A scale consists of a defined continuum or dimension along which judgments are placed. The scale may be a set of ordered categories with numbers attached corresponding increasing or decreasing amount of severity or other quantitative characteristics of the phenomena being observed. The form of a psychological scale is dependent on particular psychometric techniques; the content, however, is dependent on the underlying psychological theory as noted by Hamilton (6). Psychiatric scales are special kinds of psycho­logical scales. They derive their form from the science of psychological scales in general (about which there is a vast literature) and their content from underlying theories and notions concerning the nature of psychiatric illness thought relevant by various schools of psychiatry. Characteristics of Clinical Scales in Affective Disorders For practical purposes clinical scales in psychiatry have been divided by Garside (4) into four main types, each with distinct characteristics. The first type is the assessment or descriptive scale that can be used to assess the degree of psychiatric illness such as depression. An example of this type of scale is the Hamilton Depression Scale (HDS) (5). The second is discriminatory rather than descriptive, and may be useful in arriving at a psychiatric diagnosis; an example of this scale is the Newcastle Diagnostic Index (2). The essential feature of a descriptive scale is that individual items should correlate more highly among themselves and give rise to a general factor indicating that it is meaningful to describe patients along such a dimension, i.e., convergent validity. In diagnostic scales, it is essential that the items produce separate correlational groups, i.e., discriminant validity. While there appears to be some general agreement on empirical findings with the use of descriptive scales (12), there has been little general agreement on findings on diagnostic scales in depression, for instance. Descriptive scales have also been referred to as quantitative scales to distinguish them from qualitative ones, e.g., discriminatory or diagnostic scales (1). The other types of scales are predictive, in the sense of predicting the course and outcome of particular psy­chiatric conditions, as in Kerr et al. (8), Paykel (11), and Carney et al. (2). These latter scales will not be discussed further in any detail. There are a number of commonly used and established descriptive scales for measuring anxiety and depression (Table 1). The main uncertainty in their use centers on the question of validity. There are uncomfortably high correlations between any of the scales listed measuring depression and those designated to measure anxiety (10). For instance, the Hamilton Depression and Hamilton Anxiety Scales show intercorrelations of around 0.5 to 0.6 in the literature. This could suggest that clinical anxiety and depression are highly correlated. However, impure scales could magnify this relationship, and most of our present scales fail to distinguish anxiety from depression adequately enough. What if, for instance, the HDS was not a sufficiently pure measure of the illness termed depression. After all, it was designed on a logical empirical basis by Max Hamilton, although he did go on to examine its psychometric properties. What if the margin of error in accuracy was, say 30% or even 50%, where would that leave interpretations of amine levels or pharmacological treatment indices, when the HDS is the critical variable?
(more…)

The total scores of the scales, and individual clinical, personality, and bio­graphical items were subjected to Principal Component Analysis. A series of sequential analyses indicated the stability of the clusters derived. Figure 1 illus­trates a typical outcome of this analysis, with symptoms and personality scores included. It can be seen that the upper right-hand quadrant contains a cluster of items that are consistent with the entity depression; similarly in the upper left-hand quadrant is a cluster consistent with the term anxiety. These data were reanalyzed a number of times, leaving out a variety of items on each occasion and were also analyzed separately for sex. No matter which items were included or excluded, the basic patterns established remained invariant.
(more…)

St. Francis Day Hospital, Dublin 5, Ireland (more…)

Selection of Patients
The study was confined to hospital patients. They formed a consecutive series of primarily depressed or anxious inpatients (N = 136), day patients (N = 22), and outpatients (N = 22) who were interviewed in the psychiatric units serving the city of Newcastle-Upon-Tyne and surrounding districts. The inpatient units were sampled first, then the consecutive outpatients and day hospital patients. Limitation of time and slow accretion of these latter groups accounted for the reason why the outpatient and day patient numbers were much less than inpa­tients. The more usual method of case selection by applying diagnostic criteria was deliberately avoided. Instead, patients were selected on the basis of a sustained mood disturbance that was not considered secondary to any other psychiatric disorder. Specific criteria were used to exclude patients with schizophrenia, coarse brain disease, obsessive-compulsive neurosis, anorexia nervosa, epilepsy, alcoholism, and mental subnormality. Patients with transient mood disturbance of (more…)

Product moment correlations between these scales derived in this study are shown in Table 4. The correlations between the scales measuring anxiety and depression are still much too high. They are almost of a magnitude of those supposedly measuring depression alone. The BDI/ZDS correlations with the Hamilton are not as good as one would like. In Table 5 one can see that the correlations of the SCL-90-R in depression and anxiety are very high on all the scales. The dimensions of anxiety and depression in the SCL-90-R are not there-fore reflecting adequately different syndromes. It was found for these established scales that there were striking similarities in the mean scores, ranges, profile of item scores, and factorial structures between
(more…)

A consistent procedure was used in interviewing patients, and the information elicited was recorded in a standard format. Each patient was interviewed within 48 hours of admission. Prior to the completion of the interview schedule, a full descriptive history of each patient was taken, hospital notes scrutinized, and, if appropriate, any further sources of information sought. Documentation referring to earlier hospitalization was compiled from diverse sources, even as far back as 1929 in one case. Following the collection of clinical items, patients were asked to complete the self-rating scales in random order. Each item was completed for severity for the period covering the previous month, and over 70 clinical state items were recorded. No attempt at diagnostic categorization was made, the aim being to rate each item independently and diminish any "halo" effect. The in­terview and collection of clinical information lasted an average of 2 hours, ranging from 1 to 3 1/2 hours. At an interval of 1 year after the index episode information, an outcome was systematically collected, but is not reported here.

If we were to take an oblique factor through the depression cluster, scores of items on this factor should give us an intensive measure of depression, and likewise a factor through the anxiety cluster should give us an intensive measure of anxiety. The HDS approximates to the depression cluster and appears to be a good extensive measure of severity, but is clearly slightly askew and could be improved upon. It is fairly evident that the HAS could certainly be improved upon as it tends toward the center area. Somewhat surprisingly, the EPI-N and the 16 PF (Q4) appear to be reasonably good measures of anxiety. The numbers of items forming these clusters are variable; nonetheless, they tend to remain stable in the factor space formed by the first two components. We can, however, decide that a smaller rather than greater number of these items could more conveniently describe these clusters and might be a better measure of them. A series of Principal Component Analyses with a decreasing variety of items was undertaken. Each item was assessed visually so that only those that achieved the widest separation across the first and second factor plot and those that also showed contiguity within the separate clusters of anxiety and depressive features were included. It was electively decided to restrict the included items to current mental state and ignore biographical, personality, and other features for purposes of designing purely descriptive psychopathological scales of anxiety and depression; that is, in order to develop descriptive and intensity measures of anxiety and depression, symptom ratings alone were concentrated on in the sequential analysis with the object of deriving the most widely separated items that also remained stable in the factor space. This analysis led eventually to the selection of 26 items as the best replicable descriptive items of the depression and anxiety clusters. These clusters are shown in Fig. 3. This provides some explanation for the difficulties in interpretation posed by the intercorrelations in the literature. It would appear from these analyses that the BDI and ZDS poorly reflect depression in a descriptive sense. They more definitively correlate with the clinical anxiety items. Although they may be of value for rating mood change over time (7), they would appear to be functioning as extensive and not intensive measures of depression or mood in general. It is quite possible, therefore, for them to be sensitive to changes in mood yet not reflect specifically any depressive component. The HAS does not approximate to the center of the anxiety cluster and appears to be a less accurate measure of anxiety than is its counterpart, the HDS, of depression. These two factors are here designated as factor D (depression) and factor A (anxiety) for simplicity. One of the main objectives of this study was to design new rating scales. In this study the actual factor-derived scores of patients are frequently used rather than manipulating numerical scales scores. For application elsewhere, specific scores must be applied to the items comprising each scale. These scales correlate slightly negatively with each other in this population. In other populations selected by different criteria and subject to different selection processes, they might well be independent or positively correlated with each other. Having found these new dimensions that give an accurate mathematical representation of these clusters, we can now see how well the existing scales correlate with these dimensions. Table 6 shows the relative magnitudes of the correlations of the new dimensions or scales compared with the HDS and HAS. This measure of extensive severity is simply the extensive measure of mood referred to earlier, i.e., scores on the general factor. The new anxiety scale and the new depression scale have reason­ably good correlations with this general dimension of mood disorder. The new depression scale correlation is as good as the HDS. The correlations between the two Hamilton scales is 0.55. The new anxiety scale has a high correlation with the HAS and likewise the new depression scale has a high correlation with the HDS, yet their correlations are of a magnitude of -0.19. They are therefore more or less independent, far more so than the existing Hamilton scales. The
(more…)

In addition to ratings on these items, the scales in Table 3 were completed by the interviewer or patient as appropriate; the listing refers to the individual scales applied to consecutive patients. The numbers completing the self-rating scales
(more…)