation, but also specific suicide plans and intent to act, which constitutes a psychiatric emergency requiring immediate intervention [23,24]. Thus, utilizing a C-SSRS score of 4 or higher, rather than a score of 1 or 2 may yield more accurate predictions when assessing suicide risk.
We also examined the predictive power of past suicidal ideation and attempts on future suicide attempts or ideation with intent. The results indicated that only past-year suicide attempts predicted future attempts, with an AUC of 0.709. Notably, the SCI-2 was more predictive of future suicide risk than a past history of suicidal ideation or attempt. These findings, coupled with the limited predictive power of past suicide-related variables for future suicidal behavior [5,6], provide empirical support for the use of the SCI-2.
In the present study, all five subscales of the SCI-2 significantly predicted future suicide attempts. Among them, affective disturbances demonstrated the highest predictive power. This contrasts with findings from a study conducted in the US [11], where loss of cognitive control was identified as the strongest predictor. This discrepancy may be attributed to cultural differences between Korea and Western countries. In the alternative four-factor structure of the SCI-2 validated among Koreans, affective disturbances were combined with physical symptoms and loss of cognitive control. Notably, when examining the predictive power of these alternative subscales, the “affective, cognitive, and physical disturbances” factor emerged as the strongest predictor of future suicide attempts. This finding likely reflects cultural and linguistic characteristics of Koreans, in which emotional distress is often somatized, and cognitive and emotional experiences tend to be more integrated and less distinctly separated [29-32]. In clinical settings, individuals in Korea may also be more likely to report emotional rather than cognitive disturbances. This cultural tendency may explain the prominence of the affective disturbances subscale in predicting prospective suicide attempts in the original five-factor model.
Despite its length of the scale, the SCI-2 offers important advantages that support its clinical utility. First, the SCI-2 does not directly inquire about suicidal ideation or attempts, making it particularly useful for identifying individuals who may intentionally deny or conceal their suicidal intent due to concerns about stigma or social judgment. This indirect approach may be especially beneficial in cultural contexts such as South Korea, where suicide-related stigma remains a significant barrier to disclosure and often hinders help-seeking behavior. Second, while knowledge of past suicidal ideation or attempts provides essential context for assessing suicide risk, it offers limited guidance regarding specific treatment targets. In contrast, the SCI-2, unlike tools such as the C-SSRS, assesses modifiable symptoms, including affective, cognitive, physical, and social disturbances, thereby offering clinicians actionable information for intervention. By identifying underlying symptoms associated with suicidal risk, the SCI-2 can support individualized treatment planning and enhance suicide prevention efforts.
In determining the optimal cutoff point for the SCI-2, we considered the suicide-specific criteria suggested by Runeson et al. [28] along with Youden Index, a common method for identifying a scale’s cutoff point that emphasizes a balance between sensitivity and specificity. Runeson et al. [28] reviewed 17 suicide risk assessment tools and found that only two had a sensitivity greater than 80% and a specificity greater than 50%. Considering that suicide is an irreversible outcome, screening tools for suicide risk should prioritize sensitivity, with efforts focused on reducing Type II error (false negatives).
In this study, we first identified potential cutoff points using both the Youden Index and Runeson’s guidelines. Among those, a score of 102 had the highest sensitivity in predicting prospective suicide attempts, and was also the score where the Youden Index was highest for predicting prospective suicidal ideation with intent. For ease of use in clinical settings, we recommend a single cutoff point of 102 for identifying high risk groups for suicide. However, depending on the clinical context, less stringent criteria may be used to implement interventions for individuals at risk of suicidal ideation.
This study has several limitations. First, in this longitudinal study, we assessed SCS over the past year, rather than the past few days. Given the yearly nature of the study, this approach was used to capture suicidal crisis symptoms, which may fluctuate throughout the year. Thus, we cannot rule out the possibility of memory bias resulting from this retrospective self-reporting method. Second, the SCI-2 scale consists of 61 questions and requires significant time to administer in practice. A reliable and valid scale holds little value if it is not feasible for use in real-world settings. Therefore, further research is needed to develop and validate a shortened version to enhance its accessibility and practical utility.
The most significant aspect of this study is its use of longitudinal data to examine predictive validity. Importantly, the SCI-2 demonstrated the ability to predict future suicidal ideation with intent and attempts in a community population, which generally has a lower suicide risk than psychiatric patients. Given that more than half of individuals who die by suicide do not have a psychiatric diagnosis at the time of their death [3], and that the lifetime utilization rate of mental health services in South Korea is only 4.5% [33], identifying high-risk individuals within the community is crucial for suicide prevention. Therefore, we expect that the SCI-2, which measures suicidal crisis syndrome, will aid in the early detection and prevention of suicide in at-risk individuals.
Supplementary Materials
The Supplement is available with this article at https://doi.org/10.30773/pi.2025.0109.
Pooled estimates of predictive indices of SCI-2 cutoff scores
Notes
Availability of Data and Material
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
Conflicts of Interest
The authors have no potential conflicts of interest to disclose.
Author Contributions
Conceptualization: Gangmin Ma, Sungeun You. Data curation: Sungeun You. Formal analysis: Gangmin Ma. Funding acquisition: Sungeun You. Investigation: Sungeun You. Methodology: Gangmin Ma, Sungeun You. Project administration: Sungeun You. Supervision: Sungeun You. Writing—original draft: Gangmin Ma. Writing—review & editing: Gangmin Ma, Sungeun You.
Funding Statement
This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2020S1A5A2A03044181).
Acknowledgments
None