Validity and Reliability of the Japanese Version of the Selected Anger Expression Scale and Age , Sex , Occupation and Regional Differences in Anger Expression Among Japanese

To examine the reliability and construct validity of the Japanese version of the Anger Expression Scale among four Japanese communities, and to examine distributions of anger expression scores according to sex, age, occupation, and community, we performed a crosssectional study among 1,802 men and 3,229 women aged 20-70 in four geographic populations in 1995-97. We handed a self-administered questionnaire, which was selected from the Spielberger Anger Expression Scale, to the participants in the risk factor surveys and measured anger-in and anger-out as the anger expression scale. These scales had high internal consistency (Cronbach's alpha coefficient was 0.97-0.98 for anger-out and 0.77-0.86 for angerin) and were of almost the same structure as the original. The Pearson correlation coefficients for the anger expression scale examined in 1995 and 1996 were 0.69 for anger-out and 0 .57 for anger-in (both p<0.001). The mean scores of both anger-out and anger-in were inversely associated with age. The mean anger-out score was higher for men than for women (p<0.001) , whereas the mean anger-in score did not vary significantly between the sexes. Furthermore, the mean scores of anger-out and anger-in varied among populations and occupational groups . The present study suggests that the Japanese version of the selected Anger Expression Scale is an acceptable scale for evaluating anger expression among Japanese. J Epidemiol, 2000; 10: 118-123


INTRODUCTION
A broad range of epidemiological and clinical studies have reported that style of anger expression was one of the important risk factors for essential hypertension 1-6) and cardiovascular disease 7,8).In 1984, Spielberger et al. developed structured scaling of anger expression using 16 questions about styles of responses to anger-provoking situations, 8 for suppressed anger (anger-in) and 8 for behaviorally expressed anger (angerout) 9).This scoring system had the advantage of assessing two aspects of anger expression, whereas most previous studies focused on anger suppression [1][2][3][4][5][6].The mean score of anger-out was lower with increased age, but the anger-in score did not lower with age 10).The mean score of anger-out was higher in men than in women for Russians 11), but no sex difference was found for Americans 10).Using this scoring system, several researchers found a positive relation between anger-in and blood pressure levels and an inverse relation between angerout and blood pressure levels 9,[12][13][14].
This promising questionnaire was translated into a number of foreign languages, including Spanish, French, and

MATERIALS AND METHODS
Four geographic populations with a total survey population of 1,802 men and 3,229 women, ages 20-70 years, were included in the survey.The members of the surveyed populations were (1) residents in Ikawa, a northeast rural community; (2) residents in Kyowa, a central rural community; (3) residents in Yao, a suburb of Osaka (the second-largest city in Japan); (4) residents in Noichi, a southwest rural community.In these populations, surveillance of coronary heart disease and stroke has been conducted for more than 15 years [16][17][18][19][20].The survey was conducted between 1995 and 1997.
We handed a self-administered questionnaire on anger expression to the participants in the risk factor surveys.Eighty-seven subjects who had a history of coronary heart disease and stroke were excluded.Seventy-six questionnaires that were only partially completed were excluded.As a result, 4,868 men and women (96.8%) were included in the analyses.
To measure anger expression, we used 16 items out of 24 items in the original Anger Expression Scale developed by Spielberger et al 10).The Anger Expression Scale is based on the frequency of reactions to anger-provoking situations.Frequency was classified as almost never, sometimes, often, or always.Subclassification of anger expression was done by the direction of anger expression; anger expressed behaviorally was regarded as "anger out" and anger held in or suppressed as "anger in" and the attempts to control anger expression as "anger control ."We used eight anger-out and eight anger-in items, and we did not use eight anger-control items because no relation between anger-control and blood pressure levels has been observed 14).Scores in the eight items for either anger-out or anger-in were summed and used as anger expression scores.The Anger Expression Scale was translated from English into Japanese under the direction of a psychologist and an epidemiologist who had lived in the United States for more than two years.To assess the one-year test-retest reliability, we asked 373 men and women, randomly selected from the participants in the cardiovascular risk survey, to complete the same questionnaire again one year later.
The occupational classification was categorized as (1) employees of companies or public servants, (2) owners of small-sized companies (number of employees < 100) or shops, (3) farmers, and (4) housewives or retired employees.
Intercorrelation between anger-out and anger-in scores was examined using the Pearson correlation coefficient.Cronbach's alpha coefficient was calculated for both scales to evaluate internal consistency.The one-year test-retest reliability was assessed using the Pearson correlation coefficients.To examine a factor-based validity, all of the 16 items were subjected to an explanatory factor analysis using principal axis factor analysis.Two factors with eigenvalues of more than 1.0 were extracted and were rotated to final solution using varimax rotation.Differences in mean values of the anger expression scale among sex and age group were tested using analysis of variance.Differences among populations and occupational groups were examined using analysis of covariance controlling for age.The 0.05 level of probability was accepted as significant, and all statistical tests using probabilities were two-tailed.

RESULTS
Table 1 shows sex-specific mean values, standard deviations, and internal consistency estimates (Cronbach's alpha coefficients) of the anger-out and anger-in scores.The mean scores of anger-out and anger-in were lower with increased age.We found high internal consistency estimates for the anger-out score (the alpha coefficient=0.97-0.98)and for the anger-in score (the alpha coefficient=0.77-0.86).
There was moderate intercorrelation between anger-out and anger-in for both men and women (r=0.49and r=0.32, respectively; both p<0.001).The mean score of anger-out was significantly higher for men than for women, but the mean score of anger-in did not vary between the sexes.The Pearson correlation coefficient for the anger expression scale examined in 1995 and 1996 was 0.69 for anger-out and 0.57 for anger-in (both p<0.001).Table 2 presents factor loadings of 16 items of the anger expression scale with varimax rotation.Two factors with eigenvalues exceeding 1.0 were extracted, and factor loadings of 0.30 or higher were considered to be significant.All items for anger-out showed highest factor loadings on factor 1. Seven of the eight items of anger-in showed highest factor loadings on factor 2. The item "pout or sulk" had factor loading higher than 0.30 on both factors 1 and 2.
Residents in Yao had the highest mean anger-out score, and residents in Ikawa had the highest mean anger-in score for both men and women (Table 3).Residents in Noichi had the lowest mean anger-out and anger-in scores for both men and women.
Table 4 presents sex-specific age-adjusted mean values of   anger-out and anger-in scores among four occupational groups.
For men, owners of shops or small-sized companies and employees of companies had a higher mean anger-in score than did members of the other two occupational groups.For women, owners of shops or small-sized companies and housewives had a higher mean anger-out score than did members of the other two occupational groups.However, the mean angerout sores for men and the mean anger-in scores for women did not vary among occupational groups.

DISCUSSION
The present study showed that anger-in and anger-out scores assessed by our Japanese version of the selected items from the Anger Expression Scale had high internal consistency.Item factor analysis supported the construct validity of these two scales in both men and women much like that of the original Anger Expression Scale.The mean scores of anger-out and anger-in were inversely associated with age.The mean angerout score was higher for men than for women, whereas the mean anger-in score did not vary between the sexes.
Furthermore, the mean scores of anger-out and anger-in varied among populations and occupational groups.
In the present study, moderate to high internal consistency estimates were found: the alpha coefficient for anger-out=0.97-0.98 and for anger-in=0.77-0.86.Previous studies showed similar internal consistency estimates for American college students (the alpha coefficient for anger-out=0.73-0.75 and for anger-in=0.81-0.84),and for Japanese college students and hospital patients (the alpha coefficient for anger-out=0.78and for anger-in=0.76)9.15).Our one-year test-retest reliability study for anger-out and anger-in scores showed moderate correlation coefficients.Thus, we consider that the reliability of our version is an acceptable.
The factor structure of our Japanese version of the selected items from the Anger Expression Scale was similar to previous psychometric evaluations of the Anger Expression Scale except for the item "port or sulk" 9,11,15).The item "port or sulk" belonged to not only anger-in but also anger-out for our sample, whereas in the American and Russian samples, the item "port or sulk" belonged to exclusively anger-in 9,11).This small discrepancy on the factor structure may be due to a cultural difference in behavior rather than a translation error because previous study using another Japanese version of the anger expression scale showed a similar factor structure to our results 15).
With regard to sex difference in anger-out scores, Spielberger reported no sex difference in the American sample 10).In the Russian sample, Eckhardt et al. showed that the mean score of anger-out was higher for men than for women, which is consistent with our results 11).Our result indicates that Japanese men are more likely to express their anger than are Japanese women, much like the Russian population, while in the American population men and women express their anger equally.
We Owners of shops or small-sized companies had the highest mean value of anger-in for men and this same group had the highest mean value of anger-out for women among the four occupational groups.Recently, we reported the trends in blood pressure levels and their related factors according to occupational classification in Kyowa 23).For men, owners of shops or small-sized companies had higher mean systolic blood pressure levels than did members of other occupational groups in 1994-1997, which was attributable in part to higher alcohol intake.High anger-in levels among male owners of shops or small-sized companies are consistent with their higher blood pressure.For women, however, mean systolic and diastolic blood pressure levels did not vary among the occupational groups.
In conclusion, this study showed the reliability and validity of the anger-in and anger-out scales of the Anger Expression Scale for Japanese residents.These scales had high internal consistency and almost the same structure as the original version, even though we studied only a limited set of questions selected from the Anger Expression Scale.
Japanese 11).Validity and reliability studies have been reported for the Japanese and Russian versions 11,15) Good reliability and validity of these versions were reported, but the survey sample consisted of hospital patients and/or college student volunteers 11,15) A study of reliability and validity in community-based samples is warranted to examine distributions of anger expression scores and the relation with blood pressure levels in apparently healthy individuals, which has not been studied among Asian people.The purpose of this study are to evaluate the reliability and construct validity of the Japanese version of the Anger Expression Scale among four Japanese communities, and to examine distributions of anger expression scores according to sex, age, occupation, and community.

Table 1 .
Means, Standard Deviations, and Cronbach's Alpha Coefficients for Anger-out and Anger-in by Sex and Age Group .
* difference among age groups.*p<0 .001,difference between men and women.

Table 2 .
Factor Loadings of Anger Expression Items with Varimax Rotation.

Table 3 .
Differences in Age-adjusted Values of Anger-out and Anger-in among Population Groups.
* difference among population groups by analysis of covariance controlling for age.

Table 4 .
Differences in Age-adjusted Values of Anger-out and Anger-in among Occupational Groups.
* difference among occupational groups by analysis of covariance controlling for age.