OBJECTIVE--To develop and report the results of a system of audit of computer records in general practice. DESIGN--A retrospective audit of records in practices using the same computer system. Information about recorded preventive procedures was collected by sending the same audit program to each practice on floppy disk. Other characteristics of the practices were determined by postal questionnaire. SETTING--Forty five general practices, widely distributed in England and Wales. SUBJECTS--All 430,901 patients registered with the practices. MAIN OUTCOME MEASURES--Within each practice the percentage of patients in specified age groups for whom certain preventive procedures were recorded as having been carried out. These measures were analysed in relation to practice characteristics. RESULTS--Practice characteristics and recording rates for preventive procedures varied over a wide range. Recording rates were higher in practices with computer terminals on every doctor's desk. Only one practice achieved the new contract target of 90% coverage for recorded primary immunisations, and fewer than two thirds recorded 80% coverage for cervical cytology in the past five years. Practices holding clinics did no better than those without. Smaller partnerships and smaller doctors' list sizes were associated with better performance. CONCLUSIONS--Centrally programmed audit of computerised records is a feasible method of providing data on a regular basis for epidemiological purposes and for performance review. The fact that practices with smaller list sizes had higher levels of recorded preventive care suggests that the trend towards larger lists promoted by the new contract might militate against the intended effect of better preventive care.