Predictive validity of A-level grades and teacher-predicted grades in UK medical school applicants: a retrospective analysis of administrative data in a time of COVID-19

Establishing the predictive validity of the intercollegiate membership of the Royal Colleges of surgeons written examination: MRCS Part A

Successful completion of the Intercollegiate Membership of the Royal Colleges of Surgeons (MRCS) examination is mandatory for surgical trainees entering higher specialist training in the United Kingdom. Despite its international reputation, and the value placed on the examination in surgical training, there has been little evidence of its predictive validity until recently. In this review, we present a summary of findings of four recent Intercollegiate studies assessing the predictive validity of the MRCS Part A (written) examination. Data from all four studies showed statistically significant positive correlations between the MRCS Part A and other written examinations taken by surgical trainees over the course of their education. The studies summarised in this review provide compelling evidence for the predictive validity of this gatekeeping examination. This review will be of interest to trainees, training institutions and the Royal Colleges given the value placed on the examination by surgical training programmes.

Establishing the predictive validity of the intercollegiate membership of the Royal Colleges of surgeons written examination: MRCS part B

The Intercollegiate Membership of the Royal Colleges of Surgeons (MRCS) is a high-stakes postgraduate examination taken by thousands of surgical trainees worldwide every year. The MRCS is a challenging assessment, highly regarded by surgical training programmes and valued as a gatekeeper to the surgical profession. The examination is taken at considerable personal, social and financial cost to surgical trainees, and failure has significant implications for career progression. Given the value placed on MRCS, it must be a reliable and valid assessment of the knowledge and skills of early-career surgeons. Our first article 'Establishing the Predictive Validity of the Intercollegiate Membership of the Royal Colleges of Surgeons Written Examination: MRCS Part A' discussed the principles of assessment reliability and validity and outlined the mounting evidence supporting the predictive validity of the MRCS Part A (the multiple-choice questionnaire component of the examination). This, the second article in the series discusses six recently published studies investigating the predictive validity of the MRCS Part B (the clinical component of the examination). All national longitudinal cohort studies reviewed have demonstrated significant correlations between MRCS Part B and other assessments taken during the UK surgical training pathway, supporting the predictive validity of MRCS Part B. This review will be of interest to trainees, trainers and Royal Colleges given the value placed on the examination by surgical training programmes.

A case study of using AI for General Certificate of Secondary Education (GCSE) grade prediction in a selective independent school in England

The COVID-19 pandemic has created significant challenges for UK schools, but a time of cancelled exams and uncertainty around future examinations can provide opportunities to explore novel assessment methods. Hence, the 2020 proposal of the Ofqual algorithm which combines teachers' estimated grades and schools' historical performance seemed timely. However, the algorithmically calculated grades resulted in a public backlash and withdrawal of the proposal. While the failed Ofqual algorithm could be considered an example of AI, we do not yet have a thorough understanding of its numerical accuracy and how it performs in comparison to other AI models. This paper investigates this novel application: the potential use of a range of AI models as assessment tools in a selective, independent, secondary school in England. The following questions were examined: (1) how accurate are modern AI models in predicting GCSE exam grades? (2) what are the differences in model accuracy across subjects and can these be explained by qualitative differences in teachers' grading practices? Results indicate that while models yield acceptable mean absolute errors, individual mispredictions can be larger than desired. Subject differences highlighted that grading subjectivity is less significant in science, technology, engineering, and maths (STEM) subjects, which could explain why objective models fail to predict non-STEM grades more frequently. In summary, numerical results indicate that grade prediction could be an interesting novel application of AI, but more research is needed to reduce outliers.

BMAT's predictive validity for medical school performance: A retrospective cohort study

BACKGROUND

Although used widely, there is limited evidence of the BioMedical Admissions Test's (BMAT) predictive validity and incremental validity over prior educational attainment (PEA). We investigated BMAT's predictive and incremental validity for performance in two undergraduate medical schools, Imperial College School of Medicine (ICSM), UK, and Lee Kong Chian School of Medicine (LKCMedicine), Singapore. Our secondary goal was to compare the evidence collected with published evidence relating to comparable tools.

METHODS

This was a retrospective cohort study of four ICSM (1188 students, entering 2010-2013) and three LKCMedicine cohorts (222 students, 2013-2015). We investigated associations between BMAT Section 1 ('Thinking Skills'), Section 2 ('Scientific Knowledge and Applications') and Section 3a ('Writing Task') scores, with written and clinical assessment performance across all programme years. Incremental validity was investigated over PEA (A-levels) in a subset of ICSM students.

RESULTS

When BMAT sections were investigated independently, Section 2 scores predicted performance on all written assessments in both institutions with mainly small effect sizes (standardised coefficient ranges: ICSM: 0.08-0.19; LKCMedicine: 0.22-0.36). Section 1 scores predicted Years 5 and 6 written assessment performance at ICSM (0.09-0.14) but nothing at LKCMedicine. Section 3a scores only predicted Year 5 clinical assessment performance at ICSM with a coefficient <0.1. There were no positive associations with standardised coefficients >0.1 between BMAT performance and clinical assessment performance. Multivariable regressions confirmed that Section 2 scores were the most predictive. We found no clear evidence of incremental validity for any BMAT section scores over A-level grades.

DISCUSSION

Schools who wish to assess scientific knowledge independently of A-levels may find BMAT Section 2 useful. Comparison with previous studies indicates that, overall, BMAT seems less useful than comparable tools. Larger scale studies are needed. Broader questions regarding why institutions adopt certain admissions tests, including those with little evidence, need consideration.

Can achievement at medical admission tests predict future performance in postgraduate clinical assessments? A UK-based national cohort study

OBJECTIVE

To determine whether scores on two undergraduate admissions tests (BioMedical Admissions Test (BMAT) and University Clinical Aptitude Test (UCAT)) predict performance on the postgraduate Membership of the Royal Colleges of Physicians (MRCP) examination, including the clinical examination Practical Assessment of Clinical Examination Skills (PACES).

DESIGN

National cohort study.

SETTING

Doctors who graduated medical school between 2006 and 2018.

PARTICIPANTS

3045 doctors who had sat BMAT, UCAT and the MRCP.

PRIMARY OUTCOME MEASURES

Passing each section of the MRCP at the first attempt, including the clinical assessment PACES.

RESULTS

Several BMAT and UCAT subtest scores displayed incremental predictive validity for performance on the first two (written) parts of the MRCP. Only aptitude and skills on BMAT (OR 1.34, 1.08 to 1.67, p=0.01) and verbal reasoning on UCAT (OR 1.34, 1.04 to 1.71, p=0.02) incrementally predicted passing PACES at the first attempt.

CONCLUSIONS

Our results imply that the abilities assessed by aptitude and skills and verbal reasoning may be the most important cognitive attributes, of those routinely assessed at selection, for predicting future clinical performance. Selectors may wish to consider placing particular weight on scales assessing these attributes if they wish to select applicants likely to become more competent clinicians. These results are potentially relevant in an international context too, since many admission tests used globally, such as the Medical College Admission Test, assess similar abilities.