Abstract: Objective To analyze the quality of national English test for doctor of medicine from 2015 to 2017 using the Rasch model of item response theory (IRT) and the classical test theory (CTT), then compare the consistency of the results. Methods 40 607 papers of national English test for doctor of medicine from 2015 to 2017 were selected as the research materials. The reliabilities, difficulties and discriminations of the examinations were analyzed using the Rasch model of IRT and CTT respectively. Then the consistency of the results was analyzed. Results IRT analysis results showed that the parameters fit well with the Rasch model, the maximum information functions were all＞25.00,the estimated errors were all <0.20, and the difficulties were all between -0.70 and 0.70. CTT analysis results showed that the reliabilities of the three annual papers were all ＞0.80, the difficulties were all between 0.31 and 0.63, and the discriminations were all between 0.12 and 0.29. The results of the two theoretical analyses were highly correlated. Conclusions The three years' tests can evaluate the application abilities of candidates' medical English well. In terms of parameter estimation and error analysis, Rasch model is more accurate than CTT, and Rasch model should be used more in the subsequent examination evaluations.

Key words: National English test for doctor of medicine, Rasch model, Classical test theory