journal of anxiety disorders volume 11 issue 1 1997

Pergamon Journal of Anxiety Disorders, Vol. 11, No. 1, pp. 33-47, 1997

Copyright Q 1997 Elsevier Science Ltd Printed in the USA. All rights reserved

0887-6185/97 $17.00 + .OO

PII SO887-6185(96)00033-3

Psychometric Properties and Diagnostic Utility of the

Beck Anxiety Inventory and the State-Trait Anxiety Inventory With

Older Adult Psychiatric Outpatients

ROBERT I. KABACOFF, PH.D.,’ DANIEL L. SEGAL, PH.D.,* MICHEL HERSEN, PH.D.,~

AND VINCENT B. VAN HASSELT, PH.D.’

‘Nova Southeastern University and ‘University of Colorado at Colorado Springs

Abstract-In order to assess the psychometric properties and diagnostic utility of the Beck Anxiety Inventory (BAI) and the State-Trait Anxiety Inventory (STAI) with older adults, these measures were administered to 217 older adult outpatients with mixed psychiatric disorders. Both the BAI and STAI scales demonstrated high internal reli- abilities. The BAI demonstrated good factorial validity, with a somatic anxiety and a subjective anxiety factor emerging. In contrast, the STAI did not evidence factorial validity, with analyses failing to support presence of state and trait anxiety factors. Both the BAI and Trait Anxiety scale of the STAI demonstrated discriminant validity in separating patients with a current anxiety disorder from patients without such a disorder. However, the State Anxiety scale of the STAI did not discriminate between these groups. When used to predict presence of an anxiety disorder, no single cutting score for either the BAI or STAl proved optimal, due to tradeoffs between sensitivity and specificity. Results suggest that both the subjective subscale and total score on the BAI can be somewhat useful as a quick screening instrument in detecting presence of a current anxiety disorder for older adult psychiatric outpatients, although results were not as strong as previous findings regarding screening tests for depression in the elderly. 0 1997 Elsevier Science Ltd

Key Words - Beck Anxiety Inventory, State-Trait Anxiety Inventory, Older psychiatric outpatients, Clinical cutoffs.

Requests for reprints should be sent to Robert Kabacoff, Ph.D., Nova Southeastern University, Center for Psychological Studies, 3301 College Avenue, Fort Lauderdale, FL 33314.

33

34 R. I. KABACOFF ET AL.

Epidemiological data suggest that anxiety is a common and debilitating disorder affecting a significant number of older adults. In the general community, 5.5% of people 65 years or older meet strict diagnostic criteria for an anxiety disorder (Reiger et al., 1988), with a similar rate found among elderly medical inpatients (Rapp, Parisi, & Walsh, 1988). Rates are even higher when significant but subsyndromal levels of anxiety are assessed in older adults. For example, in a survey of community dwelling elders, Himmelfarb and Murrell (1984) found that 17% of males and 21% of females experienced anxiety symptoms of sufficient severity to warrant intervention. Prevalence rates for anxiety disorders in older adults seeking mental health services are even more substantial. Within our own clinics, almost 30% of elder clients seeking outpatient services meet diagnostic criteria for an anxiety disorder.

Given such high prevalence rates for anxiety disorders in the elderly, there is a pressing need for assessment instruments that can accurately measure anxiety related symptomatology features in older adults. To avoid the fatigue effects common in long test batteries and structured interviews, instruments that can be quickly administered and easily completed are particularly desirable. Two of the most commonly used self-report measures of anxiety are the Beck Anxiety Inventory (BAI; Beck & Steer, 1990) and the State-Trait Anxiety Inventory (STAI; Spielberger, 1985). Due to the overlap in measurement between the clinical syndromes of anxiety and depression, the BAI was specifically developed to tap anxiety symptoms with minimal presence of depressive items. The STAI was developed earlier and designed to measure and differentiate between anxiety as a state and trait variable (Oei, Evans, & Cook, 1990).

Both instruments have been widely applied by clinical researchers, and the psychometric properties of these devices have been evaluated in a number of different clinical populations. For example, the operating characteristics of the BAI and its efficiency in screening for anxiety has been investigated with adolescent inpatients (Jolly, Aruffo, Wherry, & Livingston, 1993; Kumar, Steer, & Beck, 1993), adult outpatients with mixed psychiatric disorders (Beck, Epstein, Brown, & Steer, 1988; Steer, Beck, Brown, & Beck, 1993; Steer, Ranieri, Beck, & Clark, 1993), adult inpatients with mixed psychiatric disorders (Steer, Rissmiller, Ranieri, & Beck, 1993), anxious adult outpatients (Beck & Steer, 1991; Beck, Steer, & Beck, 1993), and non-clinical college undergrad- uates (Fydrich, Dowdall, & Chambless, 1992). Similarly, the STAI is a widely employed and reliable self-report scale that has been used extensively in applied psychology practice and research (see review by Spielberger, 1985). Indeed, Spielberger (1985) noted that the revised STAI (STAI-Y) had been successfully applied to high school and college students, adults, military personnel, prison inmates, and a wide variety of psychiatric and medical patients.

Despite popularity of these devices, to our knowledge there are no studies that specifically evaluate the psychometric properties and diagnostic utility of these instruments with the elderly. Indeed, in a recent review of the literature

ASSESSMENT OF ANXIETY IN THE ELDERLY 35

concerning the behavioral assessment and treatment of anxiety in the elderly, Hersen & Van Hasselt (1992) found that existing self-report and interview- administered instruments initially developed for younger adults have not been evaluated psychometrically with older populations in terms of norms, internal consistency, reliability, factorial structure, and validity.

Determining the reliability and validity of assessment instruments for anxiety in the elderly is an essential step in the identification and treatment of anxiety related disorders within that population. Since clinicians rarely have the luxury of administering a complete structured interview schedule (which can take up to 2 hours), short and valid measures of anxiety are particularly needed.

The current study was designed to evaluate the psychometric properties and diagnostic utility of the BAI and STAI when used with older adults seeking outpatient mental health services. Each instrument was evaluated for internal reliability, factorial validity, and discriminant validity. In addition, the predictive accuracy of the BAI and STAI for identifying presence or absence of an anxiety disorder in this population was investigated. Specifically, the sensitivity, specificity, positive predictive power, negative predictive power, and hit rate for detecting presence of a DSM-III-R anxiety disorder in older adults at various cutoff scores of the BAI and STAI were determined. Finally, the study evaluated whether or not combining BAI and STAI measures improved accuracy of diagnostic prediction.

METHOD

Subjects

The present study was part of a comprehensive evaluation of social and emotional adjustment in older adult psychiatric outpatients. The sample in- cluded 217 consecutive admissions to the Nova Community Clinic for Older Adults (NCCOA), a community-based outpatient facility that provides psychiatric services for non-psychotic adults who are 55 years of age and older.

The total sample was composed of 154 (71%) women and 63 (29%) men. The mean age was 65.86 years (SD = 8.54). Almost all subjects were White (96%), 44% were married, and 99% were living in an apartment, condominium, or private home. The median SES was 3 (on a scale of 1-5) as measured by the Hollingshead Socio-Economic Scale (2-factor index) (Hollingshead, 1975).

Sixty-three (29%) of the 217 subjects met criteria for a current anxiety disorder. Of the remaining 154 subjects, 133 (61%) met criteria for various current Axis I diagnoses (other than anxiety disorder), and 21 (10%) did not meet criteria for any current Axis I diagnosis.

Instruments

Structured Clinical Interview for DSM-III-R, Patient Edition With Psychotic Screen (SCID-P). The SCID-P (Spitzer, Williams, Gibbon, & First, 1988) is a

36 R. 1. KABACOFF ET AL.

semi-structured interview schedule employed to yield current and lifetime Axis I diagnoses. The SCID was selected as the validity criterion for anxiety disorder because of its formal ties to DSM criteria (American Psychiatric Association, 1987), and highly reliable diagnoses for similar older adult outpatients and inpatients. Subjects were considered as suffering from an anxiety disorder if they met formal criteria for any DSM-III-R anxiety disorder. Evidence for the reliability of SCID-based diagnoses with a subset of the current study sample (n = 40) was provided by Segal, Kabacoff, Hersen, Van Hasselt, and Ryan (1995), who reported kappa coefficients of .73 for the broad category of anxiety disorders, and .80 for the specific diagnosis of panic disorder. In an earlier study, the SCID had been found to yield highly reliable diagnoses for the broad diagnostic category of anxiety disorders (kappa = .77) in a mixed inpatient and outpatient older adult sample (Segal, Hersen, Van Hasselt, Kabacoff, & Roth, 1993).

Beck Anxiety Inventory (BAZ). The BAI (Beck et al., 1988) is a 21-item Likert scale self-report questionnaire measuring common symptoms of clinical anxiety, such as nervousness and fear of losing control. Respondents indicate the degree to which they are bothered by each symptom. Each symptom is rated on a 4-point scale ranging from 0 (not at all) to 3 (severely, I could barely stand it), and the total scores can range from 0 to 63, with higher scores corresponding to higher levels of anxiety. Thirteen items assess physiological symptoms, five describe cognitive aspects, and three represent both somatic and cognitive symptoms. The BAI has excellent internal consistency with psychiatric outpatients (alpha = .92, Beck et al., 1988; alpha = .94, Fydrich et al. 1992). The BAI has high concurrent validity with the SCL-90-R (Derogatis, 1983) Anxiety subscale (r = .81: Steer et al., 1993), and moderate concurrent validity with the Hamilton Anxiety Rating Scale (Hamilton, 1959) in 367 outpatients with anxiety disorders (r = S6: Beck & Steer, 1991). Common cutting scores of 10 suggest mild anxiety, with 19 reflecting moderate anxiety.

State-Trait Anxiety Inventory, Form Y (STAZ). The STAI-Y (Spielberger, 1985) is a 40-item Likert scale that assesses separate dimensions of “state” anxiety (items l-20) as well as “trait” anxiety (items 21-40). Each item is rated on a 4-point intensity scale. Both STAI-Y State (S-Anxiety) and Trait (T-Anxiety) scales were developed as unidimensional measures.

Procedure

As part of the normal clinic routine, clients were administered the SCID-P, BAI, and STAI-Y as part of the standard intake battery of psychological tests given to all clients evaluated at NCCOA. All assessment instruments were completed within an initial evaluation time of two weeks, and prior to inter-


vention. All measures were administered by advanced doctoral level students, all of whom were trained in administration and scoring.

RESULTS

Characteristics qf the BAI

Cronbach’s coefficient alpha for the BAI was 0.9, with item-total correlations ranging from 0.37 to 0.69. Discriminant validity was assessed by exam- ining mean score differences between patients who met criteria for a current diagnosis of an anxiety disorder, and patients who did not meet criteria for a current anxiety disorder. As expected, a significant mean total score difference between patients with an anxiety disorder (M = 21.75, SD = 13.11) and patients without an anxiety disorder (M = 14.44, SD = 10.93) was obtained (t(215) = 4.38, p < .00001).

To investigate the factorial validity of the BAI item pool, the 21 items were submitted to a common factor analysis using an iterated principal axis factoring with squared multiple correlations as initial communality estimates, and sub- sequent promax rotation to an oblique simple structure. The first two eigenvalues were 7.73 and 1.38, with all remaining eigenvalues falling below 1.0. A two factor solution (accounting for 84% of the variance) was selected based on previous researching findings, and an examination of the scree plot. Item communalities for the two factor solution ranged from 0.23 to 0.67, with a mean of 0.43. Following factor rotation, the two factors correlated 0.59. The factor pattern matrix, consisting of standardized regression coefficients for predicting the items from the factors, and the item communalities are presented in Table 1.

As can be seen from Table 1, a high degree of simple structure was obtained. Factor I clearly consists of items describing somatic aspects of anxiety, while Factor II clearly consists of items describing subjective aspects of anxiety.

For the purpose of further investigation, the 14 items with salient loads on Factor I (items 1-3, 6-8, 11-13, 15, 18-21) were summed to form a BAI- somatic subscale, and the 7 items with salient loadings on Factor II (items 4, 5, 9, 10, 14, 16, 17) were summed to form a BAI-subjective subscale. Coefficient alpha for the BAI-somatic subscale was 0.89, while the coefficient alpha for the BAI-Subject subscale was 0.86. With regard to discriminant validity, the BAI-somatic subscale yielded a significant mean difference (t( 124.4) = 3.42, p < .OOOl) between patients with an anxiety disorder (M = 11.54, SD = 9.05) and patients without an anxiety disorder (M = 7.46, SD = 6.99). Similarly, a significant mean difference (t(215) = 4.75, p < .OOOOl) between patients with (M = 10.21, SD = 5.40) and without an anxiety disorder (M = 6.73, SD = 5.00) was obtained on the BAI-subjective subscale.


TABLE 1 FACTOR PATTERN AND COMMUNALITIES FOR BECK ANXIETY INVENTORY ITEMS

Factor Loading

Item I II h

1. Numbness or tingling .49 -.oo .24

2. Feeling hot .44 .14 .28 3. Wobbliness in legs S6 -.08 .27

4. Unable to relax .26 .46 .42 5. Fear of worst happening -.09 .86 .66

6. Dizzy or lightheaded .77 .02 .62 7. Heart pounding or racing .49 .28 .48

8. Unsteady .57 .09 .39 9. Terrified .12 .70 .60

10. Nervous .lO .67 .54

11. Feelings of choking .45 .13 .28 12. Hands trembling .55 .12 .39

13. Shaky .55 .21 .48 14. Fear of losing control .21 .54 .48

15. Difficulty breathing .44 .24 .38 16. Fear of dying -04 .50 .23

17. Scared -.ll .88 .68

18. Indigestion or discomfort in abdomen .51 -.Ol .26 19. Faint .69 -.07 .43

20. Face flushed .78 -.06 .56

21. Sweating (not due to heat) .60 .I0 .45

Note. Interfactor correlation = 0.59. Factor loadings are standardized regression coefficients.

Characteristics of the STAI-Y

Coefficient alpha for the S-Anxiety scale (state measure) was 0.92, with item-total correlations ranging from 0.49 to 0.64. Coefficient alpha for the T-Anxiety scale (trait measure) was 0.90 with item-total correlations ranging from 0.38 to 0.69. With regard to discriminant validity, the T-Anxiety scale yielded a smaller, though significant mean score difference (t(215) = 2.11, p < .05) between patients with an anxiety disorder (M = 55.92, SD = 13.35) and patients without an anxiety disorder (M = 52.64, SD = 52.65) than that found with the BAI. No significant differences were found (t(215 = 1.44, p > .05) between the anxiety disorder group (M = 53.28, SD = 13.35) and the non- anxiety disorder group (M = 50.55, SD = 13.21) on the S-Anxiety measure.

The factorial validity of the STAI-Y was investigated in a fashion similar to that used for the BAI. A common factor analysis was performed via an interated principal axis factoring, squared multiple correlations as initial communality estimates, and promax rotation of factors. The first two eigenvalues were 13.03


TABLE 2 FACTOR PATTERN AND COMMUNAL~TIES FOR STATE-TRAIT ANXIETY INVENTORY ITEMS

Item I II h2

1. I feel calm 2. I feel secure 3. I am tense 4. I am strained 5. I feel at ease 6. I feel upset 7. I am presently worrying 8. I feel satisfied 9. I feel frightened

10. I feel comfortable 11. I feel self-confident 12. I feel nervous 13. I feel jittery 14. I feel indecisive 15. I am relaxed 16. I feel content 17. I am worried 18. I feel confused 19. I feel steady 20. I feel pleasant 21. I feel pleasant 22. I tire nervous and restless 23. I feel satisfied with myself 24. I wish I could be as happy as others seem 25. I feel like a failure 26. I feel rested 27. I am calm, cool, and collected 28. I feel that difficulties are piling... 29. I worry too much over something... 30. I am happy 31. I am inclined to take things hard 32. I lack self-confidence 33. I have disturbing thoughts 34. I make decisions easily 35. I feel inadequate 36. I am content 37. Some unimportant thought runs... 38. I take disappointments so keenly... 39. I am a steady person 40. I get in a state of tension or turmoil...

.50 -.21

.61 -.lO

.Ol .63

.05 .59

.59 -.05

.03 .78 -.Ol .57 .57 -.lO

-.03 .69 .59 -.19 .70 .04 .05 .76 .Ol .73

-.09 .55 .58 -.lO .62 .06 .12 .79

-.lO .62 .69 -.Ol .68 .05 64 .02

-.32 44 .79 .09

-.20 .23 -.48 .15 .61 .09 .67 .03

-.34 .38 -.06 .37 .70 -.05

-.20 .38 -.38 .17 .78 .03 .50 .02

-.38 .21 .70 -.07

-.lO .34 -.26 .28 .52 -.08

-.38 .30

.41

.45

.39

.32

.38

.58

.33

.39

.50

.51

.46

.54

.52

.37

.41

.35 54 .45 .47 .43 .40 .44 .56 .14 .32 .33 .43 .39 .16 .52 .27 .25 .59 .27 .27 .55 .16 .22 .32 .35

Factor Loading

Note. Interfactor correlation = -0.52. Factor loadings are standardized regression coefficients.


TABLE 3 CORRELATIONS AMONG BAI AND STAI-Y SCALES

Scale 1. 2. 3. 4. 5.

1. BAI-Tot 1 .oo 2. BAI-somatic 0.93 1.00 3. BAI-subjective 0.84 0.63 1.00 4. S-Anxiety 0.52 0.46 0.50 1 .oo 5. T-Anxiety 0.44 0.36 ,047 0.72 1.00

Note. N = 217, all correlations significant at p < .OOOl.

and 2.97, and the first two factors accounted for 71% of the variance. Item communalities for the two-factor solution ranged from 0.16 to 0.58, with a mean of 0.39. Since the STAI-Y was initially developed to yield two unidimensional and correlated factors a two-factor solution was employed. Exami- nation of the scree plot supported this decision. Following factor rotation, the two factors correlated -0.52 and a high degree of simple structure was obtained. The factor pattern and item communalities are presented in Table 2.

Previous research would suggest that the first 20 items should load on one factor and represent state anxiety, while the remaining 20 items should load on a second factor and represent trait anxiety. Examination of Table 2 indicates that this clearly is not the case. Factor I consists of items worded to score in a negative direction (higher endorsement indicates less anxiety), while Factor II consists of items worded in a positive direction (higher endorsement indicates more anxiety). Thus, both Factor I and II appear to be method factors, response set artifacts.

Relationships Between Scales

Intercorrelations between the BAI total score, BAI derived subscales, and the STAI-Y scales are presented in Table 3. As can be seen from the table, the BAI-somatic and BAI-subjective subscales were each strongly positively correlated with the BAI total score and moderately positively correlated with each other. The STAI-Y scales were strongly positively correlated with each other. Correlations between the BAI scales and the STAI-Y scales were positive and moderate in size.

Predictive Accuracy

To evaluate the predictive accuracy of the BAI and STAI-Y, the sensitivity, specificity, positive predictive power, negative predictive power, and hit rate for classifying subjects into the anxiety disorder group or non-anxiety disorder group at various test cutoff scores were evaluated. Sensitivity here refers to the


proportion of correctly identified anxiety-disordered clients, while specificity denotes the proportion of correctly identified non-anxiety-disordered clients. Two other measures (positive predictive power and negative predictive power) provide important estimates of diagnostic utility because they take into account the base rate or prevalence of the disorder in question, and base rates vary widely depending on the setting, population, and specific disorder addressed (Garlos & Kline, 1988; Meehl & Rosen, 1955). Indeed, in a recent review of the problem of base rates, Elwood (1993) noted that discriminant ability of a test can only be evaluated after the base rate of a disorder for the specified population is determined. For a given base rate, the positive predictive power (probability that a score above the cutoff corresponds to an anxiety diagnosis), negative predictive power (probability that a score below the cutoff corresponds to an absence of an anxiety diagnosis), and hit rate (proportion of accurate decisions) of a test can be determined at various cutoff scores.

Results at varying cutoff scores for the BAI are presented in Table 4. As can be seen from this table, no single cutoff score provided both high sensitivity and high specificity. For example, at a cutoff score of 10 (mild anxiety), 94% of the anxiety group was correctly identified, 45% of the non-anxiety group was correctly identified, and 55% of the total sample was correctly identified. Given a score above 10, the probability of actually having an anxiety disorder was 0.30, while the probability of not having an anxiety disorder when the score was below ten was 0.97. For a cutoff score of 19 (moderate anxiety), 56% of the anxiety group was correctly identified, 72% of the non-anxiety group was correctly identified, and 69% of the total sample was correctly identified. Given a score above 19, the probability of actually having an anxiety disorder was 0.34, while the probability of not having an anxiety disorder for scores below 19 was 0.86. Again, it must be emphasized that these probabilities are highly dependent on the base rate (29%) for anxiety disorders encountered in this sample.

Predictive accuracy figures at various cutoff scores of the T-Anxiety scale of the STAI-Y were examined in a similar fashion. As with the BAI, no single cutoff score produced both high sensitivity and high specificity. In general, the T-Anxiety scale was less accurate than the BAI in detecting anxiety disorders. Since no significant mean difference between diagnostic groups was obtained for the S-Anxiety scale, its predictive accuracy was not further addressed.

Since the Subjective Anxiety subscale of the BAI demonstrated a larger b-test difference between the anxiety disorders group and the non-anxiety disorders group than either the BAI total score or the T-Anxiety scale of the STAI-Y, its predictive accuracy was investigated. As with the previous measures, high sensitivity or specificity was possible (depending on cutoff score) but not both. Diagnostic accuracy was slightly higher than found with the BAI, and required 14 fewer items (7 as opposed to 21). For a cutoff score of 8, sensitivity was .72, specificity was .62, positive predictive power was 0.33, negative predictive power was 0.90, and the overall hit rate was 64%. Predictive accuracy tables for

42 R I. KABACOFF ET AL.

the T-Anxiety scale of the STAI-Y and the Subjective Anxiety subscale of the BAI are available from the authors.

The question of whether or not anxiety measures could be combined to improve diagnostic accuracy was addressed using stepwise logistic regression. The binary dependent variable (presence vs. absence of an anxiety disorder) was the predicted variable, and the BAI total score, T-Anxiety and S-Anxiety scales were the predictor variables. After entry of the BAI (x2 = 16.14, p < .OOOl), neither T-Anxiety or S-Anxiety made a significant contribution to prediction at the p < .05 level. A second stepwise logistic regression was performed using the BAI-subjective and BAI-somatic subscales in place of the BAI-total score. After BAI-subjective was added to the equation (x2 = 18.89, p < .OOOl), none of the remaining three variables made a significant contribution at the p < .05 level. Results indicate that using the BAI-total score or BAI-subjective score alone provided maximal prediction, and that combining either with the other scales did not improve their predictive power.

DISCUSSION

Results from this study provide information on the psychometric properties and diagnostic utility of BAI and STAI-Y when used with older adult outpatients. The BAI demonstrated a high degree of internal reliability, a significant mean difference between anxiety and non-anxiety disorder groups, and evidence supporting factorial validity. Two clearly interpretable factors emerged from BAI item pool, and a high degree of simple structure was obtained. The first factor, represented by 14 items, described somatic aspects of anxiety. The second factor, represented by 7 items, described subjective aspects of anxiety. When formed into subscales using simple summation, the BAI-somatic and BAI-subjective subscales demonstrated a high degree of internal reliability and a moderate to strong inter-correlation (r = 0.67).

The STAI-Y T-Anxiety and S-Anxiety scales also evidenced a high degree of internal reliability. While the T-Anxiety scale demonstrated discriminant validity, the mean difference between anxiety and non anxiety disorder groups on the S-Anxiety scale was nonsignificant. In addition, a factor analysis of the STAI-Y item pool did not yield anticipated trait and state anxiety factors. Two interpretable factors emerged, but appeared to be method factors unrelated to anxiety constructs. The first factor was represented by items suggesting a lack of anxiety, while the second factor was represented by items suggesting the presence of anxiety. Taken together, these findings suggest a lack of construct validity for the STAI-Y in this older adult sample.

Although the BAI demonstrated slightly greater predictive accuracy than the STAI-Y, no single cutting score for predicting the presence of an anxiety disorder was found to be optimal for either scale. This result can be explained by the presence of a significant trade-off between sensitivity and specificity as one chooses various cut-off scores. The appropriate cut-off score for a given


TABLE 4

PREDICTION ACCURACY OF THE BAI AT SELECTED CUTOFF SCORES

BAI Cutoff Score Sensitivity Specificity

Positive

Predictive

Power

Negative

Predictive

Power Hit Rate

2

3

4

6

8 9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

1.00 0.08 0.22 1.00 0.27

1.00 0.14 0.23 1.00 0.32

1.00 0.17 0.24 1.00 0.34

1 .oo 0.22 0.25 1.00 0.38

0.97 0.23 0.24 0.97 0.38

0.97 0.26 0.25 0.97 0.41

0.94 0.32 0.26 0.96 0.45

0.94 0.35 0.27 0.96 0.47

0.94 0.40 0.29 0.97 0.51

0.94 0.45 0.30 0.97 0.55

0.89 0.46 0.30 0.94 0.55

0.86 0.49 0.30 0.93 0.56

0.72 0.52 0.28 0.88 0.56

0.69 0.55 0.28 0.88 0.58

0.67 0.60 0.30 0.88 0.61

0.67 0.62 0.31 0.88 0.63

0.61 0.64 0.30 0.87 0.63

0.58 0.67 0.31 0.86 0.66

0.56 0.72 0.34 0.86 0.69

0.50 0.76 0.35 0.86 0.71

0.44 0.77 0.33 0.84 0.70

0.36 0.79 0.30 0.83 0.70

0.36 0.82 0.33 0.83 0.72

0.36 0.82 0.34 0.83 0.73

0.33 0.84 0.34 0.83 0.73

0.28 0.85 0.32 0.82 0.73

0.28 0.87 0.36 0.83 0.75

0.25 0.89 0.38 0.82 0.76

0.25 0.91 0.43 0.83 0.78

0.22 0.91 0.40 0.82 0.77

0.19 0.93 0.41 0.82 0.78

0.19 0.94 0.44 0.82 0.79

0.19 0.94 0.47 0.82 0.79

0.17 0.96 0.50 0.82 0.80

0.14 0.96 0.45 0.81 0.79

0.14 0.96 0.50 0.81 0.80

0.14 0.96 0.50 0.81 0.80

0.14 0.96 0.50 0.81 0.80

0.11 0.96 0.44 0.81 0.79

0.11 0.97 0.50 0.81 0.80

0.11 0.99 0.67 0.81 0.81

0.11 0.99 0.67 0.81 0.81

0.08 0.99 0.60 0.81 0.80

0.06 0.99 0.67 0.80 0.80

(continued on next page)


TABLE ~--CONTINUED

BAI Cutoff Score Sensitivity Specificity

Positive Predictive

Power

Negative Predictive

Power Hit Rate

45 0.03 0.99 0.50 0.80 0.80 46 0.03 0.99 0.50 0.80 0.80 41 0.03 0.99 0.50 0.80 0.80 48 0.03 0.99 0.50 0.80 0.80 49 0.03 0.99 0.50 0.80 0.80 50 0.03 0.99 0.50 0.80 0.80 51 0.03 0.99 0.50 0.80 0.80 52 0.03 0.99 0.50 0.80 0.80 53 0.03 0.99 0.50 0.80 0.80 54 0.03 0.99 0.50 0.80 0.80 55 0.00 0.99 0.00 0.80 0.79 56 0.00 0.99 0.00 0.80 0.79 51 0.00 0.99 0.00 0.80 0.79 58 0.00 0.99 0.00 0.80 0.79

Note. Positive and negative predictive powers are based on a base rate of .29.

setting will depend on the relative importance of accurately identifying indi- viduals with or without an anxiety disorder, coupled with an estimate of the disorder’s prevalence. Tables 4 and 5 demonstrate this trade-off, and provide information useful in choosing an appropriate cutting score for various pur- poses.

Combining the BAI and the STAI-Y did not improve diagnostic prediction above that found for the BAI alone. In addition, the BAI-subjective subscale produced slightly better classification accuracy than either the BAI-somatic subscale or BAI total score. This finding suggests a greater relative importance of subjective/cognitive aspects of anxiety over somatic symptomatology when screening for the presence of anxiety disorders. Information for selecting appropriate cutting scores for the BAI-subject subscale can be found in Table 6.

In general, these instruments had lower diagnostic accuracies than has been reported for measures used to screen for major depression. Kogan, Kabacoff, Hersen, and Van Hasselt (1994) found the Beck Depression Inventory and the Geriatric Depressional Scale to have higher screening accuracies with older adults than those reported here. This finding is consistent with a view that the diagnosis of a major depressive episode is a more discrete and circumscribed process than the diagnosis of an anxiety disorder. Specifically, anxiety symptomatology is more likely to be found across of range of DSM diagnoses, resulting in lower predictive accuracies for these screening tests.

In summary, the BAI was found to have good psychometric properties, including internal reliability, discriminant validity, and factorial validity when use with an older adult outpatient population. The STAI-Y demonstrated


notable deficits in both discriminant and factorial validity and cannot be recommended for use with this population. Results suggest that both the BAI total score and the BAI-subjective subscale can be moderately successful as a quick screening device for the presence of an anxiety disorder. Since the BAI-subject subscale consists of seven items, it is particularly promising as a brief addition to a test battery. However, care should be exercised in selecting an appropriate cutting score, with both the expected prevalence rate and the relative importance of false positives and false negatives considered.

There are several limitations in the current study. Subjects were primarily white outpatients capable of independent living. Results may not generalize to more seriously impaired older adults, or to other ethnic groups. Additionally, this study concentrated on a discrimination between the presence and absence of an anxiety disorder, leading to the inclusion of a variety of anxiety disorders under one broad heading. Finer analyses were not possible due to sample size limitations. However, results may vary by type of anxiety disorder. Finally, older adults were not subdivided by age, again due to sample size consider- ations. Results may vary by age, even within this population.

These findings highlight the need to investigate the psychometric properties and diagnostic accuracies of popular assessment instruments when used with older adults. Test characteristics reported for younger adults may not accurately generalize to older populations. Since the value of research with older populations is dependent of the accuracy of the instrumentation employed, such investigations are quite important. Additionally, studies investigating the per- formance of these measures with older adults in other settings and with other ethnic groups are needed.

REFERENCES

American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed. rev.). Washington, DC: Author.

Beck, A.T., Epstein, N., Brown, G., & Steer, R.A. (1988). An inventory for measuring clinical anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology, 56,893-897.

Beck, A.T., & Steer, R.A. (1990). Manual for the Beck Anxiety Invenfory. San Antonio, TX: Psychological Corporation.

Beck, A.T., & Steer, R.A. (1991). Relationship between the Beck Anxiety Inventory and the Hamilton Anxiety Rating Scale with anxious outpatients. Journal of Anxiety Disorders, 5, 213-223.

Beck, A.T., Steer, R.A., & Beck J.S. (1993). Types of self-reported anxiety in outpatients with DSM-III-R anxiety disorders. Anxiety, Stress, and Coping, 6, 43-55.

Derogatis, L.R. (1983). SCL-90-R administration, scoring, and procedures manual-II. Towson, MD: Clinical Psychometric Research. Elwood, R. (1993). Psychological tests and clinical discriminations: Beginning to address the base rate problem. Clinical Psychology Review, 13, 409-419.

46 R. 1. KABACOFF ET AL

Fydrich, T., Dowdall. D.. & Chambless, D.L. (1992). Reliability and validity of the Beck Anxiety Inventory. Journal of Anxiety Disorders, 6, 55-61.

Garlos, A., & Kline, R. (1988). Understanding the accuracy of tests with cutting scores: The sensitivity, specificity, and predictive value model. Journal of Clinical Psychology, 44, 1013- 1023.

Hamilton, M. (1959). The assessment of anxiety states by rating. British Journal of Medical Psychology, 32, 50-55.

Hersen, M., Van Hasselt, V.B., & Goreczny, A.J. (1993). Behavioral assessment of anxiety in older adults: Some comments. Behavior Modification, 17, 99-112.

Hersen, M., & Van Hasselt, V.B. (1992). Behavioral assessment and treatment of anxiety in the elderly. Clinical Psychology Review, 12, 619-640.

Hollingshead, A.B. (1975). Four-factor index of social status. Unpublished manuscript. Department of Sociology, Yale University, New Haven, CT.

Himmelfarb, S., & Murrell, S.A. (1984). The prevalence and correlates of anxiety symptoms ion older adults. The Journal of Psychology, 116, 159-167.

Jolly, J.B., Aruffo, J.F., Wherry, J.N., & Livingston, R. (1993). The utility of the Beck Anxiety Inventory with inpatient adolescents. Journal ofAnxiety Disorders, 7, 95-106.

Kogan, E., Kabacoff. R., Hersen, M., & Van Hasselt, V. (1994). Clinical cutoffs for the Beck Depression Inventory and the Geriatric Depression Scale with older adult psychiatric outpatients. Journal of Psychopathology and Behavioral Assessment, 16, 233-242.

Kumar, G., Steer, R.A., & Beck, A.T. (1993). Factor structure of the Beck Anxiety Inventory with adolescent psychiatry inpatients. Anxiety, Stress, and Coping, 6. 125-131.

Meehl, P, & Rosen, A. (1955). Antecedent probability and the efficiency of psychometric signs, patterns, or cutting scores. Psychological Bulletin. 52, 194-216.

Oei, TXS., Evans, L., & Cook, G.M. (1990). Utility and validity of the STAI with anxiety disorder patients. Journal of Clinical Psychology, 29, 429-432.

Reiger, D.A., Boyd, J.H., Burke, J.D., Rae, D.S., Myers, J.K., Kramer, M., Robins, L.N., George, L.K., Karno, M., & Locke, B.Z. (1988). One-month prevalence of mental disorders in the United States. Archives of General Psychiatry, 45, 977-986.

Rapp, S.R., Parisi, S.A., & Walsh, D.A. (1988). Psychological dysfunction and physical health among elderly medical inpatients. Journal of Consulting and Clinical Psychology, 56, 85 l-855.

Segal, D.L., Hersen, M., Van Hasselt, V.B., Kabacoff, R.I., & Roth, L. (1993). Reliability of diagnosis in older psychiatric patients using the Structured Clinical Interview for DSM-III-R. Journal of Psychopathology and Behavioral Assessment, 15, 347-356.

Segal, D.L., Kabacoff, RI., Hersen, M., Van Hasselt, V.B., & Ryan, C.F. (1995). Update on reliability of diagnosis in older psychiatric patients using the Structured Clinical Interview for DSM-III-R. Journal of Clinical Geropsychology, 1, 313-321.

Spielberger, C.D. (1985). Assessment of state and trait anxiety: Conceptual and methodological issues. The Southern Psychologist, 2, 6-16.

Spitzer, R.L., Williams, J.B.W., Gibbon, M., &First, M.B. (1988). Structured Clinical Interviewfor DSM-III-R- patient version (SCID-P 6/I/88). New York: Biometrics Research Department, New York State Psychiatric Institute.

Steer, R.A., Beck, A.T., Brown, G.K., & Beck, J.S. (1993). Classification of suicidal and nonsui- tidal outpatients: A cluster-analytic approach. Journal of Clinical Psychology, 49, 603-614.

Steer, R.A., Ranieri, W.F., Beck, A.T., & Clark, D.A. (1993). Further evidence for the validity of the Beck Anxiety Inventory with psychiatric outpatients. Journal of Anxiety Disorders, 7, 195-205.


Steer, R.A., Rissmiller, D.J., Ranieri, WE, & Beck, A.T. (1993). Structure of the computer-assisted Beck Anxiety Inventory with psychiatric inpatients. Journal of Personalify Assessmenf, 60, 532-542.

journal of anxiety disorders volume 11 issue 1 1997

Documents