L.N. Dyrbye is professor of medicine and medical education, Program on Physician Well-Being, Department of Medicine, Mayo Clinic, Rochester, Minnesota; ORCID: https://orcid.org/0000-0002-7820-704X. N.M. Wittlin is PhD student, Department of Psychology, Yale University, New Haven, Connecticut; ORCID: https://orcid.org/0000-0002-0858-3576. R.R. Hardeman is assistant professor, Division of Health Policy and Management, University of Minnesota, School of Public Health, Minneapolis, Minnesota; ORCID: https://orcid.org/0000-0003-3913-5933. M. Yeazel is associate professor, Department of Family Medicine and Community Health, University of Minnesota Medical School, Minneapolis, Minnesota. J. Herrin is assistant professor of medicine, Section of Cardiovascular Medicine, Yale University School of Medicine, New Haven, Connecticut; ORCID: https://orcid.org/0000-0002-3671-3622. J.F. Dovidio is Carl Iver Hovland Professor of Psychology and professor of public health, Department of Psychology, Yale University, New Haven, Connecticut; ORCID: https://orcid.org/0000-0002-6110-8344. S.E. Burke is assistant professor, Department of Psychology, Syracuse University, Syracuse, New York; ORCID: https://orcid.org/0000-0002-6952-924X. B. Cunningham is assistant professor of health disparities, Department of Family Medicine and Community Health, University of Minnesota Medical School, Minneapolis, Minnesota; ORCID: https://orcid.org/0000-0002-3205-5538. S.M. Phelan is associate professor of health services research, Division of Health Care Policy and Research, Mayo Clinic, Rochester, Minnesota; ORCID: https://orcid.org/0000-0003-2091-6297. T.D. Shanafelt is Jeanie and Stewart Ritchie Professor of Medicine and associate dean, Stanford School of Medicine, Stanford, California; ORCID: https://orcid.org/0000-0002-7106-5202. M. van Ryn is Grace Phelps Distinguished Professor, Oregon Health & Science University School of Nursing, Portland, Oregon; ORCID: https://orcid.org/0000-0002-4258-7319.
Acad Med. 2019 Feb;94(2):217-226. doi: 10.1097/ACM.0000000000002437.
To determine baseline individual and school-related factors associated with increased risk of developing depression symptoms by year four (Y4) of medical school, and to develop a prognostic index that stratifies risk of developing depression symptoms (Depression-PI) among medical students.
The authors analyzed data from 3,743 students (79% of 4,732) attending 49 U.S. medical schools who completed baseline (2010) and Y4 (2014) surveys. Surveys included validated scales measuring depression, stress, coping, and social support. The authors collected demographics and school characteristics and conducted multivariate analysis to identify baseline factors independently associated with Y4 depression symptoms. They used these factors to create a prognostic index for developing depression. They randomly divided the data into discovery (n = 2,455) and replication (n = 1,288) datasets and calculated c statistics (c).
The authors identified eight independent prognostic factors for experiencing depression symptoms during training within the discovery dataset: age; race; ethnicity; tuition; and baseline depression symptoms, stress, coping behaviors, and social support. The Depression-PI stratified four risk groups. Compared with the low risk group, those in the intermediate, high, and very high risk groups had an odds ratio of developing depression of, respectively, 1.75, 3.98, and 9.19 (c = 0.71). The replication dataset confirmed the risk groups.
Demographics; tuition; and baseline depression symptoms, stress, coping behaviors, and social support are independently associated with risk of developing depression during training among U.S. medical students. By stratifying students into four risk groups, the Depression-PI may allow for a tiered primary prevention approach.
确定与医学生四年级（Y4）时出现抑郁症状风险增加相关的个体和学校相关的基线因素，并制定一个预测指数（Depression-PI）来分层医学生出现抑郁症状的风险。
作者分析了来自 49 所美国医学院的 3743 名学生（4732 名学生的 79%）的数据，这些学生完成了基线（2010 年）和 Y4（2014 年）的调查。调查包括测量抑郁、压力、应对和社会支持的有效量表。作者收集了人口统计学和学校特征，并进行了多变量分析，以确定与 Y4 抑郁症状独立相关的基线因素。他们使用这些因素来创建一个预测抑郁的预后指数。他们将数据随机分为发现（n=2455）和复制（n=1288）数据集，并计算了 c 统计量（c）。
作者在发现数据集中确定了八个与培训期间出现抑郁症状相关的独立预后因素：年龄；种族；族裔；学费；以及基线时的抑郁症状、压力、应对行为和社会支持。Depression-PI 将风险群体分层为四个。与低风险组相比，中风险组、高风险组和极高风险组出现抑郁的优势比分别为 1.75、3.98 和 9.19（c=0.71）。复制数据集证实了风险群体。
人口统计学特征；学费；以及基线时的抑郁症状、压力、应对行为和社会支持与美国医学生培训期间出现抑郁的风险独立相关。通过将学生分为四个风险组，Depression-PI 可能允许采用分层的一级预防方法。
