We provide here an explanation of our immediate conclusions.
What happened?
Human error caused the exam results that were released for the MRCP(UK) Part 2 exam sat on 6 September 2023 to be incorrect.
An error occurred when uploading the answer data from the digital exam provider to the MRCP(UK) database and then aligning it with the answer key held by MRCP(UK). Unfortunately, the software already contained some temporary values which should have been removed by MRCP(UK) but that did not occur.
All the data processes since 2021, when this system was started for online exams, have been reviewed and no other error has been found.
Each MRCP(UK) examination is standard-set to ensure that the passing standard of the exam is maintained for each cohort of candidates sitting the exam. This standard setting process uses statistical analysis to ensure that each version of the exam has the same level of difficulty. This process is called test equating and it ensures that each version of the exam has an equivalent level of difficulty, so all candidates are assessed at the same level.
The error described above affected not only the test equating process and derivation of the pass standard (pass mark) for this sitting of the exam but also the total number of marks awarded to each candidate.
As a result, out of 1451 candidates, the pass-fail classification for 283 was incorrect - 61 candidates who were told they had failed have passed and 222 candidates who were told they had passed have failed.
Why was it not detected at the time?
The September 2023 results did produce a somewhat higher than usual pass rate, particularly for UK graduates. This was questioned at the time. However, the subsequent investigation focussed on the distribution of candidate scores which was not found to be exceptionally unusual and the results were therefore permitted to stand.
Why has it been discovered now?
A Part 2 exam setting meeting in early 2025 reviewed the performance of anchor (marker) questions previously used in 2021. A discrepancy in some performance statistics for some questions was noticed, but, on investigation, this discrepancy had had no effect on the 2021 examination result. However, further exploration revealed the additional specific problem in September 2023, but in no other examination.
All of the details behind the error and the delay in recognition will be subject to independent external review.