Comparing performance and integrity of proctored physical, unproctored online, and proctored online MBBS exams.

BACKGROUND: The rapid pivot to online assessment during COVID-19 raised concerns about the fairness and psychometric quality of high-stakes medical examinations. This study compared performance and integrity indices across three successive modalities of The University of the West Indies Final MBBS Medicine and Therapeutics written paper. METHODS: Institutional records for every candidate in 2019 (proctored physical, n = 508), 2020 (un-proctored online, n = 529) and 2021 (physically onsite proctored online, n = 361) were analyzed retrospectively (N = 1 398). Mean scores, score dispersion, categorical outcomes, Cronbach's alpha, and item-difficulty indices were contrasted with one-way ANOVA, χ2 tests, and odds ratios. RESULTS: Examination modality significantly influenced performance. The mean (± SD) scores were highest in the unproctored online exam (2020: 68.4 ± 6.5), followed by proctored physical exam (2019: 64.2 ± 7.2) and proctored online (2021: 60.5 ± 7.3). The difference in the mean score of the three modalities was highly significant (p < 0.0001). Pass rates peaked in 2020 (92%) versus 2019 (81%) and 2021 (70%). The odds of failing the examination in 2021 were more than five times higher than in 2020 (OR = 5.652). Reliability (Cronbach's α) highest in 2019 (0.813) and 2021 (0.770) but very low in 2020 (0.307). Odds of failing in 2021 were 5.65 times higher than 2020. CONCLUSIONS: Examination modality strongly impacted on performance and reliability. Unproctored online exams led to grade inflation, reduced variability, and low reliability, while proctored formats-physical or virtual-reinstated assessment rigor, fairness, and psychometric integrity, emphasizing the need for robust invigilation in high-stakes medical assessments.