The prevalence of tobacco use in Taiwan, as in other countries with Chinese populations, is extremely high compared with western countries. To monitor and control tobacco use in Taiwan, it is crucial to have a reliable and internationally comparable instrument for surveying smoking behavior. The challenge of establishing an international surveillance system is not only to develop a standardized questionnaire but also to deal with language and cultural barriers. Although the GYTS has been established in many countries and translated into several languages such as Spanish, French, and Arabic, no version has been developed by a cross-cultural adaptation process.
The goal of translating a universal survey instrument is to achieve equivalence between the original version and its translated version [13
]. That goal is reflected in the Chinese-version GYTS, whose semantic and conceptual equivalence was obtained by carefully executing forward translation, expert review, focus group discussion by students and teachers, and back translation, as recommended for the cross-cultural adaptation process [11
Since the original GYTS assesses not only tobacco-related behaviors but also subjective attributes of smoking susceptibility, tobacco-related attitudes, and media literacy, the minimum prerequisite for acceptance of a translated tool is not only the accuracy of its translation but also its validity and reliability [13
]. Thus, the content validity and face validity of the Chinese-version GYTS were ensured by an expert committee and focus group discussion. As the nature of item questions varies, response options were scaled as either categorical or ordinal data. Therefore, the internal consistency and stability of this instrument were examined by focusing on the attitude-related subscales of smoking susceptibility, attitudes toward smoking, and media messages about smoking, whose responses have similar scaling systems. Internal consistency for these subscales was estimated by Cronbach's alpha coefficient, the most popular index of reliability for estimating the internal consistency of multiple items on a composite scale [13
]. Cronbach's alphas for these subscales were all above 0.70, indicating good internal consistency for items in these subjective domains [16
]. The stability of responses was estimated over a 2-week interval, which has been used as a test-retest approach [13
]. Similar score distributions on tests taken 2 weeks apart supported the stability of these 3 tobacco attitude-related subscales. In addition to the consistency of responses to items related to subjective attributes, high agreement was found among responses to items related to tobacco-use behaviors, indicating good reliability of the Chinese-version GYTS.
Before these results can be considered in more detail, this study has several limitations that merit attention. First, self-reported smoking behaviors cannot be directly validated because no bioassay is available for physiological evidence of smoking, e.g., serum nicotine. Second, because the data were collected anonymously, coefficients of test-retest reliability could not be estimated; instead we examined the difference in score distributions of 3 subscales between the test and retest. Third, the single focus group of participants selected mainly from the metropolitan area of Taipei may not reflect the cultural diversity of rural areas. Fourth, teachers/staff taking part in the focus group were responsible for teaching health education at their schools, but their relationship with the students was friendlier than traditional Taiwanese teacher-student relationships. We do not know whether this difference would help or hinder students from stating their opinions about tobacco behaviors. Fifth, all twelfth-grade students were excluded in the field test, which may limit generalization of the study findings.