[영문]Objective The aim of our study was to verify invariance of latent factors derived from the BMQ (Belief about Medicines Questionnaire) among Japanese adherent and non-adherent groups (adherent to medication and intentionally, unintentionally, and overlapping non-adherent groups) and to estimate mean differences of the latent factors among the groups. Setting A Japanese university hospital. Methods After administration of a cross-sectional survey, covariance structure analyses of the two-factor model were conducted. Groups that exhibited factorial invariance were identified, and structured mean analyses estimated the differences of the latent means of the factors between groups using the bootstrap method without relying on theoretical assumptions for sampling distributions. Effect size was employed as an indicator of these differences. Main outcome measure The differences in the latent means of the two factors (the necessity and concerns factors for prescribed medications) across the groups exhibiting factorial invariance, which reflect true differences between them. Results Factorial invariance was demonstrated only across adherent and unintentionally non-adherent groups. Unintentionally non-adherent patients had significantly lower latent means for the necessity factor than adherent patients, with a very close to medium effect size (-0.49; 95% CI -0.84, -0.14; bootstrap method). Conclusion A meaningful comparison of BMQ scale scores can be made between adherent and unintentionally non-adherent groups of Japanese patients.