Psychometric Properties and Measurement Invariance Across Gender and Age‑Group of the Perseverative Thinking Questionnaire–Children (PTQ‑C) in Colombia

---

Context and Objectives

The Perseverative Thinking Questionnaire-Children (PTQ-C) is a content-independent transdiagnostic measure of repetitive negative thinking and perseverative thinking patterns in child and adolescent populations. Despite its theoretical and clinical relevance in Anglophone contexts, the scale had been validated only in its original English version, limiting applicability in Spanish-speaking contexts. Perseverative thinking represents a transdiagnostic construct associated with multiple mental health conditions, including anxiety, depression, stress, and rumination-spectrum disorders, with particular developmental significance during childhood and adolescence.

This study aimed to translate and culturally adapt the PTQ-C into Spanish, comprehensively examining its psychometric properties in a large, representative sample of Colombian children and adolescents (N = 1,127, ages 8–18 years). Specific objectives included: (a) establishing internal consistency via Cronbach's alpha and McDonald's omega with 95% confidence intervals; (b) examining factor structure through exploratory (EFA) and confirmatory (CFA) factor analyses with a cross-validation strategy; (c) evaluating measurement invariance across gender and age group to demonstrate metric and scalar equivalence; and (d) determining convergent validity by correlating PTQ-C with measures of rumination, worry, depression, anxiety, stress, and experiential avoidance.

---

Method

Participants

The study included 1,127 Colombian children and adolescents (57% female, M = 11.11 years, SD = 2.73, age range 8–18 years) recruited from 9 educational institutions (4 private, 5 public) located in the Bogotá metropolitan area, Colombia. The sample represented a mixed population from diverse socioeconomic backgrounds with balanced representation of public and private schools. Participants were included if they were 8–18 years old and attended the selected schools regularly. No specific clinical exclusion criteria were applied, permitting a representative sample of the general child-adolescent population.

Design

The study employed a cross-sectional psychometric validation design with a rigorous cross-validation strategy. The total sample was randomly divided into two independent subsamples: Subsample 1 (N = 525) designated for exploratory factor analysis (EFA) and dimensionality assessment, and Subsample 2 (N = 554) dedicated to confirmatory factor analysis (CFA) of the proposed models. This methodological approach enabled both discovery and refinement of factor structure and independent confirmation thereof, strengthening validity of findings and preventing overfitting bias.

Instruments

PTQ-C (Perseverative Thinking Questionnaire-Children). A content-independent 15-item instrument on a 5-point Likert scale (0–4) designed to measure perseverative thinking regardless of content specificity. This constituted the focal instrument of the study.

PSWQ-C (Penn State Worry Questionnaire-Children). A 14-item measure, from which 3 reverse-scored items were excluded, resulting in 11 items utilized. Assessed worry specifically as a related but distinct construct from perseverative thinking.

DASS-C/DASS-21 (Depression, Anxiety, Stress Scale). For participants under 13 years, the 24-item DASS-C was administered; for participants 13 years or older, the 21-item DASS-21 was used. These scales assessed depression, anxiety, and stress symptoms as convergent validity constructs.

AFQ-Y (Avoidance and Fusion Questionnaire-Youth). A 17-item instrument measuring experiential avoidance and cognitive fusion, transdiagnostic processes related to perseverative thinking from an acceptance and commitment therapy perspective.

Statistical Analyses

Subsample 1 (N = 525) – Exploratory Factor Analysis: EFA was conducted using robust DWLS (Diagonally Weighted Least Squares) estimation with oblimin rotation. Unidimensionality was evaluated via three complementary indices: UniCo (unique common variance correlation), ECV (Explained Common Variance), and MIREAL indices to ensure data reflected a fundamentally unidimensional construct. Item discrimination was examined through corrected item-total correlations.

Subsample 2 (N = 554) – Confirmatory Factor Analysis: CFA was conducted using robust DWLS estimation to test two competing models: (1) a simple one-factor model and (2) a three-factor higher-order hierarchical model. Multiple fit indices were evaluated: Satorra-Bentler χ², RMSEA (Root Mean Square Error of Approximation) with 90% confidence intervals, CFI (Comparative Fit Index), NNFI (Non-Normed Fit Index), SRMR (Standardized Root Mean Square Residual), ECVI (Expected Cross-Validation Index), and PNFI (Parsimonious Normed Fit Index).

Full Sample (N = 1,079) – Measurement Invariance: Metric and scalar invariance of the PTQ-C was examined across gender (boys N = 462, girls N = 617) and age group (8–12 years N = 664, 13–18 years N = 284). Changes in fit indices (ΔRMSEA < .01, ΔCFI < .01, ΔNNFI < .01) served as evidence criteria for invariance.

Reliability: Cronbach's alpha and McDonald's omega were calculated with 95% confidence intervals via bootstrap procedures to establish internal consistency.

Interaction effects: Two-way ANOVA (gender × age group) was conducted to examine differential patterns in PTQ-C scores.

Convergent validity: Pearson correlations were calculated between PTQ-C and all criterion measures (PSWQ-C, DASS-C/DASS-21, AFQ-Y).

---

Results

Item Quality and Factor Structure

All 15 PTQ-C items demonstrated adequate corrected item-total correlations, ranging from .62 to .73, indicating satisfactory item discrimination. In the EFA, factor loadings were substantial, varying from .70 to .81, evidencing meaningful contributions of all items to the latent construct.

The EFA extracted a unidimensional solution explaining 60.5% of common variance. Unidimensionality indices strongly supported the factor structure: UniCo = .99 (values approaching 1.0 indicate strong unidimensionality), ECV = .94 (exceeding the .85 threshold), and MIREAL = .16 (near-zero values indicate congruence with unidimensionality). Although a three-factor higher-order model showed marginally superior fit, the three latent factors exhibited very high correlations (r = .86–.94), providing statistical and conceptual justification for retaining the more parsimonious one-factor model.

Confirmatory Factor Analysis and Model Fit

In Subsample 2, the one-factor PTQ-C model demonstrated excellent fit: χ²(90) = 194.70, p < .05; RMSEA = .046 [90% CI: .037, .055]; CFI = .99; NNFI = .99; SRMR = .038; ECVI = .46; PNFI = .85. All fit indices exceeded standard criteria (RMSEA < .06, CFI/NNFI > .95, SRMR < .08), confirming model adequacy.

Reliability

Internal consistency was excellent: Cronbach's alpha = .93 [95% CI: .92, .93] and McDonald's omega = .93 [95% CI: .92, .94]. Both coefficients substantially exceeded the .70 threshold typically required and approached the .90+ standard suggested for research instruments, indicating that the 15 items correlate strongly and mediate a single underlying dimension.

Measurement Invariance

The PTQ-C demonstrated scalar invariance (the most restrictive level) across both gender and age group. Changes in RMSEA, CFI, and NNFI between nested models were minimal (ΔRMSEA < .01, ΔCFI < .01, ΔNNFI < .01), providing evidence that measurement structure remains equivalent between boys and girls and between age groups (8–12 and 13–18 years). This finding is critical: it indicates that mean comparisons between these groups are valid and that observed differences reflect genuine latent construct differences rather than measurement artifacts.

Gender and Age Effects

A two-way ANOVA revealed a significant gender × age interaction: F = 8.22, p = .004, η² = .008. Specifically, older boys (13–18 years) showed significantly lower PTQ-C scores than younger boys (8–12 years), while older girls showed an opposite pattern with higher scores. This pattern replicates prior literature findings and suggests differentiated developmental trajectories of perseverative thinking by gender during adolescence.

Convergent Validity

The PTQ-C showed very strong correlations with conceptually related measures. The strongest correlation was with PSWQ-C (r = .76), confirming the theoretical relationship between perseverative thinking and worry. Equally robust correlations emerged with AFQ-Y (r = .75), depression (DASS-21: r = .74; DASS-C: r = .58), anxiety (DASS-C/21: r = .62–.65), and stress (DASS-C/21: r = .59–.67). These correlation patterns support convergent validity and demonstrate that the PTQ-C effectively measures a transdiagnostic construct of relevance to multiple mental health conditions.

---

Discussion and Conclusions

This study represents the first comprehensive validation of the PTQ-C in Spanish, demonstrating that the Spanish translation maintains the robust psychometric properties of the original instrument. Findings converge systematically on evidence of strong unidimensionality, confirmed factor structure, excellent reliability, and substantial convergent validity. Particularly noteworthy was the demonstration of scalar invariance, validating PTQ-C use for cross-gender and age-group comparisons in Spanish-speaking contexts.

Findings of gender-differentiated perseverative thinking by age (older boys with lower scores; older girls with higher scores) replicate patterns observed in the literature and suggest that rumination and perseverative thinking follow gender-differentiated developmental trajectories during adolescent transition, possibly related to pubertal changes, emerging gender roles, and changes in emotion regulation.

Principal conclusions:

1. The Spanish PTQ-C possesses psychometric properties equivalent to the original version, justifying its use in Spanish-speaking research and clinical practice.
2. The robust unidimensional structure permits a single total score capturing the transdiagnostic construct of perseverative thinking.
3. Measurement invariance enables valid comparisons across demographic groups.
4. Convergent validity correlations confirm the instrument's relevance across multiple clinical domains.

---

Significance and contribution

This study contributes to the field of child psychological assessment by demonstrating that the PTQ-C maintains robust psychometric properties in Spanish-speaking contexts. The instrument constitutes a valid and reliable measure of perseverative thinking in children and adolescents, with clearly unidimensional structure and confirmed measurement invariance across gender and age groups, enabling valid score comparisons across demographic subgroups in research and clinical practice settings with Spanish-speaking populations.

---

Methodological Rigor Checklist

• Adequate sample size. N = 1,127 participants, sufficient for EFA (≥300) and CFA (≥500) in independent subsamples.
• Cross-validation strategy. Random division into two subsamples for separate EFA and CFA, preventing overfitting.
• Multiple fit indices. RMSEA, CFI, NNFI, SRMR, ECVI reported; all exceeded standard criteria.
• Dimensionality assessment. UniCo, ECV, MIREAL employed to confirm unidimensionality beyond standard factor analysis.
• Reliability with confidence intervals. Alpha and omega reported with 95% CIs via bootstrap.
• Formal measurement invariance. Metric and scalar examined across gender and age with index-change criteria.
• Multifaceted convergent validity. Correlations with multiple related constructs (worry, depression, anxiety, stress, avoidance).
• Interaction analysis. Two-way ANOVA examined gender × age effects on PTQ-C scores.
• Clear participant description. Detailed demographics provided; non-clinical population in general educational sample.
• Appropriate statistical methods. Robust DWLS employed for EFA/CFA with ordinal (5-point Likert) data.

Recognized limitations:

• Self-report measures. Susceptibility to social desirability response bias.
• Lack of prior formal Spanish validation of some criterion instruments. Potential impact on convergent validity.
• Absence of clinical participants. Generalizability limited to populations without diagnosed disorders.
• Unbalanced questionnaire administration order. Possible order effects on convergent validity correlations.
• Cross-sectional design. Inability to establish predictive or causal relationships.