Given the large number of published studies using the BAI, the summary of the literature (below) focuses on core psychometric studies and studies conducted with adolescents, trauma-exposed, and diverse populations.
Numerous studies have examined the BAI’s relationship to other measures and have found evidence for its convergent and discriminant validity. The BAI has been found to correlate moderately with the Hamilton Anxiety Rating scale (Beck et al., 1988) and the State-Trait Anxiety Inventory (STAI), with no difference between correlations with Trait and State scales (Creamer et al., 1995). The BAI typically shows lower correlations with the BDI than does the STAI or other measures of anxiety, suggesting it has better discriminant validity (Creamer et al., 1995; Fydrich, Dowdall, & Chambless, 1992).
However, factor analysis combining both the BAI and STAI-State scale showed that the two scales load on different factors, suggesting that they tap different constructs (Creamer et al., 1995). A number of studies have suggested that the BAI may be tapping more physiological aspects of anxiety and may function best with anxiety disorders with a strong physiological component, such as panic disorder (Cox, Cohen, Direnfeld, & Swinson, 1996; Creamer et al., 1995). Cox et al. (1996) factor analyzed items from the BAI with items from the Panic Attack Questionnaire (PAQ). They found a 3-factor model best fit the data and was similar to an earlier 3-factor panic model. Items from the BAI and PAQ loaded on each factor. They suggested that the BAI may be measuring panic symptoms and may not tap symptoms associated with other anxiety disorders such as Generalized Anxiety Disorder, PTSD, and Obsessive-Compulsive Disorder.
Osman et al. (2002) examined the factor structure of the BAI with adolescents and suggested that the BAI taps the construct of anxious arousal but not cognitive or behavioral dimensions of anxiety. Numerous factor analyses have been conducted with the BAI. While many identify a 2-factor structure similar to that reported by Beck et al. (1988), others have found a 4- or 5- factor structure (Beck & Steer, 1991; Borden, Peterson, & Jackson, 1991; Osman, Kopper, Barrios, Osman, & Wade, 2002). A study by Creamer et al. (1995) provides a potential explanation. Maximum likelihood factor analysis using BAI data collected from normal undergrads at a time of presumably low stress (midsemester) resulted in a different factor structure than the same analysis using data collected on the same undergrads two weeks prior to exams. The data collected under more stressful conditions resulted in a factor structure similar to that identified in the original sample and in other clinical samples (Beck et al., 1988).
Numerous studies, including in other cultures, have identified a gender difference, with females scoring higher than males, in both adult and adolescent samples (e.g., Creamer et al., 1995; Jolly, Aruffo, Wherry, & Livingston, 1993; Osman et al., 2002). Osman et al. (2002) suggest that this difference suggests the need for validating the BAI separately by gender.
In a study of older adults, Wetherell & Gatz (2005) found that in normal older adult controls BAI symptoms were associated with measures of health status.
STUDIES WITH ADOLESCENTS
- Osman et al. (2002) studied the reliability, validity, and factor structure of the BAI with a group of adolescents. They included 125 boys and 115 girls aged 14-17 who were inpatients at a Midwestern state psychiatric hospital. The comparison group included 167 adolescents aged 14 to 18 from a universityaffiliated high school. Both groups were predominantly White. BAI scores differentiated between the psychiatric and comparison groups in both boys and girls. All groups, examined separately by gender, showed good internal consistency (alpha>.88), and the clinical sample showed good 1-week test-retest reliability (r=.71). BAI scores correlated with BDI scores (males: r=.58***, females r=.65***). Examination of BAI correlations with MMPI-A scales provided good evidence of convergent and discriminant validity for boys but low evidence for discriminant validity for girls (due to correlations with all MMPI-A scales). Using confirmatory factor analysis they were unable to replicate the 2-factor structure found in other investigations and instead identified a 4-factor structure using exploratory factor analysis. Further analysis identified a higher-order factor structure, which suggested that the BAI taps a single anxiety construct they termed Anxious Arousal. They suggested the BAI may be a useful screener for anxiety but other measures would be needed to comprehensively assess for anxiety.
- Kumar, Steer, & Beck (1993) evaluated the use of the BAI with 108 adolescent psychiatric inpatients aged 12-17 and reported excellent internal consistency (alpha=.91). Principal factor analysis identified 2 factors, with a factor structure similar to what is found in adult outpatients.
- Jolly et al. (1993) examined the use of the BAI with 80 adolescent psychiatric inpatients. They found excellent internal consistency (alpha=.94). BAI scores correlated moderately with the Revised Children’s Manifest Anxiety Scale (r=.58). The BAI also correlated with adolescents’ scores on the Children’s Depression Inventory (r=.49).
- Steer, Kumar, Ranieri, & Beck (1995) examined the use of BAI in a sample of 105 adolescent outpatients aged 13-17. Using principal factor analysis they found a similar factor structure as that previously found for adolescent inpatients and adult outpatients.
STUDIES WITH TRAUMA-EXPOSED INDIVIDUALS
The BAI has been used in numerous studies with trauma-exposed individuals. A PsychInfo search of “Beck Anxiety Inventory” or “BAI” AND “trauma” yielded 58 peer-reviewed journal articles (6/05).
- The BAI has been found to be sensitive to intervention effects in numerous randomized trials with individuals with diagnosed PTSD (e.g., Bryant, Moulds, Guthrie, & Nixon, 2005; Ehlers, Clark, Hackmann, McManus, & Fennel, 2005) and continued to show intervention effects at follow-up assessments.
- In a study of 205 female rape victims aged 15 and older (48% of whom were African American) who were randomly assigned to a standard postrape control condition or an intervention designed to prevent postrape distress, the BAI was sensitive to intervention effects. In addition, postexam BAI scores were associated with 6-week follow-up PTSD scores and depression symptomatology (Resnick, Acierno, Kilpatrick, & Holmes, 2005).
STUDIES WITH OTHER CULTURAL GROUPS
- Contreras, Fernanedez, Malcarne, Ingram, & Vaccarino (2004) examined the reliability and validity of the BAI and BDI in a sample of 1,110 Latino and 2,703 Caucasian undergraduate students. Scales for both groups had good internal consistencies. They also found similar factor structures for both groups, providing evidence of factorial validity, Although they used the original BDI in this study, they suggested that results would generalize to the BDI-II given the overlap between the two.
- Sanz & Navarro (2003) examined the psychometric properties of the Spanish BAI with a sample of 590 Spanish university students and found good internal consistency and a similar factor structure as found in other studies. They provided norms for the university students and separate norms for males and females because females scored higher than did males.
- Robles, Varela, Jurado, & Páez (2001) examined the psychometrics of the Mexican version of the BAI in multiple samples of individuals aged 15-80. They found good evidence of internal consistency, reliability, and convergent validity, and a similar factor structure as that found in English-speaking samples.
- Cheng, Wong, Wong, Chong, Tak-Po, Chang, Wong, Chan, & Wu (2002) examined the psychometric properties of the Chinese Version of the BAI (BAIC). They found good internal consistency and a factor structure similar to that found in English-speaking samples.
- Al-Issa, Al Zubaidi, Bakai, & Fung (2000) examined the psychometric properties of the translated Arabic BAI with a sample of 240 undergraduate students. They compared results to those found in Lebanese and Canadian students and found similar internal consistencies. Arab students scored higher than Canadian students.
- Yook & Kim (1997) examined the factorial structure of the Korean BAI. They found similar factor structures in patient and nonpatient groups. Patient groups scored significantly higher than nonpatients.
- Freeston, Ladouceur, Thibodeau, Gagnon, & Rheaume (1994) reported good internal consistency, reliability, and convergent, discriminant, and factorial validity using the French-Canadian version with Canadian university students and adults. 8. BAI scores are related to scores on the Adolescent Dissociative Experiences Scale in a sample of Turkish adolescents (Sayar, Kose, Grabe, & Murat, 2005).