Big data analysis: examination of the relationship between candidates' sociodemographic characteristics and performance in the UK's Membership of the Royal College of Physicians Part 1 examination.

Abstract

Big datasets and data analytics enable granular analyses examining group differences in performance. Our focus is on differential attainment (DA) in postgraduate College (Board) examinations. We asked: Are candidates' sociodemographic characteristics associated with performance on the UK's Membership of the Royal College of Physicians (MRCP) Part 1 after adjusting for medical school performance (MSP) and type of medical programme? This was a retrospective longitudinal cohort study of 6040 medical graduates with linked sociodemographic data in the UK Medical Education Database qualifying from a UK medical school (2012-2014) and sitting MRCP Part 1 before October 2019. Chi-squared tests established univariable associations with MRCP performance (pass/fail first sitting MRCP Part 1). Multivariable mixed-effects logistic regression identified independent explanatory factors of success, adjusted for medical school. The odds (95% CI) of passing MRCP Part 1 exams on first sitting were greater for men (OR = 1.61, CI 1.42-1.81, p < 0.001) and those on a graduate entry programme (OR = 1.44, 1.05-1.99, p < 0.001). The odds of passing were lower as age increases (OR = 0.87, 0.85-0.90, p < 0.001), for minority ethnic (OR = 0.61, CI 0.53-0.7, p < 0.001), and gateway to medicine (OR = 0.49, CI 0.27-0.90, p = 0.02) candidates. After adjusting for MSP, odds were greater for passing in men (OR = 1.62, CI 1.24-2.11, p < 0.001) and candidates with higher MSP (OR = 4.12, CI 3.40-4.96, p < 0.001). Our findings illustrate how performance on MRCP part 1 is associated with group-level social and educational factors. This DA may be due to aspects of the assessment itself, and/or the persistent nature of social and educational disadvantage.