The Brown Assessment of Beliefs Scale: Reliability and Validity

The Brown Assessment of Beliefs Scale is a seven-item clinician-administered semistructured scale designed to assess delusionality of beliefs in a broad range of psychiatric disorders. To administer the scale, the dominant belief (obsession, concern, idea, worry, or delusion) that has preoccupied the patient during the past week is first established. For obsessions, the belief and associated consequences underlying the obsessional thought must be determined. An example of an underlying belief in OCD is that AIDS could be contracted by touching any object outside the home. In body dysmorphic disorder, the belief might consist of the thought that one is among the ugliest people in the world or a greatly exaggerated view of the ugliness of minor skin blemishes. An example of a false belief in mood disorder with psychotic features is that one is half angel and half human.

The Brown Assessment of Beliefs Scale has specific probes and five anchors for each item, with descriptions corresponding to each anchor. An example of a probe is the following, which assesses conviction: “How convinced are you that you'll get sick from eating any food that was touched by anyone other than yourself? Are you certain your belief is accurate?” The score for each item ranges from 0 (nondelusional, or least pathological) to 4 (delusional, or most pathological). Ratings represent an average score for the past week.

The scale's original version contained 15 items culled from a review of the literature on delusionality (15). The items conviction, perception of others' views of beliefs, fixity of ideas, stability of beliefs over time, ego-syntonicity, and attempt to disprove beliefs were derived from the literature on delusions (8, 16) and on insight in OCD (4, 7, 11, 12). Several items similar to those in the Yale-Brown Obsessive Compulsive Scale (17, 18) were also derived from the literature on delusions: time preoccupied, time free from concerns, distress, control, and interference with functioning. Three items were derived primarily from the schizophrenia literature on delusions: need to convince others, bizarreness, and coherence (2, 3, 8). Insight—an awareness of the psychiatric etiology of the belief—has been written about extensively (19) and is included in the Scale to Assess Unawareness of Mental Disorder (10, 20). Three items were added following our pilot study: explanation of differing views, ideas/delusions of reference, and extension, which was derived from the scale of Kendler et al. (8).

The original 15-item version of the Brown Assessment of Beliefs Scale was administered to 46 subjects in a pilot study: 18 with OCD, 18 with body dysmorphic disorder, and 10 with mood disorder with psychotic features (15). Subjects were obtained from the clinical practices of the first two authors (J.L.E. and K.A.P.) and from a hospital inpatient unit. One of the authors (J.L.E., K.A.P., or D.A.B.) conducted an audiotaped interview during which the scale was administered; interrater reliability was determined by subsequent independent ratings of 10 audiotaped interviews by the other two interviewers. Interrater reliability was good to excellent. Intraclass correlations (ICCs) were 0.95 for the total score and ranged from 0.81 to 0.99 for individual items, with the exception of interference with functioning and control, which were subsequently deleted from the scale. Cronbach's alpha coefficient was 0.83, and factor analysis yielded three factors (core features, severity, and psychosis), accounting for 67% of the variance.

Subjects. The 18-item scale was administered to 50 subjects who had not participated in the pilot study; 20 subjects had OCD, 20 had body dysmorphic disorder, and 10 had mood disorder with psychotic features (eight had bipolar affective disorder, and two had major depression). Thirty subjects (60%) were women, and the mean age of all subjects was 35.2 years (SD=10.7). Subjects were referred from OCD and body dysmorphic disorder outpatient programs at two university-based hospitals, from new referrals to these programs, and from hospital inpatient units. All diagnoses were made according to DSM-III-R criteria, by a psychiatrist (J.L.E., K.A.P., or D.A.B.) with extensive clinical experience and expertise in OCD, body dysmorphic disorder, or psychotic mood disorder, on the basis of a detailed clinical interview and available records. In addition, each patient with OCD was assessed with the Yale-Brown Obsessive Compulsive Scale symptom checklist and a semistructured clinical interview to obtain additional information about OCD (21). Subjects with body dysmorphic disorder were diagnosed by using a reliable semistructured diagnostic instrument for body dysmorphic disorder (22) modeled after the Structured Clinical Interview for DSM-IV (23). Subjects were diagnosed as having psychotic mood disorder only if there was diagnostic agreement among the research psychiatrist, the admitting psychiatrist, and the treating psychiatrist based on chart review and patient interview. Patients were at various stages of treatment; therefore, their symptoms were at different levels of severity when the scale was administered. Written informed consent was obtained from all subjects.

Instrument administration. One of four raters administered the Brown Assessment of Beliefs Scale, which was audiotaped. Three of the raters (J.L.E., K.A.P., and D.A.B.) were developers of the scale, and the fourth rater (K.D.A.) was trained in its use by observing five interviews conducted by one of the scale's developers. All raters were experienced clinicians. The dominant belief (obsession, concern, idea, worry, or delusion) that had preoccupied the subject during the previous week was first established. The rater then asked the scale's specific probes.

Reliability. To assess interrater reliability, the audiotaped interviews were independently rated by the three other interviewers. All scoring was done blind to the other interviewers' ratings. Test-retest reliability was evaluated by administration of the Brown Assessment of Beliefs Scale by the same interviewer to 27 patients (16 patients with body dysmorphic disorder and 11 with OCD) on two different occasions separated by a 6-day interval. ICCs were calculated to determine reliability (24).

Discriminant and convergent validity. To assess discriminant and convergent validity, the interviewers administered the Yale-Brown Obsessive Compulsive Scale or the Body Dysmorphic Disorder–Yale-Brown Obsessive Compulsive Scale (a slightly modified version of the scale that assesses severity of body dysmorphic disorder) (25), the Brief Psychiatric Rating Scale (BPRS) (26), the Dimensions of Delusional Experience scale (8), the Scale to Assess Unawareness of Mental Disorder (10), and the Fixity of Beliefs Scale (11). Two self-report scales were completed: the Characteristics of Delusions Rating Scale (a visual analog scale that measures 11 dimensions of delusionality) (9) and the Beck Depression Inventory (27). We created a total score on the Characteristics of Delusions Rating Scale by summing the scores on each item. All ratings were obtained on the same day.

On the day the Brown Assessment of Beliefs Scale was administered, the treating clinician (who did not administer the scale) independently made a global rating of delusionality using anchors from the experimental insight question of the Yale-Brown Obsessive Compulsive Scale (0=excellent insight, 1=good insight, 2=fair insight, 3=poor insight, or 4=no insight; delusional). For patients with OCD or a psychotic mood disorder, not all of the global raters were experts in delusionality. However, the global rater for patients with body dysmorphic disorder had expertise in delusionality and extensive familiarity with the literature on delusionality.

Sensitivity to change. We conducted a separate study to assess the scale's sensitivity to change in delusionality during treatment. As part of a multicenter study of sertraline in the treatment of OCD, the Brown Assessment of Beliefs Scale was administered to 38 consecutive subjects at the study baseline visit and after 16 weeks of unblinded administration of sertraline (mean dose at 16 weeks=190 mg/day, SD=26.6, range=50–200) (28). Three of these subjects also participated in the reliability and validity study described above. Study coordinators and the principal investigators were trained in the administration of the scale by one of the authors (J.L.E.), who visited each site. The training session involved an orientation in which the design, purpose, and administration of the scale were described. To establish interrater reliability, the study investigators watched and discussed videotaped interviews of the scale's administration to two patients with OCD and subsequently independently rated a third interview.

After extensive experience with administration of the 18-item version of the Brown Assessment of Beliefs Scale, the scale was reduced to seven items because we thought that the length of the 18-item version would limit its use. In addition, we considered certain items relatively weak on the basis of their psychometric properties and the richness of the information supplied. Time occupied by beliefs, time interval free from beliefs, control, interference with functioning, and distress were deleted because these items were primarily measures of illness severity, which is assessed by other instruments, such as the Yale-Brown Obsessive Compulsive Scale, rather than core features of delusionality. In addition, correlations between these items and the total scale score were not strong (r=0.38, 0.37, 0.46, 0.27, and 0.13, respectively). The items coherence, need to convince others (pressure), and extension were dropped because of poor correlation with the total scale score (r=0.02, 0.31, and 0.09, respectively) and because they did not contribute substantial information about delusionality. Stability of beliefs over time was deleted because, although correlation of this item with the total scale score was high (r=0.68), we thought that this concept would be better assessed by prospective ratings with the scale. Although there is substantial literature on the importance of evaluating bizarreness, or plausibility, in delusional patients (29, 30), this item was not retained because of low correlation with the total score (r=0.02) and weak interrater reliability. Ego-syntonicity, a frequently discussed concept in the OCD literature, was dropped because of relatively weak correlation with the total score (r=0.32) and difficulty operationalizing and communicating this abstract concept to patients.

One item, ideas/delusions of reference, is scored but not included in the scale's total score. This item was retained because it yields important data about delusionality, particularly for certain disorders such as body dysmorphic disorder, olfactory reference syndrome, and other types of delusional disorder. However, because ideas of reference are not relevant to all disorders characterized by delusional thinking, this item's score is not included in the total score.

The final seven-item scale can be administered in only 10 to 15 minutes and includes items that have good psychometric properties and, in our view, are most relevant to the construct of delusionality. The scale's final items are conviction, perception of others' views of beliefs, explanation of differing views, fixity of ideas, attempt to disprove beliefs, insight, and ideas/delusions of reference.

ICCs were used to determine interrater and test-retest reliability (24), and Pearson's correlation was used to determine correlations between each item and the total score minus that item. Cronbach's alpha coefficient was used to evaluate internal consistency. Principal components factor analysis was conducted using orthogonal (varimax) rotation. The number of factors was based on an examination of eigenvalues greater than 1. Pearson correlations were also done to determine convergent and discriminant validity, i.e., the relationship between total score on the scale and scores on other scales.

The sensitivity and specificity of the scale in classifying patients as delusional or nondelusional, and the percentage of subjects correctly classified, were calculated for body dysmorphic disorder by using a variety of total score thresholds ranging from 14 to 20 and the “gold standard” rating from the global Yale-Brown Obsessive Compulsive Scale experimental insight question completed by the treating clinician (score of 4=delusional; score of 0–3=nondelusional). This analysis is presented only for patients with body dysmorphic disorder because the global “gold standard” ratings for body dysmorphic disorder were made by a clinician with extensive familiarity with the literature on delusions (K.A.P.).

To evaluate the sensitivity of the scale to change with treatment, a t test comparing mean baseline score and mean posttreatment score was calculated for the 38 patients with OCD who participated in the sertraline study.

ICCs demonstrated good agreement across the four raters for the total score and individual item scores (table 1). As shown in table 2, correlations between each item and the total score minus that item ranged from r=0.38 to r=0.85. Cronbach's alpha coefficient was 0.87, indicating acceptable homogeneity of the scale. The individual item test-retest ICCs ranged from 0.79 to 0.98, with a median of 0.95.

Factor analysis identified one factor with eigenvalues greater than 1 when conducted with the seven items. All seven items loaded strongly together on the factor identified, with loadings ranging from 0.48 to 0.92. This factor accounted for 56% of the variance.

As shown in table 3, the total score on the Brown Assessment of Beliefs Scale was not significantly correlated with total score on the Beck Depression Inventory or the Yale-Brown Obsessive Compulsive Scale. The Brown Assessment of Beliefs Scale correlated only weakly with the BPRS, although this correlation reached statistical significance (table 3). As a further test of discriminant validity, we analyzed correlations between the total score on the Brown Assessment of Beliefs Scale and the 18 individual items on the BPRS. All correlations were nonsignificant except for emotional withdrawal (r=0.37, p=0.008) and unusual thinking (r=0.39, p=0.005), which assesses the presence of delusions.

Convergent validity was demonstrated with the Characteristics of Delusions Rating Scale total score and the two items that assess delusional thinking on the Scale to Assess Unawareness of Mental Disorder (table 3). A high correlation was also found between the conviction item on the Brown Assessment of Beliefs Scale and conviction items on the Dimensions of Delusional Experience, the Fixity of Beliefs Scale, and the Characteristics of Delusions Rating Scale (table 3).

Discriminant and convergent validity were also assessed for each diagnosis separately. Findings for each diagnosis were similar to those already described with one exception: correlation of the total score on the Brown Assessment of Beliefs Scale with the BPRS did not reach statistical significance for the OCD group (r=0.22, p=0.36) and the mood disorder group (r=–0.47, p=0.17).

Using the “gold standard” rating from the global Yale-Brown Obsessive Compulsive Scale experimental insight question for subjects with body dysmorphic disorder, we calculated the percentage of subjects correctly classified as delusional or nondelusional by the Brown Assessment of Beliefs Scale. Total score thresholds ranged from 14 to 20. In addition to achieving a cutoff point, patients could be classified as delusional only if the first item, conviction, was scored as 4—i.e., the patient was completely convinced that the belief was accurate.

Sensitivity and specificity calculations using a variety of total score thresholds indicated that the threshold for identifying subjects judged to be delusional is a total score of 18 out of a possible total score of 24. This cutoff point, in combination with a score of 4 on the first item, conviction, correctly classified 90% (N=18) of the 20 subjects assessed (six of the six delusional patients and 12 of the 14 nondelusional patients). It produced no false negatives (100% sensitivity)—that is, every patient considered delusional by the expert rater was categorized as delusional by the scale. The specificity was 86%—i.e., two patients considered nondelusional by the expert rater were categorized as delusional by the scale.

In the sertraline study, the mean Yale-Brown Obsessive Compulsive Scale score decreased by 49% (mean baseline score=25.0, SD=3.6; mean posttreatment score=12.7, SD=5.5). The mean Brown Assessment of Beliefs Scale score decreased by 51% (mean baseline score=8.5, SD=4.2; mean posttreatment score=4.2, SD=4.4) (t=5.47, df=37, p<0.001, paired t test), indicating that the scale is sensitive to change. The correlation between mean change in total Yale-Brown Obsessive Compulsive Scale score and mean change in total score on the Brown Assessment of Beliefs Scale was 0.43 (p<0.008), indicating that improvement in degree of delusionality as measured by the scale was correlated with but not identical to improvement in severity of OCD symptoms.