Assessing the reliability of a stage of change scale

The purpose of this study was to assess the test-retest reliability of a scale measuring Prochaska's stages of change. Although structured questionnaire items are being increasingly used to segment target audiences according to Prochaska and DiClemente's stages of change, we could find only one report in the literature assessing the reliability of such scales. The unreliability of single-item or algorithm questionnaire scales might be why a number of studies show only minimal differences on some variables between individuals in different stages of change. A survey of the Perth metropolitan general population aged 16-69 years (N = 2629) was completed in August-September 1992 as part of a 3 year evaluation of the Western Australian Health Promotion Foundation. The consistency of respondents' responses was assessed across two questions measuring stages of change for the behaviours quitting smoking (n = 404), reducing alcohol consumption (n = 57) and doing more exercise (n = 704). Given the immediacy of the test-retest situation, the reliability results are moderately encouraging: kappa = 0.72, 0.73 and 0.52 for quitting smoking, reducing alcohol and doing more exercise, respectively. Health researchers should be aware of the probable moderate level of reliability if using the type of scale assessed in this study, when interpreting differences between individuals in different stages. In practice, several questionnaire items for classification purposes should be used so that internal reliability measures can be calculated. It is recommended that research be undertaken to devise more reliable scales for stages of change for the various health behaviours. It is noted that the attitude literature with respect to context and time specific intentions could be helpful in devising such scales.