Discriminative Accuracy of Affective Temperaments in Assessing Suicide Risk

Kyeong A Kang; Suyeon Lee; Seong Yoon Kim; C. Hyung Keun Park

doi:10.30773/pi.2025.0036

Psychiatry Investig > Volume 22(7); 2025 > Article

Kang, Lee, Kim, and Park: Discriminative Accuracy of Affective Temperaments in Assessing Suicide Risk

Original Article

Psychiatry Investigation 2025;22(7):796-805.

Published online: July 16, 2025

DOI: https://doi.org/10.30773/pi.2025.0036

Discriminative Accuracy of Affective Temperaments in Assessing Suicide Risk

Kyeong A Kang

, Suyeon Lee^*

, Seong Yoon Kim

, C. Hyung Keun Park

Department of Psychiatry, Asan Medical Center, Seoul, Republic of Korea

Correspondence: C. Hyung Keun Park, MD, PhD Department of Psychiatry, Asan Medical Center, 88 Olympic-ro 43-gil, Songpa-gu, Seoul 05505, Republic of Korea Tel: +82-2-3010-1657, Fax: +82-2-485-8381, E-mail: hkpark@gmail.com

^* Current affiliation: Department of Psychiatry, University of Ulsan College of Medicine, Seoul, Republic of Korea

Received January 31, 2025 Revised April 1, 2025 Accepted May 8, 2025

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Objective

This study aimed to evaluate whether individual and composite scores of affective temperaments can differentiate high-risk suicide groups among patients with major depressive disorder (MDD) and to compare their discriminative accuracy.

Methods

A retrospective review included 343 patients with MDD. Affective temperaments were assessed using the Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire, while the Depressive Symptom Inventory-Suicidality Subscale was used to classify the high-risk suicide group. Receiver operating characteristic (ROC) curve analysis assessed the discriminative performance of five individual and three composite scores of affective temperaments, with area under the curve (AUC). Logistic regression analysis was conducted to examine whether depressive temperament and composite scores independently predicted suicide risk.

Results

Depressive temperament demonstrated the highest discriminative accuracy among individual scores (AUC=0.754). All composite scores also showed good accuracy (AUC range=0.729-0.739). The results indicated that depressive temperament and three composite scores demonstrated good accuracy, with no significant differences between them. Additionally, irritable (AUC=0.660) and cyclothymic (AUC=0.650) temperaments exhibited sufficient accuracy. Logistic regression analysis indicated that only depressive temperament was a significant independent predictor of suicide risk.

Conclusion

Depressive temperament and composite scores demonstrated good discriminative accuracy for identifying high-risk suicide groups in MDD. These findings highlight the importance of assessing affective temperaments as stable traits for evaluating and predicting suicide risk. Future longitudinal, multi-center studies are needed to further explore the role of affective temperaments in predicting suicide risk among patients with MDD.

Keywords: Temperament, Suicide, Depressive disorder

INTRODUCTION

Suicide represents a significant global health concern, with approximately 800,000 deaths annually attributed to suicide, equating to one death every 40 seconds [1]. Meta-analytic findings estimate that 90% of individuals who die by suicide have a history of being diagnosed with a psychiatric disorder [2], with major depressive disorder (MDD) accounting for approximately one-half to two-thirds of these cases [3,4]. MDD is a highly prevalent psychiatric condition, with an estimated lifetime prevalence of 16% [5]. It is characterized by a depressed mood or loss of interest, accompanied by symptoms such as changes in sleep and appetite, excessive guilt, and suicidal ideation, ultimately leading to significant distress or functional impairment [6]. Given the high prevalence of MDD and its robust association with suicide, identifying individuals with MDD at elevated risk for suicide represents a significant public health challenge with critical clinical implications [3,7].

Temperament refers to an innate trait marked by stability over time and a strong genetic predisposition [8,9]. It influences early-life emotional, perceptual, and automatic behavioral responses, shaping how individuals perceive and interpret their environment [10]. Among the various dimensions of temperament, affective temperament refers to subsyndromal manifestations of mood disorders observed in patients with mood disorders and their relatives, a concept [11,12] first introduced by Kraepelin. Building on this foundation, Akiskal et al. [13] proposed five subtypes of affective temperament—cyclothymic, depressive, irritable, hyperthymic, and anxious—based on clinical observations. They suggested that distinct manifestations of affective temperament could be considered high-risk states for mood disorders [14]. Studies have shown that affective temperament exhibits stability over various follow-up periods ranging from several months to 6 years, regardless of subtype, sex, age, or symptom improvement [15,16].

There is growing interest in affective temperament as a potential factor contributing to suicide risk [13,17-19]. According to a meta-analysis, affective temperaments showed a strong positive correlation with suicide-related condition (e.g., suicide attempts, suicidal ideation, hopelessness, and death by suicide), in the following order of severity: depressive temperament, irritable temperament, cyclothymic temperament, and anxious temperament [19]. Conversely, hyperthymic temperament ratings may have a weaker association with suicidality compared to other temperament types and may potentially offer a protective effect against suicide [19]. Furthermore, combinations of affective temperament scores have also been associated with suicide. A study conducted on psychiatric inpatients across various diagnostic groups identified a significant association between the cyclothymic+depressive+irritable+anxious-hyperthymic temperament score and a lifetime history of suicide attempts or hospitalizations due to suicide risk [20]. Similarly, a study conducted on patients with mood and anxiety disorders found that the cyclothymic+depressive+irritable-hyperthymic, cyclothymic+depressive+irritable+anxious-hyperthymic, and cyclothymic+depressive+irritable temperament scores significantly distinguished individuals with suicidal behaviors from those without, outperforming individual temperament scores [17].

To our knowledge, no study has exclusively examined whether individual and composite affective temperament scores can differentiate high-risk suicide groups specifically among patients with MDD. Furthermore, the previous study [17] did not utilize standardized scales to assess suicide risk, relying solely on the presence or absence of symptoms, which limited the ability to quantify the severity of suicide risk. Therefore, the purpose of this study was to evaluate whether individual and composite scores of affective temperaments can differentiate high-risk suicide groups among patients with MDD and compare their predictive accuracy.

METHODS

Study participants

This study was conducted at the Mood Disorder and Suicidal Prevention Clinic, affiliated with the Department of Psychiatry at Asan Medical Center, a tertiary hospital located in Seoul, Republic of Korea. The clinic specializes in treating outpatients aged 18 years or older presenting primarily with mood symptoms or suicidal behavior. As part of routine clinical practice, first-visit patients aged 49 years or younger are required to complete self-report questionnaires assessing their clinical state and trait characteristics. Patients were excluded if they refused to complete the questionnaires, had difficulty understanding Korean, or were unable to complete the questionnaires due to other medical conditions. For this study, we reviewed medical records of first-visit patients diagnosed with MDD according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria who completed the questionnaires between February 2022 and September 2024. This study was approved by the Institutional Review Board of Asan Medical Center (2024-1012). The requirement for written informed consent was waived for this retrospective medical record review.

Measurements

Short version of Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire

The Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire (TEMPS-A) is a self-report questionnaire designed to assess five affective temperaments [21,22]. The short version consists of 39 items, each answered with a “yes” (1 point) or “no” (0 point) [13,23]. The subscales include 12 items for cyclothymic temperament, eight items each for depressive, irritable, and hyperthymic temperaments, and three items for anxious temperament. This study examined five individual scores (cyclothymic, depressive, irritable, anxious, and hyperthymic) and three composite scores. Composite scores were defined based on a prior study [17] as follows: composite score 1=cyclothymic+depressive+irritable-hyperthymic, composite score 2=cyclothymic+depressive+irritable+anxious-hyperthymic, and composite score 3=cyclothymic+depressive +irritable.

Depressive Symptom Inventory-Suicidality Subscale

The Depressive Symptom Inventory-Suicidality Subscale (DSI-SS) is a self-report questionnaire derived from the Hopelessness Depression Symptom Questionnaire and evaluates the severity of suicidal ideation [24]. It consists of four items assessing the frequency, intensity, controllability of suicidal ideation, and suicide-related impulses over the past 2 weeks. Each item is rated on a scale from 0 to 3. The Korean version of the DSI-SS suggests a cutoff score of 4 for identifying individuals at high risk of suicide [25].

Patient Health Questionnaire-9

The Patient Health Questionnaire-9 (PHQ-9) is a self-report questionnaire designed to assess depressive symptoms and consists of nine items [26,27]. Each item is rated on a scale from 0 to 3, with a total score ranging from 0 to 27. Higher total scores indicate greater severity of depressive symptoms [27]. Based on total scores, depressive symptom severity is categorized as follows: 0-4, minimal; 5-9, mild; 10-14, moderate; 15-19, moderately severe; and 20-27, severe [27].

Statistical analysis

Receiver operating characteristic (ROC) curve analysis was conducted to evaluate the ability of five individuals and three composite scores of affective temperaments to identify individuals at high risk of suicide. The analysis provided cutoff scores, the area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value. The AUC accuracy was classified as follows: 0.5≤AUC<0.6 (bad), 0.6≤AUC<0.7 (sufficient), 0.7≤AUC<0.8 (good), 0.8≤AUC<0.9 (very good), 0.9≤AUC (excellent) [28]. Comparisons of AUCs between individual and composite scores were performed using the DeLong’s method, and Bonferroni’s correction was applied to control for Type I error. To account for depressive symptom severity as a potential confounding factor, we conducted additional ROC curve analyses stratified by PHQ-9 scores. Participants were categorized into two groups: minimal to moderate depression (PHQ-9 score: 0-14) and moderately severe to severe depression (PHQ-9 score: ≥15). ROC curve analyses were performed separately for each group to evaluate whether the predictive performance of individual affective temperament scores and composite scores varied according to depressive symptom severity. In addition to the ROC curve analyses, we conducted binomial logistic regression analysis to examine whether depressive temperament and composite scores independently predicted suicide risk. The dependent variable was suicide risk status, while the independent variables included depressive temperament and composite scores 1, 2, and 3. Age and sex were included as covariates. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for each predictor. ROC curve analysis was conducted using MedCalc statistical software version 23.0.2 (MedCalc Inc.), and binomial logistic regression analysis was performed using SPSS for Windows version 23.0 (IBM Corp.). A significance level of p<0.05 was considered statistically significant.

RESULTS

Sociodemographic and clinical characteristics

A total of 343 patients with MDD were included in this study. The mean age of the patients was 29.37±9.10 years, and 214 (62.4%) were female. The mean scores for individual affective temperaments assessed using the TEMPS-A were as follows: cyclothymic temperament, 6.11±3.45; depressive temperament, 4.00±2.47; irritable temperament, 1.84±1.92; hyperthymic temperament, 2.17±1.94; and anxious temperament, 1.45±1.10. The mean total score on the DSI-SS was 4.90±3.49, with 228 patients (66.5%) classified as high risk for suicide (DSI-SS score ≥4). The mean total score of the PHQ-9 was 16.58±6.81, with 35.9% of participants classified as having minimal to moderate levels of depressive symptoms and 64.1% as having moderately severe to severe levels of depressive symptoms (Table 1).

Suicide risk classification based on TEMPS-A individual and composite scores

The ROC curve analysis results for individual and composite scores for affective temperaments, based on high suicide risk classification (DSI-SS), are presented in Tables 2 and 3. Among the individual scores, cyclothymic, depressive, and irritable temperament scores, as well as composite scores 1, 2, and 3, significantly identified individuals at high risk for suicide (Figure 1). Of the individual scores, depressive temperament showed the highest accuracy, with an AUC value of 0.754 (good accuracy), comparable to the three composite scores (Figure 2). The cutoff score for depressive temperament was 3, with a sensitivity of 83.33% and a specificity of 57.39%. Depressive temperament demonstrated significantly higher accuracy compared to irritable (p=0.003) and cyclothymic temperament scores (p=0.001).

Composite score 1 had a cutoff of 7, with a sensitivity of 80.70% and specificity of 55.65%. Its AUC value was 0.734 (good accuracy). Composite score 1 demonstrated significantly higher accuracy compared to the cyclothymic temperament score (p<0.001) and comparable accuracy to the depressive and irritable temperament scores, as well as composite scores 2 and 3.

Composite score 2 had a cutoff of 11, with a sensitivity of 67.11% and a specificity of 71.30%. The AUC value was 0.729 (good accuracy). Composite score 2 showed significantly higher accuracy compared to the cyclothymic temperament score (p<0.001) and comparable accuracy to the depressive and irritable temperament scores, as well as composite scores 1 and 3.

Composite score 3 had a cutoff of 8, with a sensitivity of 87.72% and a specificity of 48.70%. Its AUC value was 0.739 (good accuracy). Composite score 3 demonstrated significantly higher accuracy compared to the irritable temperament (p=0.002) and cyclothymic temperament (p<0.001) and comparable accuracy to the depressive temperament and composite scores 1 and 2.

ROC curve analysis stratified by depressive symptom severity

Minimal to moderate depression group

In the minimal to moderate depression group, similar to the findings observed in the total sample, cyclothymic, depressive, and irritable temperament scores, as well as all 3 composite scores significantly identified individuals at high risk for suicide (Figure 3 and Table 4). Among these, depressive temperament, irritable temperament, and composite scores 1, 2, and 3 demonstrated good accuracy, whereas cyclothymic temperament showed sufficient accuracy. No significant differences in accuracy were found among these predictors, except for a significant difference between composite score 3 and cyclothymic temperament (Table 5).

Moderately severe to severe depression group

In the moderately severe to severe depression group, among the individual temperament scores, only depressive temperament significantly identified individuals at high risk for suicide. However, all 3 composite scores also significantly identified high-risk individuals (Figure 4 and Table 6). The predictive accuracy of depressive temperament and the composite scores was comparable across this subgroup (Table 7). Overall, all significant predictors in this group demonstrated sufficient accuracy.

Logistic regression analysis of suicide risk prediction

To further explore the relative contributions of depressive temperament and composite scores to suicide risk, we performed a binomial logistic regression analysis (Table 8). The results indicated that only depressive temperament was a significant predictor of high-risk suicide group (OR=1.355, 95% CI [1.138-1.613], p=0.001). In contrast, composite score 1 (OR=1.080, 95% CI [0.820-1.423], p=0.582), composite score 2 (OR=0.894, 95% CI [0.699-1.143], p=0.372), and composite score 3 (OR=1.103, 95% CI [0.961-1.267], p=0.163) were not significant predictors.

DISCUSSION

This study used ROC curve analysis to evaluate whether individual and composite scores of affective temperaments could identify high-risk suicide groups among patients with MDD. The results indicated that depressive temperament and three composite scores demonstrated good accuracy, with no significant differences among them. Additionally, irritable and cyclothymic temperaments exhibited sufficient accuracy, while hyperthymic and anxious temperaments did not significantly differentiate high-risk groups. Stratified ROC curve analysis revealed that, in both the minimal to moderate and moderately severe to severe depression groups, depressive temperament and composite scores 1, 2, and 3 significantly identified individuals at high risk for suicide. Furthermore, logistic regression analysis indicated that only depressive temperament was a significant independent predictor of suicide risk, whereas the composite scores did not reach statistical significance.

A previous study involving patients with MDD, bipolar disorder, and anxiety disorders identified composite score 1 as the most effective for distinguishing individuals with suicidal ideation or behavior from those without suicidal ideation or behavior [17]. All three composite scores outperformed individual scores in discriminative performance [17]. Consistent with these findings, our study demonstrated that composite scores significantly distinguished high-risk suicide groups, while hyperthymic and anxious temperaments showed poor discriminative performance. Another study demonstrated the association between composite scores and suicide risk, findings that are similar to the results of our study [20]. In that study, conducted with psychiatric inpatients, composite score 2 was linked to lifetime suicidal attempts or hospitalizations due to suicide risk, reinforcing the relevance of composite scores for identifying high-risk individuals.

Our findings differ from prior research in the following ways. First, depressive temperament demonstrated discriminative performance comparable to composite scores for identifying high-risk suicide groups. This contrasts with a previous study [17], where composite scores were reported to outperform individual scores. There has been only one prior study [17] that investigated whether individual and composite scores of affective temperaments can distinguish high-risk suicide groups and compared the discriminative performance of each score. This has posed challenges for direct comparisons with our study. Additionally, the study included patients with various psychiatric conditions, which may have influenced results due to disorder-specific variations in affective temperaments and their impact on suicide risk. Depressive temperament may exert a greater influence on suicide risk in MDD than in other psychiatric conditions. Previous studies have reported significant positive correlations between depressive temperament and the severity of depressive symptoms [29], a known predictor of suicidal ideation [3]. Severe depressive symptoms impair executive functioning [30], leading to negative self-evaluations and impaired decision-making processes that elevate the risk of suicide [31]. Furthermore, hopelessness, which is a strong predictor of suicidal ideation [32], may be influenced by depressive temperament, as its pessimistic traits [33] might increase vulnerability to hopelessness. These findings suggest that depressive temperament plays a critical role in suicide risk among patients with MDD, exhibiting discriminative performance comparable to composite scores.

Second, this study found no statistically significant differences among the three composite scores in discriminative performance. However, the prior study [17] assessed the presence of suicidal ideation and behaviors through clinical evaluations rather than standardized measures, which may limit the validity of its findings. Furthermore, it assessed AUC values without statistical comparisons, raising the possibility that no significant differences existed among composite scores.

Third, our study predominantly included younger patients (mean age: 29.37±9.10 years), whereas the previous study analyzed an older population (44.0±13.7 years). This difference may be crucial in understanding variations in the predictive accuracy of affective temperaments. While affective temperaments have been shown to have a genetic basis, as demonstrated by studies identifying significant heritability estimates and genetic markers associated with specific affective temperaments [34,35], the influence of environmental factors on psychiatric symptoms is known to increase with age by middle adulthood [36]. Given this, it is plausible that in younger individuals, depressive temperament exerts a more direct and substantial impact on psychopathology, whereas in older populations, cumulative environmental factors may gradually surpass genetic and temperament-based predispositions [36]. This may explain why depressive temperament alone exhibited high discriminative accuracy in our younger sample. Further research is needed to investigate how age-related changes influence the impact of affective temperaments on psychopathology, particularly in relation to suicide risk.

To further explore the potential influence of depressive symptom severity, we conducted ROC curve analyses stratified by PHQ-9 scores. The results demonstrated that depressive temperament and all composite scores maintained high predictive accuracy for suicide risk across different levels of depressive symptom severity. Even in the stratified analyses, depressive temperament exhibited discriminative performance comparable to composite scores, further reinforcing its clinical utility as an independent indicator of suicide risk in MDD.

The logistic regression analysis revealed that depressive temperament was the only significant predictor of suicide risk, whereas the composite scores did not reach statistical significance. This finding contrasts with the results of the ROC curve analyses, where depressive temperament and composite scores demonstrated comparable predictive ability. One possible explanation is that while ROC curve analysis assesses classification performance at varying thresholds, logistic regression evaluates the unique contribution of each variable while controlling for others. These findings underscore the clinical utility of assessing depressive temperament as an independent marker of suicide risk, reinforcing its importance in risk assessment and intervention strategies.

Although our findings highlight the potential utility of depressive temperament and composite scores in identifying suicide risk, it is equally important to consider these temperamental traits within a broader context of interacting factors. Suicide is a complex and multifactorial phenomenon [37] shaped by genetic [38], psychological [39], and social influences [40]. Affective temperaments are not deterministic predictors of suicide [41], and various protective factors can mitigate this risk [42]. Furthermore, identifying affective temperaments may facilitate early, optimal interventions [43,44] rather than leading to labeling or stigmatization. Future research should also explore protective factors in individuals with depressive temperament to support strengths-based approaches to suicide prevention. For example, individuals with depressive temperament are known to be hard-working, realistic, and dependable [33], traits that may enhance interpersonal relationships by fostering reliability. The resulting family and social support can serve as a protective factor against suicidal risk [45].

This study had some limitations. First, as a cross-sectional study, it could not establish causality between affective temperaments and suicide risk in MDD. Second, psychiatric diagnoses were based on a clinician’s evaluation rather than structured interviews, potentially limiting diagnostic accuracy. Third, the single-institution setting may restrict the generalizability of findings. Fourth, due to its retrospective design, this study was unable to systematically collect detailed data on comorbid psychiatric conditions and psychosocial influences, which may have contributed to variations in findings. Nonetheless, this is the first study to compare the discriminative performance of individual and composite scores of affective temperaments in identifying high-risk suicide groups among patients with MDD.

In conclusion, we demonstrated that both composite and depressive temperament scores independently exhibit good predictive accuracy for suicide risk in patients with MDD. While both factors showed promising predictive capabilities, depressive temperament was found to be a particularly strong independent predictor of suicide risk, suggesting its potential clinical utility in MDD. These findings underscore the importance of assessing affective temperaments as stable traits for evaluating and predicting suicide risk. Future longitudinal, multi-center studies are needed to further explore the role of affective temperaments, particularly depressive temperament, in predicting suicide risk among patients with MDD.

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

C. Hyung Keun Park, a contributing editor of the Psychiatry Investigation, was not involved in the editorial evaluation or decision to publish this article.

Author Contributions

Conceptualization: Kyeong A Kang, C. Hyung Keun Park. Data curation: Kyeong A Kang, C. Hyung Keun Park. Formal analysis: Kyeong A Kang, Suyeon Lee, C. Hyung Keun Park. Investigation: C. Hyung Keun Park. Methodology: all authors. Writing—original draft: Kyeong A Kang, C. Hyung Keun Park. Writing—review & editing: all authors.

Funding Statement

None

Acknowledgments

None

Figure 1.

Comparison of receiver operating characteristic curve for suicidal risk evaluation using three composite scores and three single scores of affective temperaments. CYC, cyclothymic; DEP, depressive; IRR, irritable.

Figure 2.

Comparison of receiver operating characteristic curve for suicidal risk evaluation using individual scores of affective temperaments. CYC, cyclothymic; DEP, depressive; IRR, irritable; ANX, anxious; HYP, hyperthymic.

Figure 3.

Comparison of receiver operating characteristic curve for suicidal risk evaluation using three composite scores and three single scores of affective temperaments (minimal to moderate depression group). CYC, cyclothymic; DEP, depressive; IRR, irritable.

Figure 4.

Comparison of receiver operating characteristic curve for suicidal risk evaluation using three composite scores and depressive temperament (moderately severe to severe depression group). DEP, depressive.

Table 1.

Demographic and clinical characteristics of patients with MDD

Variables	MDD (N=343)
Age (yr)	29.37±9.10
Sex
Female	214 (62.4)
Male	129 (37.6)
TEMPS-A
Cyclothymic	6.11±3.45
Depressive	4.00±2.47
Irritable	1.84±1.92
Hyperthymic	2.17±1.94
Anxious	1.45±1.10
Composite score 1	9.78±6.37
Composite score 2	11.23±6.71
Composite score 3	11.95±6.15
DSI-SS	4.90±3.49
High risk	228 (66.5)
Low risk	115 (33.5)
PHQ-9	16.58±6.81
Minimal to moderate	123 (35.9)
Moderately severe to severe	220 (64.1)

Values are presented as number (%) or mean±standard deviation.

MDD, major depressive disorder; TEMPS-A, Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire; DSI-SS, Depressive Symptom Inventory-Suicidality Subscale; PHQ-9, Patient Health Questionnaire-9

Table 2.

Suicide risk versus various TEMPS-A scores: ROC curve analyses

Variables	Cutoff	High risk	Low risk	AUC	SE	p	95% CI	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Composite score 1	≥7	184	51	0.734	0.029	<0.001	0.684-0.780	80.70	55.65	78.30	59.26
Composite score 1	<7	44	64
Composite score 2	≥11	153	33	0.729	0.030	<0.001	0.679-0.776	67.11	71.30	82.26	52.23
Composite score 2	<11	75	82
Composite score 3	≥8	200	59	0.739	0.029	<0.001	0.689-0.785	87.72	48.70	77.22	66.67
Composite score 3	<8	28	56
Cyclothymic	≥6	145	46	0.650	0.031	<0.001	0.597-0.700	63.60	60.00	75.92	45.39
Cyclothymic	<6	83	69
Depressive	≥3	190	49	0.754	0.029	<0.001	0.705-0.799	83.33	57.39	79.50	63.46
Depressive	<3	38	66
Irritable	≥2	121	34	0.660	0.029	<0.001	0.607-0.710	53.07	70.43	78.06	43.09
Irritable	<2	107	81
Anxious	≥2	121	53	0.538	0.032	0.240	0.483-0.591	53.07	53.91	69.54	36.69
Anxious	<2	107	62
Hyperthymic	≥5	37	16	0.504	0.033	0.896	0.450-0.558	16.23	86.09	69.82	34.14
Hyperthymic	<5	191	99

TEMPS-A, Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire; ROC, receiver operating characteristic; AUC, area under the curve; CI, confidence interval; SE, standard error; PPV, positive predictive value; NPV, negative predictive value

Table 3.

Comparisons of AUCs from ROC curve by DeLong’s method

Variables	Composite score 1	Composite score 2	Composite score 3	Cyclothymic	Depressive	Irritable
Composite score 1	-
Composite score 2	0.312	-
Composite score 3	0.620	0.354	-
Cyclothymic	<0.001^*	<0.001^*	<0.001^*	-
Depressive	0.300	0.200	0.455	0.001^*	-
Irritable	0.007	0.014	0.002^*	0.761	0.003^*	-

Each cell contains the p-value from the pairwise comparison of AUCs between two variables. p-value less than the Bonferroni-corrected significance value of 0.0033 (0.05/15) was considered to indicate a significant difference.

^* significant findings at p<0.0033.

ROC, receiver operating characteristic; AUC, area under the curve; -, not applicable

Table 4.

Suicide risk versus various TEMPS-A scores: ROC curve analyses (minimal to moderate depression group)

Variables	Cutoff	High risk	Low risk	AUC	SE	p	95% CI	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Composite score 1	≥7	36	24	0.760	0.044	<0.001	0.675-0.832	75.00	68.00	60.00	80.95
Composite score 1	<7	12	51
Composite score 2	≥8	35	27	0.744	0.045	<0.001	0.658-0.819	72.92	64.00	56.45	78.69
Composite score 2	<8	13	48
Composite score 3	≥8	43	33	0.784	0.042	<0.001	0.701-0.853	89.58	56.00	56.58	89.36
Composite score 3	<8	5	42
Cyclothymic	≥5	36	32	0.684	0.048	<0.001	0.594-0.764	75.00	57.33	52.94	78.18
Cyclothymic	<5	12	43
Depressive	≥3	35	23	0.780	0.041	<0.001	0.696-0.850	72.92	69.33	60.34	80.00
Depressive	<3	13	52
Irritable	≥2	27	19	0.705	0.047	<0.001	0.616-0.784	56.25	74.67	58.69	72.73
Irritable	<2	21	56
Anxious	≥3	11	11	0.512	0.052	0.813	0.421-0.603	22.92	85.33	50.00	63.37
Anxious	<3	37	64
Hyperthymic	≥2	35	43	0.591	0.051	0.076	0.499-0.679	72.92	42.67	44.87	71.11
Hyperthymic	<2	13	32

TEMPS-A, Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire; ROC, receiver operating characteristic; AUC, area under the curve; SE, standard error; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value

Table 5.

Comparisons of AUCs from ROC curve by DeLong’s method (minimal to moderate depression group)

Variables	Composite score 1	Composite score 2	Composite score 3	Cyclothymic	Depressive	Irritable
Composite score 1	-
Composite score 2	0.141	-
Composite score 3	0.226	0.077	-
Cyclothymic	0.030	0.086	<0.001^*	-
Depressive	0.522	0.295	0.899	0.061	-
Irritable	0.265	0.429	0.087	0.727	0.165	-

^* significant findings at p<0.0033.

ROC, receiver operating characteristic; AUC, area under the curve; -, not applicable

Table 6.

Suicide risk versus various TEMPS-A scores: ROC curve analyses (moderately severe to severe depression group)

Variables	Cutoff	High risk	Low risk	AUC	SE	p	95% CI	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Composite score 1	≥6	154	27	0.612	0.050	0.026	0.545-0.677	85.56	32.50	85.19	33.33
Composite score 1	<6	26	13
Composite score 2	≥10	138	23	0.603	0.051	0.044	0.535-0.668	77.22	42.50	85.71	28.81
Composite score 2	<10	42	17
Composite score 3	≥8	157	26	0.621	0.049	0.013	0.553-0.685	87.22	35.00	85.85	37.84
Composite score 3	<8	23	14
Cyclothymic	≥7	98	15	0.564	0.049	0.186	0.496-0.631	54.44	62.50	86.64	23.36
Cyclothymic	<7	82	25
Depressive	≥4	129	19	0.627	0.053	0.015	0.560-0.691	71.67	52.50	87.23	29.17
Depressive	<4	51	21
Irritable	≥2	94	15	0.590	0.047	0.053	0.522-0.656	52.22	62.50	86.24	22.52
Irritable	<2	86	25
Anxious	≤0	39	4	0.551	0.047	0.270	0.483-0.618	21.67	90.00	90.69	20.34
Anxious	>0	141	36
Hyperthymic	≤4	156	37	0.506	0.048	0.896	0.438-0.574	86.67	7.50	80.83	11.11
Hyperthymic	>4	24	3

For anxious and hyperthymic temperaments, the optimal cutoff values were ≤0 and ≤4, respectively. However, the ROC curve analyses for these temperaments did not reach statistical significance, indicating that their predictive ability for suicide risk was not significant. TEMPS-A, Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire; ROC, receiver operating characteristic; AUC, area under the curve; SE, standard error; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value

Table 7.

Comparisons of AUCs from ROC curve by DeLong’s method (moderately severe to severe depression group)

Variables	Composite score 1	Composite score 2	Composite score 3	Depressive
Composite score 1	-
Composite score 2	0.284	-
Composite score 3	0.500	0.239	-
Depressive	0.703	0.533	0.867	-

Each cell contains the p-value from the pairwise comparison of AUCs between two variables. p-value less than the Bonferronicorrected significance value of 0.0083 (0.05/6) was considered to indicate a significant difference. ROC, receiver operating characteristic; AUC, area under the curve; -, not applicable

Table 8.

Logistic regression analysis for suicide risk identification

Variables	Adjusted odds ratio (95% CI)	p
Composite score 1	1.080 (0.820-1.423)	0.582
Composite score 2	0.894 (0.699-1.143)	0.372
Composite score 3	1.103 (0.961-1.267)	0.163
Depressive temperament	1.355 (1.138-1.613)	0.001
Age	0.976 (0.948-1.004)	0.094
Sex	0.654 (0.384-1.114)	0.118

Age and sex were included as covariates in the logistic regression model. Sex was coded with female as the reference category. CI, confidence interval

REFERENCES

1. Cai H, Xie XM, Zhang Q, Cui X, Lin JX, Sim K, et al. Prevalence of suicidality in major depressive disorder: a systematic review and meta-analysis of comparative studies. Front Psychiatry 2021;12:690130

2. Arsenault-Lapierre G, Kim C, Turecki G. Psychiatric diagnoses in 3275 suicides: a meta-analysis. BMC Psychiatry 2004;4:37

3. Li X, Mu F, Liu D, Zhu J, Yue S, Liu M, et al. Predictors of suicidal ideation, suicide attempt and suicide death among people with major depressive disorder: a systematic review and meta-analysis of cohort studies. J Affect Disord 2022;302:332-351.

4. Hegerl U. Prevention of suicidal behavior. Dialogues Clin Neurosci 2016;18:183-190.

5. Kessler RC, Berglund P, Demler O, Jin R, Koretz D, Merikangas KR, et al. The epidemiology of major depressive disorder: results from the national comorbidity survey replication (NCS-R). JAMA 2003;289:3095-3105.

6. American Psychiatric Association. Diagnostic and statistical manual of mental disorders, fifth edition. Arlington: American Psychiatric Association; 2013.

7. May AM, Klonsky ED, Klein DN. Predicting future suicide attempts among depressed suicide ideators: a 10-year longitudinal study. J Psychiatr Res 2012;46:946-952.

8. Lingiardi V, McWilliams N. Psychodynamic diagnostic manual: PDM-2 (2nd ed). New York: Guilford Press; 2017.

9. Park GN, Shin HK, Seok JH. [The association between clinical characteristics and mentalization capacity in patients with borderline personality disorder: a preliminary analysis of a Korean cohort]. J Korean Neuropsychiatr Assoc 2022;61:178-185. Korean.

10. Svrakic DM, Draganic S, Hill K, Bayon C, Przybeck TR, Cloninger CR. Temperament, character, and personality disorders: etiologic, diagnostic, treatment issues. Acta Psychiatr Scand 2002;106:189-195.

11. H Vazquez G, Gonda X. Affective temperaments and mood disorders: a review of current knowledge. Curr Psychiatry Rev 2013;9:21-32.

12. Aguiar Ferreira Ad, Vasconcelos AG, Neves FS, Laks J, Correa H. Affective temperaments: familiality and clinical use in mood disorders. J Affect Disord 2013;148:53-56.

13. Akiskal HS, Mendlowicz MV, Jean-Louis G, Rapaport MH, Kelsoe JR, Gillin JC, et al. TEMPS-A: validation of a short version of a self-rated instrument designed to measure variations in temperament. J Affect Disord 2005;85:45-52.

14. Gonda X, Vázquez GH. Theoretical and clinical overview of affective temperaments in mood disorders. Psicodebate 2014;14:39-58.

15. Karam EG, El Khoury E, Itani L. How stable are temperaments in the clinical setting: a pilot study. Clin Pract Epidemiol Ment Health 2016;12:59-63.

16. Kawamura Y, Akiyama T, Shimada T, Minato T, Umekage T, Noda Y, et al. Six-year stability of affective temperaments as measured by TEMPS-A. Psychopathology 2010;43:240-247.

17. Tondo L, Vázquez GH, Sani G, Pinna M, Baldessarini RJ. Association of suicidal risk with ratings of affective temperaments. J Affect Disord 2018;229:322-327.

18. Karam EG, Salamoun MM, Yeretzian JS, Mneimneh ZN, Karam AN, Fayyad J, et al. The role of anxious and hyperthymic temperaments in mental disorders: a national epidemiologic study. World Psychiatry 2010;9:103-110.

19. Vázquez GH, Gonda X, Lolich M, Tondo L, Baldessarini RJ. Suicidal risk and affective temperaments, evaluated with the TEMPS-A scale: a systematic review. Harv Rev Psychiatry 2018;26:8-18.

20. Baldessarini RJ, Innamorati M, Erbuto D, Serafini G, Fiorillo A, Amore M, et al. Differential associations of affective temperaments and diagnosis of major affective disorders with suicidal behavior. J Affect Disord 2017;210:19-21.

21. Akiskal HS, Akiskal KK, Haykal RF, Manning JS, Connor PD. TEMPS-A: progress towards validation of a self-rated clinical version of the Temperament Evaluation of the Memphis, Pisa, Paris, and San Diego Autoquestionnaire. J Affect Disord 2005;85:3-16.

22. Yu H, Yoon J, Lee CW, Park JY, Jang Y, Park YS, et al. Korean validation of the Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire. Psychiatry Investig 2022;19:729-737.

23. Choi S, Yu H, Yoon J, Jang Y, Lee D, Park YS, et al. Korean validation of the short version of the TEMPS-A (Temperament Evaluation of Memphis, Pisa, Paris, and San Diego Autoquestionnaire) in patients with mood disorders. Medicina (Kaunas) 2023;59:115

24. Metalsky GI, Joiner TE. The hopelessness depression symptom questionnaire. Cognit Ther Res 1997;21:359-384.

25. Suh S, Ryu H, Chu C, Hom M, Rogers ML, Joiner T. Validation of the Korean depressive symptom inventory-suicidality subscale. Psychiatry Investig 2017;14:577-584.

26. Choi HS, Choi JH, Park KH, Joo KJ, Ga H, Ko HJ, et al. [Standardization of the Korean version of patient health questionnaire-9 as a screening instrument for major depressive disorder]. J Korean Acad Fam Med 2007;28:114-119. Korean.

27. Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 2001;16:606-613.

28. Simundic AM. Diagnostic accuracy—part 1: basic concepts: sensitivity and specificity, ROC analysis, STARD statement. Point Care 2012;11:6-8.

29. Toyoshima K, Masuya J, Ono M, Honyashiki M, Hashimoto S, Kusumi I, et al. Effects of the interaction between affective temperaments and BIS/BAS on depressive symptoms in individuals with major depressive disorder. Int J Environ Res Public Health 2022;19:15841

30. Naismith SL, Hickie IB, Turner K, Little CL, Winter V, Ward PB, et al. Neuropsychological performance in patients with depression is associated with clinical, etiological and genetic risk factors. J Clin Exp Neuropsychol 2003;25:866-877.

31. Orsolini L, Latini R, Pompili M, Serafini G, Volpe U, Vellante F, et al. Understanding the complex of suicide in depression: from research to clinics. Psychiatry Investig 2020;17:207-221.

32. Van Gastel A, Schotte C, Maes M. The prediction of suicidal intent in depressed patients. Acta Psychiatr Scand 1997;96:254-259.

33. Parker G. Bipolar II disorder: modelling, measuring and managing (2nd ed). New York: Cambridge University Press; 2012.

34. Gonda X, Eszlari N, Torok D, Gal Z, Bokor J, Millinghoffer A, et al. Genetic underpinnings of affective temperaments: a pilot GWAS investigation identifies a new genome-wide significant SNP for anxious temperament in ADGRB3 gene. Transl Psychiatry 2021;11:337

35. Greenwood TA, Badner JA, Byerley W, Keck PE, McElroy SL, Remick RA, et al. Heritability and genome-wide SNP linkage analysis of temperament in bipolar disorder. J Affect Disord 2013;150:1031-1040.

36. Kendler KS, Eaves LJ, Loken EK, Pedersen NL, Middeldorp CM, Reynolds C, et al. The impact of environmental experiences on symptoms of anxiety and depression across the life span. Psychol Sci 2011;22:1343-1352.

37. de Faria FM, Oliveira MC, de Araújo Filho GM, Lazzaro CDS, dos Santos Gouveia ETF, Birolli MF, et al. Suicide attempts in a general hospital: a case-control. Open J Psychiatry 2023;13:518-531.

38. Edwards AC, Ohlsson H, Mościcki E, Crump C, Sundquist J, Lichtenstein P, et al. On the genetic and environmental relationship between suicide attempt and death by suicide. Am J Psychiatry 2021;178:1060-1069.

39. De Beurs D, Fried EI, Wetherall K, Cleare S, O’ Connor DB, Ferguson E, et al. Exploring the psychology of suicidal ideation: a theory driven network analysis. Behav Res Ther 2019;120:103419

40. Na PJ, Shin J, Kwak HR, Lee J, Jester DJ, Bandara P, et al. Social determinants of health and suicide-related outcomes: a review of meta-analyses. JAMA Psychiatry 2025;82:337-346.

41. Erbuto D, Innamorati M, Lamis DA, Berardelli I, Forte A, De Pisa E, et al. Mediators in the association between affective temperaments and suicide risk among psychiatric inpatients. Psychiatry 2018;81:240-257.

42. O’Brien E, Whitman K, Buerke M, Galfalvy H, Szanto K. Life-satisfaction, engagement, mindfulness, flourishing, and social support: do they predict depression, suicide ideation, and history of suicide attempt in late life? Am J Geriatr Psychiatry 2023;31:415-424.

43. Favaretto E, Bedani F, Brancati GE, De Berardis D, Giovannini S, Scarcella L, et al. Synthesising 30 years of clinical experience and scientific insight on affective temperaments in psychiatric disorders: state of the art. J Affect Disord 2024;362:406-415.

44. Walsh MA, Royal AM, Barrantes-Vidal N, Kwapil TR. The association of affective temperaments with impairment and psychopathology in a young adult sample. J Affect Disord 2012;141:373-381.

45. Zhou Q, Li N, Li CA, Zhang J. Interpersonal relationship and suicide attempt: the role of family and social relationship. J Nerv Ment Dis 2024;212:595-602.