OBJECTIVES: Primary care patient data are increasingly used for research and recording of death is important as it helps define end of follow-up and mortality may be a study outcome. This study compared recording of death in a UK primary care database with national death rates for England, Wales, Scotland and Northern Ireland (NI).  METHODS: The annual number of deaths (1990 to 2008) for each country was collected after the mortality recording quality threshold from The Health Improvement Network (THIN) database. THIN holds longitudinal anonymised primary care medical records from an increasing number of practices over time and currently contains data from more than 450 practices throughout the UK. Annual age and gender specific person time was estimated and multiplied with annual age, gender and country specific death rates to derive the expected number of deaths. Observed deaths divided by expected deaths provided the standardised mortality ratio (SMR) and Byar’s approximation formula was applied to derive 95% confidence intervals (CI). An SMR close to 1.0 indicates that recorded deaths are near national rates.  RESULTS: The average annual number of practices was 263.5 (standard deviation (SD):80.2) for England, 18.8 (SD:8.5) for Wales, 33.6 (SD:17.0) for Scotland, and 14.0 (SD:7.6) for NI. Average annual death rate per 1,000 population in THIN was 10.41 (SD:1.05) overall, 10.38 (SD:1.02) in England, 11.25 (SD:1.45) in Wales, 11.03 (SD:1.54) in Scotland and 8.29 (SD:0.77) in NI. The average annual SMR was 0.88 (SD:0.02) overall, 0.88 (SD:0.02) in England, 0.94 (SD:0.05) in Wales, 0.94 (SD:0.10) in Scotland and 0.91(SD:0.07) in NI.      CONCLUSIONS: The observed death rates were slightly lower than expected (SMR < 1) over time and in each country, especially in England. Social deprivation impacts death rates and as unadjusted for this could explain some variation. Reasons for lower rates need to be investigated further.