Dimensionality, discrimination power and difficulty of English test items: the case of graduate exam for healthcare applicants

Background & Objective: Administered by the Iranian Center for the Measurement of Medical Education, national university entrance exams are administered nationwide where English constitutes a vital section. This study aimed to assess dimensionality, discrimination power and difficulty of English test items in this graduate entrance exam. Material & Methods: This quantitative study examined 160 English test items administered to 41633 test-takers applying for graduate studies in Iranian universities of medical sciences in 2021, and reported the characteristics of test takers during three successive years (2019, 2020, and 2021). NOHARM software (version 4.0) was used to analyze the data by examining dimensionality of the tests reporting a two-parameter model. Results: Generally, female participants outnumbered the male, with a similar pattern among the admitted participants (70% females vs. 30% males). A positively significant correlation was found between participants’ Grade Point Average and English test scores (p < 0.05). In 2021, the results of four administration sessions with a high reliability (i.e. 0.92, 0.88, 0.90 and 0.91) were analyzed separately. Two dimensionality parameters (i.e., difficulty & discrimination) fitted the model while the guessing parameter did not. English tests proved to be “difficult”, with either “high” or “very high” discrimination power. Neither “easy” nor “very easy” items were found. No items were associated with “no” or “very low” discrimination power. Conclusion: Overall, the tests functioned well; however, more research is required to rigorously evaluate the exams. Improvements concerning the social and long-term effects of these tests are suggested.