Factor structure and psychometric properties of the Dysfunctional Attitude Scale – Revised in Colombian undergraduates (DAS-R)

Complete Reference

Ruiz, F. J., Suárez-Falcón, J. C., Barón-Rincón, D., Barrera, A., Martínez, A., & Peña, A. (2016). Factor structure and psychometric properties of the Dysfunctional Attitude Scale Revised in Colombian undergraduates. Revista Latinoamericana de Psicología, 48, 81-87. DOI: http://dx.doi.org/10.1016/j.rlp.2015.10.002

Theoretical Framework and Background

This study is grounded in Beck's cognitive theory of depression, which posits that dysfunctional schemas constitute the primary cognitive vulnerability for depressive episodes (Beck, Rush, Shaw, & Emery, 1979). The original Dysfunctional Attitude Scale (DAS), developed by Weissman and Beck (1978), was designed to measure these maladaptive cognitive schemas through 40 items. However, subsequent research revealed substantial inconsistencies in factor structure, with solutions varying between two and four factors depending on the sample studied.

A significant advancement emerged from the work of de Graaf, Roelofs, and Huibers (2009), who employed confirmatory factor analysis (CFA) with a large Dutch sample (N=8960) to develop a revised, shortened version of the instrument: the 17-item DAS-R. This version maintains two correlated factors: Perfectionism/Performance Evaluation and Dependency. Subsequently, Ruiz and colleagues (2015) replicated and extended these findings in Spanish population, identifying a more sophisticated hierarchical structure comprising two first-order factors plus a second-order general factor, demonstrating excellent psychometric properties.

Despite these advances in Europe, neither the original DAS nor the DAS-R had been formally validated in Latin American contexts, particularly in Colombia. This gap represented an important limitation for research and clinical practice in the region.

Study Objective

The primary purpose of this investigation was to examine the factor structure and psychometric properties of the DAS-R in a sample of Colombian undergraduate students, evaluating whether the hierarchical model identified in European studies would replicate in the Colombian sociocultural context.

Methodological Design

Participants and Demographic Characteristics: The sample consisted of 762 undergraduate students from four universities located in Bogotá, Colombia's capital. Participants ranged in age from 18 to 63 years (M=21.16, SD=3.76), reflecting a wide age range although predominantly young adults. Disciplinary composition was heterogeneous: 46% Psychology students, with the remainder from Law, Engineering, Mathematics, and Physics programs, ensuring diversity in academic training. Regarding gender, the sample included 62% women and 38% men.

Regarding history of psychological treatment, 26% of participants reported having received some form of psychological or psychiatric intervention at some point in their lives, while only 4.3% were in active treatment at the time of the study, and 2.9% reported using psychotropic medication. This information is relevant because the sample is eminently nonclinical, which constitutes an important limitation to consider in generalizing results.

Assessment Instruments: The evaluation protocol included six self-report instruments:

1. Dysfunctional Attitude Scale Revised (DAS-R): A 17-item version derived from the original Weissman and Beck (1978) instrument, adapted to Spanish by Ruiz et al. (2015). A 7-point Likert scale was used (where 7 represents "fully agree" and 1 "fully disagree"). Two factors: Perfectionism/Performance Evaluation (11 items, α=.87) and Dependency (6 items, α=.81). Total alpha in the Spanish version α=.90.

2. Automatic Thought Questionnaire (ATQ-8): A shortened 8-item version (Netemeyer et al., 2002; Spanish adaptation by Cano-García & Rodríguez-Franco, 2002) with a 5-point Likert scale, used to assess frequency of negative automatic thoughts. Demonstrated excellent internal consistency (α=.85).

3. Acceptance and Action Questionnaire (AAQ-II): A 7-item instrument (Bond et al., 2011; Spanish version by Ruiz et al., 2013) with a 7-point Likert scale, designed to measure psychological inflexibility and experiential avoidance. Showed adequate reliability (α=.88).

4. General Health Questionnaire (GHQ-12): An abbreviated 12-item version (Goldberg & Williams, 1988; Spanish adaptation by Rocha et al., 2011) with a 4-point scale, used as an indicator of general psychological distress and mental health. Internal consistency α=.88.

5. Depression Anxiety Stress Scale (DASS-21): A 21-item instrument (Antony et al., 1998; Spanish version by Daza et al., 2002) with a 4-point scale evaluating three dimensions: Depression (α=.86), Anxiety (α=.80), and Stress (α=.80).

6. Satisfaction with Life Scale (SWLS): A 5-item instrument (Diener et al., 1985; Spanish adaptation by Atienza et al., 2000) with a 7-point Likert scale to assess overall life satisfaction. Demonstrated adequate internal consistency (α=.85).

Procedure: The study was preceded by a pilot test with 10 Colombian students to confirm adequate comprehension of the DAS-R items. Instruments were administered collectively in classroom settings, with prior informed consent obtained from all participants. Following completion, debriefing was conducted with students.

Data Analysis: Confirmatory factor analysis (CFA) was conducted using LISREL 8.71 software with unweighted least squares estimation. Three alternative models were compared: (a) unidimensional model, (b) bifactorial model with correlated factors, and (c) hierarchical model with two first-order factors and a second-order general factor. Evaluated fit indices included RMSEA (root mean square error of approximation), CFI (comparative fit index), NNFI (Tucker-Lewis non-normed fit index), and ECVI (expected cross-validation index).

Differences between models were evaluated using chi-square difference tests. The Schmid-Leiman transformation (Schmid & Leiman, 1957) was applied to determine the variance explained by the general factor, utilizing Factor 9.2 software with direct oblimin rotation. Remaining analyses were conducted in SPSS 19 and included: internal consistency via Cronbach's alpha with 95% confidence intervals, corrected item-total correlations, descriptive statistics, discriminant validity via Student's t-test with GHQ-12 cutoff ≥12, and convergent-divergent validity via Pearson correlations.

Results

Factor Structure: The comparative analysis of the three models revealed that the hierarchical model provided the best fit to the data (Table 1). Specifically:

• Unidimensional model: RMSEA=.091[95% CI .085-.096], CFI=.97, NNFI=.96, ECVI=1.23[95% CI 1.11-1.35], χ²(119)=864.68
• Bifactorial model with correlated factors: RMSEA=.073[95% CI .067-.079], CFI=.98, NNFI=.98, ECVI=.88[95% CI .78-.98], χ²(118)=599.17
• Hierarchical model (two factors + general factor): RMSEA=.059[95% CI .053-.065], CFI=.99, NNFI=.98, ECVI=.65[95% CI .58-.74], χ²(117)=425.64

The chi-square difference between the bifactorial and hierarchical models was 173.53 (df=1, p<.05), indicating that the hierarchical model provided significantly superior fit. Both first-order factors showed a substantial correlation (r=.83), which theoretically justifies the existence of a second-order general factor.

The Schmid-Leiman transformation revealed that the general factor explained 72.4% of extracted variance, a figure substantially exceeding the 40-50% thresholds generally used to establish the presence of a significant general factor. In contrast, the two first-order factors explained 22.5% (Perfectionism/Performance Evaluation) and 5.1% (Dependency) of extracted variance. Factor loadings on the general factor were high for both first-order factors: .91 for Perfectionism and .92 for Dependency.

Individual item factor loadings (Table 2) ranged from .43 (item 6) to .82 (item 12), with all items showing loadings on the general factor exceeding .30, indicating significant contributions of all items to the measured construct.

Internal Consistency: The total DAS-R demonstrated excellent internal consistency (Table 3): α=.91[95% CI .90-.92], with mean M=38.27 and standard deviation SD=16.55. Regarding specific factors:

• Perfectionism/Performance Evaluation: α=.87[95% CI .86-.89], M=24.53(SD=11.15)
• Dependency: α=.81[95% CI .79-.83], M=13.73(SD=6.69)

Corrected item-total correlations fell within adequate ranges: .46-.66 for the total scale, .49-.68 for the Perfectionism factor, and .42-.65 for the Dependency factor, indicating that each item contributes significantly to the global measure without redundancy.

Discriminant Validity: The ability of the DAS-R to discriminate between participants with and without general psychological distress was evaluated (Table 4). Participants with GHQ scores ≥12 (indicating probable psychological distress) obtained significantly higher DAS-R total scores (M=43.99, SD=18.48) compared to those with GHQ<12 (M=34.52, SD=13.99), t=7.55, p<.001. The same pattern was observed for both factors: Perfectionism (t=7.28, p<.001) and Dependency (t=6.57, p<.001), demonstrating that the instrument is sensitive to general psychological distress.

Convergent and Divergent Validity: Correlation analysis (Table 5, all p<.001) confirmed theoretically expected patterns:

The DAS-R total correlated positively with indicators of distress and dysfunction: GHQ-12 (r=.31), DASS Depression (r=.42), DASS Anxiety (r=.34), DASS Stress (r=.31), Negative Automatic Thoughts ATQ-8 (r=.43), and Psychological Inflexibility AAQ-II (r=.42). It correlated negatively with Life Satisfaction SWLS (r=-.26), consistent with theory predicting that dysfunctional attitudes constitute risk factors for well-being.

Specific factors demonstrated similar validity patterns although with slightly different magnitudes: Perfectionism showed correlations of .29 with GHQ-12, .41 with DASS-Depression, .33 with DASS-Anxiety, .29 with DASS-Stress, .41 with ATQ-8, .41 with AAQ-II, and -.25 with SWLS. The Dependency factor presented comparable magnitudes: .29 with GHQ-12, .36 with DASS-Depression, .29 with DASS-Anxiety, .28 with DASS-Stress, .39 with ATQ-8, .37 with AAQ-II, and -.24 with SWLS.

Discussion and Implications

This study successfully replicated the hierarchical structure model previously identified by Ruiz et al. (2015) in Spanish population, extending its validity to the Colombian context. The statistical superiority of the hierarchical model over unidimensional and bifactorial alternatives suggests that dysfunctional attitudes operate both through specific dimensions (Perfectionism and Dependency) and through an underlying general factor that might represent a more fundamental latent construct of cognitive vulnerability.

The excellent internal consistency (α=.91) of the total DAS-R, along with adequate values for specific factors, confirms that the instrument measures internally coherent constructs. Convergent validity is supported through substantial positive correlations with measures of depression, anxiety, stress, negative automatic thoughts, and psychological inflexibility, supporting the instrument's validity as a measure of theoretically relevant dysfunctional attitudes. Divergent validity is evidenced through negative correlation with life satisfaction and discriminant validity in distinguishing between individuals with and without general psychological distress.

The existence of both a global score and specific factor scores provides clinical and research flexibility, allowing assessment of both general cognitive vulnerability and specific dimensions of dysfunction (perfectionism versus interpersonal dependency).

Limitations

The study presents several limitations that should be considered in interpreting results: (1) the sample consisted exclusively of nonclinical university students, limiting generalization to clinical populations or non-students; (2) among participants reporting previous psychological treatment, specific diagnostic information was unavailable; (3) reliance on self-report measures could introduce common method bias; (4) the relatively narrow age range (predominantly young adults) limits evaluation of factorial invariance by age; (5) prospective predictive validity for development of depressive disorders was not evaluated.

Recommendations for Future Research

It is suggested to conduct validation studies with clinically diagnosed depressive disorder samples, samples broader in terms of age range, and participants with varying educational levels. Additionally, prospective predictive validity studies, evaluation of factorial invariance by gender and age, and examination of the clinical utility of the instrument in cognitive-behavioral intervention contexts would be valuable.

---