ĐÁNH GIÁ ĐỘ KHÓ VÀ ĐỘ PHÂN BIỆT CỦA BỘ ĐỀ THI TỐT NGHIỆP SINH VIÊN Y6 – TRƯỜNG ĐẠI HỌC NGUYỄN TẤT THÀNH – TIẾP CẬN TỪ CTT ĐẾN MIRT

Abstract

Background: In addition to Classical Test Theory, Item Response Theory, particularly the two-parameter logistic model and the multidimensional model—provides deeper insights into item difficulty, discrimination, and the latent structure of examinations. Objectives: To evaluate the psychometric properties of the final examination for medical graduates at Nguyen Tat Thanh University by combining analyses from three methods. Materials and methods: A cross-sectional descriptive study was conducted on 118 sixth-year medical students, using a 120-item multiple-choice final examination. Classical test theory analysis and the Item Response Theory model were used to estimate item difficulty and discrimination parameters. A simple-structure multidimensional model was then employed to assess item characteristics within majors. Results: The test demonstrated acceptable reliability (Cronbach’s alpha = 0.769). Based on classical test theory, 15 items (12.8%) were difficult, and 40 items (34.2%) were easy. In terms of discrimination, 80 out of 120 items showed poor discriminative ability. Under the Item Response Theory analysis, 32 items achieved at least moderate discrimination (a > 0.65). A total of 57 items were classified as easy (b < –2), while only 24 were considered difficult (b > 1). In the simple-structure multidimensional model, the Internal Medicine domain performed best with 12 well-discriminating items, whereas the Surgery domain had 16 poorly discriminating items. Conclusion: The integration of those methods enables a comprehensive evaluation of the psychometric properties of examinations, thereby providing guidance for improving test quality.