Background: When a set of items is designed to measure the same construct (e.g., quality of life), item scores are often summed to represent the level of the construct. This summation method assumes that all items contribute equally to the construct and that the item scores are on an interval scale. These assumptions are problematic, especially for producing the attenuation paradox on scale reliability and validity not increased together. In recent years, the Rasch model has been developed to resolve these problems by yielding person measures that are on an interval scale. The aim of this study was to examine whether these two methods yield a similar or a different prevalence rate for job strain on workers.Methods: The Chinese-version of the Job-Content Questionnaire (C-JCQ) was used to compare the prevalence rate for job strain on workers using Rasch analysis when deriving a person measure to represent his/her level of a construct. The data were collected from 1,124 employees at a hospital in Taiwan. The dimensionality was evaluated by the principal component analysis on Rasch residuals. The prevalence of job strain was calculated. A visual presentation was displayed to use in workplaces.Results: The five-factor structure of the Job-Content Questionnaire was supported by the parallel analysis. Four types of jobs were classified using two subscales of the C-JCQ. When the summation method was used to classify workers into the four types, it yielded similar results to those from the Rasch analysis, different results were found in both the hypothesis testing and the confidence interval estimation, but same inference making. Rasch analysis has an advantage over the summation method in the treatment of missing data and resolving the attenuation paradox. To facilitate the use of the Rasch method, an Excel module, combined with the computer program WINSTEP, was developed to reveal valuable information for workers and mental health consultants.Conclusion: It is recommended that the Rasch method replace the summation method when representing the levels of latent traits for individual workers when missing data are in existence. The Rasch method is not only theoretically sound and capable of handling missing data, but the summation method might yield different results for hypothesis testing and the attenuation paradox from the results achieved by the Rasch method that is based on the alleged theoretical advantages of this approach. The Excel-VBA module is helpful in facilitating the Rasch method and reveals valuable information about workers' job strain.
