قابلیت اطمینان و ساختار عاملی پیترز و همکاران. پرسشنامه هذیان(PDI-21)

The 21-item version of the Peters et al. Delusions Inventory (PDI-21) is a commonly used tool to measure delusional ideation in the normal population. Two recent principal component analyses have concluded that the PDI-21 has a seven-factor structure. Although these studies found identical factors associated with religiosity and grandiosity, the items loading on the remaining five factors, and hence the interpretation of these, differed. Such seven-factor structures of the PDI-21 are beginning to be used in research; however, a consistent item-level seven-factor structure has not been replicated and no data have been reported to support the reliability of such factors. We administered the PDI-21 to a non-psychiatric sample (N = 493). It was found that, with the exception of religiosity/religiousness, the previously reported factors of the PDI-21 had Cronbach’s alphas of less than 0.7. After a factor analysis using principal axis factoring, parallel analysis suggested the extraction of three factors. Of these factors, only one (religiosity/religiousness) was found to be both internally reliable and meaningful. It is concluded that the PDI-21 is best used with a unidimensional scoring system and that new measures are needed to assess specific types of delusion-like beliefs in the normal population.

Delusional beliefs have been found not to be limited to pathological conditions but also to exist in the normal population (Eaton et al., 1991 and Johns and van Os, 2001). However, measurement of such sub-clinical delusional beliefs has until recently been problematic. Peters, Joseph, and Garety (1999) noted that existing psychometric instruments either examined first-rank symptoms of psychosis, which are rarely endorsed by the general populace, or asked about superstitious beliefs which are widespread and thus, hard to justify as delusional. Furthermore, the existing measures only investigated a subset of the delusional themes found in psychosis. In order to address these problems the Peters et al. Delusions Inventory was developed (Peters et al., 1999). This is a 40-item measure of tendency to delusional ideation in non-clinical samples. In order to make this tool easier to administer, an abbreviated 21-item version was subsequently developed. The 21-item Peters et al. Delusions Inventory (PDI-21; Peters and Garety, 1996 and Peters et al., 2004) was designed to measure delusional ideation in the general population incorporating measures of the distress, preoccupation and conviction associated with delusional beliefs. Its originators specifically designed the PDI-21 not to consist of a limited number of well-defined subscales with high internal reliability but instead to “sample as wide a variety of delusions as possible” (Peters et al., 2004, p. 558). Accordingly, the inventory contains items addressing a wide range of delusional beliefs. The authors note that paranoia is a central theme of the PDI-21, with multiple items measuring persecution, suspiciousness and paranoid ideation. Amongst other delusional beliefs addressed are those relating to religiosity, grandeur, reference and depersonalisation. Peters et al. (2004) performed a confirmatory factor analysis (N = 444) with a forced one-component solution of the PDI-21. Loadings on this single factor ranged from 0.31–0.63, which was taken to support the adequacy of a unidimensional scoring system for the PDI-21. Given this intended unidimensional scoring system, a principal components analysis (PCA) was not performed to investigate multifactoriality. In a separate study, Verdoux et al. (1998) administered a French translation of the PDI-21 to a sample of adults (N = 444) with no history of psychiatric disorder. A PCA using varimax rotation was performed on the PDI-21 items. A total of seven factors with eigenvalues greater than one (which explained 55.3% of the variance) were found and hence, following Kaiser’s (1960) eigenvalue rule, seven factors were extracted. These factors were labeled as persecution, thought disturbances, grandiosity, religiosity, paranormal beliefs, reference guilt, and apocalypse. Verdoux et al. then went on to examine the association between each of these factors and participants’ age and gender. Associations were reported between five of the seven factors of the PDI-21 (persecution, thought disturbances, grandiosity, religiosity and paranormal beliefs) and age, as well as an association between one factor (religiosity) and gender. One limitation of this study was that it used Kaiser’s rule to select the number of factors extracted. Tabachnick and Fidell (2007) note that Kaiser’s rule should function as a “quick first estimate” (p. 644) of the number of factors, and that this procedure may lead to incorrect estimation of the number of factors to be extracted. A second limitation was the failure to report whether the factors had adequate internal reliability. Internal reliability can be assessed through an examination of a factor’s Cronbach’s alpha values (e.g., Blumberg, 2000), and is generally accepted to be guaranteed by a Cronbach’s alpha greater than 0.7 (Bland & Altman, 1997). The failure of Verdoux et al.’s study to report such data raises questions over the reliability, and hence validity, of the reported associations between these factors and participants’ age and gender. The only other study to have examined the factor structure of the PDI-21 is that of Lopez-Ilundain, Perez-Nievas, and Otero (2006). A Spanish translation of the PDI-21 was administered to a general population sample (N = 356), and showed good overall internal reliability (Cronbach’s alpha = 0.75). Lopez-Ilundain et al. performed a PCA with varimax rotation on the data and, using Kaiser’s rule, a total of seven factors which accounted for 53.7% of the variance were identified. Both the number of factors and the percentage of the variance explained were hence consistent with the PCA of Verdoux et al. (1998). These factors were interpreted by Lopez-Ilundain et al. as experiences of influence, depressive, paranoid, grandiosity, referential, magic thinking, and religiousness. Associations were reported between five of the seven factors of the PDI-21 (experiences of influence, depressive, grandiosity, referentiality, and magic thinking) and age, as well as an association between one factor (magic thinking) and gender. However, this study shares the limitations of Verdoux et al. (1998), namely the use of Kaiser’s rule to determine factor structure and the failure to report internal reliability data for these factors. A comparison of the factors identified by Verdoux et al. (1998) and Lopez-Ilundain et al. (2006) and the PDI-21 items loading onto each of them is shown in Table 1. Both studies identified religiosity/religiousness and grandiosity as factors of the PDI-21, with the same items loading onto each factor in each study. Two other factors are similar between the two studies (persecution/paranoid and paranormal beliefs/magic thinking) albeit with some variance in items loading onto these factors. However, the remaining factors do not agree well with each other.

The mean PDI-21 total score was 49.24 (SD = 32.9). This is slightly lower than the mean of 58.9 (SD = 48.0) of the PDI-21 reported by Peters et al. (2004) and is probably due to the greater homogeneity of the student sample used here, as compared to the adult sample used by Peters et al. (2004). The mean male score of 48.94 (SD = 33.55) was compared to that for females of 49.40 (SD = 32.65) and found not to be significantly different, t(491) = 0.15, n.s. This is in line with the finding of Peters et al. (2004). Correlations between scores on the three response scales (distress, preoccupation and conviction) ranged from r = 0.87, p < 0.001 to r = 0.93, p < 0.001. Cronbach’s alpha for the PDI-21 as a whole was 0.77, suggesting satisfactory internal reliability. Cronbach’s alphas, per the data from this study, for the factors identified by Verdoux et al. (1998) and Lopez-Ilundain et al. (2006) are presented in Table 2. Only one factor in each structure, that relating to religiosity, was found to have a Cronbach’s alpha greater than 0.7.