Objective In view of the practical problems, such as the difficulty applying of the traditional deterministic indicators for performance evaluation of standardized training for resident physicians and the strong subjectivity of examiners' scoring and difficulty objectively determining of the weights, to explore anew evaluation model based on non-parametric group decision-making to enhance the scientificity, discrimination and diagnostic nature of the evaluation.

Methods Taking the skills assessment data of 200 pediatric internal medicine resident physicians in the standardized training at the First Affiliated Hospital of Bengbu Medical University examination site as samples, a non-parametric group decision-making method was designed, and a non-radial performance evaluation model without input was constructed to comprehensively calculate the assessment performance of the trainees in five dimensions: cardiopulmonary resuscitation, humanistic communication, medical history collection, physical examination, medical record writing and case analysis, and the improvement space of each indicator was quantified through relaxation variables.

Results The non-radial model measured that only 5 out of 200 trainees achieved DEA effectiveness. Among all the assessment indicators, the average score of medical record writing and case analysis was the lowest, while the average scores of the other four indicators were all above 90. The results of relaxation variable analysis showed that the proportion of the maximum improvement in medical record writing and case analysis was the highest, followed by humanistic communication and physical examination, while the cardiopulmonary resuscitation and medical history collection were relatively low. The correlation coefficient between the new method and Spearman rank of the traditional average score ranking is 0.845. Moreover, the new method eliminates the tied ranking, and its discrimination was significantly better than that of the traditional method. The Cronbach coefficient of the scores given by the two examiners was 0.827, indicating good internal consistency in the scores and reasonable endogenous weights in the model.

Conclusions The new method does not rely on subjective empowerment, can stably identify the weak links of trainees and provide quantitative improvement directions, and has scientific and promotional value. Based on the empirical results, it is suggested that the resident training bases should focus on strengthening the specialized training of medical record writing and case analysis, add real case simulation and phased feedback mechanisms, and incorporate these two abilities into the core assessment indicators and appropriately increasing their weights. At the same time, the intensive training can be carried out for humanistic communication and physical examination to comprehensively enhance the clinical comprehensive ability of resident physicians. This research provides a quantifiable decision-making tool for the precise teaching management of performance evaluation in resident training.