Validity of the LRSP Instrument: Short Version to Measure Psychopathic Traits in Mexican Adolescents

1. Introduction

The National Survey on Discrimination in Mexico indicates that 3 out of 10 adolescents aged 12 to 17 (28.1%) report having felt bullied, with a higher prevalence among those from low socioeconomic backgrounds [1]. Exposure to violence can have lasting and generalizable consequences on an adolescent’s psychosocial adjustment, defined as a lack of adaptability to the transitional stages of life, affecting their future psychosocial and occupational well-being in adulthood [2]. At the level of the nervous system, exposure to violence in adolescence leads to hyperactivity in the sympathetic nervous system, similar to experiences involving fight-or-flight responses. Furthermore, scientific evidence points to a morphological reduction in the cerebral cortex in infants exposed to constant stress resulting from exposure to violence [3,4]. Adolescence is a stage in which the individual is immersed in a series of cognitive, physical, affective, and social changes and is vulnerable to engaging in risky behaviors. It is characterized by a strong need for psychosocial belonging, involving voluntary and involuntary actions that can lead to harmful consequences, such as violent interpersonal behaviors and breaking rules, undermining the rights of others, based on a low capacity for adaptation to the social context, even resulting in criminal acts and alterations of mental health, such as antisocial personality disorder and psychopathy [5,6,7,8].

Psychopathy is characterized by impaired empathy, narcissism, lack of remorse, fear, and insensitivity to the adverse circumstances of others, with a dominant innate factor. Secondary psychopathic traits are defined by a lack of cognitive coping (management of stimulating affective information), impulsivity, irresponsibility, aggressiveness (hyperaccessibility to negative emotions), difficulty forming and maintaining interpersonal relationships, anxiety, few long-term goals, substance and alcohol abuse, risky and antisocial sexual behaviors, resulting from maladaptive factors in the environment and low-income family relationships, rejection and inadequate parental upbringing, and adverse social contexts such as continuous exposure to violence. The prevalence of these personality traits ranges from 1% to 3% of the world’s population; among adolescents, it is 2.68%, with a ratio of three males to one female [9,10,11,12,13].

Instruments developed to measure psychopathic traits in adolescents from a Latino school population are registered only in the United States. Horan et al.[14] conducted a study of invariance by ethnicity and sex in the LRSP model of primary and secondary psychopathic traits, finding adequate configural invariance, metric invariance, and concurrent and divergent validity. However, the model showed weak goodness-of-fit indices.

Other international research has validated the YPI scale [15] in Spanish and Portuguese samples from prison and school settings. In the first study, Ivanova-Serokhvostova et al. [16] reported a psychopathy factor structure comprising four dimensions (interpersonal, affective, behavioral, and antisocial) with adequate internal consistency and convergent and divergent validity. The second study observed invariance of the model across gender and prison and school settings, as well as convergent validity with a significant relationship between the instrument and symptoms of behavioral disturbance, drug and alcohol abuse, and risky sexual behavior [17]. However, both studies have significant methodological limitations; the first showed multicollinearity among factors, and the second lacked rigor in using correlation matrices and factor extraction for ordinal qualitative data.

Scientific literature on instrument validation for Latin American populations remains limited. Internationally, the Youth Psychopathic Traits Inventory (YPI) [15] is a reference instrument; however, it is lengthy (50 items) and lacks validation in Latin American populations. Furthermore, early assessment of psychopathic traits is fundamental to designing future preventive interventions [18]. Therefore, the purpose of this research was to adapt and validate the LSRP in a sample of Mexican School-attending mid- and late-adolescents, evaluating its psychometric properties in terms of construct validity, internal consistency, and convergent and divergent validity, as well as the invariance of the measurement model according to sex.

2. Materials and Methods

2.1. Participants

Through a cross-sectional study, a non-probabilistic sample of n =1010, early and late adolescents (female n = 643 and male n = 367) was evaluated with a psychopathic traits instrument, their age range was from to 14-19 years old (M = 17.35, SD =1.43 for sample 1 and, M = 17.34, SD = 1.40, years for sample 2), n = 505 participants for each sample were included in this analysis. All participants were literate (9 to 14 years of schooling) and came from urban and rural areas in Hidalgo and Mexico States, Mexico (Table 1).

2.2. Instrument

The Levenson Self-Report Psychopathic Scale (LRSP) was designed to assess psychopathic traits and has been culturally adapted and validated for the Mexican adult population in jail conditions [19]. The original instrument was developed to evaluate community and non-clinical samples [14] and consists of 26 items rated on a four-point Likert scale from Strongly Disagree to Strongly Agree. The original model comprises primary and secondary traits [20]. The Mexican adaptation includes three dimensions: egocentric (items 1, 2, 3, 4, 5, 6, 7, 8, 9, and 11; ω = 0.746), antisocial (items 15, 16, 17, 18, and 19; ω = 0.683), and affective factors (items 10, 12, 13, and 14; ω = 0.631).

2.3. Procedure

We collaborated with two educational institutions (in the States of Mexico and Hidalgo) to evaluate and collect information. The researchers obtained informed consent from adult adolescents and from the parents or guardians of minor adolescents and ultimately obtained informed assent from them.

The study adhered to the ethical principles and regulations of the Declaration of Helsinki [21] and the Mexican Health Research Law, Articles 16 and 17, thereby safeguarding participants’ privacy and confidentiality and classifying the study as low risk [22]. The study was part of a larger research project approved by the Investigation and Bioethics Committee of the Autonomous University of the State of Hidalgo (Protocol code: Uq/6V%#-yn). We reviewed the English translation using the Grammarly app.

2.4. Statistical Analysis

To establish the construct validity and dimensionality of the psychopathic instrument, we conducted a multivariate procedure that included an exploratory factor analysis (EFA) using the open-source stats FACTOR software (version 12.06.08) and an SEM analysis to confirm the theoretical model and assess the instrument’s internal consistency JAMOVI (R Integration) software was used.

For the EFA on the first sample, we conducted a sample adequacy analysis using the Kaiser-Meyer-Olkin test and Bartlett’s test of sphericity to assess whether the correlation matrix deviated from an identity matrix. To define, which factorial extraction method to use, we get and evaluate the values for each item with the asymmetry and kurtosis indexes (< 1) and all items together with Mardia’s test, this analysis suggested a non-linear model using a polychoric matrix and the factorial extraction method of robust diagonally weighted least squares (DWLS) recommended for non-normal distribution and ordinal data like Likert-type scales, finally we use Promin rotation for factor simplicity extraction [23,24].

Likewise, for EFA, we conduct a parallel analysis using minimum-rank factor analysis to reduce data and determine the number of dimensions in the psychopathic test. To decide which data model fits well with the hypothesized model, as recommended by Hu and Bentler [25] multiple fit indices were used, considering, robust mean and variance adjusted chi-square statistic, approximate fit indexes like root mean square error of approximation (RMSEA [<0.06-0.08]), goodness of fit index (absolute fit index GFI [≥0.95]), comparative fit index (incremental fit index CFI [≥0.95]), and the standardized root mean square residual (the mean absolute correlation residual SRMR[<0.08]) [26].

For the SEM model, we used the robust weighted least squares method (WLSMV), considering robust adjusted goodness indexes (chi-square test, SRMR, RMSEA, CFI, and TLI indexes); the justification for the use of this method was because of the nature of the items as ordinal variables, lack of multivariate normality, and size of the second sample. For the invariance model, we conducted a multigroup analysis to assess configural, metric, and scalar invariance, using the criteria of Cheung & Rensvold [27] and Khademi et al. [28] (ΔCFI ≤ -0.010 and ΔRMSEA ≤ 0.015). Finally, to test convergent and divergent validity, we assess composite reliability (CR), Average Variance Extracted (AVE), Maximum Shared Variance (MSV), and Average Shared Variance (ASV) [29,30].

3. Results

To ensure the equivalence of the subsamples used in the EFA (n = 505) and SEM (n = 505) analyses, we used the Solomon method to divide the data into equivalent representative subsamples (communality ratio = 0.981; [31]). Regarding sampling adequacy and the comparison of the observed covariance matrix with the identity matrix, the KMO coefficient was 0.851, and the Bartlett index (136) = 2518.1, p = 0.001, indicating that the data were adequate for exploratory factor analysis.

Multinormality of the 19 items associated with the psychopathic traits construct was assessed, yielding a Mardia multivariate skewness index (Coeff= 22.45, skewness = 1893.81, p = 1.00; Coeff = 346.6, kurtosis = 10.45, p = 0.001), indicating a non-normal multivariate distribution and therefore the use of a polychoric correlation matrix. To determine the number of factors, we used parallel low-rank approximation, which identified two underlying factors (F1 = 33.22 and F2 = 14.33) that explained 47.55% of the variance in the psychopathic traits construct. Likewise, items 4 and 10 (MSA = 0.629 and 0.667) were excluded because their sampling adequacy index (MSA) was < 0.70 [32] (Table 2).

We retain the 2-factor model excluding items 4 and 10; however, the goodness-of-fit indices suggested a better fit for the 3-factor model and the 3-factor model without items 4 and 10. This is because, in the 3-factor model, two items had overlapping factor loadings across two factors. Regarding the 3-factor model without items 4 and 10, only two items formed a dimension, failing to meet the theoretical assumption that at least three items must constitute a factor to be considered a dimension [23].

In the confirmatory phase with the second sample, we evaluated the 17-item model using the WLSMV estimator. Based on the factor loadings in this second sample, 3 items were identified and omitted because their loadings fell below 0.4.

After removing the items, the initial model fit was suboptimal because of significant residual covariance. To correct this, a sequential analysis of the modification indices (MI) and the expected change in the standardized parameter (sEPC all) was performed. This procedure revealed three inverse-order residual covariances with respect to items p2 and p3 (items associated with Machiavellian behavior [MI=49.7, sEPC =-0.493]), p8 and p11 (items associated with interpersonal manipulation MI=18.6, sEPC =- 0.435]), and p16 and p17 (items related to boredom propensity MI=22.6, sEPC =- 0.437]), due to their semantic redundancy. Resulting in a reweighted model with significant model fit indices χ²(73) = 104, p=0.009; SRMR robust = 0.042; RMSEA robust = 0.045, p =0.714 (95% CI=0.029-0.059); CFI robust =0.960 and TLI robust =0.951 (Figure 1 and Table 3).

To assess scale invariance, we conducted a multigroup confirmatory factor analysis across sex. The configural model (model 1) yielded a similar factor structure for females and males, with χ²(146) = 186, p = 0.015, CFI = 0.919, RMSEA = 0.061 (p = 0.126), SRMR = 0.054. For metric invariance (model 2), which controlled factor loadings, an adequate fit was observed (χ²(251) = 507, CFI = 0.938, TLI = 0.933, RMSEA = 0.064, SRMR = 0.085). Scalar invariance (model 3), which controlled for factor loadings, intercepts, and thresholds, also indicated a good overall fit (χ²(283) = 529, CFI = 0.941). According to Cheung et al. [27] and Khademi et al. [28], the changes in the comparative fit indices fall within the normative criteria (ΔCFI and ΔTLI = 0.01) (Table 4).

After confirming the scalar invariance of the instrument, a comparison was made between the latent means of the groups (men and women) in each of the factors, obtaining a significant difference in the latent mean for the second factor (SPT) (M = 0.106, z = 1.993, p = 0.046, 95% CI [0.002-0.210]), determining that women present higher levels in the SPT factor compared to men. With respect to the first PPT factor, no differences were observed between the sexes (M = 0.091, z = 1.456, p = 0.145).

With respect to convergent validity, the first factor, representing PPT, showed adequate composite reliability scores (CR= 0.83), exceeding the minimum criterion established by Fornell and Larcker [30], which compensates for the low average variance extracted (AVE=0.3). In contrast, the SPT factor exhibited lower composite reliability (CR = 0.64) and AVE (0.27), affecting convergent validity. Nevertheless, discriminant validity was supported because the AVE exceeded both the maximum shared variance (MSV) and the average shared variance (ASV). Table 5 presents detailed values for convergent and discriminant validity. Regarding internal consistency, the ordinal reliability coefficients were α Ordinal = 0.838 for the PPT factor and α = 0.673 for the SPT factor.

4. Discussion

The objective of this research was fulfilled by determining the psychometric properties of the psychopathic traits instrument of Levenson et al. [20], considering the version of the cultural adaptation in Mexico by Amador-Zavala et al. [19] in adolescents from rural and urban areas of Mexico, observing a bifactorial structure with adequate goodness of fit indices and acceptable internal consistency for the two factors of secondary and primary psychopathic traits (F1/ α Ordinal=0.838 and F2/ α Ordinal= 0.673), likewise, adequate convergent validity for one factor (CR = 0.83); and reduced for another (CR = 0.64), however, adequate discriminant validity in both factors.

Horan et al. [14] conducted a study similar to the present one, in which they evaluated psychopathic traits in Hispanic and Black adolescents with the LSRP test and a self-report of callousness and lack of empathy answered by parents in New York public schools, considering configural and metric invariance. The findings observed in the multigroup analysis were similar to this study, determining an equivalent invariance by sex and ethnicity (black and Latino adolescents), however, a reduced CFI index in both variables (Sex χ²/df = 1.81, CFI = 0.71, RMSEA = 0.05; Ethnicity= χ²/df = 1.77, CFI = 0.69, RMSEA = .05), with a difference in metric invariance within the critical limits (ΔCFI = 0.001 and 0.003), indicating that the factor loadings were similar between the groups with respect to sex and ethnicity, compared to our study, in the present a CFI goodness of fit index of invariance above 0.90 was obtained, adequate according to the associated literature [34].

In a similar study, Ivanova- Serokhvostova et al. [16] validated and adapted the Psychopathy Checklist, Youth Version, in a Spanish population (15-22 years old) in prison and on parole, determining a four-dimensional factor structure (interpersonal, affective, behavioral, and antisocial), showing model invariance in both conditions. The main limitation of that study was the questionable large magnitude of the covariances between the dimensions and the general construct of psychopathy (CoV = 0.83-0.96), due to the lack of a clear division between factors, indicating multicollinearity and redundancy among them, and therefore a lack of discriminant validity [29]. In the present study, the PPT and SPT factors were clearly differentiated, with a covariance of 0.44 and adequate discriminant validity.

Similarly, Pechorro et al. [17] adapted the YPI test for adolescents of Portuguese origin in public school settings and juvenile detention centers, determining an equivalence in the invariance of the model according to sex (S-Bχ²(74) = 262.5374, CFI= 0.98, RMSEA= 0.08) and the schooled vs. detention center condition (S-Bχ²(74)= 181.12, CFI= 0.98, RMSEA= 0.07). However, methodological limitations were observed, such as the lack of mention of the use of a correlation matrix and factor extraction, which was not reported but inferred by the version of the program used, which does not consider analyses for qualitative and ordinal data, unlike this study, which rigorously considered the analyses relevant to the type of sample obtained.

Regarding the results of the latent mean in this study, it was observed that adolescent women (mainly from the southern area of the country where the economy is precarious) score higher in the dimension of secondary psychopathic traits, similar to what was found by Neumann et al. [35] in a study with a large international sample of adult population (n= 33,016), observing that women in North America (including Mexico), Oceania and Western Europe present high levels of psychopathic traits associated with the Lifestyle dimension, characteristic of the SPT factor.

Regarding the study’s limitations, the instrument showed low average extracted variance for both factors (AVE < 0.50). However, the PPT factor showed adequate composite reliability (CR = 0.83), and the SPT factor showed acceptable reliability (CR = 0.64). According to Malhotra [36], a CR between 0.6 and 0.7 is acceptable if the model estimation is good. He also indicates that the CR coefficient itself is a more reliable index than the AVE. Finally, the satisfactory indices across the three levels of invariance support the model structure, given the factor loadings and intercepts.

5. Conclusions

In Mexico and Latin America, there is an increase in the prevalence of adolescents who commit violent acts, exhibit behavioral and delinquent behavior, and engage in bullying, primarily at the upper secondary and higher education levels, considering adverse contexts and social competition within the school environment. Early identification of psychopathic traits at this developmental stage will be a valuable tool for identifying these traits and creating future primary interventions, preventing their evolution into a mental disorder. This tool is also useful in research to deepen our understanding of psychopathic traits, their interactions with culture, and their impact on society [1,37].