The Personal Disturbance Scale (sAD): normative data and latent structure in a large non-clinical sample

Abstract

The Personal Disturbance Scale [sAD; Bedford & Foulds (1978) Delusions–Symptoms–States Inventory State of Anxiety and Depression. Windsor: NFER-Nelson] is widely used in diverse settings and yet there are unresolved issues concerning its psychometric properties and normative data for the English speaking version are limited. The sAD was administered to a large sample of the general adult population (N=758). Demographic variables (gender, age, years of education and occupational status) had only very modest influences on sAD scores. Tables are presented for conversion of raw scores on the Anxiety, Depression and Total scales to percentiles. The sAD scales possessed adequate convergent and discriminant validity, as demonstrated by their pattern of correlations with two other measures of depression and anxiety (the DASS and the HADS). Ten competing models of the latent structure of the sAD were derived from theoretical and empirical sources. These models were evaluated using confirmatory factor analysis. The best fitting model (CFI=0.96) had a tripartite structure, and consisted of a general factor of psychological distress/negative affectivity (all items loaded on this factor) plus orthogonal specific factors of anxiety and depression. Correlated errors specified according to previous empirical findings were permitted. The theoretical and practical implications of this latent structure are discussed.

Introduction

The Personal Disturbance Scale (sAD) is a brief (14-item) self-report measure of depression and anxiety derived from the Delusions–Symptoms States Inventory (DSSI; Bedford & Foulds, 1978). It was originally developed for use in the normal population, but has been predominantly used in the domain of general medicine (Bedford & Deary, 1997). Its popularity may be attributed to its brevity, the fact that it solely enquires about recent mental state, and the omission of items tapping personality or attitude. Moreover, the scale was developed according to the hierarchy of classes of personal illness model, and is thus argued to constitute an effective means of detecting ‘cases’ (Foulds & Bedford, 1975).

The DSSI was developed to provide a measure of personal illness which is regarded as hierarchical in nature (Foulds, 1976, Foulds and Bedford, 1975). This model posits that the high degree of comorbidity which typically exists between psychiatric disorders is systematic, and attributable to the fact that four specific levels of personal illness exist which are inclusive and non-reflexive in nature. These range from Dysthymic States (Class One), to Delusions of Disintegration (Class Four). That is, whilst individuals in Class Three are also members of Class Two and Class One, the reverse relationship does not necessarily hold. Bedford and Deary (1999) presented 36 data sets in which the percentage of subjects fitting the hierarchy of classes of personal illness model ranged from 73.3% to 97.8%. Thus, since the sAD was derived from the Dysthymic States class, the sAD should represent an effective means of detecting cases as individuals in the upper classes with more serious ‘personal illness’ should also score highly in this lower domain.

Despite its widespread use there have been relatively few studies of the psychometric properties of the sAD, particularly in non-clinical samples. Three studies which have been conducted suggest that the sAD Anxiety, Depression and Total scales possess adequate internal consistencies, with estimates ranging from 0.80 to 0.88 (Bedford et al., 1999, Christensen et al., 1999, Morgan et al., 1987). However, no study has assessed the reliability of the sAD in a sample of the British general population.

At present, interpretation of the sAD is primarily based on the use of cut-off scores. Bedford and Foulds (1978) recommend that cumulative Depression and Anxiety scores in the range zero to two be interpreted as non-personally disturbed, three to six as personally disturbed, and seven or greater as personally ill. Thus, in this regard they do not dichotomise between anxiety and depression as most self-report measures do, but rather, view each as cumulatively contributing to personal illness. However, scores of four or greater are argued to represent membership of a set; that is, whether an individual should be regarded as suffering from an anxiety and/or depressive disorder. However, updated normative data is necessary, as the size of the sample was relatively modest and its generalisability to the normal population questionable; participants consisted of “non-graduate hospital personnel, and members of recreational and educational classes.” (p. 6), i.e. a convenience sample composed predominantly of young participants.

Means and SDs for the sAD scales are available for a large general population sample in Australia (Christensen et al., 1999). However, these data cannot be used to estimate the rarity or abnormality of a given sAD score as it is to be expected that the distribution of scores in the general population would be highly positively skewed. This rules out the use of the reported summary statistics to express an individual’s standing as a z score or percentile. Christensen et al. (1999) scored each item on the sAD 1, 2, 3 or 4 instead of the test authors' recommended 0, 1, 2 or 3. However, by deducting 7 from each scale’s total score, norms directly comparable with those presented by Bedford and Foulds (1978) can be derived. This results in means of 2.26 (SD=2.89) for sAD depression, and 2.95 (SD=2.96) for sAD anxiety (N=2622). These values are higher than the original norms presented by Bedford and Foulds (1978). Analogously, norms presented for the Greek version of the sAD (Angelopoulos and Economou, 1994, Economou and Angelopoulos, 1989, Lyketsos et al., 1979) differed, often markedly, from the original British norms (Bedford & Foulds, 1978). This highlights the need to obtain updated UK norms based on a large sample.

A number of studies have examined the influence of demographic characteristics on sAD scores. Studies investigating gender effects have produced conflicting results. Whilst three studies involving non-clinical samples found that females had higher scores on both sub-scales than males (Christensen et al., 1999, Henderson et al., 1981, Lyketsos et al., 1979), no differences were reported for the original norms presented for the British population (Bedford & Foulds, 1978). The influence of other important demographic variables, notably age, education and occupation, also needs to be investigated. The only study to date which has assessed the influence of these demographics variables reported significant effects of each upon sAD scores in an Australian sample (Christensen et al., 1999). The relationships between demographic variables and sAD scores in the general population are of interest in their own right, but investigation of these relationships would also serve the very practical purpose of identifying whether normative data should be stratified. Whilst Christensen et al. (1999) did present norms broken down according to each of these categories, only means and SDs, not percentiles, were presented.

If the use of the sAD in research and clinical practice is to be optimal, then it is also necessary to delineate the underlying structure of the instrument. This is particularly important given that the existing method of scoring the scale into separate anxiety and depression scales may not adequately reflect its structure. In an influential series of papers, Clark and Watson have argued that anxiety and depression have an important shared component which they call negative affectivity (NA; Clark and Watson, 1991a, Clark and Watson, 1991b, Watson et al., 1988). NA is conceptualised as a dispositional dimension, with high NA reflecting the experience of subjective distress and unpleasurable engagement, manifested in a variety of emotional states such as guilt, anger and nervousness, and low NA represented by an absence of these feelings (Watson & Clark, 1984). Studies have supported the existence of a dominant NA dimension (Watson and Clark, 1984, Watson and Tellegen, 1985) and provide evidence that this dimension is highly related to the symptoms and diagnosis of both anxiety and depression (Brown et al., 1997, Watson et al., 1995a, Watson et al., 1995b). Thus, there are strong reasons for suggesting that the anxiety and depression scales of the sAD should simply be collapsed, as both measure the same underlying dimension, NA.

However, as Bedford and Deary (1997) note, “studies of the sAD’s psychometric structure are rare” (p. 493), particularly in non-clinical populations. Better information on its properties in the general adult population would place clinical interpretation of the sAD on a sounder footing, and would also have wider applicability given that the sAD is also used in non-psychiatric and non-medical settings to provide a measure of mood or general distress.

Factor analytic studies conducted thus far have yielded inconsistent results (Bedford and Deary, 1997, Bedford et al., 1999, Christensen et al., 1999, Shevlin et al., 1998). Bedford and Deary (1997) conducted an exploratory factor analysis (EFA) of the SAD in a psychiatric sample (N=480). A single factor model, a two factor model (with specific anxiety and depression factors), and a three factor model could be delineated. This latter model consisted of specific anxiety and depression factors, with items 2 (“Recently I have been so miserable that I have had difficulty with my sleep”) and 11 (“Recently worrying has kept me awake at night”) forming the third factor. Confirmatory factor analysis (CFA) was then used to test five competing models. The optimal model permitted a general NA factor upon which all items loaded, plus two specific anxiety and depression factors. Three Anxiety items (items 1, 11 and 13) did not load significantly on the anxiety factor, so were a posteriori constrained to load only on the shared NA dimension. Bedford and Deary (1997) permitted the item-specific variances of the two ‘sleep’ items to correlate because, “first, two items is small to define a latent trait; and second, the correlation of the error terms captures the possibility that the two items might represent a bloated specific, essentially having a similar item content.” (p. 505). Indeed, Kendler, Heath, Martin, and Eaves (1987) used EFA to investigate the latent structure of a 13-item version of the sAD, and they too found that a third factor was defined by items 2 and 11. Thus, this suggests that there is a degree of item redundancy (i.e. the two items are simply rewords of the same question).

Shevlin et al. (1998), however, criticised Bedford and Deary (1997) for failing to take parsimony into consideration. Shevlin et al. (1998) conducted CFA using data derived from elderly general population participants (N=979). A general factor model (Model 1), a correlated two-factor model (Model 2), and a tripartite model consisting of a shared general factor as well as specific anxiety and depression factors (Model 3) were tested. Model 3 was then re-tested, but with loadings found to be non-significant omitted (Model 4). Shevlin et al. (1998) argued that the fit of all these models was comparable (although, in fact, large differences in chi square were obtained). However, they also employed parsimony fit indices (which control for the number of free parameters), and argued that these indices indicated a one factor model represented the optimal fit.

In response to Shevlin et al., 1998, Bedford et al., 1999 conducted an EFA using data from a psychiatric sample (N=132). Again, one, two and three factor solutions could be delineated, corresponding almost exactly to those identified by Bedford and Deary (1997). CFA was then employed to test the fit of each competing model. A nested three factor model which permitted correlated error between items 2 and 11 proved optimal (CFI=0.97), and a significantly better fit than a one factor model (P<0.001). However, only two of the three Anxiety items constrained to load uniquely on NA in Bedford and Deary’s (1997) best-fitting model were restricted in this manner (items 13 and 11); the third item (item 1) loaded significantly on both the NA and the Anxiety factors.

Finally, Christensen et al. (1999) conducted a confirmatory factor analysis using data derived from a large Australian sample generalisable to the normal population (N=2622). A correlated two factor model was originally specified, and represented a relatively poor fit (GFI=0.91). However, the residuals of three item pairs were permitted to correlate on an a posteriori basis (items 2 and 11; 2 and 1; and 6 and 14), and resulted in substantial improvement (GFI=0.96). This model was then compared with a one factor model permitting correlated error as specified above. The correlated two factor model represented a substantially better fit.

It should be clear that a general problem with this area of research has been that many of the researchers cited have made post-hoc adjustments in the absence of strong theoretical justifications to increase model fit (Bedford and Deary, 1997, Bedford et al., 1999, Christensen et al., 1999, Shevlin et al., 1998). This is problematic because a very real possibility of capitalising on chance exists. That is, a particular model may prove optimal as a consequence of idiosyncracies in the study sampled, and thus not be replicated in subsequent analyses. Bedford and Deary (1997) for example, found that although item 1 failed to load significantly on the Anxiety factor in the optimal model, this was not the case in a subsequent study (Bedford et al., 1999). The use of such post-hoc adjustments may partially account for the fact that the studies described have presented different interpretations of the underlying structure of the sAD.

In addition, only two of the studies investigating the latent structure of the sAD employed samples drawn from non-clinical populations, and only one of these was representative of the general population (Christensen et al., 1999). Moreover, this study did not test the fit of a tripartite structure. There are strong theoretical grounds for believing that this structure will be optimal. As noted, it has been argued that anxiety and depression scales tap a common factor, NA. The tripartite theory of anxiety and depression (Clark & Watson, 1991b) extended this theory to posit that anxiety and depression nevertheless have specific components and can thus be differentiated, with anxiety uniquely related to physiological hyperarousal, and depression to low positive affectivity (PA). This theoretical position has been supported by empirical evidence suggesting that a tripartite structure is optimal for the sAD in independent clinical samples (Bedford and Deary, 1997, Bedford et al., 1999). It is necessary to test competing models of the sAD in a sample generalisable to the normal population because of the possibility that the factor structure of the measure may not be invariant across different populations.

The aims of the present study are (1) to investigate the influence of demographic variables on sAD scores in the general adult population, (2) to provide normative data for the sAD in the form of tables for converting raw scores to percentiles, and (3) to obtain estimates of the reliability of the sAD. In addition, (4) competing models of the latent structure of the sAD will be evaluated using confirmatory factor analysis. Details of the parameterisation of the models (and the theoretical, methodological and empirical considerations that guided their selection) are presented in the methods section. Finally, (5) the convergent and discriminant validity of the sAD will be assessed by correlating each sub-scale with two other independent measures of depression and anxiety.