This study was carried out in two phases: phase I involved the translation of the instrument, and phase II the validity and reliability assessment.

Instrument: Pain catastrophizing was assessed by means of the PCS-C. This is a 13-item self-report scale developed by Crombez et al.9 to assess catastrophic beliefs associated with the experience of pain in children. The PCS-C assesses 3 dimensions (magnification, rumination, and helplessness) with a 5-point response scale (“not at all”, “mildly”, “moderately”, “severely” and “extremely”). Higher scores indicate stronger catastrophic beliefs about pain (the possible score ranges from 0 to 52). The English and Catalan versions of the PCS-C have proven to be reliable and valid in community and clinical samples of children and adolescents.9,13,14

Translation process: The PCS-C scale was translated from English, which was the original language of the scale,9 into Spanish following the translation and back-translation process. Two bilingual pain research experts translated the scale from English to Spanish independently. A professional bilingual translator resolved the discrepancies between the 2 versions of the pain researchers. Later, two bilingual natives back translated the Spanish version into English, and a professional bilingual translator compared this version with the initial version proposed by Crombez et al.9 The response options for intensity rating (0=not at all, 1=mildly, 2=moderately, 3=severely and 4=extremely) were replaced by frequency options (0=never, 1=rarely, 2=sometimes, 3=often and 4=always) because it was observed that this alternative improved comprehension in Spanish, as had been the case in the Catalan version.13 The Spanish version of the PCS-C can be found in Supplementary Document I.

Ethical Considerations: The study was approved by the ethics committee of the hospital and adhered to the principles of the Declaration of Helsinki Declaration. It was conducted following the Standards for Reporting Diagnostic Accuracy Studies (STARD).17

The convenience sample included children and adolescents. The inclusion criteria were (a) ability to read and write in Spanish by both child and parents; and (b) age between 8 and 18 years (c) paediatric surgical patient classified as American Society of Anesthesiologists (ASA)16 Physical Status 3 or lower, which excluded patients with severe illnesses or those requiring urgent surgical interventions. Children with an intellectual disability that could interfere with the implementation of the study protocol, children undergoing urgent surgery and oncological patients were excluded.

We extended an invitation to participate in the study to children scheduled for surgery and their parents upon hospital admission. If the patients were interested in participating, a 45-minute interview was held with the parents, in which they were informed in detail about the study, completed the informed consent and filled out the print questionnaires. In order to participate, in addition to parental consent, children aged less than 12 years had to provide their assent, and adolescent participants their consent. Then, the children were interviewed apart from their parents so that their presence would not influence the response to the questionnaires (also in print format); this interview also lasted 30min. All interviews took place in the department of surgery of the hospital.

To carry out the validation, we assessed demographic characteristics and the following variables: the child’s previous surgeries, number of medical visits in the last year, pain intensity, assessed by the child (Faces Pain Scale-Revised [FPS-R]18) and by the parents (Parent’s Postoperative Pain Measure [PPPM]19), health-related quality of (Pediatric Quality of Life Inventory [PedsQL] version 4.0 Generic Core Scale20), functional disability (visual numeric rating scale [NRS]18) and pain interference (Patient-Reported Outcomes Measurement Information System [PROMIS] Pain Interference form21). Information on the characteristics of the scales and their psychometric properties can be found in Supplementary Document II.

The data were analysed with the SPSS version 21 (IBM SPSS Statistics) and the Mplus version 7.11 statistical software packages. The level of statistical significance was set at 5% (P<.05).

A 2-step process was used to examine the validity of the construct: (1) exploratory factor analysis (EFA) to identify the optimal factor structure and (2) confirmatory factor analysis (CFA) to confirm the theoretical factor structure. Based on the criteria established by several authors, we estimated that a sample size of at least 100 children was needed to perform an adequate EFA.22,23 To confirm the theoretical factor structure, we increased the sample size by 50 children because in cases in which the data are well conditioned, samples of 50 participants can generate reliable results.22 Thus, the CFA was performed with a total of 150 observations.

We used the Bartlett test of sphericity and the Kaiser-Meyer-Olkin (KMO) to determine whether the Pearson correlation matrix was suitable for factor analysis.23 Following recommendations, we determined the optimal number of factors based on the Kaiser criterion (eigenvalue ≥ 1; scree plot).24 In the EFA, factors were extracted with the principal axis factoring with oblique rotation method. We established that for an item to be included in a factor, it needed to have a factor loading greater than 0.4.

The CFA was performed using the maximum likelihood with robust standard errors (MLR) estimation method. We used the following goodness of fit indices: chi squared (χ2), comparative fit index (CFI), Tucker Lewis index (TLI), standardized root mean square residual (SRMR) and root mean square error of approximation (RMSEA). We applied the Hu and Bentler criteria to determine whether the fit of the model was acceptable (TLI and CFI≥0.95, SRMR and RMSEA≤0.08).25 Furthermore, modification indices were calculated to identify local mis-specified areas of the model not sensitive to the overall goodness of fit indices previously mentioned.26

We assessed the reliability of the Spanish version of the PCS-C based on its internal consistency. The internal consistency would be considered acceptable if the Cronbach α was greater than 0.70.27 We defined floor or ceiling effect as at least 15% of patients achieving the minimum or maximum score, respectively.28

We assessed convergent validity based on the Pearson correlation coefficient obtained in the comparison of the Spanish version of the PCS-C with the following clinical measures aimed at assessing pain, functional disability and quality of life in children: medical consultations in the last year; previous surgeries; FPS-R; PPPM; PROMIS Pediatric Pain Interference score; NRS for Functional Disability; PedsQL. A coefficient greater than 0.60 was considered indicative of a strong correlation; a coefficient of 0.30–0.60 of a moderate correlation and a coefficient of less than 0.30 of a weak correlation.28

We calculated the Cronbach α for the entire scale (α=0.89) and for the adjusted item-total correlations (average item-total correlations=0.583) prior to the EFA. No items were removed, since all contributed substantially to the scale.

The KMO showed that the data was suitable for factor analysis (KMO=0.866), and the Barlett test of sphericity refused the identity matrix null hypothesis: χ278=592.41 (P<.001). According to these results, it was appropriate to proceed with the EFA. The Kaiser criterion indicated that the 3 factors should be retained (see Fig. 1). In addition, the goodness of fit indices showed that the 3-factor model fit the data well: CFI=0.980; TLI=0.963; SRMR=0.035; RMSEA=0.049; 95% CI: 0.001–0.088. Table 2 presents the factor loading of each of the items.

Figure 1.

Scree plot of Spanish version of the PCS-C.

(0.05MB).

The CFA corroborated that all items loaded on the assumed theoretical factor, showing an optimal factor loading (≥0.560). Thus, we decided to fit a 3-factor CFA model with the 13 items mentioned above. The final model fit the data well: χ245=66.48 (P=.325); CFI=0.993, TLI=0.992; RMSEA=0.022, 95% CI 0.001–0.055; SRMR=0.043. Fig. 2 shows the standardized factor loadings of the final 3-factor solution.

Figure 2.

Structural equation modelling for the final Spanish version of the Pain Catastrophizing Scale-Child version (PSC-C).

(0.18MB).

The final Spanish version of the PCS-C consisted of a total of 13 items. All items were worded as direct, affirmative statements and rated on a 5-point Likert scale (0–4). Therefore, the total score can range from 0 to 52 points. The internal consistency of the scale was excellent (Cronbach α=0.904; 95% CI, 0.880−0.925), with its 3 subscales showing an internal consistency of 0.800 or greater (helplessness, 0.810 [95% CI, 0.759−0.854]; magnification, 0.800 [95% CI, 0.737−0.850]; and rumination, 0.843 [95% CI, 0.797−0.880]). Only 8.7% of the children obtained the lowest possible score on the scale, and no child obtained the highest possible score. Thus, there was no evidence of floor or ceiling effects for the final Spanish version of the PCS-C.

Table 3 presents the data on the correlation of the Spanish version of the PCS-C with all the other measures. Pain interference in the child’s life was the variable most strongly associated with the total score of the PCS-C, with a significant, positive and moderate correlation between the two (r=0.400). Thus, children with high levels of pain interference were children with higher levels of pain catastrophizing. In addition, the PCS-C exhibited a weak correlation with the other pain-related measures in child reports (FPS-R for pain intensity, r=0.217) and parental reports (PPPM, r=0.254; and NRS for Functional Disability, r=−0.186). However, quality of life was only significantly related to the PCS-C in the child version of the PedsQL, with a weak-to-moderate correlation (r=0.303). In short, children with greater catastrophizing perceived more intense pain and a greater deterioration in their quality of life, whereas their parents observed greater changes in their behaviour in association with pain. We did not find a correlation with the number of surgeries or of previous medical consultations.

The study had some limitations. It was conducted on a convenience sample, which could affect the extrapolation of the results to a different group of children. However, the study was performed at a tertiary referral hospital in Spain, which is a paediatric care reference centre and therefore serves patients from every region in Spain. The study was conducted in Spain and its findings showed that the scale had very good psychometric properties, but it is unknown whether these favourable properties would hold if the scale were applied in other Spanish-speaking countries. Finally, the sample analysed in this study consisted exclusively of patients with scheduled surgeries. Due to the complexity of the evaluation and the development of the study, we were unable to evaluate the performance of the scale in other populations, such as children with cancer-related pain or undergoing urgent surgeries. This limitation could restrict the generalization of the results, which suggest that future studies should pursue this line of research.