A multimethod validation of the frequency of suicidal ideation inventory for adolescents in Peru

Abstract

Background

Suicidal ideation assessment requires psychometrically robust measures, particularly in adolescent populations in which risk factors may differ from those of adults. This study examined the psychometric properties of the Frequency of Suicidal Ideation Inventory (FSII) in Peruvian adolescents.

Methods

A cross-sectional study was conducted among 1211 adolescents (50.3% female; M = 14 years, SD = 1.53) from grades 1 to 5 in public schools in Lima (55.3%, surveys administered in Spanish) and Cusco (44.7%, surveys administered in Quechua). Data were collected using self-administered paper-based questionnaires, which included the Frequency of Suicidal Ideation Inventory (FSII), the Generalized Anxiety Disorder Scale (GAD-7), and the Patient Health Questionnaire (PHQ-9).

Results

Confirmatory factor analysis (CFA) supported the unidimensionality of the scale with excellent fit indices [Comparative Fit Index (CFI) = 0.99, Tucker–Lewis Index = 0.99, Root Mean Square Error of Approximation = 0.017, Standardized Root Mean Square residual (SRMR) = 0.005] and high factor loadings (0.82–0.91). Complementing the CFA, Item Response Theory analyses further confirmed the structural robustness of the scale, with excellent model fit (97.6% of response patterns fit the graded response model), strong item discrimination parameters (α > 2.7), and well-targeted difficulty levels across the trait continuum. Measurement invariance testing demonstrated equivalent functioning across gender (girls vs. boys) and ethnolinguistic groups (Spanish-speaking vs. Quechua-speaking) (ΔCFI < 0.010, ΔSRMR < 0.030 at all levels). Moreover, the FSII showed strong convergent validity, with significant correlations with depressive symptoms (r = 0.68) and generalized anxiety (r = 0.65). Reliability was consistently high (ω = 0.90, H = 0.94, rxx = 0.82) and provided precise measurement at moderate-to-high levels of suicidal ideation.

Conclusions

These results support the FSII as a psychometrically sound measure for assessing suicidal ideation among adolescents in Peru, with potential utility for identifying individuals at risk.

Introduction

Suicide is one of the most serious public health problems affecting adolescents worldwide [1]. According to the World Health Organization [2], approximately 720,000 people die by suicide each year, making it the third leading cause of death among individuals aged 15 to 29 years. The burden is particularly high in low- and middle-income countries, where structural inequalities, stigma, and limited access to mental health services exacerbate risk [3]. In the Americas, suicide rates have increased substantially over the past two decades, underscoring the need for effective prevention strategies targeting young populations.

In Peru, suicide has shown a marked upward trend, rising from 518 cases in 2016 to 3211 in 2024 [4]. Adolescents and young adults are among the most affected groups, and approximately 20% of Peruvian adolescents report experiencing suicidal thoughts [5]. The risk is heightened by factors such as depression, bullying, family violence, and limited access to care, particularly in underserved regions [6]. Indigenous adolescents may face additional vulnerabilities related to poverty, marginalization, and cultural disconnection. However, research in these populations remains scarce [7].

Suicidal ideation as a critical risk factor

Suicidal behavior often originates from suicidal ideation, a key marker of vulnerability within the continuum leading to attempts or death by suicide [8]. This ideation ranges from passive wishes (e.g., a desire to disappear) to active thoughts about death as an escape or as a solution to perceived meaninglessness in life [9, 10]. Empirical evidence shows that suicidal ideation is strongly associated with suicide attempts and is a robust predictor of future suicidal behaviors [11].

Adolescence is a particularly sensitive developmental period, characterized by emotional, cognitive, and social changes that increase vulnerability to mental health problems. Among these, depression and anxiety are consistently identified as major risk factors for suicidal ideation and behavior [12–15].

Depressive symptoms in adolescents, such as irritability, social withdrawal, academic difficulties, and hopelessness, are closely linked to both the emergence and persistence of suicidal thoughts [16, 17]. Similarly, anxiety disorders contribute significantly to suicidal ideation, with evidence from Peruvian adolescents showing that both depression and anxiety predict suicidal thoughts and their potential progression to attempts [18].

These findings are consistent with contemporary theoretical frameworks, such as the Integrated Motivational-Volitional Model [19], and the biopsychosocial model [20] which describe suicidal behavior as a progressive process in which psychosocial stressors trigger ideation that may escalate into suicidal actions [21]. These models highlight early identification of suicidal ideation as a critical window for prevention, emphasizing the need for appropriate assessment tools.

Despite this, existing instruments present important theoretical and psychometric limitations. Widely used scales, such as the Beck Scale for Suicide Ideation [9] and the Columbia-Suicide Severity Rating Scale [22] assess dimensions like presence, intensity, or distress, but often overlook the frequency of suicidal thoughts, which may be crucial for identifying high-risk individuals [23]. Evidence suggests that presence or severity alone is an insufficient predictor of future suicidal behavior [24] and that suicidal ideation is dynamic and subject to recall biases [25, 26].

Importantly, longitudinal studies indicate that frequency may be the most reliable predictor of suicide risk. For instance, Miranda et al. [27] found that among several dimensions, only frequency significantly predicted subsequent suicide attempts, while Simon et al. [28] reported that frequency, rather than mere presence, showed the strongest association with risk.

Finally, a key limitation concerns cultural applicability. Several instruments have been developed in Western, high-income contexts, limiting their use in low- and middle-income countries [29]. This is particularly relevant in adolescents, whose developmental characteristics differ from those of adults. Additionally, some measures assess related but distinct constructs (e.g., intent or planning), which may compromise a precise evaluation of suicidal ideation [9, 30, 31].

The Frequency of Suicidal Ideation Inventory (FSII)

In response to the limitations previously noted, the Frequency of Suicidal Ideation Inventory (FSII) represents a promising alternative for assessing suicidal ideation. Developed by Chang and Chang [32], the FSII is a brief, unidimensional instrument specifically designed to evaluate the frequency of suicidal thoughts over the last 12 months. The scale consists of five ordinal response items, enabling a more direct, concise, and time-sensitive assessment compared to more complex, multidimensional instruments that require weighing multiple aspects of suicidal thinking.

In its initial validation in a non-clinical sample in the United States, the FSII demonstrated excellent internal consistency (Cronbach’s α = 0.96) and a unifactorial structure that explained 86.96% of the construct variance, with factor loadings above 0.85 [32]. In a subsequent cross-cultural study, the scale maintained acceptable internal consistency in samples from the United States, Hungary, Turkey, and China (α = 0.85-0.93), and showed significant correlations with both risk factors (depression, hopelessness) and protective factors (life satisfaction, optimism) related to suicide.

Additional studies have confirmed the validity of the FSII in diverse cultural contexts. In Hungary, Chang et al. [33] reported good model fit indices [Comparative Fit Index (CFI) = 0.97; Tucker–Lewis Index (TLI) = 0.95; Goodness-of-Fit Index (GFI) = 0.95; Root Mean Square Error of Approximation (RMSEA) = 0.07], as well as strong correlations with relevant psychological variables. Similarly, research in Brazil [34] and Spain [10] has supported its unidimensional structure, internal consistency (α > 0.80), and concurrent validity with both risk and protective measures related to suicide.

In Peru, Baños-Chaparro et al. [35] assessed the scale in a general adult sample, finding excellent fit indices [CFI = 0.99; Standardized Root Mean Square Residual (SRMR) = 0.03; Weighted Room Mean Square Residual = 0.37; RMSEA = 0.03], factor loadings above 0.74, and high internal consistency (ω = 0.80; H = 0.91). Moreover, the instrument demonstrated metric and scalar invariance across sex and age groups, and showed a moderate correlation with depressive symptoms, supporting its criterion validity.

While the FSII has demonstrated favorable psychometric properties in adult populations, evidence supporting its validity in adolescent samples remains limited, being a significant limitation in the literature. This gap is particularly concerning given that adolescents may interpret and respond to self-report items differently than adults due to developmental differences in cognitive maturity, emotional regulation, and social context [36]. Ensuring that the FSII items are equally comprehensible and relevant for adolescents is essential, as variations in interpretation could influence the structural validity of the instrument.

Gender and ethnolinguistic considerations in the FSII

Regarding gender differences, research indicates that suicidal ideation is more prevalent among adolescent females than males. For instance, Yeo et al. [37] found that in a large sample of South Korean 12th graders, the prevalence of suicidal ideation was 13.8% in girls compared to 9.1% in boys, suggesting a significant gender difference in suicidal thoughts. Similarly, Zhang et al. [38] confirmed that while girls might report lower rates of being bullied compared to boys, those who were bullied exhibited a significantly higher prevalence of suicidal ideation, indicating differing consequences of bullying across genders. This supports the notion that context and social experiences contribute to the psychological outcomes linked to gender.

In addition to gender, linguistic and cultural diversity may also influence how adolescents interpret and respond to items related to suicidal ideation. Although the FSII has been validated in Spanish [10], no study has examined its use in Quechua, a native language spoken in countries such as Argentina, Bolivia, Chile, Colombia, Ecuador, and Peru. Notably, Quechua is spoken by approximately 7.8 million people worldwide [39].

Furthermore, Lima (urban) and Cusco (predominantly rural and Andean) have been identified as regions with notable prevalence of suicidal ideation and behavior among adolescents, reflecting distinct sociocultural and structural risk contexts [40]. Epidemiological evidence indicates that suicidal ideation is a significant public health concern among Peruvian adolescents, with prevalence estimates approaching 8–10% in recent national analyses [41] and up to 20% in school-based samples [42]. In urban contexts such as Lima, adolescents may be exposed to stressors related to socioeconomic inequality, population density, and interpersonal violence, which have been associated with suicidal ideation in local studies [43]. In contrast, rural Andean contexts such as Cusco present additional vulnerabilities, including geographic isolation, limited access to mental health services, and culturally mediated patterns of emotional expression and help-seeking, which may influence both the experience and reporting of suicidal thoughts.

Cultural and linguistic factors may affect both the interpretation of scale items and the willingness to disclose suicidal thoughts. For instance, expressions related to death or suicide can carry distinct connotations depending on regional norms, and cultural stigma may influence the openness of adolescents to report such experiences. In Peru, a culturally diverse country shaped by Hispanic, Andean, and Indigenous influences, ensuring the cultural and linguistic equivalence of assessment tools like the FSII is especially important.

The concept of measurement invariance refers to an instrument’s ability to measure a construct in the same way across different groups (e.g., gender, language). In other words, it ensures that respondents interpret the items and the underlying latent variable similarly [44]. While the FSII has demonstrated invariance across gender and age groups in adult populations [10], no study has yet tested its measurement invariance in adolescents or across ethnolinguistic groups, such as Spanish and Quechua speakers.

Verifying the measurement invariance of the FSII in diverse adolescent populations is essential for ensuring valid, culturally sensitive assessment of suicidal ideation. Doing so would help address important gaps in the literature and support the development of equitable mental health screening practices.

The present study

Within this framework, the present study aimed to analyze the psychometric properties of the FSII in a sample of Peruvian adolescents, using both Classical Test Theory and Item Response Theory (IRT) approaches. Additionally, we sought to estimate the reliability of the scale through various coefficients (Cronbach’s α, McDonald’s ω, and IRT-based reliability), to ensure the internal consistency of the instrument within the target sample.

Subsequently, the study assessed the convergent validity of the FSII by examining its association with standardized measures of depression and anxiety, given the well-established link between these affective symptoms and suicidal ideation in adolescence.

Finally, the measurement invariance of the scale was analyzed based on gender (boys vs. girls) and ethnolinguistic groups (Quechua speakers vs. Spanish speakers), to determine whether the scale measures the construct of suicidal ideation equivalently across culturally and linguistically diverse subgroups. With these analyses, this research aimed to contribute to the validation of a technically robust and culturally sensitive tool for assessing suicide risk in Peruvian adolescents.

Method

Participants

The study initially included 1260 adolescents. A total of 49 cases were excluded due to missing responses in the sociodemographic questionnaire (e.g., age, sex), and incomplete responses on the FSII items. The final sample consisted of 1211 adolescents, including 609 females (50.3%) and 602 males (49.7%), with a mean age of 14 years [standard deviation (SD) = 1.53], enrolled in 1st grade to 5th grade of secondary school. A total of 670 participants (53.3%) were from Lima Province, and 541 (44.7%) were from Cusco.

Participants were recruited using a non-probabilistic convenience sampling strategy, based on accessibility and institutional collaboration with participating schools.

The sample size was estimated based on the hypothesized factorial model of the FSII (five items, one factor), considering four parameters: an expected CFI of 0.95, an average factor loading of 0.60, statistical power of 99%, and a significance level of . The minimum required sample size was 611 participants. The final study sample exceeded this recommendation.

Instruments

Frequency of Suicidal Ideation Inventory (FSII; [32])

This tool is designed to assess the frequency of suicidal thoughts over the past 12 months using five items that ask how often adolescents: [1] thought about hurting themselves [2], felt their life was not worth living [3], wondered what would happen if they ended their own life [4], thought about committing suicide, and [5] wished they did not exist. Responses are rated on a 5-point ordinal response scale ranging from 1 (never) to 5 (almost every day).

For the present study, the Spanish-language version of the FSII was used [35]. To ensure linguistic and conceptual equivalence for Quechua-speaking participants, a rigorous translation methodology was employed. The adaptation process started with forward translation by a bilingual expert proficient in both Spanish and Quechua. This initial translation was subsequently back-translated into Spanish by an independent translator to verify semantic accuracy. The back-translated version was systematically compared with the original instrument to identify and resolve any discrepancies. To further refine the translation, the scale was pilot tested with 30 bilingual students who provided feedback on item clarity and cultural relevance. Based on their comments, minor wording adjustments were made to improve comprehension while preserving the original meaning. Finally, a panel of bilingual community experts evaluated the adapted measure for cultural appropriateness and linguistic accuracy, yielding a finalized Quechua version.

Generalized Anxiety Scale (GAD-7; [45, 46])

This is a brief self-report instrument designed to detect and assess the presence and severity of generalized anxiety symptoms over the past two weeks. It consists of seven items (e.g., “Over the last two weeks, how often have you been bothered by the following problems? Not being able to stop or control worrying”). Responses are rated on a 4-point ordinal response scale ranging from 0 (not at all) to 3 (nearly every day). For Quechua-speaking participants, the validated Quechua version of the GAD-7 was used [45]. In the present study, the scale demonstrated good internal consistency (Cronbach’s α = 0.89).

Patient Health Questionnaire (PHQ-9; [47, 48])

This is a brief self-report questionnaire designed to detect, diagnose, and assess the severity of depressive symptoms. It consists of nine items (e.g., “Over the last two weeks, how often have you been bothered by any of the following problems? Feeling down, depressed, or hopeless”) rated on a 4-point ordinal response scale ranging from 0 (not at all) to 3 (nearly every day). For Quechua-speaking participants, the validated Quechua version of the PHQ-9 was used [47]. In the present study, the scale showed good internal consistency (Cronbach’s α = 0.88).

Procedure

A cross-sectional study was conducted in public schools located in Lima and Cusco, using a paper-based self-report questionnaire administered to a sample of secondary school students.

Schools were selected through a convenience sampling procedure, based on their accessibility and willingness to participate. The final sample included seven schools in Lima and six in Cusco. The initial step involved obtaining authorization from the relevant educational authorities to carry out the assessments. Once approval was granted, arrangements were made with teachers at each participating school to schedule appropriate times for administration.

The assessments were conducted during regular school hours and were supervised by a trained psychologist, assisted by two research assistants who were present to address any questions or concerns. Each evaluation session lasted approximately 50 min.

Given the sensitive nature of assessing suicidal ideation, procedures were implemented to ensure participants’ safety. Following data collection and scoring, participants with elevated levels of suicidal ideation were identified and referred to the corresponding school personnel to facilitate appropriate follow-up and referral to available mental health or counseling services. Additionally, the presence of a trained psychologist during administration ensured that immediate support could be provided if any participant experienced distress during the assessment.

Ethical topics

Participants provided informed assent, and their parents gave written informed consent before participation. No incentives were offered. The study adhered to the ethical principles outlined in the Declaration of Helsinki and received prior approval from the Ethics Committee of the Universidad Científica del Sur (CIEI-CIENTÍFICA), under approval code 101-2024-PRO99.

This work was funded by CONCYTEC through the PROCIENCIA program as part of the call “E041-2024-04: Proyectos de Investigación en Ciencias Sociales 2024-04” under contract PE501088191-2024, with support from Universidad Científica del Sur (DGIDI).

Data Analysis

All the analyses were performed using RStudio software (version 4.3.2). In the first stage, descriptive characteristics were examined for all three instruments, exploring measures of central tendency (e.g., mean) and dispersion (including SD, skewness, and kurtosis). Additionally, relationships between items were examined through a polychoric correlation matrix, and the corrected item-test correlation (r_it > 0.30) was reported [49].

Subsequently, structural validity (applied only to the FSII) was analyzed by means of confirmatory factor analysis (CFA), performed using the lavaan package with the Weighted Least Squares Mean and Variance-adjusted estimator (WLSMV), recommended for ordinal data [50]. Model fit was assessed using the following indices: CFI > 0.95, TLI > 0.95, RMSEA < 0.08, and SRMR < 0.08 [51].

For internal consistency estimation (FSII only), the Bayesrel package was used to compute the Bayesian omega coefficient (ω), with 95% credible intervals [52]. The Hancock–Mueller function was applied to calculate coefficient H, along with empirical reliability (r_xx), estimated from an IRT perspective using the mirt package [53, 54].

Additionally, factorial invariance (FSII only) was evaluated through multigroup CFA to verify model stability across gender and ethnolinguistic group [55]. This procedure progressively imposed equality constraints on parameters (configural, factor loadings, intercepts, and residual errors). Fit assessment was based on changes in indices: ΔCFI < 0.010 and ΔSRMR < 0.030 [56].

For validity assessment associated with other variables, a covariance-based structural equation modeling approach was implemented for suicidal ideation, depressive symptoms, and generalized anxiety, using the lavaan package, with the robust maximum likelihood estimator, suitable for continuous variables. Good-of-fit criteria included incremental indices (CFI and TLI > 0.95) and absolute indices (RMSEA and SRMR < 0.08) [51]. Effect sizes were interpreted as small (0.10), moderate (0.30), and large (0.50) [57].

From an IRT perspective (FSII only), the mirt package was used to fit a two-parameter logistic model, considering the discrimination parameter (*a*), which reflects an item’s ability to differentiate latent trait (θ) levels (adequate values > 1; [58], and the difficulty parameter (β), representing response thresholds (β1 to β4) relative to θ. IRT assumptions were verified: [1] Unidimensionality by means of CFA [2], Local independence using the LD-X² statistic [3, 59] Monotonicity via Mokken’s nonparametric model, evaluating the critical statistic (crit < 0.40) and scalability coefficients Hj (per item) and H (total scale), with expected values > 0.30 [60, 61].

Item information curves and test information curves were generated. Global model fit was examined using the C² statistic (suitable for ordinal IRT models), alongside CFI and TLI (> 0.90) [62–64]. Acceptable fit thresholds were RMSEA₂ = 0.089 and SRMR = 0.05, while excellent fit required RMSEA₂ = 0.05 and SRMR = 0.027 [65]. Local item fit was reviewed via the S-X² index and RMSEA.S-X² (< 0.06) [66, 67]. Finally, person-level fit was evaluated using the Zh index, with a threshold of ± 2 to identify atypical responses [68].

Results

Descriptive analysis

The sociodemographic characteristics of the participants are presented in Table 1. Regarding parental educational attainment, both mothers and fathers most frequently reported having completed secondary education, followed by lower levels of schooling.

Table 1.

Sociodemographic characteristics of the participants

Variables	N (%)
Gender
Females	602 (49.7%)
Males	609 (50.3%)
Grade
First	240 (19.8%)
Second	270 (22.3%)
Third	267 (22%)
Fourth	186 (15.4%)
Fifth	248 (20.5%)
Province
Lima	670 (55.3%)
Cusco	541 (44.7%)
Father’s education level
No education	39 (3.2%)
Incomplete primary	106 (8.8%)
Complete primary	182 (15%)
Incomplete secondary	150 (12.4%)
Complete secondary	406 (33.5%)
Incomplete technical	31 (2.6%)
Complete technical	134 (11.1%)
Incomplete university	30 (2.5%)
Complete university	69 (5.7%)
Mother’s education level
No education	110 (9.1%)
Incomplete primary	207 (17.1%)
Complete primary	120 (9.9%)
Incomplete secondary	128 (10.6%)
Complete secondary	453 (37.4%)
Incomplete technical	49 (4%)
Complete technical	72 (5.9%)
Incomplete university	31 (2.6%)
Complete university	41 (3.4%)
Current parental status
Living together	852 (70.4%)
Separated	275 (22.7%)
Father widowed	12 (1%)
Mother widowed	44 (3.6%)
Father unknown	18 (1.5%)
Mother unknown	3 (0.2%)
Both parents unknown	3 (0.2%)
Both parents deceased	4 (0.4%)
Father’s occupation
Works	1104 (91.2%)
Studies	1 (0.1%)
Studies and works	16 (1.3%)
Neither works nor studies	14 (1.2%)
Other	23 (1.9%)
Don’t know	53 (4.4%)
Mother’s occupation
Works	700 (57.8%)
Studies	6 (0.5%)
Studies and works	25 (2.1%)
Neither works nor studies	11 (0.9%)
Housewife	447 (36.9%)
Other	22 (1.9%)
Pride of native language
Yes	1195 (98.7%)
No	16 (1.3%)

In terms of occupation, the majority of fathers were employed, whereas mothers were more commonly engaged in paid work or domestic activities. The distribution of participants by sex, grade, and province, as well as other family characteristics, is also detailed in Table 1.

As shown in Table 2, the highest mean was observed for item 3 of the FSII, while the lowest was found for item 4. Regarding the SD, item 3 presented the highest variability, whereas item 1 showed the lowest. Skewness and kurtosis values were within acceptable ranges (± 1.5), indicating no severe departures from normality. Additionally, the corrected item-total correlations were adequate (r_it > 0.30), and the polychoric correlation matrix revealed positive associations among items and no evidence of multicollinearity (r > 0.90).

Table 2.

Descriptive analysis and polychoric correlation matrix of the FSII items

Items	Descriptive analysis					Correlation matrix
	M	SD	g1	g2	r it	1	2	3	4	5
1	2.07	1.31	0.99	−  0.27	0.70	–
2	2.17	1.34	0.85	−  0.53	0.81	0.74	–
3	2.31	1.40	0.73	− 0.80	0.72	0.68	0.74	–
4	2.03	1.38	1.08	− 0.23	0.76	0.72	0.79	0.72	–
5	2.24	1.44	0.81	− 0.77	0.80	0.73	0.82	0.75	0.81	–

M = mean; SD = standard deviation; g1 = Skewness; g2 = kurtosis; r_it = corrected item-total correlation

Evidence based on internal structure

The unidimensional structure of the FSII proved to be appropriate, with excellent fit indices: CFI = 0.99, TLI = 0.99, RMSEA = 0.017 [90% CI 0.001, 0.046], and SRMR = 0.005. Additionally, standardized factor loadings ranged from 0.82 to 0.91 (Fig. 1).

Fig. 1.

Internal structure of the FSII

Measurement invariance

Table 3 presents the results of the factorial invariance analysis using the WLSMV estimator. At each level of hierarchical constraint, the difference in ΔCFI was less than 0.010 and the difference in ΔSRMR was below 0.030, indicating adequate construct equivalence across gender and ethnolinguistic groups.

Table 3.

Measurement invariance of the FSII by gender and ethnolinguistic groups

Group	X2(df)	p	CFI	SRMR	∆CFI	∆SRMR
Gender
Configural	17.04 (10)	0.073	1	0.009	–	–
Metric	30.08 (14)	0.007	1	0.037	0.000	0.028
Scalar	45.40 (28)	0.020	1	0.017	0.000	0.020
Strict	45.40 (28)	0.020	1	0.017	0.000	0.000
Ethnolinguistic groups
Configural	40.01 (10)	0.001	0.999	0.009	–	–
Metric	91.51 (14)	0.001	0.999	0.026	0.000	0.017
Scalar	97.18 (28)	0.001	0.999	0.013	0.000	0.013
Strict	97.18 (28)	0.001	0.999	0.013	0.000	0.000

CFI Comparative Fit Index; SRMR Standardized Root Mean Square Residual

Evidence based on the relationship with other variables

Figure 2 shows the results of the structural relationship model, which presented a good fit to the data (CFI = 0.97, TLI = 0.96, RMSEA = 0.034 [90% CI 0.031, 0.038], SRMR = 0.032). Based on Cohen’s classification [57] of correlations around r ≈ 0.50 as large, suicidal ideation was positively and statistically significantly correlated with depressive symptoms (r = 0.68, p = 0.001) and generalized anxiety (r = 0.65, p = 0.001), indicating large positive correlations.

Fig. 2.

Structural relationship model among suicidal ideation, depressive symptoms, and generalized anxiety in Peruvian adolescents

Evidence based on the item response theory

The assumptions for the IRT model were met: unidimensionality was supported by the CFA results, local independence was verified using the LD-X² statistic with standardized values within ± 1.96 (range = − 0.177 to − 0.115), and monotonicity was confirmed, with no significant violations detected (crit = 0). Adequate values were also found for the Hj coefficient (range = 0.632 to 0.705) and the overall scalability coefficient H = 0.67. Overall, the model fit was satisfactory (C² = 3.651, df = 5, p = 0.600, RMSEA = 0.001 [90% CI 0.001, 0.033], SRMR = 0.053, CFI = 0.999, TLI = 0.999). In addition, the Zh index showed that 97.6% of participants displayed response patterns consistent with the IRT model, indicating a high level of model fit and suggesting strong validity of the responses obtained. At the item level, all items showed acceptable fit values (RMSEA.S-x² < 0.06) (see Table 4).

Table 4.

Discrimination, difficulty, and general fit parameters of the FSII items

Items	Item parameters					Item Fit
	a	β 1	β 2	β 3	β 4	S-x2(df)	RMSEA.S-x2
1	2.746	− 0.002	0.595	1.066	1.591	113.62 (42)	0.038
2	3.953	− 0.110	0.458	0.913	1.398	62.18 (36)	0.025
3	2.787	− 0.262	0.411	0.881	1.355	193.98 (42)	0.055
4	3.758	0.175	0.649	1.000	1.335	117.14 (40)	0.040
5	4.357	− 0.058	0.442	0.810	1.158	39.30 (36)	0.009

a = discrimination parameter; b = difficulty parameter; S-x² = fit index; df = degrees of freedom; RMSEA.S-x² = root mean square error of approximation

Regarding the estimated parameters: for the discrimination parameter (α), all items demonstrated adequate discrimination of the latent trait θ (α > 1), with Item 5 showing the highest value (see Table 4). For the difficulty parameter (β), the probability of endorsing successive response options increased progressively with higher levels of θ. As shown in Fig. 3, the item and test information curves indicate that Item 5 provided the most information, and overall, the scale is more precise at higher levels of the latent trait θ.

Fig. 3.

Item information function (Section A) and scale information function (Section B)

Internal consistency

Reliability was assessed using Bayesian estimation, structural equation modeling, and IRT. Specifically, the posterior mean of McDonald’s omega coefficient (ω) was 0.90 [95% Credibility Interval: 0.89 – 0.91]. In addition, the H and r_xx coefficients showed adequate values for the FSII (H = 0.94, r_x= 0.82), respectively.

Discussion

In recent years, suicidal ideation has become an increasingly important factor to explore in adolescents, given its association with mental health risks and long-term psychological outcomes.

The present study examined the internal structure of the FSII in a sample of Peruvian adolescents.

The CFA results demonstrated excellent model fit, which exceeded conventional goodness-of-fit thresholds. The findings align with previous studies [10, 32–35], which have reported a unidimensional structure with excellent fit indices in the FSII tool. This consistency suggests that suicidal ideation, as operationalized by the FSII, may represent a relatively stable and coherent latent construct across cultural contexts. This supports the conceptualization of suicidal ideation as a unidimensional experience in early and middle adolescence, facilitating its use in screening contexts where brevity and clarity are essential.

Additionally, the IRT analysis reinforces the precision and sensitivity of the scale. From a clinical perspective, the ability of the FSII to discriminate across different levels of the latent trait is particularly relevant, as it enables not only the identification of high-risk individuals but also the detection of subtle variations in ideation frequency. This is crucial in preventive contexts, where early-stage identification may interrupt the progression toward more severe suicidal behaviors. Regarding the association of suicidal ideation and other variables (i.e., anxiety and depression symptomatology), our findings align with existing literature demonstrating a robust association between anxiety and suicidal ideation in adolescents [69]. A growing body of evidence suggests that anxiety symptoms not only correlate with but may also predict the emergence of suicidal thoughts. For instance, Chiu et al. [70] found that baseline social anxiety symptoms significantly predicted suicidal ideation two years later, even after controlling for initial suicidal tendencies. This underscores the role of anxiety as an independent risk factor beyond preexisting ideation. Further supporting this link, O’Neil Rodriguez and Kendall [71] identified key transdiagnostic mechanisms, such as impulsivity and distress intolerance, that may exacerbate the risk of suicidal ideation among youths with anxiety disorders. Their work highlights how maladaptive coping strategies, often prevalent in anxious adolescents, may contribute to a heightened vulnerability to suicidal thoughts.

Longitudinal studies further reveal that anxiety disorders can accelerate the progression from ideation to more severe outcomes. Wolk et al. [72] observed that suicidal ideation frequently emerges years before actual plans or attempts, particularly in adolescents with clinical anxiety. This pattern emphasizes the critical need for early screening and intervention in this population to disrupt the trajectory toward self-harm.

Our results corroborate with previous research demonstrating a strong association between suicidal ideation and depressive symptomatology. For instance, Ibrahim et al. [73] identified depression as the strongest predictor of suicidal ideation among Malaysian youth, with depressive symptoms significantly increasing the likelihood of suicidal thoughts. Similarly, Orri et al. [74] found that depressive symptoms were consistently linked to both passive and active suicidal ideation across diverse adolescent populations.

Understanding this relationship has critical implications for clinical practice and prevention strategies. Targeted therapeutic interventions addressing depressive symptoms in adolescents may help mitigate the risk of suicidal ideation. This is particularly salient given that untreated depression substantially elevates the likelihood of developing suicidal thoughts, as evidenced by Kwon et al. [75]. Thus, early psychiatric intervention is essential to disrupt this risk trajectory.

Regarding measurement invariance, this study evaluated the equivalence of the FSII between male and female Peruvian adolescents. The results are in line with previous research conducted in Spain, where strict invariance was also established [10]. Furthermore, our findings demonstrated that participants who completed the Quechua version in Cusco and the Spanish version in Lima understand the construct equivalently, with strict invariance achieved. Notably, no previous study has evaluated measurement invariance across different language adaptations of the scale. These results support the hypothesis that FSII is a robust measure across diverse linguistic and cultural contexts, with its items being consistently interpreted.

The reliability indices further support the robustness of the instrument. High internal consistency, particularly within a Bayesian framework, suggests that the FSII items capture a well-defined construct with minimal measurement error. However, internal consistency alone does not guarantee temporal stability, which should be addressed in future studies.

From a clinical perspective, these findings are particularly relevant for the early detection of suicide risk in populations with limited access to mental health services. The fact that the FSII performs equally well in Quechua, a language spoken by nearly 4 million people in Peru [76], enables its implementation in mental health services in rural areas and Indigenous communities, where linguistic and cultural barriers often limit access to accurate assessments. This is crucial in countries such as Peru, where adolescent suicide rates in Andean regions exceed the national average, according to reports from the Ministry of Health [5, 77].

This study provides compelling evidence for policymakers to consider the FSII in multicultural healthcare frameworks. The findings justify its inclusion in screening protocols for bilingual schools, community health centers, and prevention programs, ensuring that interventions reach vulnerable populations without methodological distortions. Furthermore, the study sets a precedent for future adaptations into other Indigenous languages (such as Aymara or Amazonian languages), a necessary step toward reducing mental health disparities.

Despite its relevance, this study has limitations that must be considered when interpreting the results. First, the cross-sectional design prevents the assessment of temporal stability. Relatedly, the absence of test–retest reliability data limits conclusions regarding the consistency of FSII scores over time. Future research should incorporate longitudinal designs to examine both stability and predictive validity. Second, the sample did not include adolescents with a documented history of suicide attempts, which limits the generalizability of the findings to high-risk clinical settings. Moreover, the sample was restricted to public-school adolescents from specific regions of Peru, which may limit the generalizability of the findings. Additionally, the use of a non-probabilistic convenience sampling strategy may have introduced selection bias and reduced the representativeness of the sample. Future research should prioritize probabilistic or stratified sampling designs and include more diverse populations (e.g., private-school students and out-of-school youth).

Third, the reliance on self-report measures introduces potential sources of bias, including social desirability and the underreporting of suicidal ideation. Given the sensitive and stigmatized nature of suicide, particularly in Indigenous and rural contexts, participants may have minimized or concealed their experiences due to fear of social judgment, cultural norms surrounding emotional expression, or mistrust in assessment processes. These factors may lead to an underestimation of the true prevalence and severity of suicidal ideation in the sample. Future studies should consider incorporating multi-method assessment strategies (e.g., clinical interviews, informant reports, or behavioral indicators) to mitigate these biases and enhance the validity of findings.

Lastly, although the study addresses a critical gap in the literature by focusing on Quechua and Spanish-speaking adolescents, questions remain about how specific cultural variables, such as Andean worldviews regarding death, the role of spirituality, or the stigma associated with mental health, may influence responses to the instrument. These cultural dynamics may also interact with response biases, shaping how adolescents interpret and endorse items related to suicidal ideation. Future research should incorporate mixed methods (quantitative and qualitative) to explore these dynamics and refine the items according to the region’s linguistic and sociocultural particularities.

Despite these limitations, the study provides robust preliminary evidence for the cross-cultural validity of the FSII, positioning it as a promising tool for assessing suicide risk in diverse adolescent populations. Its rigorous implementation could facilitate early identification of at-risk individuals, thereby optimizing prevention strategies in multicultural settings such as Peru.

Author contributions

DM-M conceived the study, participated in its design, reviewed the statistical analysis, drafted and edited the manuscript. JB-CH performed the statistical analysis and reviewed the final manuscript. JG-A coordinated data collection and contributed to manuscript writing. DV-Z participated in manuscript writing and provided feedback on the study design. NW critically reviewed the manuscript and provided substantive revisions. All authors reviewed and approved the final version of the manuscript.

Funding

This work was funded by CONCYTEC through the PROCIENCIA program as part of the call “E041-2024-04: Research Projects in Social Sciences,” under contract PE501088191-2024. Likewise, the work was co-financed by Universidad Científica del Sur through the office of the DGIDI.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Participants provided informed assent, and their parents gave written informed consent prior to participation. No incentives were offered. The study adhered to the ethical principles outlined in the Declaration of Helsinki and received prior approval from the Ethics Committee of the Universidad Científica del Sur (CIEI-CIENTÍFICA), under approval code 101-2024-PRO99.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.