Development of the Spanish version of the Parents' Self-Stigma Scale

Abstract

Background and Objectives

It is known that those who regularly interact with people with mental illness, such as family members, caregivers and mental health professionals, can be subjected to public stigmatization and that they may eventually develop self-stigmatization. Despite the relevance of parental self-stigma for the upbringing and treatment of children with psychiatric problems, only one instrument has been developed to identify it, the Parents' Self-Stigma Scale (PSSS). The lack of a similar instrument in Spanish motivated the present study, with the aim of developing a Spanish version of the PSSS.

Methods

After translating the PSSS, it was administered to two samples of parents of children who were treated consecutively in child-adolescent mental health centres in Reus and Valencia. The Reus sample was subjected to Exploratory Factor Analysis (EFA) and the Valencia sample to Confirmatory Factor Analysis (CFA), taking as reference the factor load matrix obtained in Reus. Temporal stability was estimated by calculating the ICC between the results obtained in two administrations of the questionnaire separated by four weeks. To estimate convergent validity, the correlation of the questionnaire score with the scores of the Rosenberg Self-Esteem Scale and the Schwarzer General Self-Efficacy Scale was calculated.

Results

The EFA showed the existence of three factors, “Bad Father”, “Self-blame” and “Self-shame”, which confirms the structural equivalence of the Spanish version and the original PSSS. Likewise, it was confirmed that the Spanish version is temporarily stable and valid.

Conclusions

The results obtained show that the Spanish version of the PSSS is semantically and psychometrically equivalent to the original PSSS, and that it has acceptable temporal stability and convergent validity.

Introduction

It is known that those who regularly interact with people with mental illness, such as family members, caregivers and mental health professionals, can be subjected to public stigmatization.¹ Parents of children with mental health problems may be blamed for their children´s problems. This form of stigmatization, if assumed and internalized by the parents, may lead to so called parental self-stigma.² Parental low self-esteem and low perceived self-efficacy,^2,3 as well as cultural factors,^4,5 can help facilitate this process.

Parental self-stigma is associated with high levels of anxiety and depression, sadness, embarrassment, shame, guilt and fear, that it may require professional help⁴ It affects the parent-child relationship, increases aggressive behaviour in children and deteriorates their social skills, contributing to social isolation⁶ and reducing community participation.⁷ In relation to treatment, feelings of shame and guilt often inhibit help-seeking behaviour^8,9 and hinder parents' willingness to actively collaborate with the therapist.¹⁰

In view of the extent and importance of the impact of self-stigma on the health and well-being of parents and their children, and particularly considering the impact of self-stigma on help-seeking and parental collaboration in treatment, it would be useful to have an instrument to identify self-stigmatized parents early in the treatment of their children to offer them help if necessary.

A systematic search in Web of Science and Scopus revealed the existence of six original questionnaires^10-15 but only one of them, the Parents' Self-Stigma Scale (PSSS), applies specifically to parents, excluding other family members. Furthermore, the content of the questionnaire fits in well with previous qualitative research regarding parental self-stigma.¹⁶

The PSSS¹⁴ is a self-administered questionnaire of 11 items scored on a Likert scale of frequency ranging from ‘never’ (1) to ‘almost all the time’ (5). To construct the questionnaire, a number of drafted items were initially obtained from a group of parents affected by self-stigma and later developed according to Participatory Action Research methodology. The questionnaire was validated using an Australian community sample of 424 parents of children aged 4–12, obtained through advertisements and online surveys using the Qualtrics platform. To explore the factor structure of the PSSS, the data were randomly split into two subsets. An Exploratory Factor Analysis (EFA) was performed using the first subset and the Partial Confirmatory Factor Analysis (PCFA) was performed using the second subset. As mentioned before, EFA revealed the existence of three factors: Self-Blame, Self-Shame, and Bad Parent, which were confirmed by PCFA in the second subset. Convergent validity was calculated as the correlation between the PSSS scores and two scales that measured self-esteem and empowerment.

Currently, there is no equivalent instrument to the PSSS available in Spanish. The availability of such an instrument would not only increase the understanding and awareness of parental self-stigma and foster research in this area but also provide the opportunity to use it in routine clinical work in child mental health centres to identify parents with self-stigma and provide them with help as necessary. Furthermore, such an instrument would facilitate the development and evaluation of therapeutic interventions when used in clinical trials. Consequently, the aim of the present study is to develop an instrument equivalent to PSSS by translating it into Spanish and, similarly to Eaton’ study, estimating its factor structure, internal consistency and convergent validity. In addition, its temporal stability will be estimated.

Method

Translation and adaptation of PSSS-S

After obtaining authorization from the scale's original author, one bilingual translator translated the questionnaire into Spanish and another, without knowledge of the original questionnaire, translated the Spanish version into English. The back-translated version was then sent to the original author for approval. Once approval was obtained, the questionnaire was administered to 40 parents find out their emotional reaction and level of understanding of the questions. The resulting questionnaire will be referred to as PSSS-S, a Spanish version of the Parents’ Self-Stigma Scale.

Sample

In total, 292 parents of children with mental disorders were interviewed. To estimate the reliability and internal structure, a sample was taken from parents whose children were in treatment in two child mental health centres in Reus (n = 141). Following this, to estimate the validity and confirm the factor structure of the questionnaire, another sample of parents from another child mental health centre in Valencia were interviewed (n = 111). All samples were collected by consecutive sampling in each of the centres. In all cases, samples included biological parents who voluntarily accepted to participate in the study and provide signed informed consent. Parents who were unable to fill the questionnaire because of language or mental problems were excluded. Sample sizes were calculated based on the minimum number of participants needed to carry out the factorization of a questionnaire with 11 items, with a ratio of 10 participants per item (n = 110).¹⁷ The study was approved by the corresponding committee on bioethics.

Procedure

Two mental health nurses interviewed each parent from Reus’ centres and Valencia's centre, and after explaining the purpose of the study and obtaining their informed consent, the questionnaire was handed over to one of the parents for self-administration of the questionnaire.

To assess the questionnaire's temporal stability, a subsample of the parents from Reus was re-interviewed after four weeks. To estimate convergent validity, the parents from Valencia were given, in addition to the questionnaire, the Spanish version of the Rosenberg Self-Esteem Scale^18,19 and Schwarzer General Self-Efficacy Scale,^20,21 two variables frequently associated to parental self-stigma.⁴ Information was also obtained and recorded on parents’ gender, age, and years of schooling, as well as their children's gender, ages, and years of progression of the disorder.

Analysis

The overall self-stigma score was obtained by summing the scores of all PSSS-S items. Items 5 (‘It is not my fault that my daughter has her problem’), 10 ('I am as good a mother as I can be'), and 11 (‘I am a good mother, whatever others say’) were reverse scored. To estimate test-retest reliability, the Intraclass Correlation Coefficient (ICC) was calculated between the overall questionnaire scores obtained before and after the administration of the PSSS-S to the subsample of 30 parents. SPSS v.25® software was used to calculate the ICC and describe parent variables.

The Reus sample was used to investigate the questionnaire's factor structure. First, exploratory factor analysis was conducted, with a matrix of polychoric correlations, due to the ordinal character and lack of symmetry of the items (the Mardia test for symmetry and kurtosis indicated non-normality), with factor extraction by unweighted least squares and PROMIN rotation.²² A parallel analysis was performed to establish the number of factors to be retained.

The reliability of the retained factors was calculated. Using bootstrapping, 95% confidence intervals (95% CI) for the model measures were calculated, as well as the factor loadings of the questionnaire items.

Confirmatory factor analysis was then performed using the data from the Valencia sample, taking as a reference the factor loadings matrix obtained in the EFA of the Reus sample. The load matrix was semi-specified, with non-zero values in the coefficients identified with values greater than 0.30 for each factor and zero values in the rest. All analyses were conducted using FACTOR v. 10.9.01^23-27

To evaluate the relationship between the questionnaire scores (overall and for each of the subscales identified with the EFA) and the demographic variables and progression of the disorder, the Spearman correlation coefficient was calculated (with parents’ and children's ages, parents’ years of schooling, and years of disorder progression), and the existence of significant differences according to gender for parents and children were verified using Student's t-test. We considered p < 0.05 as a criterion of statistical significance.

Finally, we proceeded to estimate the convergent validity of the PSSS-S by using the Valencia sample to calculate the Pearson (r) correlation coefficients between the PSSS-S score and the scores obtained from the Self-Esteem Scale and the General Self-Efficacy Scale. Missing cases were excluded from the related analysis. In general, a level of p < 0.05 was established to detect significant results.

Results

Translation and adaptation

After obtaining the approval of the author of the PSSS, the Spanish version (PSSS-S) was administered to 40 parents. All but six agreed to complete the questionnaire.

Table 1 shows the average ages of parents and children, as well as parents’ average years of schooling and the average years of disorder progression for children. Both samples comprise middle-aged adults, the majority of whom were women, with an educational level of 12 years of schooling. The majority of the children were male, with an average age between 10 and 12 years. The only notable difference between the two samples was that, for the children in the Reus sample, the age and the duration of their mental health disorder was, on average, one and a half years longer than those in the Valencia sample.

Table 1.

Parents’ and children's basal data in Reus and Valencia samples, two provinces of Spain chosen for the study.

Parents’ basal data
Gender (Reus/Valencia)	Mean age (SD)		Mean years of schooling (SD)
	Reus’ sample	Valencia's sample	Reus’ sample	Valencia's sample
Male (43/25)	45.1 (5.8)	46.8 (5.0)	13.1 (4.6)	13.6 (3.5)
Female (98/85)	43.0 (6.1)	43.8 (8.4)	12.4 (3.7)	12.7 (3.7)
Total (141/110)	43.6 (6.1)	43.4 (8.5)	12.6 (4.0)	12.8 (3.7)

Children's basal data
Gender (Reus/Valencia)	Mean age (SD)		Mean time since diagnosis (SD)
	Reus’ sample	Valencia's sample	Reus’ sample	Valencia's sample
Male (90/85)	10.7 (3.3)	10.0 (4.4)	6.9 (3.8)	5.3 (3.4)
Female (51/26)	13.3 (3.2)	10.6 (4.6)	5.4 (4.1)	3.9 (2.6)
Total (141/111)	11.7 (3.5)	10.1 (4.4)	6.4 (4.0)	4.9 (3.3)

Note: SD: Standard deviation.

Regarding the principal diagnosis, 70% of the children were diagnosed with hyperkinetic disorder (ICD 10 F90), pervasive developmental disorder (ICD 10 F84), eating disorders (ICD 10 F50), or reaction to severe stress and adjustment disorders (ICD 10 F43). None of the remaining diagnoses reached 5% of the total sample, including psychotic disorders.

The 30 subjects randomly selected to carry out the reliability analysis presented almost no differences in the means and standard deviations of the scores from before to after (mean: 23.3 before vs. 23.2 after; standard deviation: 3.7 before vs. 3.1 after). The ICC between the before and after scores was 0.76 (95% CI: 0.55, 0.88).

The EFA (Table 2) presented fair adequacy, with a Kaiser−Meyer−Olkin measure of 0.74 (95% CI: 0.70, 0.79) and a significant Bartlett statistic (p < 0.001). Parallel analysis suggested a three-factor solution, with a total explained variability of 70.2%. This solution coincided with the one obtained by the eigenvalue criterion greater than 1. The model's goodness-of-fit indices were RMSEA < 0.001 (95% CI: 0.000, 0.051; below the 0.05 limit to be considered a good fit); Chi-square value (25 gl) = 18.7, non-significant (adequate fit); NNFI = 0.999 (95% CI: 0.989, 1.722); and IFC = 0.999 (95% CI: 0.995, 1.328; NNFI and IFC values above 0.99 are considered to be an excellent fit).

Table 2.

Rotated factorial loadings in the model obtained with exploratory factor analysis (EFA) using PSSS-S^∞ (Reus’ sample, n = 141).

	Factor 1 (F1) (Bad Parent)		Factor 2 (F2) (Self-blame)		Factor 3 (F3) (Self-shame)
Item^a	Load	95% CI	Load	95% CI	Load	95% CI
My child has his/her problem because of meSoy la(el) causante del problema de mi hija(o)	−0.007	(0.244, 0.346)	0.821*	(0.619*, 1.099*)	−0.06	(0.449, 0.193)
The way I have raised my child has contributed to his/her problemLa manera en la que he educado a mi hija(o) ha contribuido a su problema	0.143	(0.052, 0.346)	0.679*	(0.107*, 0.827*)	0.112	(0.019, 0.351)
I feel guilty that my child has his/her problemMe siento culpable de que mi hija(o) tenga su problema	−0.029	(0.274, 0.114)	0.910*	(0.803*, 1.067*)	−0.161	(0.474,0.016)
I deserve to be blamed for my child's problemMerezco que se me culpe del problema de mi hija(o)	−0.049	(0.288, 0.173)	0.798*	(0.458*, 0.992*)	0.135	(0.085, 0.386)
It is not my fault that my child has his/her problem ^bNo es mi culpa que mi hija(o) tenga su problema^b	0.050	(0.115, 0.393)	0.519*	(0.284*, 0.743*)	−0.111	(0.482, 0.080)
I am not a good enough parentNo soy una(un) madre(padre) lo suficientemente buena.	0.385*	(0.000, 0.680*)	0.225	(0.027, 0.517)	0.187	(0.026, 0.463)
I am embarrassed that my child has his/her problemMe da apuro ser la(el) madre(padre) de una(un) niña(o) con problemas	0.141	(0.060, 0.511)	−0.225	(0.703,0.090)	0.822*	(0.649*, 1.139*)
I am ashamed that my child has his/her problemMe avergüenza ser la(el) madre(padre) de una(un) niña(niño) con problemas	−0.191	(0.921, 0.048)	0.146	(0.143, 0.451)	0.738*	(0.122, 0.997*)
I am self-conscious about being a parent of a child with problemsMe siento incómoda(o) conmigo misma(o) por ser la(el) madre(padre) de una(un) niña(o) con problemas	0.022	(0.174, 0.375)	0.129	(0.062, 0.401)	0.780*	(0.205, 0.986*)
I am the best parent I can be ^bSoy todo lo buena(buen) madre(padre) que puedo ser^b	0.613*	(0.000, 0.912*)	−0.030	(0.183, 0.110)	0.227	(0.064, 0.508)
I am a good parent, no matter what others say ^bSoy una(un) buena(buen) madre(padre), digan lo que digan los demás^b	1.017*	(1.000*, 1.067*)	0.074	(0.085, 0.199)	−0.216	(0.494,0.036)

Note:

a

In italics: Text of the items in Spanish.

b

items scored inversely.

∞

Spanish version of the Parents' Self-Stigma Scale (Escala de autoestigma de los padres).

⁎

Factorial loadings above 0.30.

CI Confidence interval.

Table 3 presents the rotated factor loads of the obtained model. The correlations between factors were 0.27 (95% CI: 0.12, 0.99) for F1 with F2, 0.42 (95% CI: 0.25, 0.77) for F1 with F3, and 0.31 (95% CI: 0.16, 0.73) for F2 with F3. The factors presented elevated ORION (‘Overall Reliability of fully-Informative prior Oblique N-EAP scores’) consistencies from 0.86 (95% CI: 0.52, 0.94) for F1, 0.99 (95% CI: 0.84, 1.00) for F2, and 0.90 (95% CI: 0.62, 0.96) for F3, all above 0.80.

Table 3.

PSSS-S rotated factor loadings of model obtained with Valencia's sample (n = 107^a).

	Factor 1 (F1) (Bad Parent)		Factor 2 (F2) (Self-blame)		Factor 3 (F3) (Self-shame)
Item	Load	95% CI	Load	95% CI	Load	95% CI
My child has his/her problem because of me	−0.154	(2.651,0.038)	0.914*	(0.768*, 3.775*)	−0.342	(2.455,0.153)
The way I have raised my child has contributed to his/her problem	0.124	(0.014, 0.388)	0.642*	(0.416*, 0.933*)	0.135	(0.012, 0.388)
I feel guilty that my child has his/her problem	−0.148	(0.328,0.027)	0.839*	(0.725*, 1.015*)	−0.054	(0.225, 0.083)
I deserve to be blamed for my child's problem	0.051	(0.016, 0.366)	0.503*	(0.226*, 0.999*)	−0.023	(0.194, 0.104)
It is not my fault that my child has his/her problem ^b	0.070	(0.006, 0.354)	0.364*	(0.192*, 0.671*)	−0.157	(0.415,0.014)
I am not a good enough parent	0.221	(0.001, 0.577)	0.244	(0.070, 0.564)	0.202	(0.082, 0.409)
I am embarrassed that my child has his/her problem	0.045	(0.008,1.185)	−0.177	(1.141,0.007)	0.855*	(0.742*, 1.041*)
I am ashamed that my child has his/her problem	0.006	(0.069, 0.224)	0.109	(0.027, 0.505)	0.964*	(0.776*, 1.001*)
I am self-conscious about being a parent of a child with problems	−0.061	(0.566, 0.009)	0.052	(0.057, 0.455)	0.932*	(0.797*, 1.004*)
I am the best parent I can be ^b	0.928*	(0.518*, 1.016*)	−0.020	(0.172, 0.017)	−0.007	(0.039, 0.107)
I am a good parent, no matter what others say ^b	0.797*	(0.303*, 1.001*)	0.036	(0.004, 0.218)	−0.064	(0.205,0.010)

Note:

a

Four subjects did not complete the questionnaire and were eliminated.

b

items scored inversely.

⁎

Factorial loadings above 0.30.

CI: Confidence interval.

The factor model obtained with the Reus sample was confirmed by PCFA using the Valencia sample. A PCFA with a semi-specific factor load coefficient matrix was performed. With this procedure, the congruence or similarity is checked with a model for which the factor loads are 0 for the low load items (< 0.30) in each factor of the reference model, without specifying any specific value for the high load items (≥ 0.30) in the factors of the reference model.

The rotated factor loading of the model obtained with Valencia´s sample is shown in Table 3. The model presented a fair fit, with a Kaiser−Meyer−Olkin measure of 0.66 (95% CI [0.61,0.75]) and a significant Bartlett statistic (p < 0.001). The parallel analysis also suggested a three-factor solution, with a total explained variability of 67.9%, somewhat lower than that of the previous EFA. The model's goodness-of-fit indices were RMSEA = 0.001 (95% CI [0.000,0.093]); χ2 (25 gl) = 10.1, non-significant; NNFI = 0.999 (95% CI [0.914, 1.979]); and IFC = 0.999 (95% CI: 0.961, 1.445).

Table 4 shows estimated congruencies between the data and the semi-specified theoretical model for each studied variable. The estimated overall congruencies for each of the factors were 0.883 (95% CI [0.806, 0.967]) for F1, 0.933 (95% CI [0.908, 0.976]) for F2, and 0.960 (95% CI [0.716, 0.993]) for F3. According to the interpretation of these coefficients,²⁸ values in the range of 0.85−0.94 suggest a reasonable similarity between the model factors derived from the data and the factors of the specified theoretical model, while values greater than or equal to 0.95 imply that the factors compared can be considered equal. In this case, all factors were congruent above 0.85 and could be considered reasonably equivalent. When observing the congruencies by variables, the item 'I am not a good enough parent' was the only one that presented a congruency index lower than 0.85.

Table 4.

Data congruence coefficients with the 3-factor model semi-specified as from the EFA results of Valencia's sample (n = 107).

Item	Congruence coefficients	95% CI
My child has his/her problem because of me	0.925	(0.320, 0.980)
The way I have raised my child has contributed to his/her problem	0.962	(0.304, 0.999)
I feel guilty that my child has his/her problem	0.983	(0.915, 0.999)
I deserve to be blamed for my child's problem	0.994	(0.968, 1.000)
It is not my fault that my child has his/her problem ^b	0.905	(0.311, 0.999)
I am not a good enough parent	0.572	(0.001, 0.998)
I am embarrassed that my child has his/her problem	0.978	(0.499, 0.999)
I am ashamed that my child has his/her problem	0.994	(0.879, 1.000)
I am self-conscious about being a parent of a child with problems	0.996	(0.871, 1.000)
I am the best parent I can be ^b	1.000	(1.000, 1.000)
I am a good parent, no matter what others say ^b	0.996	(0.944, 1.000)

Note: EFA Exploratory Factor Analysis.

CI Confidence interval.

b

Items scored inversely.

Table 5 shows the factor loadings of the PSSS-S calculated for the whole sample (248 subjects, used in the previous factor analyses). The factor structure is coincident with that previously obtained, and the values of the factor loadings present few differences with the previous ones The correlations between factors were 0.30 (95% CI: 0.18, 0.51) for F1 with F2, 0.11 (95%CI: −0,03, 0.32) for F1 with F3, and 0.36 (95% CI: 0.19, 0.55) for F2 with F3. The consistencies varied from 0.84 for F1, 0.89 for F2, and 0.91 for F3, all above 0.80.

Table 5.

PSSS-S rotated factor loadings of model obtained with the whole sample (n = 248^a).

	Factor 1 (F1) (Bad Parent)		Factor 2 (F2) (Self-blame)		Factor 3 (F3) (Self-shame)
Item	Load	95% CI	Load	95% CI	Load	95% CI
My child has his/her problem because of me	−0.089	(−0.240, 0.045)	0.913*	(0.800*, 1.064)	−0.243	(−0.512,−0.077)
The way I have raised my child has contributed to his/her problem	0.145	(0.027, 0.277)	0.667*	(0.514*, 0.772*)	0.141	(0.026, 0.261)
I feel guilty that my child has his/her problem	−0.122	(−0.233, −0.026)	0.884*	(0.780*, 0.987*)	−0.094	(−0.212, 0.019)
I deserve to be blamed for my child's problem	0.022	(−0.127, 0.192)	0.668*	(0.458*, 0.825*)	0.045	(−0.111, 0.196)
It is not my fault that my child has his/her problem ^b	0.059	(−0.095, 0.241)	0.490*	(0.325*, 0.645*)	−0.167	(−0.346, 0.018)
I am not a good enough parent	0.321*	(0.104*, 0.486*)	0.269	(0.096, 0.433)	0.193	(0.050, 0.346)
I am embarrassed that my child has his/her problem	0.099	(−0.003, 0.246)	−0.212	(−0.345, −0.057)	0.873*	(0.675*, 0.994*)
I am ashamed that my child has his/her problem	−0.082	(−0.257, 0.072)	0.056	(−0.193, 0.225)	0.863*	(0.626*, 0.979*)
I am self-conscious about being a parent of a child with problems	−0.043	(−0.233, 0.081)	0.166	(0.031, 0.329)	0.831*	(0.639*, 0.936*)
I am the best parent I can be ^b	0.806*	(0.635*, 0.969*)	−0.023	(−0.135, 0.071)	0.040	(−0.059, 0.166)
I am a good parent, no matter what others say ^b	0.854*	(0.690*, 0.999*)	0.060	(−0.032, 0.154)	−0.137	(−0.272, −0.027)

Note: ^a Four subjects did not complete the questionnaire and were eliminated.

b

items scored inversely.

⁎

Factorial loadings above 0.30

CI Confidence interval.

Table 6 shows the basic descriptive characteristics of the average scores obtained by the parents for the total questionnaire and subscales, obtained as an average of the item scores. By performing a repeated-measures ANOVA, with Greenhouse−Geisser correction, it was found that mean subscale scores were significantly different (p < 0.001) for the whole sample. After making two-to-two comparisons using Student's t-tests and Bonferroni correction for multiple comparisons, significant differences (p < 0.01) were found between pairs of subscales. These differences were maintained by gender.

Table 6.

Means and 95% confidence intervals (95% CI) of PSSS-S total and subscale scores by sex.

Subscale	Sex
	Male (n = 43)		Female (n = 98)		Total (n = 141)
	Mean	95% CI	Mean	95% CI	Mean	95% CI
Bad-Parent (F1)	2.15	(1.89, 2.40)	2.18	(1.98, 2.38)	2.17	(2.02, 2.33)
Self-Blame (F2)	1.65	(1.44, 1.80)	1.90	(1.74, 2.06)	1.82	(1.69, 1.95)
Self-Shame (F3)	1.18	(1.08, 1.28)	1.16	(1.08, 1.24)	1.17	(1.11, 1.23)
Total	1.66	(1.52, 1.79)	1.78	(1.66, 1.89)	1.74	(1.65, 1.83)

Note: F1, F2, F3: Self-stigma factors.

CI: Confidence interval.

When calculating correlations between the average scores for the total questionnaire and subscales and parent's age, parent's years of schooling, child's age, and years of progression of the disorder, only the subscale F3, self-shame, showed a significant correlation (p < 0.05) with parent's age.

Cronbach's alphas for the self-esteem scale (alpha = 0.71) and general self-efficacy scale (alpha = 0.91) were calculated. The convergent validity of the PSSS-S was then estimated by calculating the Spearman correlation coefficients between the scores of the PSSS-S and the General Self-efficacy Scale (−0.27, p = 0.005) and Self-esteem scale (−0.21, p = 0.032). Both correlations were highly significant in the expected direction.

Discussion

The present study found that PSSS-S, a Spanish version of PSSS, can be considered psychometrically and semantically equivalent to the original questionnaire. In addition, the PSSS-S was also shown to have fair test-retest reliability.

It was necessary to repeat the translation and back-translation process several times with different translators until achieving, with the original author's help (Kim Eaton), a Spanish version of the PSSS (PSSS-S) that was semantically equivalent to the English version.

Only six parents refused to collaborate, and none of those who completed the questionnaire reacted to it negatively. Most parents completed the questionnaire in approximately 11 min. Thus, the questionnaire appears to be acceptable for its target population.

This study was based on a clinical sample because the objective of the study was to develop an instrument to be used in clinical situations. This may explain some of the differences found with Eaton's study, where a community sample was used.

The structural equivalence of the PSSS-S to the PSSS was confirmed, despite the large differences in the samples used in Australian and Spanish studies, as well as the cultural differences that separate both countries, which attests to the robustness of the findings. In the Spanish sample, 70% of the parents were middle-aged women with a low educational level; 70% of the children were preteens, diagnosed with hyperkinetic, pervasive developmental, or eating disorders, and with four or more years of treatment. In the Australian sample, 97% of the parents were women and generally younger and more educated than the Spanish sample, while the children were younger, and most who were diagnosed with hyperkinetic disorders had a shorter duration of treatment (11.8% had four or more years of treatment). The Spanish sample was drawn from users of public mental health centres, while the Australian sample was drawn from parents of children with a mental health disorder who were invited to participate, after distributing the survey to parent support groups across Australia and New Zealand. It could explain some of these differences.

Regarding the effect of culture, previous reports have shown that self-blame and self-shame are emotional states particularly sensitive to cultural differences.⁵ Regarding parents’ self-stigma scores, the bad parent subscale was the highest, distantly followed by self-blame and self-shame. This pattern was identical to that obtained with the PSSS;²² however, the scores were much higher for Australian parents. The analysis of the relationship between mean PSSS-S scores and factor subscales in relation to parent and child characteristics showed that only the self-shame subscale had a significant positive correlation with the age of parents of both genders. This finding is contrary to Eaton who found that self-stigma scores decreased with parent age and time since diagnosis, but was in line with a recent meta-analysis.²⁹

The test-retest reliability estimated for the questionnaire was adequate, with the ICC above the recommended limit of 0.75. ³⁰ Unlike the general use of Pearson's correlation coefficient to represent the correlation of scores between the test and retest, we used the ICC to estimate agreement, a much more demanding concept than a mere correlation expressing instrument reproducibility.³¹ Regarding EFA, the obtained model had excellent goodness-of-fit indices. The sample size was adequate to estimate with accuracy all indices, factor loads, correlations between factors, and congruencies with the Eaton model.

Factor analysis models have been used, which considered the categorical ordinal and non-symmetrical nature of the items, using a poly-correlation matrix. This could explain the good fit rates obtained for the EFA models. The factor structure of the PSSS-S corresponded to that obtained by Eaton through factorization of the PSSS. Only the item ‘I am not a good enough parent’ presented an insufficient degree of congruence with the PSSS after confirmatory analysis, which means that this is the most dubious item of the PSSS-S in conforming to the proposed theoretical model, although it has an almost acceptable level of congruence. This same item was also considered doubtful in Eaton's analysis, although it was ultimately retained, so as not to lower the scale's internal consistency coefficient.

The three factors found in the present study - Bad Parent, Self-blame and Self-shame were significantly and positively correlated with each other, but with rather moderately sized coefficients, suggesting a moderate degree of overlap. The correlation between the bad parent and self-shame subscales presented the highest coefficient, in contrast to Eaton's study, in which the correlation between the bad parent and self-blame subscales was found to be higher. It is possible that this is due to cultural differences in the two samples. According to Benedict (2011),³² cultures may be divided into cultures of shame, where social control is based on honour and pride, and social exclusion follows when norms are violated, and cultures of guilt, where unwanted behaviour is controlled by feelings of guilt. In a culture of guilt, like in Australia,^33,34 it is well understood that the belief that one is a bad parent is associated with feelings of self-blame, while in Spanish culture, which can be classified as a culture of honour ³⁵ the belief that one is a bad parent is associated with feelings of self-shame.

The present study represents an advancement in the development of an instrument to measure the self-stigma of parents of children with mental health problems, adapting the PSSS to a different culture from the Australian one, and adding an estimate of its temporal stability.

However, some limitations affecting this study should be noted. It seems possible that children's diagnosis could influence the degree of stigma experienced by their parents. The sample of this study contains various pathologies but not in sufficient number to explore the relationship between the child's diagnosis and the parent's self-stigma. Future studies may analyse the particular effect of certain pathologies on the degree and content of stigma. It is possible that pathologies with greater visibility or social weight could contribute more to the development of feelings of shame or guilt, for example, disruptive symptomatology or drug addiction, while other pathologies could evoke more compassionate attitudes, for example, Down's syndrome. In evaluating the composition of the sample, it should be noted that non-biological parents were excluded for the sake of sample homogeneity. However, it could be that the adoptive parents do not experience the same degree of stigma as the biological parents and this could be the object of examination in future research.

Conclusion

This paper describes the development of a Spanish version of the Parents' Self-stigma Scale. It shows that the Spanish version is semantically and psychometrically equivalent to the original scale and also that it has an acceptable test-retest reliability. Regarding content, the three factors found in this study - Bad Parent, Self-blame and Self-shame - are the same as those found as those found in Eaton's study. These results consolidate and expand the psychometric characteristics of the PSSS and add to its intercultural validity.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

Research ethics and patient consent

This study was approved by Comissió Clínica d'Investigació del Hospital Universitari Institut Pere Mata (Ref. 25/01/2019, PR_01–21-2019_02_Tassara) and the Comitè Ètic d'Investigació en Medicaments (CEIm) del Institut d'Investigació Sanitària Pere Virgili IISPV. Both entities ensure that research involving human subjects is carried out in accordance with the Declaration of Helsinki and in compliance with current legislation on biomedical research. (Ley 14/2007, 03/06). All participants voluntarily accepted to participate in the study and provided signed informed consent according to the recommendations of the European General Data Protection Regulation (GDPR).

Consent for publication

All authors have read and approved the final manuscript.

Acknowledgements

We acknowledge the help provided by Kim Eaton in translating the questionnaire.