We describe a new method for active post-marketing surveillance of vaccine safety based on patient records. We studied the association between diphtheria/tetanus/pertussis (DTP) vaccination and febrile convulsion, and between measles/mumps/rubella (MMR) vaccination and febrile convulsion and idiopathic thrombocytopenic purpura (ITP) in five district health authorities in England by linking vaccination records with computerised hospital admission records. We found an increased relative incidence for convulsions 0-3 days after DTP vaccination. The effect was limited to the third dose of vaccine for which the attributable risk (all ages) was 1 in 12,500 doses. Completion of vaccination by 4 months instead of 10 months after the change in the UK to an accelerated immunisation schedule may have resulted in a 4-fold decrease in febrile convulsions attributable to DTP vaccine. 67% of admissions for a convulsion 6-11 days after MMR vaccination were attributable to the measles component of the vaccine (risk 1 in 3000 doses). An excess of admissions for a convulsion 15-35 days after MMR vaccination was found only for vaccines containing the Urabe mumps strain (1 in 2600 Urabe doses). There was a causal association between MMR vaccination and ITP resulting in admission 15-35 days subsequently; there was no evidence of a mumps strain-specific effect. The estimated absolute risk of 1 in 24,000 doses was 5 times that calculated from cases passively reported by clinicians. This finding emphasises the need for active surveillance of adverse events. The record linkage method that we used is an effective way to identify vaccine-attributable adverse events.