Toward Understanding of Suicidality in a Spanish Clinical Population: Validation of the European Spanish Version of the Interpersonal Needs Questionnaire

Abstract

Introduction:

Suicidal ideation (SI) represents one of the most prominent predictors of suicidal behavior (SB). The Interpersonal Needs Questionnaire (INQ) was developed from the Interpersonal Theory of Suicide (ITS) to assess the 2 core drivers of SI proposed by the theory. Despite the relevance of suicide-related ideations and ITS, there is a lack of psychometric measures validated in clinical Spanish population that adequately evaluate SI components of ITS. Thus, the main aim of the study was to validate INQ-10 in a Spanish clinical sample including the genuine cultural and linguistic characteristics of European Spanish.

Methods:

315 participants were included in the analyses; 149 of them consulted mental health services for the presence of suicide-related behaviors. A series of exploratory and confirmatory factor analyses were carried out to identify the factor solution. Bivariate and multivariate analyses were used to analyze psychometric properties. Finally, sensitivity and specificity properties were explored through receiver-operating characteristic analyses which also provided the cut-off values of the questionnaire.

Results:

An 8-item version demonstrated a good fit to the 2-factor solution. Likewise, this 8-item version showed good psychometric properties. Sensitivity and specificity indices of the version validated as well as the calculated cut-off points were excellent.

Conclusions:

The current results demonstrate the utility of an 8-item INQ European Spanish version as a valid measure of the current SI in Spanish clinical population. In addition, the validated form reflects the theoretical framework on which it was built.

Introduction

Suicide is a major public health problem worldwide [1]. Traditionally, Spain has been one of the European countries with the lowest rates of suicide, although rates have worryingly increased over the last few years [2]. In addition, suicide-related behaviors rates could increase significantly in the context of the COVID-19 pandemic [3]. Problematically, there remains an important deficit in prevention and treatment programs for suicidal behavior (SB) in Spain [4].

The study of human behavior has led to the development of diverse theories that have tried to understand and predict the occurrence of certain behavioral phenomena. In that sense, the Interpersonal Theory of Suicide (ITS) is one of the most prominent theories of SB [5]. The ITS identifies 3 core constructs related to SB: (i) intractable perceived burdensomeness, (ii) intractable thwarted belongingness, and (iii) capability for suicide.

According to the theory, all 3 components must be present for the occurrence of SB. Specifically, perceived burdensomeness and thwarted belongingness together are proposed to lead to the development of suicidal ideation (SI), which then interacts with capability for suicide to result in lethal or near-lethal SB [6].

The Interpersonal Needs Questionnaire (INQ) was developed to measure perceived burdensomeness and thwarted belongingness [7]. Hill et al. [8] identified 5 different versions of the INQ in 2015: (i) a 25-item version [9, 10], and shorter versions with a subset of items from the INQ-25, including (ii) an 18-item version [11–15]; (iii) a 15-item version [7, 16]; (iv) a 12-item version [17–19]; and (v) a 10-item version [8, 20–23]. Hill et al. [8] reported that the INQ-15 and INQ-10 demonstrated the best model fit. However, the authors recognized that the INQ-10 held advantages related to administration time as well as better predictive validity [8].

To the best of our knowledge, since the study by Hill et al. [8] was published, at least 3 new INQ versions have been developed. In particular, (i) a version with a shortened response scale adapted to elderly people [24]; (ii) a short version with 8 items [25]; and (iii) a 10-item version with refinement of the INQ response scale from a 7-point Likert-type scale to a 4-point Likert-type scale [26].

Previous publications have analyzed and studied the psychometric properties of the INQ in Hispanic populations. First, Silva et al. [27] published the results of a translation and validation of the INQ-15. The authors reported that the INQ-15 did not show especially good fit, but an INQ-9 version did [27]. On the other hand, a group of Spanish researchers have also examined the psychometric properties of the INQ-12 [28]. The authors reported a suboptimal model fit for the 12-item version although the fit for a 10-item version proved adequate. Both validations differ not only in the number of items but also in the items selected; in fact, they share only 4 items. Furthermore, in neither validation do the authors report clinical information of the participants or test the specificity and sensitivity of their respective validations.

One of the main objectives of the present work was to be able to analyze the sensitivity, specificity, and clinical utility of a European Spanish language version of the INQ in a clinical population. On the other hand and given the lexical and grammatical variability of Spanish and the cultural differences that may exist [29], we aimed to include the genuine cultural and semantic characteristics of the European Spanish language.

Due to the advantages of shorter scales (e.g., in clinical use), and the apparently good characteristics shown by the 10-item version, we sought to validate and analyze the psychometric properties of this version [8, 20–23]. Based on previous publications, we hypothesized that the INQ-10 would show a 2-factor solution and adequate psychometric properties. In addition, we hypothesized that perceived burdensomeness and thwarted belongingness would independently predict current SI, in line with the ITS assumptions.

Materials and Methods

Participants

A total of 334 participants were recruited initially. Participants who conformed the clinical sample were recruited from the Virgen del Rocío University Hospital (Seville, Spain) and the University Hospital Germans Trías i Pujol (Barcelona, Spain). In addition, healthy adults willing to participate comprised university students from Universitat Autònoma de Barcelona as well as volunteers who were contacted and enlightened about study objectives using Google Forms through e-mail lists and professional WhatsApp, following a snowball technique. All data were anonymized.

The total sample was divided into 2 subsamples to carry out exploratory factor analyses (EFA) and confirmatory factor analyses (CFA) in independent samples allowing a more rigorous examination of the psychometric properties of the questionnaire [30]. Inclusion criteria to be included in the first subsample were (1) the absence of SI over the last month or score <11 on the Beck Scale for Suicide Ideation (BSS) [31]; (2) the absence of a history of SB over lifetime; and (3) native or demonstrated competency in Spanish. Exclusion criteria were (1) age under 18 years and (2) no demonstrated competency in Spanish. The inclusion criteria for the second subsample were (1) the presence of SI during the last month, demonstrated by scores >11 on the BSS [31]; (2) having made a suicide attempt in the last month; and (3) native or demonstrated competency in Spanish. Exclusion criteria were (1) age under 18 years and (2) no demonstrated competency in Spanish. Nineteen participants (5.69%) were removed from the analyses for not fulfilling inclusion criteria.

Participants who met inclusion criteria and none of the exclusion criteria were invited to participate in the study. All participants provided informed consent. The research protocol was approved by the local Ethics Committees of Germans Trías i Pujol University Hospital (PI-19-213) and Virgen del Rocio University Hospital (PI-1378-N-18).

Translation Process

Copyright permission was obtained to reproduce the INQ items in Spanish. Following Muñiz et al.’s [32] recommendations, the INQ was initially translated (instructions and items) into European Spanish by the first author of this article. Taking into account the linguistic, cultural, and psychological characteristics of the specific population to which the test is translated and adapted is one of the main recommendations made by the test translation and adaptation guides [33]. Three independent Spanish clinicians separately reviewed this initial draft for grammar and fluency. The resulting version of the instrument was then translated back into English by a native bilingual English-Spanish speaker who was blind to the English version. There was a consensus meeting between the first author of the present article and the second author to discuss apparent item discrepancies between the original INQ (English version) and the back translation. The final version was sent to Van Orden et al. [7] (the first author of the INQ-150 item version) who confirmed that the back-translated version corresponded appropriately with the English version, as well as that the back-translated items reflected the 2 factors related to SI from the ITS.

Linguistic Differences of the European Spanish Version versus the Latin American Spanish Version

European Spanish language represents a significant change in terms of the Spanish used in the Silva et al. [27]version. European Spanish refers to the Spanish spoken in the territory belonging to the current Kingdom of Spain. One of the main differences involves the elimination of the subject pronouns that were included in all items of the Silva et al. [27] version (e.g., yo pienso [“I think”] vs. pienso [also means “I think”]). The Royal Spanish Academy emphasizes that the use of subject pronouns should be limited to avoid ambiguity, when the verb forms coincide or there are several possible references. Otherwise, these translations may be completely redundant [34].

On the other hand, we have replaced the expression “gente” [“people”]” with “personas” [the literal translation of which is “persons” although the correct translation is “people”]. As was noted by the Diccionario Panhispánico de Dudas, the word “gente” is used differently in South America than in Spain. For example, in South America, “gente” is used in the sense of “people,” while in Spain it is used more as a collective name. Moreover, “gente” in South America is used also as an adjective (usually positive) which is not the case in Spain [35].

Additionally, the expression “in my life” translated by Silva et al. [27] as “en mi vida” [“in my life”] has been translated in European Spanish as “de mi vida” [“in my life”]. The preposition “en,” which is a literal translation of the English preposition “in,” denotes in Spanish place, time, or mode, while the preposition “de” denotes belonging.

Finally, Silva et al. [27] made a literal translation of the temporal reference that starts each of the INQ-15 items. However, from our point of view, the addition of the preposition “durante” in all of the items included in the questionnaire facilitates the understanding of the temporal moment to which the items refers for European Spanish speakers.

Measures

Sociodemographic variables such as gender, age, ethnicity, socioeconomic status, and level of education were recorded using a questionnaire created ad hoc for the study (See Table 1). In addition, the following instruments were administered.

Table 1.

Sociodemographic and clinical characteristics

	Total sample (n = 315)	Subsample 1 (n = 166)	Subsample 2 (n = 149)
Age, years, mean±SD	38.51±13.81	36.13±12.90	41.17±14.33
Gender (men), n (%)	112/(35.56)	57/(34.3)	61/(40.94)
Ethnicity, n (%)
Caucasian	303 (96.2)	162 (97.6)	141 (94.6)
Gypsy	5 (1.6)	0 (0)	5 (3.4)
Asian	1 (0.3)	1 (0.6)	0 (0)
Latin Americans	2 (0.6)	1 (0.6)	1 (0.7)
Black African	2 (0.6)	1 (0.6)	1 (0.7)
Arab	2 (0.6)	1 (0.6)	1 (0.7)
Civil status (married), n (%)	82 (26)	44 (26.5)	38 (25.5)
Socioeconomic status (low), n (%)	107 (34)	24 (14.5)	83 (55.7)
Education (primary), n (%)	70 (22.2)	16 (9.6)	54 (36.2)
INQ-10 Total score, mean±SD	28.81±16.79	17.22±7.67	41.72±14.62
Perceived burdensomeness	12.54±9.70	6.22±2.48	19.60±9.89
Thwarted belongingness	16.26±9.13	10.99±6.26	22.12±8.20
Beck Scale for suicide ideation, mean±SD	19.68±5.78	8.01±0.89	20.29±5.25
SELS-S, mean±SD	48.59±22.76	35.97±17.34	62.65±19.64
Social loneliness	15.50±9.67	10.13±6.70	21.49±8.93
Familiar loneliness	14.03±9.42	8.73±5.73	19.94±9.23
Romantic loneliness	19.06±10.06	17.11±10.75	21.22±8.79
BDI-II, mean±SD	19.60.16.64	7.40±7.84	33.19±12.89
BNSG-S, mean±SD	79.99±19.84	91.02±15.27	67.72±16.96
Autonomy	15.90±4.73	17.62±3.69	13.99±5.02
Competence	26.23±9.05	31.39±7.26	20.48±7.21
Relationship	37.86±9.10	42.01±6.46	33.25±9.42
ISEL, mean±SD	118.58±23.17	130.89±15.51	104.87±22.62
Appraisal	36.17±7.21	39.57±4.59	32.38±7.70
Tangible	31.41±6.98	34.37±5.04	28.11±7.37
Self-esteem	29.80±6.37	33.01±4.57	26.21±6.18
Belonging	21.21±5.43	23.94±3.84	18.16±5.33

SD, standard deviation.

Interpersonal Needs Questionnaire-10 Item [20–23].

The INQ-10 was developed from the 25-item INQ [9, 10]. The questionnaire measures the 2 main components of the interpersonal theory of suicide [5] related to SI: perceived burdensomeness and thwarted belongingness. Participants rate how true each statement is for them recently on a 7-point Likert scale (1 = not at all true for me to 7 = very true for me). Higher scores reflect higher levels of perceived burdensomeness and thwarted belongingness. The INQ-10 has demonstrated adequate psychometric characteristics [8].

Beck Scale for Suicide Ideation [36].

The BSS is a 21-item measure that assesses passive and active suicide ideation. The last 2 items are not part of the total BSS score. Each item response is graded according to suicidal intensity on a 3-point scale ranging from 0 to 2. Higher scores reflect higher levels of SI. The BSS has been previously used in a Spanish population demonstrating good psychometric properties [31].

Social and Emotional Loneliness Scale Short (SELS-S) [37].

The SELS-S was developed to evaluate different dimensions of loneliness (i.e., emotional and social). Three factors have been found to comprise the scale: (i) social loneliness, understood as lack of friends as well as social relationships; (ii) familial loneliness, which reflects lack of family support; and finally (iii) romantic loneliness, which reflects the lack of an intimate relationship [37]. The scale is composed of 15 items on a 7-point Likert scale (1 = strongly disagree to 7 = strongly agree). The Spanish version has demonstrated good psychometric properties [38].

Beck Depression Inventory-II (BDI-II) [39].

The BDI-II is a 21-item measure for assessing the presence and intensity of depressive symptoms. Items range from 0 to 3, with a maximum total score of 63. The ninth item of the questionnaire evaluates the existence of SI. In the current study, the scale shows high internal consistency, with a Cronbach’s alpha coefficient of 0.91 and a Spearman-Brown split-half coefficient of 0.83 [40].

Basic Needs Satisfaction in General Scale (BNSG-S) [41].

The BNSG-S is a 16-item measure of satisfaction for the need for autonomy, competence, and relationships. Items are rated on a 7-point Likert scale (1 = not at all true to 7 = very true). Higher scores are indicative of greater needs satisfaction. There is a Spanish validated version which shows good psychometric properties [42].

Interpersonal Support Evaluation List (ISEL) [43, 44].

The ISEL is composed of 40 items rated on a 4-point Likert scale (1 = definitely false to 4 = definitely true), which assesses the perceived availability of 4 separate dimensions: (i) belonging or companionship support; (ii) appraisal, described as emotional support or someone to talk to about problems; (iii) self-esteem maintenance through social positive comparison statements when comparing oneself to others; and (iv) tangible, that is, aid material or instrumental support. The ISEL can also produce a total score, with higher scores reflecting more interpersonal support.

Self-Injurious Thoughts and Behaviors Interview (SITBI) [45].

To evaluate suicide attempts, we used the module of the SITBI designed to evaluate suicide attempts. The SITBI is comprised of 5 modules that correspond to 5 types of SITBs: (i) SI, (ii) suicide plans, (iii) suicide gestures, (iv) suicide attempts, and (v) non-suicidal self-injury. The SITBI has been validated in a Spanish population, and its psychometric properties demonstrated to be adequate [46]. The questions included in the SITBI capture the recommendations made by Silverman et al. [47, 48] insofar as they appear to be specific, selective, and clear regarding what is meant by “suicide attempt.”

Statistical Analyses

Statistical analyses were conducted using the Statistical Package for the Social Sciences (SPSS), ver. 25 [49]. Analysis of Moment Structure (AMOS) [50] was also used for structural equation modeling analyses.

The Kolmogorov-Smirnov test examined the normality of variable distribution. The first aim of the study was to analyze the number of factors that best define the INQ-10. For this purpose, parallel analysis was initially conducted to confirm the number of factors. In particular, rawpar.sps script developed by O’Connor [51] was used to perform these analyses. Moreover, EFAs were conducted to analyze the factor solution. Before EFA, the Kaiser-Meyer-Olkin (KMO) test and Bartlett’s test of sphericity were conducted to evaluate the factorability. As the KMO measure of sampling was adequate and the significance of Bartlett’s test of sphericity was <0.05, EFAs proceeded. Following factor analysis recommendations, EFAs were performed with principal axis factoring. This extraction method is recommended when the assumption of normality is severely violated or for a relatively simple factor pattern [52, 53]. Extracted factors were rotated by varimax rotation. This rotation is the most recommended when the number of components is relatively low [54]. For selection of items corresponding to each of the factors, item factor loadings must be >0.50, and at least double its loading on the other factor [55, 56]. The reliability of items in each factor was examined by Cronbach’s α.

CFA were also conducted. Several model fit indices were used to examine the goodness-of-fit of the model with the given dataset: χ² mean/degree of freedom, root mean square error of approximation (RMSEA), standardized root mean squared residual (SRMR), normed fit index, comparative fit index, and the Tucker-Lewis index.

Procedures proposed by Fornell and Larcker [57] were used to analyze reliability as well as convergent and discriminant validity. We calculated Composite Reliability (CR), average variance extracted (AVE), maximum shared squared variance (MSV), and average shared squared variance (ASV) to compute reliability as well as convergent and discriminant validity.

Moreover, convergent validity was also analyzed using correlational analyses between the scores of the 2 scales of the INQ-10 with the results obtained for the BSS, SELS, BNSG-S, and ISEL. Evidence of validity based on differential scores was obtained also. For this, the mean scores of the INQ subscales of subsample 1 and subsample 2 were compared. Given the significant differences between the groups in age (41.17 ± 14.33 vs. 36.15 ± 12.90; U = 10,266.50, p = 0.001), Analysis of Covariance (ANCOVA) was used to explore the possible differences between the groups adjusting by age.

Predictive validity was calculated using binary logistic regression. Dependent variable was the presence/absence of SI measured by BBS which was dichotomized using cut-off scores established previously in a Spanish population [31]. Predictor variables were: (i) perceived burdensomeness, (ii) thwarted belongingness, and (iii) depressive symptomatology. These variables were selected to examine predictive power of INQ factors including relevant cofounders such as depressive symptomatology and demographic variables. Three regression models were done, using in each of the regression analyses the following independent variables: (i) perceived burdensomeness and thwarted belongingness measured by the INQ, (ii) demographic parameters, and depressive symptomatology measured by the BDI-II, and (iii) depressive symptomatology in conjunction with INQ factors. Variance inflation factor was calculated for each regression model and all variance inflation factor values were <3; thus, the assumption of multicollinearity was not violated [58]¹.

Sensitivity is understood as a positive predictive value (i.e., the presence of SI) while specificity is negative predictive value (i.e., the absence of SI). Receiver operating characteristic (ROC) curve analyses were performed to evaluate the accuracy of the INQ-10. In particular, a ROC curve is a graphical display that plots sensitivity estimates against one minus the specificity of a marker for all possible values. Area under the ROC curve (AUC) is the performance of a marker. Higher AUC values indicate better marker performance [59].

Moreover, ROC analyses allow determination of the optimal cut-off point. To obtain the optimal cut-off point value we used the Index of Union, a method based on the value of the AUC. This method defines the optimal cut-off point as the value at which sensitivity and specificity are closest to the value of the AUC, and the absolute value of the difference between sensitivity and specificity values is at its minimum [60]. State variable was defined as BSS total score >11.

Results

Sample Characteristics

315 (94.31%) of the initial 334 participants were included in the analyses. Of the participants included in the study, 197 (62.54%) were women, while 118 (37.46%) were men, aged 18–80 years (38.51 ± 13.81). Sample 1 consisted of 166 participants (109 women, 57 men), aged 20–70 years (36.13 ± 12.9). Sample 2 consisted of 149 patients (88 women, 61 men), aged 18–80 years (41.17 ± 14.33), 90 of whom (60.4%) had made a suicide attempt over the last month and 59 (39.6%) had consulted mental health services for the presence of SI. Full demographic and clinical characteristics of the study sample are given in Table 1.

Exploratory Factor Analyses

Data from subsample 1 were used to carry out EFA. The Kolmogorov-Smirnov test was statistically significant (p < 0.001) for all INQ-10 items, rejecting the null hypothesis for a normal distribution of scores. The number of factors was first explored using parallel analysis. The comparison between the eigenvalues of the source data and the eigenvalues of random datasets suggested a retention of 2 factors (shown in Fig. 1a). In relation to EFA, KMO revealed good sampling adequacy (KMO = 0.73), and Bartlett’s test of sphericity was 540.75 (p < 0.001), supporting the implementation of the EFA. The 2-factor solution accounted for 55.9% of the variance; factor one (perceived burdensomeness) contained the first 5 items (39.6% of the variance) and items 6–10 made up the second factor (thwarted belongingness), accounting for 16.3% of the variance. The analyses of the factor loadings revealed that item 5 saturated at a level <0.50. Moreover, item 9 loaded clearly onto factor 2, but its factor loading was not twice the loading on the first factor. Therefore, the EFA was newly executed removing items 5 and 9 from the new analyses.

Fig. 1.

a, b Graphs of the parallel analysis of INQ-10.

The exclusion of the items improved the EFA. In fact, the percentage of the variance explained was higher, from 55.9% to 63%. KMO and Bartlett’s test supported the exploratory analyses and all the items showed adequate factor loadings. The factor loadings of each of the items are provided in Table 2. The comparison between the eigenvalues of the source of the data and the eigenvalues of the random data made in the parallel analyses re-affirmed a 2-factor solution (shown in Fig. 1b).

Table 2.

Exploratory factor analysis

Items	INQ-10		INQ-8
	factor 1	factor 2	factor 1	factor 2
1	0.73	0.19	0.71	0.20
2	0.91	0.10	0.92	0.11
3	0.74	0.08	0.76	0.12
4	0.80	0.19	0.81	0.19
5	0.43	0.27	(−)	(−)
6	0.24	0.59	0.27	0.62
7	0.07	0.84	0.08	0.86
8	0.27	0.62	0.23	0.55
9	0.42	0.50	(−)	(−)
10	0.01	0.85	0.02	0.89
Variance explained, %	39.6	16.3	42.7	20.3
	55.9		63
KMO test	0.74		0.73
Bartlett’s sphericity test	701.99 (p < 0.001)		540.75 (p < 0.001)

Factor loading >0.5 are in bold.

In order to analyze internal consistency of the subscales, Cronbach’s alpha coefficient was calculated. For the perceived burdensomeness subscale, the reliability estimate was 0.79, and was 0.75 for the thwarted belongingness subscale.

Confirmatory Factor Analyses

Once the EFA was conducted, a CFA was performed on subsample 2 to confirm the internal structure of the INQ formed by 8 items. The initial results showed a nonoptimal RMSEA estimation (0.10) as well SRMR estimation (0.08). Given that, modification indices were used to identify possible discrepancies between the proposed and estimated model. These indices indicated a potential significant correlation between the uniqueness’s of items 3 and 4 (Modification index = 11.89). CFA was repeated but this time freeing the correlation between items 3 and 4. This time, fit indices showed good model fit: χ² mean/degree of freedom = 1.79; RMSEA = 0.07; SRMR = 0.07; normed fit index = 0.94; comparative fit index = 0.97; Tucker-Lewis index = 0.95. The standardized factorial coefficients of the model are shown in Figure 2. The structure found mirrors the version proposed by Bryan et al. [23].

Fig. 2.

CFA; standardized solution.

CR was used to examine the reliability of the subscales. CR = 0.91 for factor 1 (perceived burdensomeness) and CR = 0.79 for factor 2 (thwarted belongingness). The AVE value for the first factor was 0.71 and >0.051 for the second factor. The values of the statistics to test the presence of discriminant validity were MSV = 0.08 and ASV = 0.04.

Spearman correlations between the INQ-8 factors and the different total scores of the SELS-S, BNSG-S, and ISEL were as follows: significant positive correlations were found between perceived burdensomeness and the SELS (r = 0.62, p < 0.001), familial loneliness (r = 0.65, p < 0.001), social loneliness (r = 0.55, p < 0.001), and romantic loneliness (r = 0.25, p < 0.001). In relation to the BNSG-S, the correlations were as follows: total score (r = −0.61, p < 0.001), autonomy (r = −0.45, p < 0.001), competency (r = −0.53, p < 0.001), and relationship (r = −0.55, p < 0.001). Finally, correlations between thwarted belongingness and ISEL were as follows: total score (r = −0.60, p < 0.001), tangible (r = −0.50, p < 0.001), appraisal (r = −0.49, p < 0.001), self-esteem (r = −0.60, p < 0.001), and belonging (r = −0.56, p < 0.001).

Thwarted belongingness was related positively to the SELS-S total score (r = 0.63, p < 0.001), familial loneliness (r = 0.62, p < 0.001), social loneliness (r = 0.68, p < 0.001), and romantic loneliness (r = 0.24, p < 0.001). The thwarted belongingness subscale maintained negative significant correlations with the BNSG-S (r = −0.67, p < 0.001), autonomy (r = −0.42, p < 0.001), competency (r = −0.58, p < 0.001), and relationship (r = −0.67, p < 0.001), as well as with the ISEL total score (r = −0.70, p < 0.001), tangible (r = −0.63, p < 0.001), appraisal (r = −0.58, p < 0.001), self-esteem (r = −0.63, p < 0.001), and belonging (r = −0.65, p < 0.001).

Construct validity of INQ-8 was also checked by analyzing possible differences between subsample 1 and sub-sample 2 using ANCOVA. The results of the statistic showed significant differences between the groups in perceived burdensomeness (14.68 ± 8.41 vs. 4.68 ± 1.83; F_2,312 = 117.35, p < 0.001) as well as in thwarted belongingness (17.25 ± 6.80 vs. 8.98 ± 5.50; F_2,312 = 70.91, p < 0.001).

SI as measured by the BSS was significantly related to perceived burdensomeness (B = 1.23; CI = 1.17–1.30; p < 0.001) and thwarted belongingness (B = 1.11; CI = 1.06–1.17; p < 0.001) using only both factors as predictors. The model was significant (χ² = 176.69, p < 0.001) and the amount of variance explained ranged from 42.9% (Cox and Snell R square) to 59.1% (Nagelkerke R square). The second binary regression model was carried out using as independent variables demographic and depressive symptomatology. Results indicated that only depressive symptomatology was significantly associated with ideation (B = 1.16; CI = 1.13–1.20; p < 0.001). The model was significant (χ² = 208.84, p < 0.001), and the amount of variance explained ranged from 48.5% (Cox and Snell R square) to 66.7% (Nagelkerke R square). Finally, in the third regression model, constructed with the 2 INQ factors and total BDI score, the significant predictors of ideation were perceived burdensomeness (B = 1.13; CI = 1.06–1.20; p < 0.001) and depressive symptomatology (B = 1.11; CI = 1.08–1.15; p < 0.001). The model was significant (χ² = 228.73, p < 0.001) and the amount of variance explained ranged from 51.6% (Cox and Snell R square) to 71% (Nagelkerke R square).

Finally, sensitivity and specificity properties of the INQ-8 subscales were analyzed using ROC analyses. Regarding Factor 1 (perceived burdensomeness) the results were as follows: AUC = 0.89 (CI 0.84–0.93), with 84.7% sensitivity and 85.3% specificity for an optimal cut-off point value of 6.5. Factor 2 (thwarted belongingness) showed an AUC of 0.81 (CI 0.76–0.86) as well as 73.9% sensitivity and 73.5% specificity for an optimal cut-off point value of 13.5 (shown in Fig. 3).

Fig. 3.

ROC curve.

Discussion

The main findings our results supported were as follows: (i) a 2-factor model for the INQ, (ii) adequate psychometric properties, including reliability and validity; as well as (iii) a good sensitivity and specificity for the optimal cut-off points identified for the INQ subscales of perceived burdensomeness and thwarted belongingness.

EFA revealed that items 5 and 9 did not clearly load as expected onto the first factor and second factor, respectively. So, both items were removed and EFA as well as parallel analyses were repeated, now without the fifth and ninth items from INQ-10. The results of the INQ with 8 items showed good accuracy in all the statistics included in the EFA and the new parallel analyses re-confirmed a 2-factor solution.

One possible explanation for why the fifth item from INQ-10 did not load onto perceived burdensomeness could be because it was the only item which highlights that the existence of the person worsens the lives of others. In contrast, the rest of the INQ-10 items making up the first factor emphasize that it is the absence of the person which would improve the lives of their loved ones. Furthermore, the differences between American and Spanish culture could explain this inconsistency. In fact, INQ was originally developed in the USA which has been recognized as a country with very strong tendency to emphasize the individual, self-reliance, independence, and personal goals. However, Spain is far more moderate than USA in terms of individualism [61]. For example, identity in Spain is often deeply influenced by the groups to which one belongs. Spanish people usually live in closely knit networks of family and friends, generating more social connectedness, interdependence, and in-group goals than in more individualistic societies such as the USA [62]. This fact may facilitate experience of situations that disconfirm the ideas that their presence makes things worse (reality testing). In that sense, the presence of more interdependence between individuals could influence the way in which the sense of belonging is formed in less individualistic cultures.

The low factor loading of item number 9 of INQ-10 could also be due to cultural reasons. The original item “These days, I often feel like an outsider in social gatherings” was translated as “…a menudo me siento ajeno en las reuniones sociales.” Although “reuniones sociales” is grammatically correct and understandable, it may not fully capture the meaning of the ITS. In Spain, “reuniones sociales” is rarely used. Instead, a sentence such as “…a menudo me siento ajeno cuando me relaciono en grupo” could better capture the meaning of the theory and reflect it better in European Spanish.

The CFA confirmed a two-factor solution. However, items 3 and 4 had to be allowed to correlate to improve CFA indices [63]. A correlation between those items was also reported by Park and Kim [64] in a Korean adaptation of INQ-15. The results obtained from parallel EFA and CFA analyses support the two-factor model of the INQ-8 Spanish version and is in line with the expected factor loadings for the items.

Internal consistency was estimated for the 2 factors using Cronbach method with subsample 1. The values reported were >0.70, which demonstrates good internal consistency [65]. Likewise, CFA indices were also used to check reliability, and convergent and discriminant validity. Convergent validity was considered to be confirmed when AVE >0.5 [66]. Moreover, CR values >0.7 indicate good scale reliability [67] and discriminant validity was established in that MSV and ASV were both lower than AVE [68]. The indexes AVE, CR, MSV, and ASV resulting from our analyses meet the criteria specified above indicating that INQ-8 items present adequate reliability as well as convergent and discriminant validity.

Factor 1 (i.e., perceived burdensomeness) correlated significantly with the dimensions included in SELS-S, especially with the familial loneliness dimension, while factor 2 (i.e., thwarted belongingness) maintained significant relationships with all dimensions contained in BNGS-G dimensions, that is, autonomy, competence, and relationship. However, in line with previous work, the divergent validity analyzed using Spearman’s correlations was not as robust. In fact, we found that both perceived burdensomeness and thwarted belongingness correlated significantly with all the dimensions of SELS-S although perceived burdensomeness correlated more strongly with familial loneliness in SELS-S and autonomy in BNGS-S, while factor 2 (i.e., thwarted belongingness) with social loneliness (SELS-S), competency, and relationship from BNSG-S and information, instrumental, self-esteem as well as belonging dimensions from ISEL.

The differing results from divergent validity using Spearman’s correlations have been explained by Park and Kim [64] as a result of cultural differences between individualistic and collectivistic cultures or due to the inadequacy in the selection of measures to analyze discriminant validity. As was noted by Van Orden et al. [7] these results could reflect the broad relation of social disconnection across a variety of psychological experiences. In addition, Van Orden et al. [7] also suggested that poorer discriminant validity could be due to the fact that the “need to belong,” the measurement of which the items that make up the thwarted belongingness factor were developed, is a “fundamental human motivation” [69]. This fact could affect the divergent validity because the “need to belong” would impact on many psychological experiences and behaviors, making it difficult to find psychological constructs that are not related to it [7].

Significant differences in factor 1 and 2 of the INQ-8 between subsample 1 and 2 were found. ANCOVA showed significant differences between the subsamples in both factors indicating that the INQ-8 is a sensitive measure to the presence of SI. Moreover, logistic regression models demonstrated the concurrent predictive validity of the questionnaire. It is worth noting that even when the factors of the INQ-8 are included in combination with depressive symptomatology, perceived burdensomeness was the main predictor of SI, quite consistent with the larger literature on the theory [70].

There has been an interesting debate regarding the role of perceived burdensomeness and thwarted belongingness as predictors of SI. In fact, differing results have been published regarding not only whether these 2 factors are or not predictors of suicidality but also which of them turn out to be more significant in the prediction of SI [70]. Our results are in agreement with those which found that perceived burdensomeness is the main predictor of current SI [71]. It seems that perceived burdensomeness could have a higher impact on collectivist cultures than on individualistic ones [72]. It has been argued that selfidentity in collectivist cultures might be more dependent on social relationships and could be translated into personal inability since personal fulfillment in these cultures depends more on the group than in more individualistic societies in which personal achievement is more emphasized [62, 64].

ROC analyses revealed good performance of the INQ-8 Spanish version subscales of perceived burdensomeness and thwarted belongingness. These results are in agreement with previous publications which reported that INQ-15 may be useful in the detection of SI [73]. In fact, our results are in line with those presented by Mitchell et al. [74] who reported that perceived burdensomeness had larger effects on SI than thwarted belongingness.

Some limitations have to be noted. First, the samples included in the study were relatively small. However, one of the main advantages of the study is that subsample 2 is composed of a clinical population. Second, relevant information such as hopelessness was not measured. According to the ITS, hopelessness could have a relevant role in the perception of perceived burdensomeness and thwarted belongingness as stable and unchanging. Finally, the selection of measures to analyze discriminant validity could have been inadequate. Nevertheless, CFA supported good discriminant validity of the questionnaire.

One of the periods with the highest suicide risk is after clinical contact [75]. In this regard, in most cases, suicide risk evaluations depend on the clinical judgements made by clinicians [76]. In fact, it has been suggested that psychometric tools could help clinical decision-making differentiating those people with a high risk of SB [77]. Thus, from our point of view, INQ-8 European Spanish Version is an empirically derived suicide risk tool that can assist clinicians in the assessment of 2 core factors related to SB such as perceived burdensomeness and thwarted belongingness.

In conclusion, the INQ-8 Spanish (online suppl. material 1, 2; for all online suppl. material, see www.karger.com/doi/10.1159/000519792) demonstrated good psychometric properties and could be a useful instrument for use in suicide risk assessment to evaluate current SI in Spain and other areas in which European Spanish is spoken. In addition, the proposed cut-off points may help clinicians interpret the questionnaire’s scores. Future research is necessary to refine response options of possible future versions of the questionnaire as well as to replicate the results of the present study.

Data Availability Statement

Data are available in Canal-Rivero, Manuel (2021), “Interpersonal needs questionnaire. Spanish validation,” Mendeley Data, V1, doi: 10.17632/hnx3sfdf38.1 [78].