2026-10-14T15:05:17Z https://www.scidb.cn/oai

10.57760/sciencedb.37662 2026-09-28T10:45:29Z

2026-09-28 Dataset of Interpretable Automated Decision Analysis System for Anesthesiology Residents doi:10.57760/sciencedb.37662 en Data Volume:Internal Training Set: 999 original records (440 non- feedback texts)External Test Set: 817 original records (508 non-blank feedback texts)Total: 1,816 records, 98 valid feedback textsData Collection Method:This data is sourced from the daily teaching evaluation systems of the standardized residency training (SRT) programs in the anesthesiology departments two Grade A tertiary hospitals. When trainees complete their rotations, evaluating teachers (attending physicians and above) submit written feedback comments on the SRT physicians through the hospital's teaching management system The research team exported the original records from the system, and two anesthesiology experts with over 5 years of teaching experience independently annotated the utility dimensions. Discrepancies were through discussion to reach a consensus.Annotation Framework:Based on the feedback utility theoretical framework and combined with feedback quality research in the field of medical education, a three- evaluation system of "Relevance-Specificity-Correctiveness" was constructed:- Relevance: Whether the feedback content is related to the cultivation of clinical competence- Specificity: Whether contains specific behavioral descriptions rather than general evaluations- Correctiveness: Whether it provides actionable improvement suggestionsThe three dimensions are comprehensively used to determine a four-level utility grade: Effective,, Ineffective, and Irrelevant.Data Anonymization Description:All personally identifiable information (names of evaluating teachers and SRT physicians) has been replaced with a uniform identifier***). The dataset does not contain any information that can directly identify individuals, complying with the "Personal Information Protection Law" and medical research ethics requirements.The Excel spreadsheet contains total of 8 columns of data from A to H, with the meanings of each column as follows:| Column | Field Name | English Description | Data Type | Value Description| Column A | Serial number | Serial number | Integer | Record number, incrementing from 1 || Column B | Evaluator | Evaluator | String (masked | Anonymized teacher identifier, uniformly displayed as "***" || Column C | Resident physician | Resident physician | String (masked) | Anonymized trainee identifier, uniformly displayed "***" || Column D | Feedback text | Feedback text | String | Written comments from the teacher to the trainee (may be blank) || Column E | Relevance Relevance | Binary (0/1) | Whether the feedback content is related to clinical competence || Column F | Specificity | Specificity | Binary (0/1) | the feedback contains specific behavioral descriptions || Column G | Correctiveness | Correctiveness | Binary (0/1) | Whether the feedback contains actionable improvement suggestions || Column H | Utility level | Utility level | Categorical | Final utility level determined comprehensively by the three dimensions E/F/G |Potential Application Value:1. Providing a largescale annotated dataset for medical education feedback quality research2. Supporting the development and validation of NLP models for Chinese medical education texts3. Providing a data-driven decision-making for teaching quality improvement in SRT bases4. Promoting cross-institutional and cross-disciplinary medical education evaluation Feedback literacy; Competency-based medical education; Narrative feedback; Residency training; Interpretable AI lifeng yao EMBARGO https://creativecommons.org/licenses/by-nc/4.0/ dataset Science Data Bank