Development and Validation of a Short-Form Version of the Brazilian State-Trait Anxiety Inventory

The Brazilian State-Trait Anxiety Inventory has been widely used to measure state (STAI-S) and trait (STAI-T) components of anxiety. The present work developed and evaluated a short version of the STAIS and STAI-T in a large sample of Brazilian subjects. The first study selected short-form scales. The second one employed factor analyses to investigate whether each of the two short-form scales presented a well-defined and balanced structure with both anxiety-present and anxiety-absent factors. Results indicated that the STAI-S and STAI-T could be reduced to six items (STAI-S-6 and STAI-T-6, respectively) without sacrificing their psychometric properties. The findings suggest that STAI-S-6 and STAI-T-6 may be employed in situations where time restraints make the use of full-length forms unfeasible.

The State-Trait Anxiety Inventory (STAI) is one of the most used self-report measures of anxiety in research and clinical settings across different cultures (Lonner & Ibrahim, 1989).This instrument was developed by Spielberger, Gorsuch, and Lushene (1970) to provide an operational measure of two different components of anxiety: state and trait.State anxiety refers to a transitory emotional state characterized by subjective feelings or tension that may vary in intensity over time.Trait anxiety refers to a relatively stable disposition to respond to stress with anxiety and a tendency to perceive a wider range of situations as threatening (Cattell & Scheier, 1961).Accordingly, the STAI is composed of two different scales: one related to state anxiety (STAI-S) and one related to trait anxiety (STAI-T).
The STAI-S scale has 20 items that are answered on a 4-point Likert scale based on "how you feel right now, at this moment."Half of these items are positively worded (e.g., "I feel calm"), and the other half are negatively worded (e.g., "I feel tense").The STAI-T scale also has 20 items that are answered on a different 4-point Likert scale based on "how do you generally feel."Seven of these items are positively worded (e.g., "I feel secure"), and the other 13 items are negatively worded (e.g., "I worry too much over something that really doesn't matter").
Although factor analysis studies have been controversial (Bieling, Antony, & Swinson, 1998), they have revealed that both STAI scales have two factors (Barker, Barker, & Wadsworth, 1977;Endler & Magnusson, 1976;Endler, Magnusson, Ekehammer, & Okada, 1976;Gaudry & Poole, 1975;Loo, 1979;Spielberger, Vagg, Barker, Donham, & Wetsberry, 1980).One of the STAI-S factors is related to the presence of anxiety (S-anxiety present) and includes all 10 negatively worded items.The other factor of this scale is associated with the absence of anxiety (S-anxiety absent) and includes all 10 positively worded items.Similarly, the STAI-T scale also has an anxiety-present factor (T-anxiety present) associated with the 13 negatively worded items and an anxiety absence factor (T-anxiety absent) associated with the seven positively worded items.
The original STAI, termed STAI-X, was revised by Spielberger, Gorsuch, Lushene, Vagg, and Jacobs (1983).This new version, referred to as STAI-Y, improved the specificity of the anxiety measure by replacing some of the original STAI items related to depression.The STAI-Y also improved the structure of the STAI-T scale by achieving a better factor structure balance between anxiety-present and anxiety-absent items.Although the STAI-Y appears to have better psychometric properties compared with the STAI-X, both instruments appear to be comparable for anxiety assessment given that the correlation between them ranges between .96 and .98(Spielberger et al., 1983).
Administering and scoring extended scales can take a considerable amount of time.Moreover, completing a long instrument might be tiresome and can lead to measurement error attributable to wrong or unanswered items (Schmidt, Le, & Ilies, 2003).Consequently, developing a shortened version of a full-length scale is an important issue in research and clinical settings.
Previous studies consistently provided evidence that the STAI-S scale can be reduced to an abbreviated form.Marteau and Bekker (1992) developed the first short-form version of this scale.To preserve the S-anxiety present and S-anxiety absent dimensions of this scale, Marteau and Bekker (1992) selected three anxiety-present items and three anxiety-absent items.This 6-item short-form STAI-S scale produced scores similar to those observed with the full 20-item scale and has been widely used in clinical and basic research (Macaluso et al., 1996;Rose, Humm, Hey, Jones, & Huson, 1999;Ubhi et al., 1996).
In Brazil, only the full-length STAI-X was initially translated into Portuguese by Biaggio and Natalício (1979).Since then, several studies were conducted to validate both the STAI-S and STAI-T scales in Brazilian populations (Andrade, Gorenstein, Vieira Filho, Tung, & Artes, 2001;Fioravanti, Santos, Maissonette, Cruz, & Landeira-Fernandez, 2006;Pasquali, Pinelli, & Solha, 1994).However, no studies have attempted to develop a shortened version of the STAI-S and STAI-T scales.Therefore, the purpose of the present study was to develop a short form of these two STAI scales with acceptable psychometric properties in a Brazilian population.
Typically, the construction of a short-form version of an existing instrument considers the internal consistency and reliability based on item-total correlation.However, short-form scales must have the same factor structure of the full-length scale (Smith, McCarthy, & Anderson, 2000).Following this guidance, a first study was performed to select the items with the best item-total correlation representing the STAI factor structure.A second study employed factor analysis to investigate whether the factor structure of the short-form STAI scales are equivalent to their respective full-length scales in a different subset of subjects.

Participants
A group of 4.455 subjects from five different samples participated in this study.The five samples included 1.173 high school students from Rio de Janeiro, 1.240 high school students from Niterói, 1.328 university students from Rio de Janeiro, 317 university students from Santa Catarina, and 397 employees of a public hospital and private company, both in Rio de Janeiro.Each of these samples was randomly divided into two subsets to perform the two studies of the present report.Table 1 presents the age and gender distribution of these five samples used in the first and second studies.Instrument Both studies used the 20 state and 20 trait anxiety items from the STAI-X and the six new items from each scale of the STAI-Y, providing a 26 item state scale and a 26 item trait scale.The 12 new items were translated into Portuguese by two translators and then back-translated to English by two other translators who had not seen the original items.This back translation was then compared with the original English items by one of the authors.This took into consideration linguistic and semantic equivalence between translations.Afterward, a pilot study indicated that all subjects were able to perfectly understand the 12 new items of the STAI scales.
The state scale required the participant to describe how he or she feels at "this very moment" in relation to 26 items presented on a 4-point Likert intensity scale: 1 = "not at all," 2 = "somewhat," 3, = "moderately," 4 = "very much."The trait scale required the participant to describe how he or she "generally" feels in relation to the 26 items presented on the same 4-point Likert scale.

Procedure
All subjects were invited to participate in the study and were informed of the nature of the study.High school and undergraduate students answered the instrument in a group setting that varied between 10 and 50 participants during their regular class schedule.Hospital and company employees answered the instrument alone during customary working hours.Questionnaires were completed with no time limit.None of the participants demonstrated difficulty in understanding either the instructions or any of the items.Subjects took between 15 to 30 min to complete the scales.No payment or course credits were offered to the subjects.

Statistical Analysis
The purpose of the first study was to choose the optimal short forms of the STAI-S and STAI-T scales.The item selection procedure was based on the statistical methodology reported by Marteau and Bekker (1992).According to this procedure, items from the STAI-S and STAI-T scales were ranked according to their corrected item-total correlation scores.Based on this parameter, an equal number of anxiety-present and anxiety-absent items were selected to create 10-, 8-, 6-, and 4-item forms of the STAI-S and STAI-T scales.Subsequently, internal consistency of each of the four short-form versions was assessed by calculating their respective Cronbach a coefficients.Accepted Cronbach α coefficients are usually higher than .7 (Nunnally & Bernstein, 1994).Pearson correlation coefficients between each of the four short-form and full-length scales were calculated to evaluate the similarity between the short-form and fulllength scales.A correlation value within .9 is generally acceptable as a good indication of proportionality between scales (Kline, 1993).The second study evaluated in a new subset of subjects the internal consistency and factor structure of the short version of the STAI-S and STAI-T scales derived from the first study.Factor structure was evaluated by a principal axis factor analysis with varimax rotation.The number of factors was determined based on (a) the number of eigenvalues greater than 1, (b) examination of the scree plot of the eigenvalues against the number of factors, and (c) the interpretability of the factors.

Study 1. Choosing the Optimal STAI-S and STAI-T Short-form Scales
STAI-S.To select the best items of the STAI-S scale, items were ranked according to their corrected item-total correlation coefficients (Table 2).With the exception of item 10, all items presented values above the .3cut-off criteria suggested by Nunnally and Bernstein (1994), indicating the good association of these items with the total score of the STAI-S.
Table 3 depicts the Cronbach α coefficients for each of the four short-form scales with balanced positively and negatively worded items.Table 3 also presents the correlation between scores from each of these four shortform scales with the scores of the full-length STAI-S scale.As expected, both the Cronbach's α coefficients and shortform and full-length scale correlation coefficients were proportional to the number of items from each of the shortform scales.According to these results, the STAI-S-6 presented the lower number of items with acceptable Cronbach α coefficients and good correlation coefficients with its full-length form.(anxiety-present items: 9, 21, 20, 11, and 12; anxiety-absent items: 13, 7, 25, 16, and 10).As shown in Table 4, the 26 items of the STAI-T scale could be reduced to a 6-item scale with good Cronbach a coefficients and accep-table correlation coefficients with its full-length form.
Table 5 presents the Cronbach α coefficients and correlations with the full-length STAI-T scale of the 10item (STAI-T-10), 8-item (STAI-T-8), 6-item (STAI-T-6), and 4-item (STAI-T-4) scales with the best balanced anxiety-present and anxiety-absent items of the STAI-T scale with higher corrected item-total correlation The factor structure of the STAI-S-6 scale was evaluated by a principal axis factor analysis with varimax rotation.Analysis of eigenvalues and scree plot and interpretability of the factors indicated a two-factor solution.The rotated factor loading for this two-factor solution is presented in Table 6.The first factor was responsible for 39.8% of the variance, with an eigenvalue of 2.4.This factor was consistent with the anxiety-present interpretation and incorporated all three items related to this construct.The other factor explained 17.9% of the variance, with an eigenvalue of 1.1.This factor was consistent with an anxiety-absent interpretation, which includes all three items associated with this construct.STAI-Trait.The factor structure of the STAI-T-6 scale was also evaluated by a principal axis factor analysis with varimax rotation.Again, analysis of eigenvalues and scree plot and interpretability of the factors indicated a twofactor solution.Table 6 presents the rotated factor loading for a two-factor solution.A well-defined structure of two factors (anxiety-present and anxiety-absent) was found.The first factor was responsible for 41.4% of the variance, with an eigenvalue of 2.5.The other factor explained 14.2% of the variance, with an eigenvalue of 0.86.As shown in Table 7, all six items had salient loadings in a single factor exclusively.

Normative Data
To provide normative data for the STAI-S-6 and STAI-T-6, all cases from study 1 and 2 were combined.Male and female scores in both scales are presented separately because of the fact that females presented statistically higher scores compared with males in both scales (STAI-S-6: t 4453 = 6.67, p < .001;STAI-T-6: t 4453 = 6.49, p < .001).Table 8 summarizes the means, standard deviations (SD), and percentile distributions according to the gender of the subjects.

Discussion
The present study presents the first attempt to develop and validate a short-form version of the Brazilian STAI.Based on the present results, both the STAI-S and STAI-T could be reduced to 6-item short-form scales without jeopardizing their psychometric properties.The STAI-S-6 (see Appendix A, top) presented acceptable internal consistency and a well-defined and balanced structure with both anxiety-present and anxiety-absent factors.These results are consistent with previous reports that also found good psychometric properties in a STAI-S short form (Koizumi et al., 1998;Marteau & Bekker, 1992;Micallef et al., 1998;van der Bij et al., 2003).
Findings from the present work also indicated that the Portuguese STAI-T-6 (see Appendix A, bottom) had an adequate internal consistency coefficient and a welldefined factor structure with three anxiety-present items and three anxiety-absent items.To our knowledge, this is the first evidence that the full version of the STAI-T scale can be reduced to a 6-item scale.Therefore, the present STAI-T-6 also appears to be a reliable and valid instrument to measure trait anxiety.
Importantly, reducing the length of the STAI eliminated subtle items associated with depression.In fact, several reports indicated that one of the factors of the full version of the STAI-T is related to depression (Andrade et al., 2001;Bieling et al., 1998).The fact that both STAI scales can be shortened to six items and still preserve their psychometric properties represents a valuable alternative in research and clinical settings.Short-form scales have several advantages.For example, they tend to reduce response bias, which is more likely to occur with long and time-consuming scales.Moreover, short forms are much easier to use, especially in clinical contexts where time pressure precludes the use of full-length scales.
Although the present study concluded that both STAI-S and STAI-T could be reduced to six items without sacrificing their psychometric properties, it is important to acknowledge that there were several limitations that can be addressed in future research.For example, convergent validity of the STAI-S-6 and STAI-T-6 could be investigated if the study had another anxiety instrument.In the same vein, it would be important to use a depression scale, such as the Beck Depression Inventory, so that the STAI-S-6 and STAI-T-6 divergent validity could be evaluated.Moreover, the fact that the five samples were composed mainly by young subjects might have lessened the representativeness of the sample which, in turn, could limit the generalizability of present findings.Finally, the presence of a clinical sample diagnosed with anxiety disorders would be particularly important to evaluate the factor structure of the STAI-S-6 and STAI-T-6 in a psychopathological context.A clinical sample would also allow describing cut-off scores which would help to employ the short form of these two STAI scales as screening instrument for anxiety disorders.Therefore, future work taking into consideration these variables will certainly help to further evaluate the utility of the STAI-S-6 and STAI-T-6 scales.

Table 1
Age and Gender Distribution of the Five Samples Used in the Present Study

Table 2
Corrected Item-total Correlations of the STAI-S Scale.Correlations in Bold Indicate the Five Higher Positively (P) and Negatively (N) Worded Items

Table 3
Cronbach a Coefficients and Full-length and Short-form Version Correlations of Four Short Versions of the STAI-S Scale

Table 5
Cronbach α Coefficients and Full-length and Short-form Version Correlations of Four Short-form Versions of the STAI-T Scale

Table 6
Principal Axis Factor Analysis Loading of the STAI-S-6 Scale Items Following Varimax Rotation

Table 7
Principal Axis Factor Analysis Loading of the STAI-T-6 Scale Items Following Varimax Rotation