This study aims to identify the extent of terminal digit bias in routinely recorded blood pressures (BP) across a number of different general practices and report on changes in terminal digit bias over a 10-year period. It also explores the effect this may have had on the mean recorded BP in this population. BP records were taken from The Health Improvement Network database containing anonymized patient records from information entered by UK general practices in the financial years 1996-1997 to 2005-2006. The proportion of measurements ending in zero and the mean BP readings were calculated for each practice and for each year of data. Over this 10-year period the percentage of systolic BPs with zero terminal digits fell from 71.2 to 36.7% and mean recorded BP fell from 152.3 to 145.3 mm Hg. Correcting the BPs to remove terminal digit bias indicates a 2-3 mm Hg underestimation of the mean population systolic BP over this period. The between-practice variation in the percentage of zero terminal digit readings increased from 3.5 to 6.5 s.d. Although it is welcome to see a reduction in terminal digit bias, it is worrying to see the increase in variation between practices. There is evidence that terminal digit bias may lead to potential misclassification and inappropriate treatment of hypertensive patients. The increase in variation observed may therefore lead to an increased variation in the quality of care given to patients.