Assessing Eating Disorder Thoughts and Behaviors: The ...

C 2006) Cognitive Therapy and Research, Vol. 30, No. 5, October 2006 ( DOI: 10.1007/s10608-006-9003-3

Assessing Eating Disorder Thoughts and Behaviors: The Development and Preliminary Evaluation of Two Questionnaires Myra J. Cooper,1,4 Gillian Todd,2 Rachel Woolrich,1 Kate Somerville,1 and Adrian Wells3 Published online: 31 October 2006

This paper describes the development of two measures. The first is designed to assess eating disorder-related automatic thoughts; the second is designed to assess a wide range of eating disorder-related behaviors. Principal components analysis identified three dimensions of thoughts: positive thoughts about eating, negative thoughts about eating, and permissive thoughts. Principal components analysis also identified six dimensions of behavior related to: shape and weight, bingeing, dieting, food, eating, and overeating. Both measures possess promising psychometric properties, including good construct and criterion-related validity. Both successfully discriminated eating disorder patients from dieting and non-dieting groups. The two measures may be useful additions to those currently available to researchers (and clinicians) interested in eating disorders. KEY WORDS: eating disorders; cognition; behaviors.

INTRODUCTION Cognitive theories of eating disorders, both bulimia nervosa (Fairburn, Cooper, & Cooper, 1986) and anorexia nervosa (Fairburn, Shafran, & Cooper, 1999; Garner & Bemis, 1982), are less developed than theories in many other disorders, particularly anxiety disorders (Wells, 1997). Two recent papers highlight the areas in which there is room for progress and development (Cooper, 2005; Cooper, Wells, & Todd, 2004). The former is a review paper, assessing the state of current cognitive theory in eating disorders; the latter proposes a revised cognitive model of bulimia nervosa. Ideas for the future in both papers include the development of reliable and valid

1 Isis

Education Centre, University of Oxford, Oxford, United Kingdom. of Cambridge, Cambridge, United Kingdom. 3 University of Manchester, Manchester, United Kingdom. 4 Correspondence should be directed to Myra J. Cooper, University of Oxford, Warneford Hospital, Oxford OX3 7JX, United Kingdom; e-mail: [email protected]. 2 University

551 C 2006 Springer Science+Business Media, Inc. 0147-5916/06/1000-0551/1

552

Cooper, Todd, Woolrich, Somerville, and Wells

measures of cognitions hypothesized to be important in these disorders, particularly as applied to revised cognitive models of eating disorders (e.g., Cooper et al., 2004). One problem is that current cognitive theories of eating disorders do not generally specify in detail the content of the thoughts and behaviors that maintain either disorder. At the same time, there are relatively few good self-report questionnaire measures of cognitions relevant to eating disorders. Cognitive theories also tend to focus primarily on diagnostic behaviors and not other commonly observed behaviors (e.g., in relation to food). While there are more measures of eating disorder-related behaviors, these are generally confined to single constructs and tend to reflect key diagnostic features such as dieting and binge eating. Currently, there is no multidimensional self-report questionnaire measure that incorporates both types of behaviors—those linked to key diagnostic features and those that are not but that may be important theoretically and that are observed phenomenologically. The current study aims to redress this balance, by including relevant behaviors beyond those associated only with key diagnostic features. Recently, we have suggested that three types of thought are important, particularly in the maintenance of bulimia nervosa. These are positive thoughts about eating, negative thoughts about eating, and permissive thoughts (Cooper et al., 2004; Cooper, Todd & Wells, 2000). These thoughts are all examples of “automatic thoughts” or “the moment to moment, unplanned thoughts . . . that flow through our minds throughout the day” (see Padesky & Greenberger, 1995, p. 5). Positive thoughts about eating concern the perceived benefits of eating (e.g., “If I eat, it will stop the pain”). Negative thoughts about eating concern the perceived adverse consequences (e.g., “I’ll get fat”). Permissive thoughts are those that make it easy to eat or keep eating (e.g., “One more bite won’t hurt”). Positive thoughts may also be important in anorexia nervosa (Cooper et al., 2004), where they may help maintain failure to achieve an adequate food intake and contribute to “drive for thinness.” In theoretical work (see Cooper, 2003; Cooper et al., 2004) we have proposed that the three different types of automatic thought act in a vicious circle to maintain disturbed eating behavior. In bulimia nervosa, for example, eating is triggered by positive thoughts about eating. These typically occur in the context of a negative self-belief. Once eating starts negative thoughts are activated. The dissonance between the positive and negative thoughts (and associated distress) is then resolved by permissive thoughts, and (binge) eating occurs. As the vicious circle revolves, eating (bingeing) is typically followed by negative self-beliefs—thus linking back into the start of the vicious circle and helping to maintain the behavior. Eating (bingeing) also serves to maintain the problematic thoughts (and their associated emotions) through a direct effect on thoughts, feelings, and also interoception. An important aspect of the model is that the relative balance of positive and negative thoughts is important in determining the presence of eating/bingeing, vomiting, and restricting. Thus, in bulimia nervosa, vomiting follows bingeing when the negative thoughts predominate. The importance of weight and shape to self-esteem is subsumed in the model in the notion of “underlying assumptions”—the “if . . . then” beliefs that

ED Thoughts and Behaviors

553

are thought to link negative or core self-beliefs with the maintaining factors.5 The difference between the behavior of eating and bingeing in our model is merely one of degree—although the reader will note that the underlying cognitions involved may differ in normal and abnormal eating—an area for much further research and a topic discussed in some detail in Cooper (2003). We have also suggested that several different behaviors are involved in the maintenance of bulimia nervosa (and also anorexia nervosa), including those that focus or remove attention from weight and shape, and behaviors that involve detailed attention to food and eating (Cooper et al., 2000, 2004). These may include, for example, wearing baggy clothes to avoid drawing attention to weight and shape, eating diet foods, reading cookery books and magazines, and putting only very small amounts of food onto a fork or spoon at a time. Currently, while a great deal of attention is given to the key “diagnostic” behaviors, other behaviors that may also help maintain the eating disorder, such as those mentioned above, have received less attention. The “eating” subscale for example, draws attention to the unusual behaviors those with eating disorders may display around food. Currently, these are not diagnostic but may, nevertheless, play a role in maintaining the disorder. Compared to other models of bulimia nervosa and anorexia nervosa (e.g., Fairburn et al., 1986; 1999; Garner & Bemis, 1982) and generic eating disorder models (e.g., Vitousek & Hollon, 1990), the model of bulimia nervosa we propose identifies several different types of automatic thought. It includes a detailed formulation of binge eating that highlights the inter-relationship of cognition, emotion, behavior, and physiology, which together operate in a vicious circle that incorporates all the elements of a good cognitive formulation (Butler, 1998). The model incorporates different levels of cognition, including both automatic thoughts, underlying assumptions, and core (negative self) beliefs. Unlike other models, it also explicitly incorporates behaviors related to eating, food, weight, and shape that are not core or diagnostic symptoms. These behaviors, like bingeing, act via positive or coping beliefs, to decrease arousal associated with negative self-beliefs. As such they play a significant role in the maintenance of the eating disorder. Overall, and unlike other models, our model attempts to explain not only the maintenance but also the development of the eating disorder. Empirical evidence, from our own work and that of others, supports the new model of bulimia nervosa that we propose (for a summary, Cooper, 2003; see Cooper et al., 2004), and there is also some evidence relevant to the usefulness of the model in anorexia nervosa (for a summary, see Cooper, 2005), where the different types of cognition and nondiagnostic behaviors may play a part in the maintenance of the disorder. Currently, few instruments are available that measure different content dimensions of negative automatic thoughts, or different subtypes of behavior relevant to eating disorders. A search of the literature revealed three self-report questionnaire measures that might be termed measures eating disorder-related “automatic thoughts,” using the definition of Padesky and Greenberger (1995). 5 These

are not discussed further here as the current paper focuses more on automatic thought level cognitions.

554


These are measures developed by Franko, Zuroff, and Rosenthal (1986); Phelan (1987); and Clark, Feldman, and Channon (1989). However, only one was developed using factor analysis or principal components analysis (Phelan, 1987) and, in this study, the three factors identified do not map onto a clear theory of eating disorders. Neither of the other two studies distinguished different types of thought. Factor analysis (and principal components analysis) has certain advantages—including the ability to identify, sort, and define in a systematic (and statistically powerful) way themes and dimensions that are theoretically relevant. Eating disorder-related behaviors assessed with self-report questionnaires are largely confined to binge eating (e.g., Gormally, Black, Daston, & Rardin, 1982; Hawkins & Clement, 1980) and dieting or restrained eating (e.g., the Dutch Eating Behavior Questionnaire-Restraint Scale, van Strien, Fritjers, Bergers, & Defares, 1986). However, all of these measures also include nonbehavioral items, e.g., “How much are you concerned about your binge eating?” There are also one or two relevant sub-scales, for example of the Eating Disorder Inventory, (Garner, Olmsted, & Polivy, 1983), but these are also not purely behavioral measures. The same is also true of measures that assess weight and shape-related behaviors (e.g., the Body Shape Questionnaire, Cooper, Taylor, Cooper, & Fairburn, 1987). These are generally measures of concern and satisfaction rather than measures of specific behaviors. Moreover, in addition to diagnostic behaviors, several important dimensions of behavior that may help maintain eating disorders (anorexia nervosa; Fairburn et al., 1999, and bulimia nervosa; Cooper et al., 2000) are not currently assessed at all in the literature—e.g., food and eating-related behaviors. Currently, no one multidimensional self-report questionnaire measure incorporating all the different diagnostic and non-diagnostic dimensions, and describing only behaviors, exists. Such a measure, which would tap a number of the dimensions identified as relevant to treatment, for example, based on the new bulimia nervosa model (e.g., Cooper et al., 2000), would be particularly useful in evaluating the impact of such a treatment, as well as testing aspects of the theoretical links proposed between cognition, emotion, and behavior. The aim of the current study therefore was to develop multidimensional self-report questionnaire6 measures of negative automatic thoughts and behavior. Two measures were developed. The first was a measure of automatic thoughts. We aimed to develop a measure of negative thoughts starting with the three categories (positive, negative, and permissive thoughts) that we have identified theoretically. The second was to develop a multidimensional measure specifically designed to assess behaviors and not also other constructs. In addition to diagnostic behaviors, non-diagnostic behaviors likely to be important in the maintenance of eating disorders were included. The study was therefore relevant both to evidence for the constructs suggested by the theory (particularly true of the measure of negative thoughts), and to the development of useful multidimensional assessment measures (particularly true of the single, but broad measure of relevant behaviors), which might also, of course, be useful in further development of theory and treatment. 6 Copies

of the two questionnaire measures and scoring keys can be obtained from the senior author.

ED Thoughts and Behaviors

555

METHOD Participants In the main study, 258 female university students and staff (mean age 25.7 years, SD = 8.1, range 17–63) participated. All were volunteers, recruited while attending lectures, societies, and clubs, or individually. In addition, discriminant validity was assessed in two samples of patients and dieters. Study 1 compared 14 female patients with anorexia nervosa (AN) (mean age 31.3 years, SD = 10.3), 17 nonsymptomatic dieters (mean age 29.3 years, SD = 5.7), and 18 non-dieting controls (mean age 29.8 years, SD = 8.3). Study 2 compared 12 female patients with bulimia nervosa (BN) (mean age 27.5, SD = 6.3), 18 non-symptomatic dieters (mean age 28.1 years, SD = 5.1), and 20 nondieting controls (mean age 28 years, SD = 4.9). The patients were recruited through their primary clinician or through the research database of a self-help organization. The non-clinical controls were recruited in the same way as those in the main sample, by asking hospital and University staff to volunteer, and through slimming and gym classes. Item Selection Items for the two questionnaires were generated by two clinicians (both trained cognitive therapists) experienced in the assessment and treatment of patients with eating disorders. Both clinicians drew on their experience of treating patients using cognitive therapy based on the model and treatment outlined by Cooper and colleagues (Cooper et al., 2000). Typical thoughts were identified to sample three theoretical themes: positive thoughts about eating, negative thoughts about eating, and permissive thoughts. Typical behaviors were identified to sample five logical themes: dieting, bingeing, shape and weight, interest in food, and specific eating-related behaviors. The latter included the typical unusual behaviors, such as slowed eating and cutting food into tiny pieces, often seen in those with an eating disorder. Fiftyseven items were chosen to sample the themes for the Thoughts Questionnaire, and 90 were chosen for the Behaviors Questionnaire. For the Thoughts Questionnaire belief in each item was rated on a Likert (0–100) scale. End points were anchored at “I do not usually believe this at all” and “I am usually completely convinced that this is true.” For the Behaviors Questionnaire a Likert (0–100) scale was anchored at “never” and “always,” and participants were asked to rate “how often you find yourself behaving in the way described.” The items were therapist generated (as is usual in the development of eating disorder self-report measures, e.g., Eating Disorder Inventory, Garner, 1991; Eating Disorder Examination—Questionnaire, Fairburn & Beglin, 1994), but both clinicians had extensive experience of reviewing patientgenerated thought and other records in therapy, and drew on this in generating the items. Measures In order to assess the construct validity of the two questionnaires, participants in the main study also provided information on height and weight and completed self-report measures of symptoms associated with eating disorders and depression.

556


These measures were: the Eating Attitudes Test (EAT: Garner & Garfinkel, 1979); Body Shape Questionnaire (BSQ: Cooper et al., 1987); Dutch Eating Behavior Questionnaire–Restraint subscale (DEBQR: Van Strien et al., 1986), and the Beck Depression Inventory (BDI: Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). The EAT is a widely used 40-item measure of the symptoms of eating disorders. It has good psychometric properties, including reliability (Cronbach’s α coefficient = 0.94 for combined anorexia nervosa and control samples) and validity (range of correlations with other eating disorder measures = 0.42–0.81) (Williamson, Anderson, Jackman, & Jackman, 1995). The BSQ is a 34-item measure of concerns about body shape. It has good concurrent validity (correlations with EAT range from 0.35 to 0.61, and satisfactory discriminant validity (Cooper et al., 1987) and excellent internal consistency (Cronbach’s α coefficient = 0.94, Evans & Dolan, 1993). The DEBQ-R measures dietary restraint with 10 items. It has good internal consistency (Cronbach’s α coefficient > 0.90), good construct validity, and high test-retest reliability (r = 0.92) (Allison, 1995). The BDI is a 21-item measure of depression symptomatology. It has good psychometric properties, including high concurrent validity and internal consistency (Cronbach’s α coefficient = 0.86) (Beck, Steer, & Garbin, 1988). Participants in the two discriminant validity studies also provided information on height and weight and completed the EAT and BDI. The self-report measures were selected based on their general applicability to both bulimia nervosa and anorexia nervosa and because they are all widely used in the evaluation of eating disorder-related psychopathology. As a first step, it was considered important to validate the new measures against general rather than specific measures (e.g., of binge eating or other cognitive measures) because administration of the extensive battery of measures needed to achieve the latter was not feasible in a single study. Procedure Participants in the main study completed the questionnaires individually, either at the time of recruitment or at home, returning them before a subsequent lecture or meeting. Each participant in the two studies designed to assess discriminant validity was interviewed individually to determine (confirm or exclude) DSM-IV diagnosis of an eating disorder using the Structured Clinical Interview for DSM-IV (SCID: Spitzer, Williams, & Gibbons, 1996). The interviewers were postgraduate clinical psychology students with extensive training in diagnostic interviewing, including with patients with eating disorders. All patients had already received a psychiatric diagnosis from a qualified clinician; thus the primary aim here, for the patients, was to confirm that this was accurate. No patient was found to have been misdiagnosed. The questionnaires were then completed. RESULTS Descriptive Statistics—Clinical Samples Body Mass Index [BMI = weight(kg)/height(m)2 ], EAT, and BDI scores for the samples in the two discriminant validity studies can be seen in Table I.

ED Thoughts and Behaviors Table I.

557

Mean Scores (and SDs) for AN and BN Patients, Dieters, and Controls on Descriptive Measures Study 1: Comparison of AN, Dieter, and Control groups

AN N = 14

BMI EAT BDI

Dieters N = 17 Controls N = 18

Mean

SD

Mean

SD

Mean

SD

16.2 56.1 29.3

1.7 24.3 15.3

25.5 9.8 3.7

4.0 8.7 3.3

20.5 3.9 2.5

1.5 2.8 1.9

Study 2: Comparison of BN, Dieter, and Control groups BN N = 12 SD

Dieters N = 17 Controls N = 20

Mean

SD

SD

Mean

SD

6.3 21.6 15.7

26.1 14.7 3.8

4.1 8.9 3.9

22.0 3.8 3.2

1.8 3.1 2.1

24.1 61.9 34.1

Note. BMI, Body Mass Index; EAT, Eating Attitudes Test; BDI, Beck Depression Inventory. AN, Anorexia Nervosa patients; BN, Bulimia Nervosa patients. SD, Standard deviation.

Thoughts Questionnaire Reliability Analysis Using the three theoretical themes (positive thoughts about eating, negative thoughts about eating, and permissive thoughts), reliability analyses were run to reduce the number of items sampled by each theme. Eighteen items with low itemtotal correlations (