Paper: Mood and Anxiety in the Medically Ill

The measurement of anxiety and mood on a dimension of severity of affective states should be considered analogue to the measurement of blood pressure in the health screening setting. Consequently, we have to speak about primary and secondary anxiety or depression in the same way as we speak about primary or essential hypertension (without a medical explanation) versus secondary hypertension (when caused by various somatic disorders). With this background, the rating scales for anxiety and mood should be considered as quantifiable as the measurement of hypertension.

The term ‘somatization’ in psychosomatic medicine can also be an indicator of severity of affective states. If complaints about symptoms of pain seem not to be secondary to medical disorders but rather as an indicator of primary affective states, it is important to identify accompanying symptoms of anxiety or mood. Indeed, pain is a purely subjective experience which to a varying degree takes both sensations (localized in different parts of the body) and anxiety or mood (unpleasant states) into account. From a narrow medical point of view, physicians often consider anxiety or depression as a misleading factor in the diagnostic process. However, it is important to use a monistic, phenomenological approach by adding somatization and symptoms of anxiety or depression within the affective dimensions and hereafter to make a dualistic analysis when diagnosing primary versus secondary disorders.

With this background, this review on rating scales for anxiety and depression in the medically ill is developed to be of practical use at the phenomenological level of measuring severity of affective states. The symptoms are thus considered for their validity in their relation to the clinical syndrome (anxiety or depression). At the dualistic level of diagnosing the underlying disorder, the clinical syndrome precedes etiological reflections [1]. In other words, we consider a syndrome as an expression of a disordered function which might be produced by one or more disease processes.

The practical use of rating scales or questionnaires in psychosomatic medicine as captured in this review is on the one hand to describe syndromes by symptoms interpreted by psychometric analyses, and, on the other hand to make etiological considerations for mode-specific interventions in anxiety and depression [2].

The Hospital Anxiety and Depression Scale (HADS) was actually constructed to help physicians identify disorders of anxiety and/or depression in the medically ill patient [3]. The underlying consideration was to select symptoms of anxiety and depression which were as ‘psychic’ as possible to avoid symptoms with overlapping ‘somatic’ associations to medical disorders. Therefore, with regard to the 7 items in the HADS covering the syndrome of anxiety, only one single item (‘butterflies in the stomach’) was an expression of somatic anxiety, and among the remaining anxiety items the psychic anxiety symptoms in the HADS were being tense or wound up, worry thoughts, frightened feelings, restlessness, and panic. As regards the 7 items in the HADS covering the syndrome of depression, once more only one single item (‘less care of my appearance’) was an expression of somatic depression. However, the remaining items did not cover the psychic symptoms of depression but rather positive well-being feelings such as enjoying things, able to laugh at the funny side of things, being cheerful, looking forward with enjoyment to things, enjoy a good book.

When screening for depression in the medically ill patient, we therefore need to distinguish between measuring lack of well-being and measuring the syndrome of depression. Measuring well-being is a screening for health-related quality of life and thus falls outside the scope of this review. In the HADS, we have no item covering depressed mood or lack of interests in the daily functioning. The syndrome of depression as well as the syndrome of anxiety each has to cover the clinical theories of these states as shown in table 1. In primary and secondary depression (e.g. depression secondary to medical disorders), the three items in Beck's negative triad of depression [4] should be included (depressed mood or hopelessness, guilt feelings or feelings of failure, and lack of interests or helplessness). Likewise, in primary and secondary anxiety, Spielberger's psychic state of anxiety [5] should be included [nervousness, tension, worry, apprehension and fearfulness (panic)]. The psychic anxiety items in the Spielberger state anxiety scale can also be found, as indicated in table 1, in the briefer version of Taylor's manifest anxiety scale which was derived by their clinical validity from the items in the Minnesota Multiphasic Personality Inventory [6].

In contrast to the HADS, Kellner's Symptom Questionnaire [7], which is based on the same principle with symptom versus well-being items, does have Beck's negative triad in the depression subscale (see online suppl. appendix). Therefore, the Symptom Questionnaire is included in this review, whereas the HADS is not. It is, however, important to evaluate the standardization of the existing scales for anxiety and depression also in the medically ill patients.

OSM

There are many versions of even the most frequently used rating scales for anxiety and depression, and even when using the original standardization of these scales in primary affective disorders, this fact still gives rise to problems (see appendix). An illustrative example is the most archetypical scale for depression, the Hamilton Depression Scale (HAM-D) which during its 50 years in clinical use has grown into more as a family of scales than one single instrument [8]. In this review, the evolutional modifications of the HAM-D will be treated, including recent attempts to develop self-rating scales for the HAM-D. In the field of anxiety, the Hamilton Anxiety Scale (HAM-A) is also the archetypical anxiety scale, and its modification with corresponding self-rating versions will also be illustrated.

The modifications of the HAM-D and HAM-A have been made both on the basis of clinical validity (correspondence to experienced psychiatrists’ global assessments as index of validity) and on the basis of psychometric validation analyses.

The clinical validity of a depression rating scale refers to how the total scores are standardized into the clinical categories of no depression, mild depression, moderate depression, and severe depression. Thus, the HAM-D₁₇ version has been standardized both in primary depression and in depression secondary to stroke, i.e. post-stroke depression. A score of 7 or less on HAM-D₁₇ equals no depression or remission. A score of 13 equals less than major depression, while a score of 18 or more equals moderate or major depression. A score of 25 or more equals severe depression.

The psychometric validation analysis refers to the ability of a scale to measure the same dimension in different groups of patients, e.g. primary depression or post-stroke depression, or in the same group of patients during a trial with weekly assessments. Therefore, the psychometric validation analysis is an important factor when using the total score of a scale for standardization.

The psychometric validation analysis used to evaluate whether the total score is a sufficient statistic for standardization is the item response theory model [8].

A sufficient statistic signifies that the total score provides us with information about the relation between the various symptoms on the dimension under examination. In depression, the symptoms of depressed mood (hopelessness), lack of interests or work functioning (helplessness), and guilt feelings (worthlessness) should be placed with increasing severity distances on the dimension of depression. The nonparametric item response analysis is the Mokken analysis and the one-parametric item response analysis is the Rasch analysis [8].

Figure 1 shows the item response theory model analysis of the HAM-D subscale (HAM-D₆) which has been found to have an adequate clinical as well as psychometric validity when measuring depressive states in both primary depression [9] and in post-stroke depression [10]. The three items within Beck's theory of depression (table 1), depressed mood, lack of interests, and guilt feelings, have different prevalence as illustrated in figure 1. Hence, depressed mood and lack of interests have the highest prevalence (i.e. are present even in the milder forms of depression), whereas guilt feelings have a rather low prevalence (i.e. are present in the more severe forms of depression). However, depressed patients with a positive score on guilt feelings also need to have a positive score on the items of depressed mood and lack of interests. If not, the patient is not to be considered as a depressed patient. This requirement is the basic rule in the item response theory analysis. This hierarchy of the prevalence of the six items in figure 1 has been found to emerge when testing males versus females, young patients versus elderly patients.

The items in HAM-D₆ (fig. 1) can be considered as ‘psychic’ symptoms of depression apart from the item of tiredness and pains (general somatic symptoms). In a medical sense, the symptom of tiredness is a systemic symptom, i.e. it has no localization in the body, in contrast to pains. Thereby pains are more ‘somatic’ than tiredness. The diagnostic specificity of tiredness is low compared to pains in the medically ill patient. When placed in the order of prevalence as shown in figure 1, depressed mood, and tiredness or pains are important indicators of depression severity. However, depressed mood and tiredness or pain can be important indicators of cancer severity if ascribable to the cancer itself. It is, therefore, important to differentiate between psychic and somatic symptoms of depression or anxiety when evaluating the medically ill patient.

The psychometric analysis relevant for discrimination between psychic depression and neurovegetative depression at the level of symptoms is the Principal Component Analysis in which the bi-directional factor or component, via item loadings, can identify the negative loadings versus the positive loadings, thereby subdividing the HAM-D or HAM-A into two subscales [1].

Table 2 shows the rating scales for measuring severity of the states of depression and anxiety selected for this overview, also indicating which of these scales are to be found in the appendix. The platforms of this review are the HAM-D and HAM-A; however, the new developments (e.g. the Inventory for Depressive Symptomatology, IDS [11]) and self-rating scales with reference to HAM-D and HAM-A have been considered as well.

Accepting [2] that the inescapably clinical nature of a diagnosis is to seek to distinguish between states of depression or anxiety in which a trigger is present (secondary to psychosocial stressors or to medical disease), and those which seem to arise out of nowhere (primary depression or primary anxiety), the etiological issues are collected in table 3, including bipolar versus unipolar depression.

Reliability statistics are conventionally considered to provide consistency and stability of the rating scales as a platform on which the validation procedures can be evaluated. However, we need to know the clinical validity as well as to which extent the total score is a sufficient statistic before we can perform reliability statistics.

As discussed elsewhere [12], the item response theory model implies that rating scales with high validity, e.g. HAM-D₆, also have high inter-rater agreement. In both clinician-administered scales (e.g. HAM-D and HAM-A) and in patient-administered questionnaires, reliability in the form of Cronbach's coefficient a is used to express internal consistency. However, the length of the scale is part of the a coefficient, so that briefer scales have lower internal consistency simply due to having fewer items. Test-retest reliability refers to stability of total scores over repeated testings. For symptom scales, the time interval between the test-retest evaluation is crucial because scales with a brief time frame like Spielberger's State anxiety scale by their nature give a more fluctuating total score than depressive states.

It is beyond the scope of this review to consider the various anti-anxiety or anti-depressive therapies. However, with reference to primary versus secondary hypertension, we can say that the rating scales for anxiety and depression selected in this review are as quantifiable as the clinical measures for antihypertensive drugs with regard to validity and reliability.

The HAM-D₁₇ is the archetypical scale for the measurement of depression severity. The first version of the HAM-D was released in 1960 [13], and over the years several attempts to improve the definitions of the seventeen items have been made [14]. Most of these attempts have been made with reference to the 1967 version [15] in which 8 items are scored on a 3-point Likert scale (0-2), while 9 items are scored on a 5-point Likert scale (0-4). In this version, the total score ranges from 0 (no depression) to 52 (maximum depression). The standardization of these versions of the HAM-D₁₇ is: 0-7 (no depression), 8-12 (doubtful depression), 13-17 (mild depression), 18-24 (major depression), and 25-52 (severe depression).

Experience with other symptom scales such as the Brief Psychiatric Rating Scale [16] has shown that most psychiatrists are able to score on Likert scales from 0 to 6, where 0 = no symptoms, 1 = doubtful, 2 = very mild, 3 = mild to moderate, 4 = moderate, 5 = marked to severe, and 6 = extreme. Paykel [17] developed a version of the HAM-D with Likert scales from 0-6 on all the items (Clinical Interview for Depression, CID). An overview of the CID has recently been released [18]. However, it is not possible to extract the HAM-D₁₇ from the CID. In the Clinical Interview for Depression and Related Syndromes (CIDRS) which is based on the CID [19], it is possible to derive the HAM-D₁₇ with Likert scales from 0 to 6 for all the items. The etiological issues from section F in CIDRS are shown in table 3.

From the CIDRS, it is also possible to extract the Montgomery Åsberg Rating Scale for Depression (MADRS) [20]. However, the MADRS derived from the CIDRS has Likert scales from 0 to 6 with anchored definitions, in contrast to the original MADRS in which the uneven points in the Likert scale (1, 3, 5) have no definitions, i.e. are empty boxes.

Validation studies on the HAM-D₁₇, in which clinical validity was evaluated using experienced psychiatrists as index of validity and psychometric validation was performed using item response theory models (total score being a sufficient statistic), have shown that only six of the seventeen items (fig. 1) are acceptable [19]. Table 4 shows the HAM-D₆ with the corresponding IDS₆. The rank-order from top to bottom in table 4 shows how each item provides information about the dimension of depression severity covered by it. Thus, the three top-listed items (depressed mood, lack of interests or involvement, and fatigue) cover the mild to moderate dimension of depression, i.e. are present in mild to moderate cases. The next three items (anxious mood, guilt feelings and psychomotor retardation or slowing) are especially present in the more severe cases. However, patients with positive scores on these down-listed items in table 4 also have to have positive scores on the three top-listed items (if not, then the patient is not a typical depressed patient).

Table 3 shows how each of the items is scored on the original HAM-D version (0-2/0-4), and on the CIDRS version (0-6). On the IDS₆, the quantifier is a Likert scale from 0 to 3.

As discussed by Marzolf [1], the syndrome is a central tendency which is best thought of in statistical terms. As regards the individual items in the syndrome, each item may become clinically significant only when it exists in an amount above a certain critical point [1]. It is therefore very important to use rating scales with definite anchoring points, and this is the reason why a scale such as the MADRS is not included in this review.

The IDS has both a clinician version corresponding to the HAM-D but with a different quantifier (as shown in table 5), and a patient version (self-rating version, IDS-SR).

Max Hamilton himself was not interested in developing a self-rating version of his scales (neither the HAM-D nor the HAM-A). Attempts to develop a self-rating version of HAM-D covering all 17 items have not been successful, but the version corresponding to HAM-D₆ has been found acceptable [21, 22]. From the Symptom Check List (SCL-90) [23], we have derived an SCL₆ corresponding to HAM-D₆ [24].

The Major Depression Inventory (MDI) is a self-rating scale designed to cover the ten depression symptoms in ICD-10 [25] (see appendix). By combining two of the items into one single item of guilt feeling (item 4 = lack of self-confidence and item 5 = guilt) the MDI also covers the nine depression symptoms in DSM-IV major depressive episode. In a study on interferon-a-induced depression in patients with chronic hepatitis C [26], we obtained a baseline rate of 6% for major depression using the MDI. However, after 12 weeks of therapy 33% developed major depression on MDI [26]. A principal component analysis on the patients at the 12 weeks visit identified a general component (all items in the MDI correlated positively) and a second factor, a bi-factor according to Marzolf [1]. In this second factor, the five psychic depression items in MDI (depressed mood, lack of interests, lack of self-confidence, guilt feeling, and suicidal thoughts) were negatively loaded, while the neurovegetative items in MDI (tiredness, sleep, concentration, agitation, and appetite) were positively loaded.

The DMS-IV algorithm for major depression states that depressed mood and/or lack of interests have to be present. Using the total score of MDI in psychosomatic medicine should be supplemented by a subscale of the psychic versus neurovegetative items.

The HAM-A is the most frequently used scale worldwide for the measurement of anxiety. This scale was developed by Max Hamilton in 1959 [27]. The original 1959 version included 13 items, but was revised by Hamilton in 1969 [28]. This 1969 version of the HAM-A included 14 items, as the item of somatic general symptoms was subdivided into muscular and sensory symptoms. Unfortunately, many handbooks on rating scales [e.g. 29, 30] recommend the 1959 version despite the fact that since the 1970s all intervention studies have used the 1969 version, which was also the one recommended by Hamilton himself [31] (see appendix).

From a practical psychometric point of view, the HAM-A₁₄ illustrates how principal component analysis can be used in the validation procedure. Hamilton used principal component analysis in his 1959 and 1969 versions of the HAM-A to show that the first factor or component is a general factor and the second factor or component is a bipolar or dual factor with negative loadings on the psychic anxiety items and positive loadings on the somatic anxiety items.

table 6 shows the results obtained by Pichot et al. [32] when performing a principal component analysis on the baseline dataset of 411 patients with anxiety disorder. The first factor or component was a general factor because all items are positively correlated. The second factor or component was dual with negative loadings on the psychic anxiety items and positive loadings on the somatic anxiety items (table 6). This result is very much in harmony with Hamilton's own findings. Moreover, Pichot et al. [32] showed that a rotation of the factors did not provide more information.

Rasch [33] developed the first one-parameter item response theory model for the purpose of testing to what extent the total score of a rating scale is a sufficient statistic because he had realized [34] on the one hand that a general factor in itself is no argument for adding up items on a scale in order to measure a dimension such as intelligence or depression, and on the other hand that the many ways of performing factor rotations are without scientific basis (a matter of trial and error until the investigator feels satisfied).

The identification of the separation of psychic versus somatic anxiety symptoms in the HAM-A (table 6) was the background on which Rickels et al. [35] demonstrated that imipramine was superior to diazepam on the psychic anxiety symptom but not on the somatic. The total score on HAM-A is often too biased as to the somatic anxiety symptoms in patients with anxiety disorders. In patients with primary depression, the psychic anxiety symptom is much more present than the somatic anxiety symptoms [36, 37].

When validating the HAM-A with reference to the work by Snaith et al. [38], we showed that the six selected items of the HAM-A (table 7) fulfilled the item response theory model [39]. The factor of psychic anxiety symptoms (table 6) includes such items as depressed mood and sleep which are much more associated with depressive disorders than with anxiety disorders.

In the medical conditions associated with anxiety (e.g. hyperthyroidism or hypertension), it is the psychic anxiety symptoms that are most valid for the evaluation of specific anti-anxiety interventions. With regard to anxiety-inducing drugs, the ß-adrenergic agonists used in the treatment of asthma are among the most well-known anxiogenic drugs.

Over the last decades, many psychologists have defined anxiety in terms of the patients’ score on a self-rating scale found in Taylor's Manifest Anxiety Scale [6] or in Spielberger's State Trait Anxiety Scale [5]. As discussed by Spielberger et al. [5], five of the items can be considered as core items of psychic anxiety, covering nervousness, worry, apprehension, fearfulness and tension (table 1). These items are included in the HAM-A₆ (table 7).

Among comprehensive anxiety scales are the Symptom Checklist (SCL-90) and Kellner's Symptom Questionnaire (SQ-92) [7]. Both of these self-rating scales include subscales measuring anxiety, depression and hostility. The SCL-90 is based on symptoms to be rated by the patients themselves on a Likert scale from 0 to 4, whereas SQ-92 is actually a checklist scale, i.e. each item is scored yes or no. Moreover, the SQ-92 has items measuring well-being analogue to the HADS. In both the anxiety and depression subscales of SQ-92, six items are well-being questions (see appendix). All the symptom-formulated anxiety items in SQ-92 cover psychic anxiety symptoms including the five anxiety symptoms Spielberger has found most important (nervous, tense, worried, panicky, feeling that something bad will happen). In a placebo-controlled study on the anti-anxiety effect of propanolol, the SQ-92 anxiety subscale was found more sensitive than the HAM-A in discriminating between active treatment and placebo [40], probably because the HAM-A includes too many somatic anxiety items.

The 17 items of the SQ-92 that measure depressive symptoms cover the psychic depression symptoms (see appendix).

Like the SQ-92, the SCL-90 has been developed with reference to factor analytic studies. Most studies employing SCL-90 have by use of principal component analysis identified a general factor with eigenvalues around 5 times higher than the second factor or component. However, most studies have then performed different rotations (orthogonal or oblique) with rather diverse outcomes. In a recent study with principal component analysis on 555 patients with anxiety disorders consecutively assessed at the Day Hospital of the Mental Health Centre North Zealand [41], we identified as the first component a general factor (in which all 90 items were positively correlated) and a second component which was a dual factor separating depression (negative loadings) and anxiety (positive loadings). The anxiety items of SCL-90 corresponded to HAM-A₁₄ (see appendix), whereas the SCL-6 items for depression were among the negative-loaded items.

In the Sourcebook of Adult Assessment Strategies [29], there are very few scales for depression (HAM-D, the Melancholia Scale for clinicians, the Center for Epidemiological Studies as well as the Zung Self-Rating Depression Scale for patient ratings), while there are around 15 self-rating scales for anxiety.

Scales for specific anxiety profiles such as panic, social phobia and obsessive-compulsive states can be adequately covered by global items rather than symptom scales. When describing panic attacks, Freud himself [42] noticed that one single symptom, e.g. ‘my heart pounds’, often dominates the clinical picture. Therefore, panic attacks are most practically measured by asking globally (on the verge of an attack versus a clear attack). Similarly, for obsessive thoughts or compulsive behavior, the time occupied or spent on these factors within a waking day is the clinically most valid measure.

The Anxiety Profile Inventory (API), which has been adopted for use in Denmark when screening for anxiety disorders in the family doctor setting, is shown in the appendix together with the interview version for the family doctor when assessing patients with common mental disorders in the GP setting.

The use of definite anchoring points in rating scales or questionnaires has special importance in the API scale and in the CIDRS-GP version. Thus, in the CIDRS-GP, a score of 4 or more on the suicidal items should result in the interviewer acting on this information. As regards item 14, a score of 4 or more has diagnostic consequences.

Taylors Manifest Anxiety Scale, Spielberger’s original state anxiety scale as well as Kellner's SQ-92 scale include items with both negative and positive formulations (symptoms versus well-being). This approach was used to ensure that the patients did not automatically complete these questionnaires ‘without consciousness’. Among the few to study the validity of using both negative and positive phrasing in anxiety patients, Eysenck has shown with his Neuroticism questionnaire that the negative (symptom-like) formulation is more valid. We have shown that the sensitivity of anxiety scales in discriminating between active treatment and placebo treatment is much higher for the negatively formulated items [43]. In the Kellner SQ-92 (see appendix), most of the items in the anxiety or depression subscales are negatively formulated, and he has shown that the anxiety scale has a high sensitivity in discriminating between propanolol and placebo [40].

When measuring anxiety or depression in the setting of the medically ill patient, we should stop trying to be physicians of the body and should recognize that we are physicians of the psyche (mind). There seems to be little doubt that there are such clinical phenomena as anxiety or depression for which the meaningful dimension is the degree of severity. In this review, we have focused on such dimensions of severity in anxiety and depression that at quantifiable, like the measurement of hypertension in the medical setting. Consequently, we can speak of primary and secondary anxiety and depression in the same way as we speak of primary and secondary hypertension.

The psychopathological symptomatology captured by the Hamilton Anxiety and Depression Scales can be clinimetrically validated for quantifiability by item response theory analysis. Such validated rating scales for anxiety and depression also have adequate reliability. The psychometric term sensitivity is not an extra characteristic, as this is part of the validation procedure.

We can consider the HAM-D₆ and HAM-A₆ as the most valid measures as to the core items of the states of depression and anxiety, respectively. However, the IDS₃₀ is relevant when discriminating between typical and atypical depression, while the HAM-A₁₄ and the SCL-20 are the most relevant scales when discriminating between psychic and somatic anxiety. The CIDRS-GP should be considered when covering the common mental disorders in the general practice setting.

As discussed elsewhere, the time has come to evaluate when a state of depression or anxiety is a clinical case for treatment, taking the primary or secondary etiology into account. The etiological issues covered by the CIDRS section (see appendix) should be consulted in this respect.