Shaw D L, Martz D M, Lancaster C J, Sade R M
Department of Psychiatry and Behavioral Sciences, Medical University of South Carolina, College of Medicine (MUSCCM), Charleston 29425-0742, USA.
Acad Med. 1995 Jun;70(6):532-6. doi: 10.1097/00001888-199506000-00015.
Interviews are commonly used to measure noncognitive traits of medical school applicants. The present study investigated the influence of knowledge of applicants' cognitive abilities on interviewers' ratings of noncognitive traits.
Academic and demographic predictors of interview ratings of applicants' noncognitive traits were examined at the Medical University of South Carolina College of Medicine during two years: 1992, when applicants' Medical College Admission Test (MCAT) total scores and undergraduate grade-point averages (GPAs) were available to interviewers; and 1993, when MCAT and GPA data were not available. In 1992, 226 applicants met study criteria (i.e., they received ratings from three interviewers in addition to having MCAT and GPA data on file); in 1993, 245 applicants met the criteria. Step-wise regression analyses were conducted to measure the influences of seven independent variables on applicants' interview ratings. Two-way analyses of variances and t-tests were used to determine the effects of gender of applicants and interviewers. Cronbach's alpha coefficients were used as measures of interviewers' reliability.
GPA was the best predictor for both years but accounted for double the amount of variance in interview ratings in 1992 (15.7%) compared with 1993 (7.4%). The reliability coefficients for the interviewers were .496 for 1992 and .473 for 1993.
If the goal of the medical school admission interview is to assess noncognitive traits independently from academic skills, the authors recommend that MCAT and GPA data not be available to interviewers during interviews. The authors also found that gender and race influenced interview ratings in accordance with affirmative-action goals. Finally, the authors found that interview scores were only moderately reliable across different interviewers. They discuss ways to increase their reliability.
面试常用于评估医学院申请者的非认知特质。本研究调查了申请者认知能力信息对面试官非认知特质评分的影响。
在南卡罗来纳医科大学医学院对申请者非认知特质面试评分的学术和人口统计学预测因素进行了为期两年的研究：1992年，面试官可获取申请者的医学院入学考试（MCAT）总成绩和本科平均绩点（GPA）；1993年，无法获取MCAT和GPA数据。1992年，226名申请者符合研究标准（即除了有MCAT和GPA存档数据外，他们还获得了三位面试官的评分）；1993年，245名申请者符合标准。进行逐步回归分析以测量七个自变量对申请者面试评分的影响。使用双向方差分析和t检验来确定申请者和面试官性别的影响。使用克朗巴哈系数作为面试官评分信度的指标。
GPA在这两年都是最佳预测因素，但1992年其在面试评分中所占方差比例（15.7%）是1993年（7.4%）的两倍。1992年面试官的信度系数为0.496，1993年为0.473。
如果医学院入学面试的目标是独立于学术技能评估非认知特质，作者建议在面试期间面试官不应获取MCAT和GPA数据。作者还发现性别和种族按照平权行动目标影响面试评分。最后，作者发现不同面试官的面试分数信度仅为中等。他们讨论了提高信度的方法。
