Training experience across UK medicine: deanery differences versus site variation-General Medical Council National...

The General Medical Council's (GMC) National Training Survey (NTS) is widely used to benchmark postgraduate training experience in the UK. However, for actionable interpretation, it is necessary to distinguish deanery-level patterns from variation that is more locally specific to individual sites and specialities.

We analysed the 2025 GMC NTS post-speciality-by-site extract for medicine, benchmarked against all medical speciality posts. Each row represented an aggregated site × speciality × indicator cell. The primary metric was a gap score, defined as the site mean minus the national mean. Analyses followed a prespecified descriptive framework: deanery × indicator profiling, indicator-level comparisons with effect sizes and false discovery rate correction, intraclass correlation coefficients (ICCs), site-level correlation and volatility analyses, a suppression audit, outcome-category comparison, and principal component analysis (PCA)/cluster analysis.

The NTS data included 26,190 scored observations across 16 deaneries, 31 medical specialities, and 18 indicators. The mean deanery gap ranged from +3.23 in the North East to -3.13 in the East of England. The North East scored above the national average in 17 of 18 indicators, with the largest positive gaps in regional teaching (+7.34), rota design (+4.55), and reporting systems (+4.44). The East of England showed the largest negative gap in study leave (-10.17). After correcting for multiple testing, 17 of 18 indicators differed across deaneries; local teaching was the only indicator that did not differ. The largest deanery-level effect sizes were noted for regional teaching, clinical supervision out of hours, teamwork, and reporting systems. However, ICCs were modest overall, with the highest ICC observed for regional teaching (0.0973), indicating that a majority of the variance remained outside deanery-level clustering. Adequate experience correlated strongly with overall satisfaction (Spearman ρ = 0.848). Study leave showed the greatest site-to-site variation [standard deviation (SD) = 17.82]. Suppression for cells with fewer than three responses affected 21,060 of 47,442 rows (44.4%). Official NTS outcome categories also differed across deaneries (χ ² = 597.9, df = 60, p < 0.001).

Training experience in UK medicine varies across deaneries and between individual sites. The findings support a dual improvement approach-deanery-wide attention to domains with stronger regional patterning, such as regional teaching and out-of-hours clinical supervision, alongside site-level review of highly variable domains such as study leave and rota design. PCA clustering and the resilience index should be interpreted as hypothesis-generating. Interpretation is also limited by the use of aggregated survey data and substantial suppression due to small cell counts.