의과대학 중퇴에 미치는 영향 - 입학시 보는 시험(test at admission) vs 전적대학 성적

의과대학 중퇴에 미치는 영향 - 입학시 보는 시험 vs 전적대학 성적 

Medical school dropout - testing at admission versus selection by highest grades as predictors

Lotte O’Neill,1,2,3 Jan Hartvigsen,1,2,4 Birgitta Wallstedt,3 Lars Korsholm1 & Berit Eika5

배경

의과대학 중퇴자에 대한 입학시험(admission test)의 효과를 다룬 연구는 매우 극소수이다. 이 연구의 주 목적은 비성적기준 입학시험(non-grade based admission testing)과 성적 기준 입학(grade-based admission)이 차후 의과대학 중퇴자와 어떤 관계가 있는지 알아보고자 하는 것이다.

방법

이번 전향적 코호트 연구는 University of Southern Denmark에서 2002–2007년에 입학한 총 1544명으로 구성된 6개의 코호트를 대상으로 하였다. 절반의 입학생은 이전 학업 성적 성과(prior achivement of highest grades)로 입학하였으며 (전형 1), 나머지 반은 비성적기준 시험(non-grade-based admission test)를 통과해야 했다.(전형 2)

학생들에 대한 사회적 예측 변인들 (부모 중 의사가 있는지, 출신지, 부모 특성, 부모와 같이 사는지, 부모의 교육 수준은 어떠한지)도 같이 조사되었다. 최종적으로 보고자 하는 것은 학생이 입학 2년 후에 어느 정도나 중퇴를 하게 되는가였다. Multivariate logistic regression analysis를 사용하였다.

결과

'전형 2' 학생들이 2년 내에 중퇴할 상대위험(relative risk)이 더 낮았다(odds ratio 0.56, 95% confidence interval 0.39–0.80). 입학 전형과 Qualifying examination의 종류, 그리고 프로그램에 대한 우선순위가 최종 모델에 유의미하게 기여한 요인들이었다. 사회적 예측 변인들은 중퇴자 뿐만 아니라 전형 2 입학시험 점수도 예측하지 못했다.

결론

입학시험을 통해서 학생을 선발하는 것은 중퇴자에 대한 독립적, 보호적 효과가 있다.

---

Powis 등의 연구에 따르면 21개의 서로 다른 입학 면접의 subscale과 subscore를 의과대학 중퇴의 예측인자로 놓고 case-control study를 했을 때, 이 중 단 하나만이 중퇴와 상관관계가 있었다. (number of negative comments assigned by interviewers to the subscale for supportive and encouraging behaviour)

Powis et al. examined 21 different admission interview subscales and subscores as predictors of dropout in a case–control study, and found only one of these to be significantly associated with dropout.2 The number of negative comments assigned by interviewers to the subscale for supportive and encouraging behaviour was found to be associated with dropout (OR = 1.65, 95% confidence interval [CI] 1.01–2.70).2

Aust N Z J Med. 1992 Dec;22(6):692-8.

The structured interview as a tool for predicting premature withdrawal from medical school.

Powis DA, Waring TC, Bristow T, O'Connell DL.

Source

Faculty of Medicine, University of Newcastle, NSW.

반대로 Urlings-Strop 등은 비성적기준 입학시험(non-grade based admission test)에 대한 조금 더 긍정적인 사례를 보고했는데, 선발된(selected) 학생들이 추첨(lottery-admitted)으로 입학한 학생들보다 학교에 남을 가능성이(relative risk가) 2배 높았다. 이 연구는 또한 전적대학의 우수한 GPA로 입학한 세 번째 그룹(direct access)에 대한 언급도 했지만 direct access그룹과 선발된(selected) 그룹을 비교하지는 않았다.

By contrast, Urlings-Strop et al. presented a more optimistic case for the protective effect of non-grade-based admission tests on student dropout.3 They found that selected students were more than twice as likely as lottery-admitted control subjects to remain in school (relative risk [RR] 2.58, 95% CI 1.59–4.17; p = 0.000).3 This study also referred to a third admission group (‘direct access’) consisting of students with the highest pre-university grade point averages (GPAs), but did not report a comparison of dropout rates between the ‘direct access’ and ‘selected’ groups.3

 

Med Educ. 2009 Feb;43(2):175-83. doi: 10.1111/j.1365-2923.2008.03267.x.

Selection of medical students: a controlled experiment.

Urlings-Strop LC, Stijnen T, Themmen AP, Splinter TA.

Source

Institute of Medical Education and Research, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, The Netherlands. l.urlings-strop@erasmusmc.nl

---

Analysis

Descriptions of all variables delivered by the USD Admission Office and Statistics Denmark were scrutinised to check for changes in data collection methods. The variables set forth in the research protocol were either prepared for analysis (educational ⁄USD data) or generated by merging various datasets (social ⁄ Statistics Denmark variables). Missing data for the social variables generated from Statistics Denmark data were categorised with the nonevent ⁄ reference category and summary tables were produced.

Variables were then examined for co-linearity and zero cells before analyses, by inspection of matrix graph plots, 2 · 2 tables and boxplots. 

Individual predictors of dropout were then examined with univariate logistic regression analyses, and variables with p < 0.1 were eventually included in the multivariate models. 

Multivariate logistic regression was used to analyse the dichotomous outcome of dropout ⁄ non-dropout. Post-estimation diagnostics of models consisted mainly of checking linearity assumptions and influential data points. 

Additivity was assumed because we did not want to risk overfitting models by including interactions , due to the relatively large number of potential predictors and the modest number of dropouts. 

The linearity of the age variable was checked by inspecting LOWESS (locally weighted scatterplot smoothing) smoothed plots, with the logit transformed probability of dropout on the y-axis against age on the x-axis. 

Influential cases or cases for which the model fitted poorly were identified by inspection of deviance residuals, leverage and Pregibon’s delta-beta influence statistic.5 Influential cases were inspected to establish whether they were outliers on any of the predictor variablesin order to assess whether or not they should be removed from analysis.

---

Key findings

The aim of this study was to examine whether admission strategy was independently associated with dropout while controlling for relevant educational and socio-demographic variables. 

Participation in an admission test, general gymnasium exam and being enrolled on a first-priority programme were all independently associated with a reduced risk of dropout. 

By contrast, socio-demographic variables had little or no independent influence on dropout (Tables 2 and 3). This study is, to the best of our knowledge, the first published study to compare admission testing and pure grade-based admission on the outcome of dropout in medical education.

---

Med Educ. 2011 Nov;45(11):1111-20. doi: 10.1111/j.1365-2923.2011.04057.x.

Medical school dropout--testing at admission versus selection by highest grades as predictors.

O'Neill L, Hartvigsen J, Wallstedt B, Korsholm L, Eika B.

Source

Institute of Sports Science and Clinical Biomechanics, Faculty of Health Sciences, University of Southern Denmark, Odense, Denmark. ldyhrberg@health.sdu.dk

Abstract

CONTEXT:

Very few studies have reported on the effect of admission tests on medical school dropout. The main aim of this study was to evaluate the predictive validity of non-grade-based admission testing versus grade-based admission relative to subsequent dropout.

METHODS:

This prospective cohort study followed six cohorts of medical students admitted to the medical school at the University of Southern Denmark during 2002-2007 (n=1544). Half of the students were admitted based on their prior achievement of highest grades (Strategy 1) and the other half took a composite non-grade-based admission test (Strategy 2). Educational as well as social predictor variables (doctor-parent, origin, parenthood, parents living together, parent on benefit, university-educated parents) were also examined. The outcome of interest was students'dropout status at 2 years after admission. Multivariate logistic regression analysis was used to model dropout.

RESULTS:

Strategy 2 (admission test) students had a lower relative risk for dropping out of medical school within 2 years of admission (odds ratio 0.56, 95% confidence interval 0.39-0.80). Only the admission strategy, the type of qualifying examination and the priority given to the programme on the national application forms contributed significantly to the dropout model. Social variables did not predict dropout and neither did Strategy 2admission test scores.

CONCLUSIONS:

Selection by admission testing appeared to have an independent, protective effect on dropout in this setting.

© Blackwell Publishing Ltd 2011.

PMID:

 

21988626

 

[PubMed - indexed for MEDLINE]