2. Method
2.1. Participants
The sample consisted of 178 subjects with a diagnosis of cancer, inhabitants of the city of Salamanca, Spain (the distribution according to cancer types is shown in
Table 1). 89 women (50%) and 89 men (50%), ranging in age from 17 to 81 years (
M = 45.5;
SD = 16.5), participated. Almost half of the participants placed themselves within a medium socioeconomic level (48.5%), with the other half split between the lower-middle (24.7%) and upper-middle (25.3%) levels. In addition, most of the participants had university studies, both undergraduate (39.3%) and master’s or doctoral degrees (22.5%).
2.2. Procedure
For the adaptation of the instrument, two different translations of the items were made, one more literal than the other, in order to compare them and observe which of the two was better understood in Spanish. These two translations were back-translated into English by an independent native English speaker and bilingual in Spanish, who had no prior knowledge of the original text, to ensure that the items retained their original meaning. Thus, we decided on the most literal translation of these items, since when comparing the accuracy and equivalence between the two versions, it was observed that there were no differences between the Spanish translation and the original version of the scale. The items in Spanish can be found in
Appendix A.
After this, an online administration questionnaire was prepared using the Google Forms application.
Given the difficulty of contacting cancer patients, a non-probabilistic snowball sampling was carried out, through close cases of people with cancer, which helped to find more people affected by this disease. In this way, a wide sample of cancer patients was reached, since the questionnaire was sent not only to individuals, but also to workers of hospitals, associations, residences or day care centers, who were able to disseminate it among a larger number of subjects.
Once they received the questionnaire, the subjects had access on the first page to an explanation of what the questionnaire consisted of, as well as the option to give their consent to the confidential treatment of their data for research purposes. Once the conditions were accepted, the subjects had access to the second page of the questionnaire, where they had to complete the socio-demographic data requested. After this, the subjects were able to answer the different items that included the questionnaire.
Finally, given that the questionnaire was sent to healthy persons or persons affected by other types of diseases, and despite the fact that it was described in the introduction that it was focused on persons affected by cancer, we proceeded to eliminate those data of healthy persons or persons non affected by cancer. In this way, a total of 27 non-relevant cases were eliminated from an initial sample of 205 subjects.
The ethics committee of the Complutense University of Madrid audited the ethical issues for this study.
2.3. Measures
2.3.1. Sociodemographic Characteristics
An ad hoc survey was administered in which participants were asked about their age, sex, educational level, employment status, socioeconomic status, and health problems.
2.3.2. Brief Trust/Mistrust in God Scale
The Brief Trust/Mistrust in God Scale (
Rosmarin et al. 2011) consists of six items divided in both factors, three of which assess trust in God (e.g., “God loves me immensely”) and three of which assess mistrust in God (e.g., “God hates me”). This is a Likert scale with five response alternatives (from “not at all” to “very much”) were participants indicated their degree of belief. Thus, the higher the score, the greater the degree of trust or mistrust in God, depending on the subscale. The Cronbach’s alpha coefficient is 0.94 for the Trust factor and 0.88 for the Mistrust factor.
2.3.3. Positive and Negative Affect Schedule (PANAS)
The Positive and Negative Affect Schedule (PANAS) (
Watson et al. 1988), adapted to Spanish by
López-Gómez et al. (
2015), has been widely used to measure affect and emotions. The PANAS includes two subscales, one for positive affect and the other for negative affect, with ten items each (e.g., “Satisfied with oneself” and “Distressed”, respectively), for a total of twenty items. It is a five-point Likert scale where the participants must indicate whether he/she has felt each of the affects, from 1 (“not at all or very slightly”) to 5 (“very much”). In each subscale, a total score ranging from 10 to 50 is obtained, where a higher score indicates a greater presence of that type of affect. In this version, Cronbach’s alpha coefficient is 0.92 for the Positive Affect subscale and 0.88 for the Negative Affect subscale. In the case of our sample, the results were also satisfactory, both for the positive affect subscale (α = 0.88) and the negative affect subscale (α = 0.86). This scale has already been used to analyze the relationship between trust in God and positive and negative emotions, with satisfactory results (
Fadardi and Azadi 2017).
2.3.4. Multidimensional Scale of Perceived Social Support
This scale developed by
Zimet et al. (
1988), and adapted to Spanish by
Landeta and Calvete (
2002) and
Ruiz et al. (
2017), consists of twelve items that evaluate the level of social support perceived by individuals, and has three different dimensions (family, friends, and significant others). The scale is presented in Likert format with seven response alternatives, with a value of 1 “Strongly disagree” and 7 “Strongly agree”. However, for the present study only the items of the significant others dimension were used, as they represent an even more abbreviated version to assess perceived social support without focusing on any specific source of support (e.g., “There is a person who is around when I am in a difficult situation”). This subscale also yields very positive reliability data (α = 0.94). For the present sample, although the reliability is lower, it is equally adequate (α = 0.80). With the use of this scale we expect to find significant positive associations between trust in God and social support, as in the study by
Maselko et al. (
2011) mentioned in the introduction.
2.4. Data Analysis
First, descriptive statistics for the items were calculated. Then, reliability of the scale was analyzed through an internal consistency analysis using Cronbach’s alpha statistic. In addition, an exploratory factor analysis (EFA) was performed to provide evidence on the structural validity of the translated instrument: First, we assessed the adequacy of the data, via KMO estimate and Bartlett’s sphericity test. Before selecting the extraction method, we tested for Mardia’s multivariate coefficients (
Mardia 1970). The criterion for deciding how many factors to retain was a parallel analysis, comparing the empirical eigenvalues with the mean eigenvalues of 500 random correlation matrices. To facilitate interpretation, we allowed correlation between the factors, using a Promax oblique rotation.
Convergent and discriminant validity was evaluated through a correlation analysis using Pearson’s
r coefficient. In the original version
Rosmarin et al. (
2011) correlated the subscales of trust and mistrust in God with measures of depression, anxiety and distress. In this case, the correlation of both subscales was performed with the positive emotions and negative emotions subscales of the PANAS scale, as well as with the “significant others” dimension of the Multidimensional Scale of Perceived Social Support. Positive, significant correlations were considered evidence of convergent validity, and no correlation were considered evidence of discriminant validity. Level of significance, α, was set at 0.05. Some statistical analyses were carried out using base R and two packages: for Mardia’s multivariate normality analysis (MVN,
Korkmaz et al. 2014); and for Bartlett’s sphericity test, KMO estimator, and parallel analysis (psych,
Revelle 2018). We ran the rest of analyses using SPSS 25.
4. Discussion
Validity is concerned with establishing evidence for the use of an instrument in a particular setting and with a particular population (
Morgan et al. 2001). This study evaluated the psychometric properties of the Brief Trust/Mistrust in God Scale for its use in the Spanish context, in order to validate the scale.
Good reliability results (as internal consistency) were obtained for both the trust (α = 0.95) and mistrust (α = 0.86) subscales, data quite similar to those of the original scale (α = 0.90 and α = 0.85, respectively) from
Rosmarin et al. (
2011). The data showed a strong internal consistency in accordance with the published literature on the subject (
Henson 2001;
Campo-Arias and Oviedo 2008).
The significant positive correlations observed between the scores of the trust in God subscale and those of positive emotions and social support, and those of the mistrust in God subscale and negative emotions, provide evidence of convergent validity of this measure. More evidence in this sense are the negative correlations between the scores of the subscale of mistrust in God and positive emotions and social support, as well as those of the subscale of trust in God and negative emotions. In other words, we can conclude that trust in God is associated with higher levels of positive emotions and social support, and lower levels of negative emotions, with the opposite happening in the case of mistrust in God, results that are similar to those found by
Rosmarin et al. (
2011). Similarly, other research has found the same type of relationships between trust/mistrust in God and social support (
Maselko et al. 2011), as well as positive and negative emotions, including affective disorders such as depression, anxiety, stress (
Rosmarin et al. 2009a,
2009b,
2010;
Krumrei et al. 2013;
Fadardi and Azadi 2017).
These findings have several implications in the area of health care. First, in the face of a threatening life event such as a cancer diagnosis, the ability to make sense of the event by having trust in God can help preserve a person’s positive emotional state (
Fadardi and Azadi 2017). This, given the positive association between positive emotions and mental and physical health, might lead one to think that promoting trust in God within medical practice in those patients with religious and/or spiritual needs could be beneficial. However, we must be cautious with these conclusions because both our research and previous literature establish relationships between trust in God and positive emotions, but no causal ones. Therefore, more research is needed to find out if there is any other variable that mediates this relationship.
Similarly, the association between trust in God and greater social support invites us to think that given the relationship between the latter variable and health, the implications for medical practice may be important. Nevertheless, this relationship needs to be studied further, since other studies show the possibility that trust in God makes people less likely to rely on material or social resources to cope with their problems (
Fadardi and Azadi 2017).
In short, we believe that inquiring into the models of
Koenig et al. (
2012) can help to improve the understanding of the relationship between religious or spiritual variables, such as trust/mistrust in God, and emotions or social support, in order to understand how this whole web of connections can affect health.
Regarding to structural validity of this Spanish version, a two-factor model emerged from the item correlations in an EFA. This model replicates the structure of the original scale by
Rosmarin et al. (
2011). This coincidence with the model originally proposed suggests the robustness of the theoretical model.
The limitations of the study are largely due to the difficulty in achieving a representative sample. Although we believe that we have worked with a good number of participants, it is true that data protection laws make it difficult for associations and organizations working with cancer patients to collaborate with this type of studies. Therefore, this should be considered in order to find alternative ways of accessing large samples of oncology patients in the future. In addition, our results can only be applied to patients with cancer, so studies with other populations would be necessary to assure the generalization of the use of this instrument.
On the other hand, we should mention that in this study no significant differences were found in terms of the gender of the participants. However, authors such as
Cetrez (
2011) found that women present a greater number of religious behaviors than men, being more involved in their faith. In other words, the existence of differences in religiosity and spirituality between men and women have been documented, so we believe that on future occasions this variable should be taken into account.
Moreover, in relation to data analysis, we could have used a confirmatory factor analysis (CFA) to analyze the data, given that this is a new application of an existent measurement tool. However, we decided to use EFA instead. The reason was two-fold: Firstly, the two-factor model of Trust and Mistrust in God is not anchored in a broader measurement theory (i.e., a theoretical model), and thus EFA would be preferred (
Hair et al. 2015 p. 603). Secondly,
Rosmarin et al. (
2011) explored the subjacent structure of the original six items scale in two distinct samples recruited from USA and Canada, comprised of Christian (mostly Protestants) and Jewish participants respectively. Then, they confirmed the two-factor model in a third sample of Jewish participants from the USA. The same authors discussed that the measure could be appropriate for other monotheistic religious backgrounds with a personal God concept, but this concept “may vary considerably both across and within traditions”. (p. 258). As we used a Spanish adaptation of the scale, and applied it to a sample of oncologic patients from a Catholic Christian background in Spain, we considered appropriate to explore the scale as if it was a brand new measure. Further studies could confirm the two-factor structure in similar settings in Spain, adding construct validity, and allowing the development of a theoretical model throughout countries, cultures, and religions.
Likewise, on future occasions it would be appropriate and necessary to study the psychometric properties of the scale in other Spanish-speaking religious contexts, given that Spain is a country with a deep-rooted Christian tradition and other religious traditions may have been overlooked. This idea is consistent with that proposed by
Hall et al. (
2008) regarding the need to take the context into consideration in this type of research. In addition, following this line, the appropriate wording of the items is particularly important so that they are adapted to the context in which the instrument is to be applied, as proposed by
Zwingmann et al. (
2011).
Finally, the Spanish adaptation of the Brief Trust/Mistrust in God Scale has proved to be a valid and reliable instrument for the measurement of the constructs trust and mistrust in God in the Spanish oncology population. Thus, we believe that this version of the scale represents an advance in the study of the relationships between religion, spirituality and health, by providing a measure for a construct closely linked to a better understanding of these relationships.